Theory and Objectives of Air Dispersion Modelling | Air Pollution | Wound

Modelling Air Emissions for Compliance MME 474A Wind Engineering

December 2003
THEORY AND OBJECTIVES OF AIR DISPERSION MODELLING


Robert Macdonald, Ph.D., P.Eng.
Department of Mechanical Engineering
University of Waterloo
Waterloo, ON N2L 3G1


1. Introduction and Objectives

Air pollution models are routinely used in environmental impact assessments, risk analysis and
emergency planning, and source apportionment studies. In highly polluted cities such as Athens,
Los Angeles and Mexico, regional scale air quality models are used to forecast air pollution
episodes – the results from these models may initiate compulsory shutdown of industries or
vehicle restrictions. The various roles served by air pollution models, which cover a broad range
of scales from local to global, lead to distinct modelling requirements. The focus of this review
will be on the near-field impact (< 10-20 km) of industrial sources. The emphasis is on
Gaussian-plume type models for continuous releases, which are at the core of most U.S.
Environmental Protection Agency (EPA) regulatory models.

Nowadays the term “air pollution model” usually refers to a computer program, but in the past it
has also included hand calculations or use of charts and tables from simple handbooks. A
dispersion model is essentially a computational procedure for predicting concentrations
downwind of a pollutant source, based on knowledge of the emissions characteristics (stack exit
velocity, plume temperature, stack diameter, etc.), terrain (surface roughness, local topography,
nearby buildings) and state of the atmosphere (wind speed, stability, mixing height, etc.). Figure
1 illustrates the flow of information in a generic air pollution model. The basic problem is to
predict the rate of spread of the pollutant cloud, and the consequent decrease in mean
concentration. The model has to be able to predict rates of diffusion based on measurable
meteorological variables such as wind speed, atmospheric turbulence, and thermodynamic
effects. The algorithms at the core of air pollution models are based upon mathematical
equations describing these various phenomena which, when combined with empirical (field)
data, can be used to predict concentration distributions downwind of a source.

The modern science of air pollution modelling began in the 1920’s when military scientists in
England tried to estimate the dispersion of toxic chemical agents released in the battlefield under
various conditions. This early research is summarized in the groundbreaking textbook by Sutton
(1953). Rapid developments in the 1950’s and 1960’s, including major field studies and
advances in the understanding of the structure of the atmosphere, led to the development of the
first regulatory air pollution models in the U.S. The textbooks by Pasquill (1974) and Stern
(1976) review much of the research and theory up until the mid 1970’s. However, the
proliferation of air pollution research and models to date has made it necessary to read
specialized journals and conference proceedings to keep up with developments. This is not
practical for all model users, and so the present workshop has been designed to help bridge the
gap between the basic concepts of dispersion theory and the sophisticated theories used in
advanced USEPA models such as ISC3 and AERMOD-PRIME. This paper reviews the
1
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
fundamentals of Gaussian plume modelling as normally presented in an undergraduate air
pollution course. Many of the key concepts and algorithms incorporated into advanced air
dispersion models are briefly explained.

2. Model Requirements and Model Selection

There are several competing requirements in the design of an air pollution model. A model must
capture the essential physics of the dispersion process and provide reasonable and repeatable
estimates of downwind concentrations. This generally requires detailed knowledge of source
characteristics, terrain and meteorology, but it is also desirable to keep these input requirements
to a minimum, and simplicity is an important asset in any model. All models should have a fully
documented account of the equation algorithms used and their conversion into valid software
(i.e., traceability). Regulatory models must also undergo extensive quality assurance, including
the evaluation of the model under several scenarios using benchmark data. Standard statistical
procedures have been developed for expressing the uncertainty and variability of the predicted
results when comparing them to measured concentrations (e.g., Hanna, 1989).

In choosing an air dispersion model, several levels of model are available, with progressively
increasing levels of mathematical sophistication, input data requirements and user expertise
required. At the low end of the scale are the gross screening models, which require only a
hand-held calculator, nomograph, or spreadsheet. They may treat only one source at a time (e.g.,
a single elevated stack) and provide some sort of worst-case prediction based on relatively
primitive meteorological information. It is often wise to apply such a model prior to using the
more advanced models, where the flow of information is more difficult to follow.

Next on the scale of model complexity are intermediate models, usually PC-based, which may
include varying meteorology (wind speed and stability classes) and more sophisticated source
information. Many early EPA models fall in this category, including the SCREEN3 model.

Advanced models require a desktop PC or workstation. They require extensive data sets for
meteorology and emissions, and include multiple source types - point, area and volume. They
may also include additional features such as complex terrain, flow around buildings and layered
atmospheric structure. Modern models incorporate the most up-to-date treatment of the
atmosphere such as Monin-Obukhov similarity theory. Some examples of advanced models are
the EPA models ISC3, AERMOD and CALPUFF, the British Model ADMS (Carruthers et al.,
1994) and the Danish model OML (Berkowicz et al., 1987)

Specialized models are often used for predicting dispersion of special hazardous materials, such
as military models used in chemical/biological defense. Heavy gas dispersion models are used
by the chemical process industries to model the behaviour of rogue or accidental releases of
dense gases or vapours. These models may require extensive thermodynamic information to
account for release conditions. Models such as SLAB and DEGADIS (Dense Gas Dispersion
Model) models are typical of this family.

Although the input data requirements and level of sophistication increase with the more
advanced models, a more complex model does not necessarily lead to predictions that are more
2
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
accurate. As the number of input variables goes up in the advanced models, the room for input
data error increases. In addition, the level of user understanding must increase to make proper
use of the model.

3. Example Gross Screening Analysis

It is often useful to perform a simple screening analysis before applying a more refined computer
analysis. A gross screening analysis will quickly identify the order of magnitude of the expected
concentrations and may even show that no problem exists, in which case more advanced
modelling is unnecessary.

A useful formula for estimating worst case mean concentrations downwind of a point source is
the following equation suggested by Hanna et al. (1996):

9
10
wc
wc wc
Q
C
UH W
= (1)
where:
Q = source strength or emission rate of gas or particulate [kg/s]
C
wc
= worst case concentration [µg/m
3
]
U = worst case wind speed at height z = 10 m, usually 1 m/s
W
wc
= worst case cloud width [m]
(usually assume W = 0.1x, where x is distance from the source)
H
wc
= worst case cloud depth
(usually assume H = 50 m in worst case)

This equation is essentially a statement of the conservation of pollutant mass, but it illustrates
many of the basic parameter dependencies in dispersion modeling. Referring to Figure 2, we
assume a uniform concentration in the plume passing through the downwind plane HW.
Equation (1) follows from the fact that the flux of pollutant through any plane must equal the
source rate Q. Equation (1) illustrates several important dependencies that should be satisfied by
all plume models:
1. The mean concentration is inversely proportional to mean wind speed.
2. The mean concentration is directly proportional to the release rate.
3. The mean concentration is inversely proportional to the plume cross-sectional area.

As an example of the above, suppose a small amount (1-kg) of ammonia is released over a period
of 30 minutes in an accidental release. Assuming a light wind of 1 m/s, does this release pose
any risk to the occupants of a hospital located 5 km downwind? For this example the estimate of
the plume width is W
wc
= 0.1 × 5000 = 500 m, thus,

9 9
3
min min
10 (10 / ) (1 /1800 )
22.2
1 / 50 500
wc
Q g kg kg s g
C
UH W m s m m m
µ µ ×
= = =
× ×


3
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
This concentration is equivalent to 0.032 ppm, and is 1500 times below the personal exposure
limit (PEL) associated with negative health effects due to prolonged exposure to ammonia.
Therefore, we can safely say that there is no risk. In such a case there is also no need to perform
advanced modelling to assess the risk.

4. The Diffusion Equation and the Gaussian Plume Model

By performing a mass balance on a small control volume, a simplified diffusion equation, which
describes a continuous cloud of material dispersing in a turbulent flow, can be written as:

y z
dC dC d dC d dC
U K K
dt dx dy dy dz dz
| |
| |
+ = + +
| |
\ .
\ .
S (2)

where: x = along-wind coordinate measured in wind direction from the source
y = cross-wind coordinate direction
z = vertical coordinate measured from the ground
C(x,y,z) = mean concentration of diffusing substance at a point (x,y,z) [kg/m
3
]
K
y
, K
z
= eddy diffusivities in the direction of the y- and z- axes [m
2
/s]
U = mean wind velocity along the x-axis [m/s]
S = source/sink term [kg/m
3
-s]

Analytical solutions to this equation for the case of dispersion of passive pollutants in a turbulent
flow were first obtained in the 1920’s by Roberts (1923) and Richardson (1926). The eddy
diffusivities (K
y
and K
z
) are a way of relating the turbulent fluxes of material to the mean
gradients of concentration:
' ' , ' '
y
C C
v c K w c K
y z
∂ ∂
= − = −
∂ ∂
z
z
. (3)

Here primed coordinates refer to the turbulent fluctuations of terms about their mean values; for
eample, etc. Typically in the atmosphere ( ) ', ( ) ', c t C c u t U u = + = +
y
K K > , which
explains why the cross-section of a plume often takes on an elliptic shape.

A term-by-term interpretation of Equation (2) is as follows:

dC dC
U
dt dx
+ Time rate of change and advection of the cloud by the mean wind.

,
y
d dC
. K etc
dy dy
| |
|
\ .
Turbulent diffusion of material relative to the center of the pollutant cloud.
(The cloud will expand over time due to these terms.)

4
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
S Source term which represents the net production (or destruction) of
pollutant due to sources (or removal mechanisms).

Equation (2) is grossly simplified, since several assumptions are made in its derivation:

1. The pollutant concentrations do not affect the flow field (passive dispersion).
2. Molecular diffusion and longitudinal (along-wind) diffusion are negligible.
3. The flow is incompressible.
4. The wind velocities and concentrations can be decomposed into a mean and
fluctuating component with the average value of the fluctuating (stochastic)
component equal to zero. Mean values are based on time averages of 10-60 minutes.
5. The turbulent fluxes are linearly related to the gradients of the mean concentrations as
in Equation (3).
6. The mean lateral and vertical wind velocities V and W are zero, so we have also
restricted our analysis to steady wind flow over an idealized flat terrain.

The Gaussian plume model, which is at the core of almost all regulatory dispersion models, is
obtained from the analytical solution to Equation (2). For a continuous point source released at
the origin in a uniform (homogeneous) turbulent flow the solution to Equation (2) is:

2 2
( , , ) exp exp
4 ( / ) 4 ( / ) 4
y z y z
Q y
C x y z
z
K x U K x U x K K π
| |
| | − −
=
|
|
|
\ .
\ .
. (4)

Unfortunately, the turbulent diffusivities K
y
and K
z
are unknown in most flows, and in the
atmospheric boundary layer K
z
is not constant, but increases with height above the ground. In
addition, K
y
and K
z
increase with distance from the source, because the diffusion is affected by
different scales of turbulence in the atmosphere as the plume grows (Figure 3). Despite these
limitations, the general Gaussian shape of Equation (4) is often. If we define the following
Gaussian parameters:

2
y y z z
2
x x
K and K
U U
σ σ = = . (5)

then the final form of the Guassian plume equation, for an elevated plume released at z = H
p
is:

2 2
2
2 2
( ) ( )
( , , ) exp exp exp
2 2 2 2
p p
p y z y z z
z H z H
Q y
C x y z
U π σ σ σ σ σ
2
( | | | | | − +
= − − + −
|
( | |
| |
|
|
(
\ . \ \ . .
¸ ¸
(6)

In this expression a second z-exponential term has been added to account for the fact that
pollutant cannot diffuse downward through the ground at z = 0. This “image” term can be
visualized as an equivalent source located at z = -H
p
below the ground, and is further discussed
in Section 6 below.

5
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
Equation (6) is the well-known Gaussian plume equation for a continuous point source (Turner,
1994). The plume height H
p
is the sum of the actual stack height H
s
plus any plume rise ∆H due
to initial buoyancy and momentum of the release (Figure 4). The wind speed U
p
is taken to be
the mean wind speed at the height of the stack. Since we are normally interested in
concentrations at the ground (where the receptors such as people are), we set z = 0 to obtain,

2
2
2
( , , 0) exp exp
2 2
p
p y z y z
H
Q y
C x y z
U
2
π σ σ σ σ
| | | |
= = − −
|
|
\ . \ .
|
|
(7)

If we furthermore set H
p
= 0 we get the vertical distribution due to a ground-level source:

2 2
2
( , , ) exp exp
2 2
y z y z
Q y
C x y z
U
2
z
π σ σ σ σ
| |
| |
= −
|

|
\ .
\ .

|
, (8)

It turns out this latter model is not very accurate. For a ground-level release, the vertical profile
varies more like ex , rather than the Gaussian form e
1.5
p( ) z −
2
xp( ) z − , due to the large vertical
variations of the diffusivity K
z
near the ground (van Ulden, 1978). A more general, non-
Gaussian model, which allows for the vertical variation of K
z
and the vertical variation of the
velocity profile can be written as,

2
2
( , , ) exp ( )
2 2
y p y
Q y
C x y z f z
U σ π σ
| |
= −
|
|
\ .
, (9)

Here f(z) is a normalized function which describes the vertical distribution of material in the
plume. To ensure conservation of mass, such models must satisfy:


, y z
CUdydz Q =
∫∫
, (10)
where the integration is taken over the y-z plane, perpendicular to the plume axis.

5. Determination of the Gaussian Plume Parameters σ
y
and σ
z


In order to evaluate equations (6) and (7), we require the Gaussian plume parameters σ
y
and σ
z
.
These

have been measured as a function of distance from the source in numerous field studies.
The most common tabulated data are the Pasquill-Gifford sigmas, which are based primarily on
the Project Prairie Grass study in the U.S. (Barad, 1958). The P-G sigmas were formulated by
Pasquill (1961) and Gifford (1961) for low-level releases over relatively smooth terrain at
distances of a few thousand meters from a source.

The plume parameters σ
y
and σ
z
are driven by atmospheric turbulence and are influenced by the
state of convection in the atmosphere. Atmospheric turbulence is greatly enhanced by
6
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
convective motions due to heating of the earth’s surface. On clear nights, it is suppressed by
cooling of the ground due to outgoing radiation. In order to relate the state of atmospheric
convection to simply observable parameters, Pasquill developed a simple quantitative rating
scheme consisting of six stability classes ranging from highly convective [A] to highly stable
flow conditions [F]. These classes are summarized in Table I. The resulting Pasquill-Gifford (P-
G) σ
y
and σ
z
curves under varying conditions of stability are shown in Figure 5.

Table I. Relationship Between Wind Speed, Pasquill-Gifford Stability Class and Monin-
Obukhov Length [Hanna, Drivas and Chang, 1996]

Description P-G Stability
Class
Time of
Day/Condition
Wind Speed U M-O Length
L
MO
Very Unstable A Sunny Day < 3 m/s -10 m
Unstable B or C ↓ 2-6 m/s -50 m
Neutral D Cloudy or Windy > 3-4 m/s |L| > 100 m
Stable E ↓ 2-4 m/s + 50 m
Very Stable F Clear Night < 3 m/s +10 m


For use in regulatory Gaussian plume models, analytic expressions have been fitted to the
standard P-G sigma curves. In Appendix A gives the algebraic equations used to calculate σ
y

and σ
z
in the ISC3 model are provided. These sigma data can be applied for releases over flat,
rural terrain. However, dispersion in the urban environment usually produces greater rates of
spread than these field data expressions. For urban dispersion, a second set of curves sometimes
called the McElroy-Pooler (1968) sigmas, based on tracer releases in a large U.S. city, are used.
These are incorporated in the SCREEN3 and ISC3 models when the user selects the urban terrain
option in running the software.

The grouping of atmospheric stability into six discrete classes is done in most simple regulatory
dispersion models. However, if more detailed information is available, such as directly
measured wind velocity fluctuations, then it is possible to relate the plume sigmas directly to
these turbulent fluctuations using the statistical theory of diffusion. For example, Draxler (1976)
provides the following relationships:


v
y y y
x f x
U
θ
f
σ
σ = ⋅ = ⋅ σ (11a)

w
z z z
x f x
U
φ
f
σ
σ = ⋅ = ⋅ σ , (11b)

where σ
v
and σ
w
are the root-mean-square fluctuations in transverse and vertical velocities (v
and w), and σ
θ
and σ
ϕ
are the standard deviations of the wind vector azimuth and elevation
angles (in radians). The distance x is measured from the source. The functions f
y
and f
z
are unity
7
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
close to the source but are decreasing functions of x. Thus, if actual measurements of the wind
velocity fluctuations are available, σ
y
and σ
z
can be calculated directly from (11a) and (11b). An
approach similar to this is used in the AERMOD dispersion model to calculate the dispersion
parameters from atmospheric turbulence (Cimorelli et al., 1998).

6. Plume Reflection at the Ground and at Elevated Inversions

When a plume is discharged from an elevated stack, it will spread vertically until its lower edge
reaches the ground. Until this happens, the term:
2
exp( ( ) / 2 )
p
z H
2
z
σ − + in equation (6), does
not make a significant contribution to concentrations above the ground. However, as σ
z

increases, the plume will eventually be reflected at the ground. The virtual source at z = -H
p
will
then begin to contribute to the aboveground concentrations.

Similarly, if a strong inversion layer is located at some height z
i
above the stack, then the plume
will have difficulty expanding vertically and will effectively be “trapped” between the inversion
and the ground. Plume reflection will occur in this case at both the ground and the inversion
layer. To account for the initial reflection at the inversion, an image source is placed at
z = 2z
i
−Hp as illustrated in Figure 6. This image source will also be reflected at the ground (to
ensure ∂C/∂z|
z = 0
= 0), so a further image source is placed at z = 2z
i
− H
p
below the ground.
Similarly, the plume image source at z = H
p
below the ground requires an image source above
the inversion at z = 2z
i
+ H
p
to ensure ∂C/∂z|
z = zi
= 0. This process of reflecting image sources
can go on indefinitely, but only the first few terms are usually required in the vertical distribution
function. The vertical distribution will be non-Gaussian, and we can write:

2
2
( , , ) exp ( , , , )
2 2
p i z
p y z y
Q y
C x y z f z H z
U
σ
π σ σ σ
| |
= −
|
|
\ .
, (12)
where:
2 2
2 2
2 2
2
2 2
2 2
( ) ( )
( , , , ) exp exp
2 2
( 2 ) ( 2 )
exp exp
2 2
( 2 ) ( 2 )
exp exp
2 2
p p
p i z
z z
i p i p
z z
i p i p
z z
z H z H
f z H z
z z H z z H
z z H z z H
σ
σ σ
σ σ
σ σ
(
| | | − +
= − + −
( |
|
(
\ . \
¸ ¸
(
| | | | − + + −
+ − + −
( |
|
(
\ . \ .
¸ ¸
(
| | | − − + +
+ − + −
( |
|
(
\ . \
¸ ¸
2
|
|
|
.
|
|
|
|
|
.
(13)

In practice, the vertical concentration profile eventually becomes uniformly distributed
throughout the mixing layer after which the concentration can be approximated by (Turner,
1994):

8
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
2
2
( , , ) exp
2 2
y p y i
Q y
C x y z
U z σ π σ
| |
=

\ .

|
|
p
(13)

This gives reasonable results when σ
z
≥ z
i
. This formula also describes the distribution of
material during a fumigation episode, which is the brief period when an elevated plume initially
above an inversion is mixed downward by convective turbulence as the depth of the mixed layer
reaches the height of the plume due to heating at the ground (Turner, 1994).

A buoyant plume is never perfectly reflected by an inversion layer and partial plume trapping is
allowed in more advanced models such as AERMOD, where only a fraction f
p
of the plume
(where f
p
≤ 1) remains in the convective boundary layer (CBL). The remaining fraction (1-f
p
) is
allowed to penetrate the inversion and to escape temporarily, but eventually reappears in the
CBL if convective conditions raise the inversion cap and diffuse the mass that has escaped back
to the ground. The total concentration C
c
at a receptor is then given by:

C C
c d r
C C = + + (14)
where:
C
d
= direct source concentration contribution (due to downward dispersion of
material from the stack)
C
r
= indirect source concentration contribution (due to primary image source
above z
i,
. AERMOD includes a delay to mimic the lofting behaviour of a
buoyant plume)
C
p
= penetrated source concentration distribution, for material that initially
penetrates the elevated version

Each of these terms has associated image sources to satisfy the symmetry boundary conditions
(∂C/∂z = 0) at z = 0 and z = z
i
. Further details on this approach are found in the AERMOD
model formulation document by Cimorelli et al. (1998). During stable conditions, the plume is
modelled using only the direct source contribution and reflection at the ground, without
“trapping”. AERMOD is one of the few models that allows for non-Gaussian distributions of
plume material in the vertical.

6. The Plume Advection Velocity and the Wind Speed Velocity Profile

In the lowest part of the earth’s boundary layer (the surface layer), wind speed increases with
increasing height and has strong gradients near the ground. In homogeneous terrain, under
conditions of neutral atmospheric stability, the wind speed is found to vary logarithmically with
height (Panofsky and Dutton, 1984):


*
0
( ) ln
u z
z κ
| |
=

\ .
U z (15)
|

9
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
The friction velocity u
*
is related to the frictional resistance that the ground exerts on the wind.
It is typically about 10% of the wind speed at z = 10 m. The surface roughness length z
0
is a
measure of the aerodynamic roughness of the ground, and is typically 3-10% of the height of the
surface obstacles (trees, houses, crops, etc.). The von Karman constant κ is about 0.4.

The logarithmic wind profile is not easy to work with for dispersion calculations. It is often
approximated by a more simple power law of the form,

10
( )
10
p
z
U
| |
=
|
\ .
U z , (16)
The power law coefficient p increases with increasing surface roughness. For different types of
terrain, Table II gives the approximate roughness height z
0
and profile exponent p. Some typical
wind profiles are shown schematically in Figure 7. Equation (15) is valid only for neutral
stability conditions. The exponent p increases dramatically with increasing atmospheric stability
Irwin, 1979). The effect of stability on the power law exponent is shown in Appendix B.


Table II. Surface Roughness and Power Exponent for Wind Flow
Over Various Terrains in Neutral Stability
Terrain z
0
[m] p
Lake or smooth sea 10
-4
0.07
Sandy desert 10
-3
0.10
Short grass 0.005 0.13
Open grassland 0.02 0.15
Root crops 0.1 0.2
Agricultural areas 0.2-0.3 0.24-0.26
Parkland/Residential Areas 0.5 0.3
Large Forest/Cities 1.0 0.39


Since windspeed varies with height, it is not obvious what advection speed one should assume
when using the Gaussian plume model. In practice, for an elevated source one usually takes U
p

to be the mean wind speed at the stack height. Equations (15) or (16) can be used to calculate
this velocity from the standard wind speed U
10
at 10m height. For a ground level source, one
typically uses the wind speed U
10
for the plume advection velocity, although a more precise
plume advection velocity for a ground source is given by (van Ulden, 1978):

*
0
0.6
ln
z
p
u
z
U
σ
κ
| |
=

\ .
|
(17)

This is just the log-law wind speed evaluated at a height z = 0.6σ
z
. Because of the dependence
on σ
z
, it can be seen that as the plume grows vertically, it moves at progressively higher speeds.



10
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
6.1 Effect of Atmospheric Stability on the Mean Wind Profile

Along with increased turbulence, one of the effects of atmospheric convection is to modify the
shape of the mean velocity profile. A recommended velocity profile function for use in stable
and near-neutral conditions is given by (Hanna et al., 1996)

*
0
( ) ln 4.5
0.4
MO
u z z
U z
z L
(
| |
= +
|
\ .
¸ ¸
(
. (18)

Here L
MO
is the Monin-Obukhov length scale, defined as:


3
3
*
*
0.4 ' '
0.4
p a
MO
f
a
c T u
u
L
gH w T
g
T
ρ −
= =

(19)
where: ' '
f p
H c w T ρ = = vertical heat flux from ground [W/m
2
]
T
a
= absolute air temperature [
0
K]
g = gravitational constant [9.8 m/s
2
]
c
p
= heat capacity of air [287 J/kg-
0
K]

Typical values of L
MO
in various stability conditions are listed in Table I. In a stable atmosphere,
z = L
MO
defines the altitude above which the mechanical production of turbulence is suppressed
through the action of negative buoyancy. Thus, when z < L
MO
, the mechanical generation of
turbulence is dominant. In a convective atmosphere (L
MO
< 0), /
MO
z L is a measure of the ratio
of turbulence production by convection to mechanical production of turbulence. The height
MO
L then defines the boundary between flow levels where mechanical turbulence due to
friction dominates (
MO
z L < ) and levels where convective turbulence dominates the flow
(
MO
z L > ).

Equation (18) can also be approximated by a power law, with the recognition that in this case the
exponent p will be a function of z
0
and L
MO
. Typical exponents are given in Appendix B.

The type of wind profile given in equation (18) is used in the advanced AERMOD dispersion
model. AERMOD has a meteorological preprocessor called AERMET which calculates the
values of u
*
and L
MO
for a given flow. These are then used in calculating mixing height,
temperature and velocity profiles in a convective (CBL) or stable (SBL) atmospheric boundary
layer. These factors are then used in the determination of the Gaussian plume parameters σ
y
and
σ
z
. This is a much more sophisticated approach then the traditional methods using charts which
only allow for discrete stability classes and roughness categories. The use of Monin-Obukhov
scaling of meteorological variables is one of the criteria which differentiates these “advanced”
dispersion models from their predecessors.

11
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
7. Stack-Tip Downwash and Plume Rise

Most large-scale releases of pollutant are ejected from stacks with initial exit velocity w
s
and
enough initial buoyancy (due to excess temperature of the effluent) to cause the plume to rise
significantly before it is bent over by the wind. On a windy day, the initially vertical plume is
quickly bent over by the wind as shown in Figure 9. Because of the relative velocity of the
plume and the wind, the plume boundary is very turbulent. The rising plume entrains ambient air
and cools before it eventually reaches a maximum height dependant on the ambient environment.
Various empirical equations and mathematical models have been proposed for calculating the
rise of stack gas plumes in the atmosphere due to this initial momentum and buoyancy.

7.1 Stack-Tip Downwash

If the stack discharge velocity is very small, the stack effluent may be caught in the aerodynamic
wake behind the stack and drawn downward (Figure 8). This genrally occurs when the stack exit
velocity w
s
is less then 1.5 times the mean wind speed U
s
at the top of the stack. One way to
account for this effect is to introduce an effective stack height '
s
H which may is reduced by the
amount of downwash according to the following simple equation (Briggs, 1974):
' 4 1.5
s
s s s
s
w
H H R
U
| |
= − −

\ .
|
. (20)
Here H
s
is the physical stack height and R
s
is the stack radius. If w
s
> 1.5U
s
, then it is assumed
that there is no downwash the term in brackets is set to zero.

7.2 Buoyancy and Momentum Fluxes

In most EPA dispersion models, in order to calculate the plume rise the momentum flux and
buoyancy flux parameters are required. These parameters are based on initial exit conditions as
follows:
Momentum Flux:
2 2 2 2
s a
m s s s s
a s
T
R w R
T
ρ
ρ
| |
= =
|
\ .
F w [m
4
/s
2
] (21)

Buoyancy Flux:
2 2
s a s
b s s s s
a s
T T
w R gw R
T
ρ ρ
ρ
| | − −
= =

\ .
a
|
F g [m
4
/s
3
] (22)

By making several simplifying assumptions about the flow, it is possible to derive simple,
closed-form solutions of the equations of motion for the buoyant plume. These include:

1. The wind speed U
s
remains constant above the stack.
2. The plume is fully “bent over” (Figure 9) for its entire trajectory.
3. The plume is advected by the mean wind, with U
p
= U
s.

4. The plume density ρ
p
and ambient density ρ
a
are equivalent, except in terms
involving the difference ρ
a −
ρ
p
which are used to calculate the buoyancy (i.e., the
classical Boussinesq approximation).
12
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003

The resulting plume rise trajectory in a neutrally stable atmosphere is then given by (Davidson,
1989):

( )
1/ 3
3 3 3
0 m b 0
H H H H ∆ + ∆ + ∆ − ∆H ∆ = , (23)

In equation (23),
m
H ∆ is the rise component due to initial momentum,

1/ 3
1/ 3
1/ 3
2 2/ 3
3
m
m
p
F
H x
U β
| |
∆ =
|
\ .
, (24)

b
H ∆ is the rise component due to initial buoyancy,

1/ 3
1/ 3
2/ 3
2
3
2
b
b
p
F
H x
U β
| |
|
\ .
∆ = , (25)

and
0
H ∆ is a constant which accounts for the initial size of the plume,

1/ 2
0
1
a s
s
s p
T w
H R
T U β
| |
|
|
\ .
∆ = . (26)

The coefficient β is a measure of the rate at which ambient air is entrained by the plume. Both
Briggs (1984) and Davidson (1989) recommend using β = 0.6, based on matching the model to
observed plume trajectories.

7.3 Final Plume Rise

Equations (23), (24) and (25) suggest that a buoyant plume will rise indefinitely. However, as a
hot plume rises, it cools as it entrains ambient air. The plume will eventually reach an elevation
where its internal temperature is the same as the ambient air and cannot rise further because the
potential temperature θ
a
increases with height at the top of the mixing layer. If the plume kept on
rising, it would be cooler than the surrounding air and will experience negative buoyancy which
pushes it down. This effect can be seen in Figure 9. Thus in a stable atmosphere, the plume rise
must be limited. The final rise is determined by the following atmospheric stability parameter


2
a
a
g
z
S N
θ
θ

= =

. (27)

The characteristic frequency N is called the Brunt-Vaisalla frequency and it is the natural
frequency of wave-like motions in the atmosphere. It is found that the maximum rise of a
buoyant plume in a stable atmosphere is then given by (Briggs, 1984):
13
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003

1/ 3
2
2.66
b
p
F
H
U N
| |

\ .
∆ = , (28)
|
|

Here U
p
and N
2
are calculated at the final plume height. If the atmosphere is close to neutral,
and equation (28) over predicts the final plume rise. In such a case the plume is
eventually broken up by atmospheric eddies and a different formulation is necessary, involving
the friction velocity u
2
0 N →
*
. For example, in the AERMOD model the following equation is used:

3/ 5
* *
1.2 1.2
b
s
p p
F F
H H
U u U u
| | |
∆ = +
|
|
\ . \
b
|
|
|
.
. (29)

Equation (28) is also found to over predict final rise in a stable atmosphere with calm winds. In
that case the plume rise is calculated from (Briggs, 1984):


1/ 4
3
4
b
F
H
N
| |
∆ =
|
\ .
. (30)
AERMOD takes the final plume rise to be the minimum of equations (28)-(30), which ensures a
conservative analysis – i.e., maximum ground level concentrations.

The final rise in an unstable, convective atmosphere with turbulence is generated by solar
heating of the ground depends on the following heat flux parameter (Briggs, 1975):


3
*
*
' '
a M
gw T u
H
T L κ
= =
O
. (31)
In this case the final rise is given by:


3/ 5
2/ 5
*
4.3
b
p
F
H
U

| |
|
|
\ .
H ∆ = . (32)

This result is based on the idea that the plume rise will terminate when the dissipation rate in the
plume has decayed to that of the surrounding turbulent air. This is called the “plume break-up
model” (Briggs, 1975).

Prior to reaching the final plume rise, it is correct to use equation (23) in all stability conditions.
In calculating the final plume rise using any of the above models, the plume velocity U
p
must be
evaluated at the final rise height. Since this speed is unknown a priori, some sort of iteration is
required. Most simple regulatory models simply use the wind speed at stack exit U
s
. For a
particular plume, there can be considerable error in applying the individual plume rise formulae,
since they are based on an aggregate average of many observed plumes. A factor of two error in
14
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
the final rise height is quite possible, so it is often best to choose the minimum final rise height
from a variety of formulas to ensure a conservative estimate of the resulting ground-level
concentration.

7.5 Buoyancy-Induced Dispersion

As a plume rises to its final rise height, it entrains air and is spread out by turbulent eddies.
During the rise phase, the plume size can be shown to increase linearly with rise height,

0 p
R R H β = + ∆ . (33)

Here R
0
is an effective source radius, which is related to the physical source diameter by
(Davidson, 1989),

1/ 2
0
a s
s
s s
T w
R R
T U
| |
=
|
\ .
, (34)
and β is the usual air entrainment parameter (β ≈ 0.6). The initial rate of plume spreading is
greater than that due to ambient turbulence alone. In order to account for the enhancement of the
plume spread due to the buoyant rise phase, for example, the SCREEN3 model calculates
effective plume spread parameters downwind of the stack according to:

1/ 2
2
2
3.5
ye y
H
σ σ
| |

| |
= +

|

\ .
\ .
|
|
(35a)
1/ 2
2
2
3.5
ze z
H
σ σ
| |

| |
= +

|

\ .
\ .
|
|
. (35b)

Here ∆H is the plume rise, and σ
y
and σ
z
are the P-G sigmas calculated for a point source at the
stack (x = 0). A similar approach is used in ISC3 and AERMOD.

8. Initial Source Size and Virtual Source Methods

The tabulated P-G plume sigmas given in Appendix A are based on an ideal point source of
pollutant. In practice, we are often interested in area or volume sources of pollution with finite
initial lateral and vertical dimensions. In addition, nearby buildings can enhance the initial
growth of plumes and can produce an effective volume source in their wakes. As shown in
Figure 10, it is possible to account for any initial source dimensions by projecting a hypothetical
point source some distance upstream of the actual area/volume source. This is done by first
calculating the effective initial plume width σ
y0
and plume height σ
z0
due to the area or volume
source. The upstream virtual source distances x
y0
and x
z0
are then calculated using the P-G
sigma formulas. For example, suppose the P-G equation for calculating σ
z
for a point source is
given by

15
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
, (36)
b
z
ax σ =

then the virtual source distance x
z0
is given by:

1/
0
0
b
z
z
x
a
σ
| |
=
|
\ .
. (37)

At later stages of plume development downstream of the source we would calculate σ
z
from

. (38)
0
(
b
z
a x x σ = + )
z

This is equivalent to assuming that the plume originates as a point source a distance (x + x
z0
)
upstream. Table III gives recommendation for initial spread parameters based on the dimensions
of the actual volume sources. These methods are used in the ISC3 dispersion model.


Table III. Procedures for Estimating Initial Lateral Dimensions and Initial Vertical Dimensions
for Volume and Line Sources (ISC3 Model)
Source Type σ
y0
Calculation σ
z0
Calculation
Surface-based volume source Length of side ÷ 4.3 Vertical dimension ÷ 2.15
Elevated source Length of side ÷ 4.3 Vertical dimension ÷ 4.3
Line source represented by
series of volume sources
Length of side ÷ 2.15 Vertical dimension ÷ 2.15 if
ground level line source,
Vertical dimension ÷ 4.3 if
elevated line source


It can be shown that the plume sigmas in Table III are equivalent to choosing the Gaussian
plume so that it will have a concentration of 10% of its peak value at the physical edges of the
source. This idea is shown schematically in Figure 10.

Although ISC3 uses a virtual source method, AERMOD differs in its treatment of the initial
source size. In AERMOD, the initial variance of plume as given in Table III is added to the
predicted plume variance for a point source without virtual displacements:


2 2 2
0 y y y, ps
σ σ σ = + , (39)

In this equation:
0 y
σ = the initial plume size

, y ps
σ = the plume size assuming an initial point source

y
σ = the resultant plume size (including effect of initial source size).

16
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
8.1 Area Sources and Urban Pollution Problems

Often one is interested in calculating the cumulative effect of numerous small sources (small
industries, residential heating, vehicles, etc.) that are distributed over a large area. In such cases,
the source rate is best expressed as an average pollutant flux per unit area [kg/m
2
-s]. Consider a
rectangular area source that has crosswind dimension L
y
and along-wind dimension L
z
. The
concentration downwind can be calculated by applying equation (8) to each infinitesimal area
source dy×dz and then integrating over the whole area:

/ 2
2 2
2 2
0 / 2
2
2
0
1
( , 0, ) exp exp
2 2
2 1
exp
2 2 2
y
z
y
z
L
L
y z z y L
L
y
z z y
q z y
C x z dy dx
U
L
q z
erf dx
U
π σ σ σ σ
π σ σ σ

(
| |
| |
( = − −
|
|
|
( \ .
\ .
¸ ¸
| |
| |
= − |
|
|
\ .
\ .
∫ ∫

, (40)

The resulting integral can be expressed in terms of the error function erf(x)
1
. For most
problems, the area integration must be performed numerically. However in a large city where L
y

is large compared to σ
y
, the error function is approximately one. In addition, if we are interested
only in ground level concentrations (z = 0), then the concentration is given by the simple
expression:

1
0
2
( , 0, 0)
z
L
z
C x dx σ
π

=

(41)
In addition, if the receptor is within the boundaries of the area source of interest, we can replace
L
z
by the distance x from the upstream edge. For the simple case where , we get the
result given by Gifford and Hanna (1970) for concentrations due to urban area sources:
b
z
ax σ =


1
2
(1 )
b
x
C
a b π

=

. (42)

Assuming b ≈ 0.75 for typical urban dispersion, the dependence of the concentration on x (the
size of the city) is weak, because the lower layers of air over a city tend to be well mixed. Of
course, if there is a vertical limit to mixing due to an elevated inversion, then there will be a
buildup of concentration in the atmosphere, and a simple box model calculation yields (for
σ
z
> z
i
):

1
The error function erf(x) occurs frequently in probability theory and diffusion problems, and is defined as:
2
0
2
( )
x
u
erf x e du
π

=

. It is a measure of the area under the Gaussian distribution function.
17
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003

p i
q x
U z
≈ C . (43)

This last result explains why air quality in the center of large cities is generally very poor on
calm days with a low-level capping inversion that traps the pollutants near the ground.

9. Complex Terrain Algorithms

Although the derivation of the Gaussian Plume model assumes ideal conditions such as an
infinite, flat, homogeneous area, the Gaussian plume model is often used to predict
concentrations at receptors in complex, elevated terrain. There are several ways to account for
these effects, and the method applied depends somewhat on the atmospheric stability class and
whether the “elevated simple terrain” or “complex terrain” option is selected in the model.
SCREEN3 and ISC3 are both similar in their approach to terrain modelling. The approach used
in AERMOD is more sophisticated. In all these models, only the vertical distribution function is
affected. With the exception of the CALPUFF model, there is no mechanism for lateral
deflection of the plume due to terrain in these standard regulatory air dispersion models.

9.1 Elevated Simple Terrain

In SCREEN3, the elevated simple terrain option is chosen when the terrain height at the receptor
is above the stack base elevation, but below the release height. The net plume height above sea
level, which includes the effect of plume rise (i.e., H
p
= H
s
+ ∆H), is then kept constant, i.e. it
does not follow the terrain contours. Instead, the plume height above ground at a point receptor
(x,y) downwind depends on the local elevation z(x,y). The modified plume height above the
receptor is calculated as

'
p p s
( , ) H H z z x y = + − (44)
where:
H
p
′ = the effective plume height above the receptor
H
p
= uncorrected plume height above the base of the stack
z
s
= elevation of stack base
z(x,y) = elevation of terrain at the receptor location

If the receptor is located above the ground, then the effective plume height above the receptor is
reduced by the receptor height above ground (the so-called “flagpole” height), and also by the
local terrain elevation above the stack base. This is shown in Figure 11.

9.2 Complex Terrain

When the complex terrain option is selected in ISC3, the plume is assumed to be terrain
following. The effective plume stabilization height above the ground at a point (x,y) is then
given by:

18
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
' (1
p s t
)
t
H H H f = + ∆ − − H , (45)

where: H
s
= stack height
∆H = plume rise assuming flat terrain
H
t
= terrain height of the receptor location above the base of the stack.(= z(x,y)- z
s
).

The terrain adjustment factor is stability dependent:

for stability categories A-D 0.5
t
f =
for stability categories E, F . 0
t
f =

In neutral and unstable conditions (categories A-D), the plume stabilization height is adjusted to
partially follow the terrain. However, under stable conditions the plume height is not deflected
by the terrain. The plume axis is assumed to remain fixed at the plume stabilization height above
mean sea level. If the plume encounters an obstacle such as a hill that is above this height, then
the plume can impact on the obstacle, leading to high concentrations. As a result of the terrain
adjustment factor (f
t
= 0.5), during unstable and neutral conditions, the plume height relative to
the stack base is deflected upwards by an amount equal to half the terrain height as it passes over
complex terrain, Figure 12. This concept is based on a recommendation by Egan (1975), who
found from the results of wind tunnel experiments and potential flow theory that the wind
streamlines never fully follow the underlying terrain.

9.3 AERMOD Approach to Complex Terrain

In the AERMOD approach to complex terrain, the terrain features are first processed by
AERMAP (the terrain preprocessor), which assigns an effective height of terrain h
c
to each
receptor point (x
r
, y
r
, z
r
). In order to calculate the concentrations at downwind receptors located
on elevated terrain, AERMOD performs two concentration calculations corresponding to two
extreme plume states. These are: 1) a horizontal plume as under very stable conditions such that
the flow remains horizontal (and may impact on a large hill); and 2) a plume that completely
follows vertical terrain (terrain following state), so that its centerline height above the local
terrain is equivalent to the plume height above flat terrain. The concentration at a receptor C(x
r
,
y
r
, z
r
), is then calculated as a weighted sum of these two contributions,

C f , (46)
,
(1 )
T t h s t t
C f C = × + − ×
,s

where: = total weighted concentration
T
C
C = concentration due to the horizontal plume state
, h s
C = concentration due to the terrain following plume state
, t s
(i.e., the plume rises with the terrain)
f
t
= terrain weighting factor.

19
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
The weighting factor in this case is a function of the fraction of plume mass that lies below the
dividing streamline height H
c
2
. In this case, the parameter f
t
is in the range 0.5 – 1, which means
that the plume never completely approaches the terrain-following state. There is always some
contribution due to the horizontal plume state. For further details on this approach, the reader is
referred to the model formulation document by Cimorelli et al. (1998).

10. The Effect of Buildings and Building Wakes on Plume Dispersion

Many industrial stacks are located on top of buildings or in plant sites where there are large
buildings nearby. One of the major challenges in regulatory dispersion modeling is to account
for the effects of buildings on the near-field dispersion of a plume. There is often a single, large
structure that dominates the scene, such as a nuclear reactor building. Most research on this
problem had been done for releases on or nearby individual buildings.

The dispersion patterns around isolated buildings are generated by several flow features and are a
function of obstacle shape, approach flow turbulence and wind direction. The time-averaged
flow field around a rectangular building in a turbulent shear flow is shown in Figure 13, taken
from Hosker (1979). Within this flow, which is very unsteady, there are five fundamental
regions to consider:

1. A displacement zone, where the incident wind is first influenced by the building
2. A region of separated flow over the upstream edges (roof cavity zone)
3. A ground-based “horseshoe” vortex system that wraps around the base of the building
4. A wake cavity of recirculating flow behind the building.
5. A slowly decaying wake with reduced mean velocity and enhanced turbulence.

The flow field in figure 13 results when the wind blows normal to a face. When the wind
approaches the building on a diagonal, a strong elevated trailing vortex may be generated.

As a rule of thumb, if a nearby source is released from a height greater than two-and-one-half
times the building height, then there is usually no significant influence of the building on the
dispersion (Hanna et al., 1982). For sources nearer the building, the regions of flow separation,
trailing vorticity and enhanced turbulence as shown in Figure 14 will likely affect the plume.

10.1 Separated Flow

Separation occurs at the sharp upwind edges of bluff obstacles embedded in a turbulent flow.
The regions of separated flow are characterized by low wind speed, high turbulence, high
velocity gradients, and flow reversal. To quantify the size of the separated flow regions around a
building, a representative building length scale can be defined as (Wilson, 1979):


2/ 3 1/ 3
B S L
L B B = , (47)

2
The dividing streamline H
c
is the theoretical streamline height which divides the flow into two layers, one which
remains horizontal and one which rises over the terrain feature. This concept comes from experiments involving
flow over hills in stratified flow.
20
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003

where: B
S
= smaller of upwind building face dimensions H or W
B
L
= larger of upwind building face dimensions H or W.

Depending on the length of the building, the roof top cavity may or may not reattach before the
rear edge of the building. The length and height of the rooftop recirculation cavity is estimated as
(Wilson, 1979):

0.9 , 0.22
c B c B
L L H L = = . (48)

Therefore the roof cavity will reattach to the roof as in Figure 14 if L
c
< L, where L is the along-
wind length of the building.

The exact dimensions of the recirculation region downwind of an obstacle depend somewhat on
the intensity and scales of turbulence in the approaching flow. However, for a broad range of
building dimensions, the mean length of the “wake bubble” can be calculated using the following
formula (Fackrell, 1984):


0.3
1.8( / )
, 0.3 3
1 0.24( / )
R
L L W H L
H H W H H

| |
= ≤
|
+
\ .
.0 ≤ (49)

For a building in the shape of a cube, L
R
/H ≈ 1.5. Within the wake bubble, the low speed flow
recirculates and the instantaneous wind velocities fluctuate randomly so that any pollutants are
rapidly dispersed throughout the cavity region. There is a characteristic delay time before any
entrained pollutant is released to the outer flow (the “residence time”).

Downwind of the wake cavity the wind begins to return to the conditions of the approach flow,
but with reduced mean speed and enhanced turbulence. This generally leads to a greater rate of
dispersion for plumes that are entrained in the wake. The crosswind velocity defect, which is
approximately bell-shaped, extends to 10-20H downwind, depending on the W/H ratio of the
building. It is shorter for narrow buildings, longer for wide buildings (Peterka et al., 1985).

10.2 Frontal Eddy and Trailing Vorticity

When a bluff body is placed in a shear flow, there is a tendency for a frontal eddy to form at the
upwind face. Because the stagnation pressure increases with height above the ground, there is a
downward flow induced on the lower 2/3 of the upwind obstacle face as seen in Figure 14. The
frontal eddy trails around the sides of the obstacle and creates a trailing vortex with
longitudinally oriented vorticity (Figure 13). This vortex can induce a “downwash” flow that
draws the plume from rooftop stacks downward.

At acute angles to the wind, the rooftop leading edges of the building can produce longitudinally
oriented vortices that transport the higher velocity air from above down into the central portions
of the wake. This effect can extend some 50-100H downwind of the building (Hosker, 1984),
21
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
and can contribute to downwash of plume material into the wake, and increased lateral
spreading.

10.3 Effect of Building Downwash and Wake Flow on Plume Dispersion

There are several available models to account for the enhanced dispersion of effluent caused by
buildings located near a stack. Most of these are empirical correlations based on the results of
controlled wind tunnel experiments and more limited field experiments. In many simple models
for rooftop stacks, the plume rise due to momentum is calculated at a distance of two building
heights downwind, ignoring the effect of the building. If this plume height is less than some
criteria (e.g., some multiple of the building height) then the plume is assumed to enter the wake
cavity. In cases where the source is right on or very near the building, the modified plume rise
algorithm due to Schulman and Scire (1980) can be used to calculate the initial rise.

If a plume is entrained in a building wake, the SCREEN3 model calculates a cavity concentration
using the following formula (Hosker, 1984),

c
c f
Q
C
K A U
= , (50)
where: Q = emission rate
A
f
= building cross-sectional area normal to wind (H×W)
U = reference wind speed (typically at 10 m height)
K
c
= non-dimensional concentration constant (= 1.5).

Since the size of the wake cavity depends on the building orientation relative to the wind, the
cavity dimensions should be calculated for at least two extreme building orientations in order to
get a reasonable bound on the cavity concentration estimate.

When a plume is fully entrained into a building wake, it starts with an initial size approximately
equal to the building cross-sectional area A
f
. This building-enhanced dispersion can be
accounted for by virtual source displacements:


0
0 0
' (
' (
y y y
z z z
0
)
)
x x
x x
σ σ
σ σ
= +
= +
(51)

The virtual distances x
y0
and x
z0
are chosen so that the initial value of the dispersion parameters
at the rear face of the building (σ′
y0
and σ′
z0
) are some fraction of the building width and height,
respectively. A simple model recommended by Turner (1969) is:


0 0
' , '
4.3 2.15
y z
W H
σ σ = = . (52)

These virtual source estimates are illustrated in Figure 10.

22
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
For building-affected dispersion calculations in the wake, ISC3 and AERMOD use the Huber-
Snyder (1982) model to calculate the enhanced dispersion for sources entrained in the wake.
Neither of these models will predict concentrations directly in the wake cavity (x < 3H), however
SCREEN3 can be used for that. The Huber-Snyder model has different expressions for the near
wake (3 < x/H < 10) and far wake (10 < x/H) regions. In the near wake, the dispersion
parameters depend on the building dimensions, with:


' 0.35 ( 3 ) /15
' 0.7 ( 3 ) /15
y
z
W x H
H x H
σ
σ
= + −
= + −
(53)

In the far wake, the P-G sigmas are used with virtual source displacements x
y0
and x
z0
chosen so
as to match the plume dimensions at x = 10H. This is done in SCREEN3 and ISC3 models. In
AERMOD, which uses a more complicated method of calculating the plume sigmas, the plume
variances are added:


2 2 2 2 2
'
yT y y zT z z
and
2
' σ σ σ σ σ σ = + = + , (54)
where:

2
yT
σ = total variance of the plume

2
y
σ = Pasquill-Gifford variance calculated for a point source

2
'
y
σ = Huber-Snyder variance evaluated at x = 10H

Special cases may arise when applying equation (53). If W >>H (a squat building) then H
replaces the value of W in calculating σ
y
′. For a very tall building (H >>W), W replaces H in the
σ
z
′ equation, otherwise unrealistically high vertical dispersion is predicted. Further details on
incorporating the H-S model are described in the ISC3 Model documentation.

The Huber-Snyder enhanced dispersion model is only applied if the plume height, evaluated at
two building heights downstream, is less than H + 1.5×min(W,H), where min(W,H) is the lesser
of the building frontal dimensions. The ISC3 model modifies only σ
z
if the plume height is
greater than 1.2 building heights. If the building-affected sigmas from Equation (53) are less
than the P-G sigmas at the same distance downwind, then the latter are used.

In ISC3 and AERMOD, if the stack height is less than H + 0.5×min(H,W), a more sophisticated
algorithm due to Schulman and Scire (1980) is used to calculate the downwash effect because
the initial rise is reduced by the building-enhanced dilution. This model is described in some
detail in the original reference and in the model formulation documents for ISC3 and AERMOD.

The overall effect of enhanced dispersion in building wakes is to decrease the maximum ground-
level concentrations for low level releases or to increase the ground level concentrations for
elevated sources. Using extensive wind tunnel and field data, Fackrell (1984) found that the
differences between the measured and predicted concentrations using the Huber-Snyder and
other simple building-affected dispersion models was often less than a factor of two or three.
23
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
This is within the expected limits of accuracy for most Gaussian plume models (Beychok, 1995).
However, none of the models seems to cope well with acute wind angles, where the trailing
vortices can induce severe downwash and increased ground level concentrations even for more
elevated sources (H
s
> 2.5H). One promising approach is that used by the ADMS model
(Carruthers et al., 1994). In this treatment of buildings, a rooftop plume is partitioned so that
only a fraction of the plume is entrained in the wake. This part is then accounted for by locating
a virtual source upwind (Figure 15). The part of the plume that is not entrained is treated as an
elevated point source. The net concentrations downwind are found by adding the two
concentration distributions due to the wake and elevated source.

11. Summary

Although the Gaussian plume model (GPM) is based upon many simplifying assumptions about
the dispersion process, it is applied to a wide array of dispersion scenarios and some form of this
model is adopted in most regulatory air pollution models for continuous releases. In order to
extend the applicability of the GPM to realistic scenarios, the U.S. EPA models make use of
several special algorithms or semi-empirical corrections to account for the various effects. These
include the influence of atmospheric stability, plume trapping below elevated inversions,
fumigation, non-uniform wind profiles, dry or wet deposition, stack-tip downwash, buoyancy-
induced dispersion, finite initial source dimensions, complex terrain and the influence of
buildings. A brief summary of the algorithms used to incorporate these various features has been
provided in this review.

As shown in this review, many of the algorithms in the advanced EPA models are based on
simplified physical models of the various dispersion processes, combined with empirical data.
The modifications to the basic GPM make extensive use of wind tunnel and measured field data.
Because of strong peer reviews and model validations studies, the resulting model codes are
quite robust and can be used in a wide variety of situations combining many separate effects.

All of the EPA codes, such as SCREEN3, ISC3, AERMOD and CALPUFF, have been run
through extensive physical audits, sensitivity analyses, and quality assurance studies using
benchmark data in order to justify their use in environmental assessment. An understanding of
the fundamental concepts used in these models is important for the most intelligent use to be
made of the models. This background knowledge is required to ensure that the most sensible
choices are made in all aspects of the data input stages, selection of model options and the
interpretation of results. To become an “expert” user, it is essential that one read the model
formulation documentation and some of the references in this document.

24
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
12. References

Barad, M.L. (1958). Project Prairie Grass: A Field Program in Diffusion. Geophysical
Research Papers No. 59, Vols. I and II. Report AFCRC-RT-58-235. USAF Cambridge
Research Laboratory, MA.

Berkowicz, R., H.R. Olesen and U. Torp (1987). The Danish Gaussian Air Pollution Model
(OML): Description, Test and Sensitivity Analysis in View of Regulatory Applications. In Air
Pollution Modelling and its Application V (ed. De Wispelaere, Schiermeier and Gillani).

Beychok, M.R. (1995). Air Dispersion Modeling: Keep Predicted Values in Perspective.
Environmental Engineering World, Nov.-Dec., 16-18.

Briggs, G.A. (1974). Diffusion Estimations for Small Emissions. US Atomic Energy
Commission Report ATDL-106. US Atomic Energy Commission, Oak Ridge, Tennessee.

Briggs, G.A. (1975). Plume Rise Predictions. In Lectures on Air Pollution and Environmental
Impact Analysis, American Meteorological Society, Boston, Mass.

Briggs, G.A. (1984). Plume Rise and Buoyancy Effects. Chapter 8 in Atmospheric Science and
Power Production (ed. D. Randerson). Publication no. DOE/TIC-27601, U.S. Dept. of Energy,
Washington, D.C.

Carruthers, D.J., R.J. Holroyd, J.C.R. Hunt, W.S. Weng, A.G. Robins, D.D. Apsley, D.J.
Thomson and F.B. Smith (1994). UK-ADMS: A New Approach to Modelling Dispersion in the
Earth’s Atmospheric Boundary Layer. J. Wind Eng. Ind. Aero. 52, 139-153.

Cimorelli, A.J., S.G. Perry, A. Venkatram, J.C. Weil, R.J. Paine, R.B. Wilson, R.F. Lee and
W.D. Peters (1998). AERMOD: Description of Model Formulation, 113 pp. Available at EPA
Support Centre for Regulatory Air Models: www.epa.gov/scram001.

Davidson, G.A. (1989). Simultaneous Trajectory and Dilution Predictions from a Simple
Integral Plume Model, Atmospheric Environment 23, 341-349.

Draxler, R.R. (1976). Determination of Atmospheric Diffusion Parameters. Atmospheric
Environment 10, pp. 99-105.

Egan, B.A. (1975). Turbulent Diffusion in Complex Terrain, in Lectures on Air Pollution and
Environmental Impact Analysis (D. Haugen, ed.), Amer. Meteor. Soc., Boston, Mass.

EPA (1995). SCREEN3 User’s Guide. US EPA Report EPA-454/B-95-004. 51 pp.

EPA (1995). User’s Guide for the Industrial Source Complex (ISC3) Dispersion Model. Volume
II – Description of Model Algorithms. US EPA Report EPA-454/B-95-003B.

25
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
Fackrell, J.E. (1984). An Examination of Simple Models for Building Influenced Dispersion.
Atmospheric Environment 18, pp. 89-98.

Fackrell, J.E. (1984). Parameters Characterising Dispersion in the Near Wake of Buildings. J.
Wind Eng. Ind. Aero. 16, 97-118.

Gifford, F.A. (1960). Atmospheric Dispersion Calculations Using the Generalized Gaussian
Plume Model. Nuclear Safety 2, 56-59.

Gifford, F.A. and S.R. Hanna (1970). Urban Air Pollution Modelling. Presented at 1970 Int. Air
Pollution Conf. of Int. Union of Air Poll. Prevention Assocs., Washington, D.C.

Hanna, S.R. (1989). Confidence Limits for Air Quality Model Evaluations as Estimated by
Bootstrap and Jacknife Resampling Methods. Atmospheric Environment 23, 1385-1398.

Hanna, S.R., G.A. Briggs and R.P. Hosker (1982). Handbook on Atmospheric Dispersion.
Publication no. DOE/TIC-11223, U.S. Dept. of Energy, Washington, D.C. 102 pp.

Hanna, S.R., P.J. Drivas and J.C. Chang (1996). Guidelines for Use of Vapor Cloud Dispersion
Models. AIChE/CCPS, 345 East 47
th
St., New York, NY 10017, 285 pp.

Hosker, R.P. (1979). Empirical Estimation of Wake Cavity Size Behind Block-Type Structures.
In Fourth Symposium on Turbulence, Diffusion and Air Pollution, pp. 603-609. Reno Nevada,
Jan. 15-18. Amer. Meteor. Soc., Boston, Mass.

Hosker, R.P. (1984). Flow and Diffusion Near Obstacles. Chapter 7 in Atmospheric Science and
Power Production (ed. D. Randerson), pp. 241-326. Publication no. DOE/TIC-27601, U.S.
Dept. of Energy, Washington, D.C.

Huber, A.H. (1984). Evaluation of a Method for Estimating Pollution Concentrations Downwind
of Influencing Buildings. Atmospheric Environment 25A, 1237-1249.

Huber, A.H. and W.H. Snyder (1982). Wind Tunnel Investigation of the Effects of a
Rectangular-Shaped Building on Dispersion of Effluent from Short Adjacent Stacks.
Atmospheric Environment 16, 2837-2848.

Irwin, J.S. (1979). A Theoretical Variation of the Wind Profile Power-Law Exponent as a
Function of Surface Roughness and Stability. Atmospheric Environment 13, 191-194.

McElroy, J.L. and F. Pooler (1968). The St. Louis Dispersion Study. U.S. Public Health
Service, NAPCA Report AP-53.

Meroney, R.N. (1982). Turbulent Diffusion Near Buildings. In Engineering Meteorology (ed.
E.J. Plate), pp. 481-525. Elsevier Publishing Co., Amsterdam.

26
Modelling Air Emissions for Compliance MME 474A Wind Engineering
December 2003
27
Panofsky, H.A. and J.A. Dutton (1984). Atmospheric Turbulence: Models and Methods for
Engineering Applications. Wiley, New York. ISBN 0471 057142.

Pasquill, F. (1961). The Estimation of the Dispersion of Windborne Material. Meteorol. Mag.
90, 33-49.

Pasquill, F. (1974). Atmospheric Diffusion, 2
nd
Edition. Ellis Horwood Limited, Chichester.
ISBN 0 85312 015 3.

Peterka, J.A., R.N. Meroney, and K.M. Kothori (1985). Wind Flow Patterns about Buildings. J.
Wind Eng. Ind. Aero. 21, 21-38.

Roberts, O.F.T. (1923). Proc. Roy. Soc. (London), A104, p. 640.

Schulman, L.L. and J.S. Scire (1980). Buoyant Line and Point Source (BLP) Dispersion Model
User’s Guide. Document P-7304, Environmental Research and Technology Inc. (1980).

Stern, A.C. (1976). Air Pollution: Volume I - Air Pollutants, Their Transformation and
Transport. Academic Press, New York. ISBN 0 12 666601 6.

Sutton, O.G. (1953). Micrometeorology. McGraw-Hill, London. 333 pp.

Turner, D.B. (1969). Workbook of Atmospheric Dispersion Estimates. Publication no. 999-AP-
26, U.S. Department of Health, Education and Welfare.

Turner, D.B. (1979). Atmospheric Dispersion Modeling: A Critical Review. J. Air Poll. Control
Assoc. 29, 502-519.

Turner, D.B. (1994). Workbook of Atmospheric Dispersion Estimates: An Introduction to
Dispersion Modeling. 2
nd
Ed. CRC Press, London. ISBN 1 56670 023 X.

van Ulden, A.R. (1978). Simple Estimates for Vertical Diffusion from Sources Near the
Ground. Atmospheric Environment 12, 2125-2129.

Wilson, D.J. (1979). Flow Patterns Over Flat Roofed Buildings and Application to Exhaust
Stack Design. ASHRAE Transactions 85, 284-295.

Wilson, D.J. (1997). Airflow Around Buildings. Chapter 15 in ASHRAE Fundamentals.
ASHRAE , 1791 Tullie Circle, N.E., Atlanta, GA 30329. ISBN 1 883413 45 1

Modelling Air Emissions for Compliance

MME 474A Wind Engineering December 2003

fundamentals of Gaussian plume modelling as normally presented in an undergraduate air pollution course. Many of the key concepts and algorithms incorporated into advanced air dispersion models are briefly explained. 2. Model Requirements and Model Selection There are several competing requirements in the design of an air pollution model. A model must capture the essential physics of the dispersion process and provide reasonable and repeatable estimates of downwind concentrations. This generally requires detailed knowledge of source characteristics, terrain and meteorology, but it is also desirable to keep these input requirements to a minimum, and simplicity is an important asset in any model. All models should have a fully documented account of the equation algorithms used and their conversion into valid software (i.e., traceability). Regulatory models must also undergo extensive quality assurance, including the evaluation of the model under several scenarios using benchmark data. Standard statistical procedures have been developed for expressing the uncertainty and variability of the predicted results when comparing them to measured concentrations (e.g., Hanna, 1989). In choosing an air dispersion model, several levels of model are available, with progressively increasing levels of mathematical sophistication, input data requirements and user expertise required. At the low end of the scale are the gross screening models, which require only a hand-held calculator, nomograph, or spreadsheet. They may treat only one source at a time (e.g., a single elevated stack) and provide some sort of worst-case prediction based on relatively primitive meteorological information. It is often wise to apply such a model prior to using the more advanced models, where the flow of information is more difficult to follow. Next on the scale of model complexity are intermediate models, usually PC-based, which may include varying meteorology (wind speed and stability classes) and more sophisticated source information. Many early EPA models fall in this category, including the SCREEN3 model. Advanced models require a desktop PC or workstation. They require extensive data sets for meteorology and emissions, and include multiple source types - point, area and volume. They may also include additional features such as complex terrain, flow around buildings and layered atmospheric structure. Modern models incorporate the most up-to-date treatment of the atmosphere such as Monin-Obukhov similarity theory. Some examples of advanced models are the EPA models ISC3, AERMOD and CALPUFF, the British Model ADMS (Carruthers et al., 1994) and the Danish model OML (Berkowicz et al., 1987) Specialized models are often used for predicting dispersion of special hazardous materials, such as military models used in chemical/biological defense. Heavy gas dispersion models are used by the chemical process industries to model the behaviour of rogue or accidental releases of dense gases or vapours. These models may require extensive thermodynamic information to account for release conditions. Models such as SLAB and DEGADIS (Dense Gas Dispersion Model) models are typical of this family. Although the input data requirements and level of sophistication increase with the more advanced models, a more complex model does not necessarily lead to predictions that are more

2

but it illustrates many of the basic parameter dependencies in dispersion modeling. the level of user understanding must increase to make proper use of the model.1x. suppose a small amount (1-kg) of ammonia is released over a period of 30 minutes in an accidental release. usually 1 m/s = worst case cloud width [m] (usually assume W = 0. In addition. The mean concentration is directly proportional to the release rate. Cwc = 109 Q (109 µ g / kg ) × (1kg /1800 s ) µg = = 22. 3.1 × 5000 = 500 m. As an example of the above.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 accurate. in which case more advanced modelling is unnecessary. we assume a uniform concentration in the plume passing through the downwind plane HW. (1996): 109 Q Cwc = UH wcWwc where: Q Cwc U Wwc Hwc = source strength or emission rate of gas or particulate [kg/s] = worst case concentration [µg/m3] = worst case wind speed at height z = 10 m. A useful formula for estimating worst case mean concentrations downwind of a point source is the following equation suggested by Hanna et al. the room for input data error increases. thus. The mean concentration is inversely proportional to the plume cross-sectional area. A gross screening analysis will quickly identify the order of magnitude of the expected concentrations and may even show that no problem exists. As the number of input variables goes up in the advanced models. Assuming a light wind of 1 m/s. Equation (1) follows from the fact that the flux of pollutant through any plane must equal the source rate Q.2 3 UH minWmin 1m / s × 50m × 500m m 3 . Equation (1) illustrates several important dependencies that should be satisfied by all plume models: 1. Example Gross Screening Analysis It is often useful to perform a simple screening analysis before applying a more refined computer analysis. does this release pose any risk to the occupants of a hospital located 5 km downwind? For this example the estimate of the plume width is Wwc= 0. 2. where x is distance from the source) = worst case cloud depth (usually assume H = 50 m in worst case) (1) This equation is essentially a statement of the conservation of pollutant mass. The mean concentration is inversely proportional to mean wind speed. Referring to Figure 2. 3.

The Diffusion Equation and the Gaussian Plume Model By performing a mass balance on a small control volume. which describes a continuous cloud of material dispersing in a turbulent flow. can be written as: dC dC d  dC  d  dC  +U =  Ky +  Kz +S dt dx dy  dy  dz  dz   where: x y z C(x.) 4 . Therefore. dy  dy    (The cloud will expand over time due to these terms. ∂y w'c ' = −K z ∂C .y.z) Ky. c(t ) = C + c '. ∂z (3) Here primed coordinates refer to the turbulent fluctuations of terms about their mean values. a simplified diffusion equation.axes [m2/s] = mean wind velocity along the x-axis [m/s] = source/sink term [kg/m3-s] Analytical solutions to this equation for the case of dispersion of passive pollutants in a turbulent flow were first obtained in the 1920’s by Roberts (1923) and Richardson (1926).032 ppm.and z. Turbulent diffusion of material relative to the center of the pollutant cloud. Typically in the atmosphere K y > K z . we can safely say that there is no risk.z) [kg/m3] = eddy diffusivities in the direction of the y. which explains why the cross-section of a plume often takes on an elliptic shape. and is 1500 times below the personal exposure limit (PEL) associated with negative health effects due to prolonged exposure to ammonia. etc. 4. etc. for eample. In such a case there is also no need to perform advanced modelling to assess the risk. u (t ) = U + u '. The eddy diffusivities (Ky and Kz) are a way of relating the turbulent fluxes of material to the mean gradients of concentration: v 'c ' = −K y ∂C .y. Kz U S (2) = along-wind coordinate measured in wind direction from the source = cross-wind coordinate direction = vertical coordinate measured from the ground = mean concentration of diffusing substance at a point (x. d  dC  Ky . A term-by-term interpretation of Equation (2) is as follows: dC dC +U dt dx Time rate of change and advection of the cloud by the mean wind.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 This concentration is equivalent to 0.

The pollutant concentrations do not affect the flow field (passive dispersion). but increases with height above the ground. This “image” term can be visualized as an equivalent source located at z = -Hp below the ground. the turbulent diffusivities Ky and Kz are unknown in most flows. If we define the following Gaussian parameters: σ y = 2K y x U and σ z = 2 K z x . In addition. 2. 4. so we have also restricted our analysis to steady wind flow over an idealized flat terrain. U (5) then the final form of the Guassian plume equation. z ) = exp   exp  . since several assumptions are made in its derivation: 1. Ky and Kz increase with distance from the source. and in the atmospheric boundary layer Kz is not constant. Mean values are based on time averages of 10-60 minutes. 6. The mean lateral and vertical wind velocities V and W are zero. Despite these limitations. the general Gaussian shape of Equation (4) is often. Molecular diffusion and longitudinal (along-wind) diffusion are negligible. The flow is incompressible. y . y . Equation (2) is grossly simplified.  4K ( x /U )  4π x K y K z y  4K z ( x /U )    (4) Unfortunately. 5. The Gaussian plume model. is obtained from the analytical solution to Equation (2). z ) = exp  − 2  exp  −  + exp  −   (6)    2σ  2π U pσ yσ z 2σ z2  2σ z2   y        In this expression a second z-exponential term has been added to account for the fact that pollutant cannot diffuse downward through the ground at z = 0. 5 . which is at the core of almost all regulatory dispersion models. For a continuous point source released at the origin in a uniform (homogeneous) turbulent flow the solution to Equation (2) is:     Q − y2 −z2 C ( x. The turbulent fluxes are linearly related to the gradients of the mean concentrations as in Equation (3). 3.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 S Source term which represents the net production (or destruction) of pollutant due to sources (or removal mechanisms). for an elevated plume released at z = Hp is:  y2    ( z − H p )2   ( z + H p )2   Q C ( x. because the diffusion is affected by different scales of turbulence in the atmosphere as the plume grows (Figure 3). and is further discussed in Section 6 below. The wind velocities and concentrations can be decomposed into a mean and fluctuating component with the average value of the fluctuating (stochastic) component equal to zero.

The plume height Hp is the sum of the actual stack height Hs plus any plume rise ∆H due to initial buoyancy and momentum of the release (Figure 4). Atmospheric turbulence is greatly enhanced by 6 . rather than the Gaussian form exp(− z ) . y . y . 1994). perpendicular to the plume axis.5 2 varies more like exp( − z ) . nonGaussian model. (Barad.  2σ  π U σ yσ z y   2σ z   Q (8) It turns out this latter model is not very accurate. we require the Gaussian plume parameters σy and σz.S. A more general.z (10) where the integration is taken over the y-z plane. 1958). C ( x. For a ground-level release. To ensure conservation of mass. such models must satisfy: ∫∫ CUdydz = Q . the vertical profile 1. 5.  H p2   y2  exp  − 2  exp  − 2  C ( x.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 Equation (6) is the well-known Gaussian plume equation for a continuous point source (Turner. y. The wind speed Up is taken to be the mean wind speed at the height of the stack. 1978). y. These have been measured as a function of distance from the source in numerous field studies. z ) =  2σ  2π U pσ y y   (9) Here f(z) is a normalized function which describes the vertical distribution of material in the plume. which are based primarily on the Project Prairie Grass study in the U. The most common tabulated data are the Pasquill-Gifford sigmas.  y2  Q exp  − 2  f ( z ) . z = 0) =  2σ   2σ z  π U pσ yσ z y     Q If we furthermore set Hp= 0 we get the vertical distribution due to a ground-level source: (7) C ( x. due to the large vertical variations of the diffusivity Kz near the ground (van Ulden. which allows for the vertical variation of Kz and the vertical variation of the velocity profile can be written as. z ) =  y2   z2  exp  − 2  exp  − 2  . Since we are normally interested in concentrations at the ground (where the receptors such as people are). The plume parameters σy and σz are driven by atmospheric turbulence and are influenced by the state of convection in the atmosphere. Determination of the Gaussian Plume Parameters σy and σz In order to evaluate equations (6) and (7). we set z = 0 to obtain. The P-G sigmas were formulated by Pasquill (1961) and Gifford (1961) for low-level releases over relatively smooth terrain at distances of a few thousand meters from a source.

Pasquill developed a simple quantitative rating scheme consisting of six stability classes ranging from highly convective [A] to highly stable flow conditions [F]. However.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 convective motions due to heating of the earth’s surface. dispersion in the urban environment usually produces greater rates of spread than these field data expressions. On clear nights. Pasquill-Gifford Stability Class and MoninObukhov Length [Hanna. The grouping of atmospheric stability into six discrete classes is done in most simple regulatory dispersion models. The resulting Pasquill-Gifford (PG) σy and σz curves under varying conditions of stability are shown in Figure 5. The distance x is measured from the source. Relationship Between Wind Speed. 1996] Description Very Unstable Unstable Neutral Stable Very Stable P-G Stability Time of Wind Speed U Class Day/Condition A Sunny Day < 3 m/s B or C D E F ↓ Cloudy or Windy ↓ Clear Night 2-6 m/s > 3-4 m/s 2-4 m/s < 3 m/s M-O LMO -10 m -50 m |L| > 100 m + 50 m +10 m Length For use in regulatory Gaussian plume models. Table I. In order to relate the state of atmospheric convection to simply observable parameters. Draxler (1976) provides the following relationships: σy = σz = σv σw U U x⋅ f y = σθ x ⋅ f y x⋅ fz = σφ x⋅ fz . In Appendix A gives the algebraic equations used to calculate σy and σz in the ISC3 model are provided. However. The functions fy and fz are unity 7 . These sigma data can be applied for releases over flat. analytic expressions have been fitted to the standard P-G sigma curves. These classes are summarized in Table I. (11a) (11b) where σv and σw are the root-mean-square fluctuations in transverse and vertical velocities (v and w). For example. then it is possible to relate the plume sigmas directly to these turbulent fluctuations using the statistical theory of diffusion. are used. These are incorporated in the SCREEN3 and ISC3 models when the user selects the urban terrain option in running the software. it is suppressed by cooling of the ground due to outgoing radiation. a second set of curves sometimes called the McElroy-Pooler (1968) sigmas. such as directly measured wind velocity fluctuations. Drivas and Chang. and σθ and σϕ are the standard deviations of the wind vector azimuth and elevation angles (in radians). For urban dispersion. based on tracer releases in a large U. if more detailed information is available. rural terrain.S. city.

Plume reflection will occur in this case at both the ground and the inversion layer. y . but only the first few terms are usually required in the vertical distribution function. Until this happens. as σz increases. This image source will also be reflected at the ground (to ensure ∂C/∂z|z = 0 = 0). an image source is placed at z = 2zi −Hp as illustrated in Figure 6. Plume Reflection at the Ground and at Elevated Inversions When a plume is discharged from an elevated stack. The vertical distribution will be non-Gaussian. To account for the initial reflection at the inversion. if actual measurements of the wind velocity fluctuations are available. the plume image source at z = Hp below the ground requires an image source above the inversion at z = 2zi + Hp to ensure ∂C/∂z|z = zi = 0. Similarly.  2σ  2π U pσ yσ z y   (12)   ( z − H p )2   ( z + H p )2   f ( z . H p . σy and σz can be calculated directly from (11a) and (11b). so a further image source is placed at z = 2zi − Hp below the ground. Thus.σ z ) . and we can write: C ( x. 6.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 close to the source but are decreasing functions of x. Similarly. H p . 1994): (13) 8 . the vertical concentration profile eventually becomes uniformly distributed throughout the mixing layer after which the concentration can be approximated by (Turner. then the plume will have difficulty expanding vertically and will effectively be “trapped” between the inversion and the ground. z ) = where:  y2  Q exp  − 2  f ( z . it will spread vertically until its lower edge 2 2 reaches the ground.. the term: exp( −( z + H p ) / 2σ z ) in equation (6). This process of reflecting image sources can go on indefinitely. 1998). does not make a significant contribution to concentrations above the ground. The virtual source at z = -Hp will then begin to contribute to the aboveground concentrations. zi . the plume will eventually be reflected at the ground. An approach similar to this is used in the AERMOD dispersion model to calculate the dispersion parameters from atmospheric turbulence (Cimorelli et al.σ z ) = exp  −  + exp  −    2σ z2  2σ z2           ( z − 2 zi + H p ) 2   ( z + 2 zi − H p ) 2   + exp  −  + exp  −      2σ z2 2σ z2          ( z − 2 zi − H p ) 2   ( z + 2 zi + H p ) 2   + exp  −  + exp  −      2σ z2 2σ z2        In practice. However. if a strong inversion layer is located at some height zi above the stack. zi .

under conditions of neutral atmospheric stability. y . In homogeneous terrain. 1984): U ( z) = u*  z  ln   κ  z0  (15) 9 . AERMOD includes a delay to mimic the lofting behaviour of a buoyant plume) Cp = penetrated source concentration distribution. the wind speed is found to vary logarithmically with height (Panofsky and Dutton. where only a fraction fp of the plume (where fp ≤ 1) remains in the convective boundary layer (CBL). for material that initially penetrates the elevated version Each of these terms has associated image sources to satisfy the symmetry boundary conditions (∂C/∂z = 0) at z = 0 and z = zi. (1998). This formula also describes the distribution of material during a fumigation episode. During stable conditions. The remaining fraction (1-fp) is allowed to penetrate the inversion and to escape temporarily.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 (13)  y2  Q exp  − 2  C ( x. which is the brief period when an elevated plume initially above an inversion is mixed downward by convective turbulence as the depth of the mixed layer reaches the height of the plume due to heating at the ground (Turner. the plume is modelled using only the direct source contribution and reflection at the ground. AERMOD is one of the few models that allows for non-Gaussian distributions of plume material in the vertical. without “trapping”. 1994).. but eventually reappears in the CBL if convective conditions raise the inversion cap and diffuse the mass that has escaped back to the ground. The Plume Advection Velocity and the Wind Speed Velocity Profile In the lowest part of the earth’s boundary layer (the surface layer). The total concentration Cc at a receptor is then given by: Cc = Cd + Cr + C p where: (14) Cd = direct source concentration contribution (due to downward dispersion of material from the stack) Cr = indirect source concentration contribution (due to primary image source above zi. wind speed increases with increasing height and has strong gradients near the ground. z ) =  2σ  2π U pσ y zi y   This gives reasonable results when σz ≥ zi . 6. A buoyant plume is never perfectly reflected by an inversion layer and partial plume trapping is allowed in more advanced models such as AERMOD. Further details on this approach are found in the AERMOD model formulation document by Cimorelli et al.

For a ground level source.3 Large Forest/Cities 1.6σ z  ln   κ  z0  (17) This is just the log-law wind speed evaluated at a height z = 0.4. The surface roughness length z0 is a measure of the aerodynamic roughness of the ground. The exponent p increases dramatically with increasing atmospheric stability Irwin.1 0.24-0. and is typically 3-10% of the height of the surface obstacles (trees. It is typically about 10% of the wind speed at z = 10 m. For different types of terrain.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 The friction velocity u* is related to the frictional resistance that the ground exerts on the wind. Equation (15) is valid only for neutral stability conditions. Because of the dependence on σz.  10  p (16) The power law coefficient p increases with increasing surface roughness.26 Parkland/Residential Areas 0.13 Open grassland 0.10 Short grass 0.15 Root crops 0.6σz. The effect of stability on the power law exponent is shown in Appendix B. Equations (15) or (16) can be used to calculate this velocity from the standard wind speed U10 at 10m height. etc.5 0.2 Agricultural areas 0.005 0.0 0. it moves at progressively higher speeds.39 Since windspeed varies with height. houses.  z  U ( z ) = U10   . The von Karman constant κ is about 0. 10 . Surface Roughness and Power Exponent for Wind Flow Over Various Terrains in Neutral Stability Terrain z0 [m] p -4 Lake or smooth sea 10 0. It is often approximated by a more simple power law of the form.3 0. Table II. it is not obvious what advection speed one should assume when using the Gaussian plume model. it can be seen that as the plume grows vertically.02 0. 1978): Up = u*  0. In practice.2-0. crops. Table II gives the approximate roughness height z0 and profile exponent p. The logarithmic wind profile is not easy to work with for dispersion calculations.07 Sandy desert 10-3 0. for an elevated source one usually takes Up to be the mean wind speed at the stack height. Some typical wind profiles are shown schematically in Figure 7. one typically uses the wind speed U10 for the plume advection velocity. although a more precise plume advection velocity for a ground source is given by (van Ulden. 1979).).

the mechanical generation of turbulence is dominant. This is a much more sophisticated approach then the traditional methods using charts which only allow for discrete stability classes and roughness categories. 11 . Typical exponents are given in Appendix B. These are then used in calculating mixing height. temperature and velocity profiles in a convective (CBL) or stable (SBL) atmospheric boundary layer. when z < LMO. Thus.4 g Ta (19) where: H f = ρ c p w 'T ' Ta g cp = vertical heat flux from ground [W/m2] = absolute air temperature [0K] = gravitational constant [9. These factors are then used in the determination of the Gaussian plume parameters σy and σz. z = LMO defines the altitude above which the mechanical production of turbulence is suppressed through the action of negative buoyancy. AERMOD has a meteorological preprocessor called AERMET which calculates the values of u* and LMO for a given flow. with the recognition that in this case the exponent p will be a function of z0 and LMO.4 gH f 3 u* = − w 'T ' 0. z / LMO is a measure of the ratio of turbulence production by convection to mechanical production of turbulence. 1996) U ( z) = u*   z  z  ln   + 4. In a stable atmosphere. 0. A recommended velocity profile function for use in stable and near-neutral conditions is given by (Hanna et al. Equation (18) can also be approximated by a power law..8 m/s2] = heat capacity of air [287 J/kg-0K] Typical values of LMO in various stability conditions are listed in Table I.1 Effect of Atmospheric Stability on the Mean Wind Profile Along with increased turbulence. one of the effects of atmospheric convection is to modify the shape of the mean velocity profile. The use of Monin-Obukhov scaling of meteorological variables is one of the criteria which differentiates these “advanced” dispersion models from their predecessors.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 6.4   z0  LMO  (18) Here LMO is the Monin-Obukhov length scale. The type of wind profile given in equation (18) is used in the advanced AERMOD dispersion model. In a convective atmosphere (LMO < 0).5 . The height LMO then defines the boundary between flow levels where mechanical turbulence due to friction dominates ( z < LMO ) and levels where convective turbulence dominates the flow ( z > LMO ). defined as: LMO = 3 − ρ c pTau* 0.

the classical Boussinesq approximation).5Us.Modelling Air Emissions for Compliance 7. The plume density ρp and ambient density ρa are equivalent. 2. Us   (20) Here Hs is the physical stack height and Rs is the stack radius. 1974):  w  H s ' = H s − 4 Rs 1. 3.5 − s  . in order to calculate the plume rise the momentum flux and buoyancy flux parameters are required. the stack effluent may be caught in the aerodynamic wake behind the stack and drawn downward (Figure 8). it is possible to derive simple. Because of the relative velocity of the plume and the wind. Various empirical equations and mathematical models have been proposed for calculating the rise of stack gas plumes in the atmosphere due to this initial momentum and buoyancy. Stack-Tip Downwash and Plume Rise MME 474A Wind Engineering December 2003 Most large-scale releases of pollutant are ejected from stacks with initial exit velocity ws and enough initial buoyancy (due to excess temperature of the effluent) to cause the plume to rise significantly before it is bent over by the wind. 12 . closed-form solutions of the equations of motion for the buoyant plume. These include: 1. The plume is advected by the mean wind. The wind speed Us remains constant above the stack. The rising plume entrains ambient air and cools before it eventually reaches a maximum height dependant on the ambient environment. One way to account for this effect is to introduce an effective stack height H s ' which may is reduced by the amount of downwash according to the following simple equation (Briggs. On a windy day. This genrally occurs when the stack exit velocity ws is less then 1. If ws > 1.2 Buoyancy and Momentum Fluxes In most EPA dispersion models. then it is assumed that there is no downwash the term in brackets is set to zero. 7. These parameters are based on initial exit conditions as follows: Momentum Flux: ρ  T Fm = ws2 Rs2  s  = ws2 Rs2 a Ts  ρa  Fb = gws Rs2 [m4/s2] (21) Buoyancy Flux: T −T  ρs − ρa = gws Rs2  s a  [m4/s3] ρa  Ts  (22) By making several simplifying assumptions about the flow. the plume boundary is very turbulent. 7. with Up = Us..5 times the mean wind speed Us at the top of the stack. the initially vertical plume is quickly bent over by the wind as shown in Figure 9. The plume is fully “bent over” (Figure 9) for its entire trajectory.1 Stack-Tip Downwash If the stack discharge velocity is very small. 4. except in terms involving the difference ρa − ρp which are used to calculate the buoyancy (i.e.

 3  ∆H b =  2   2β  1/ 3 Fb1/ 3 2 / 3 x . This effect can be seen in Figure 9. It is found that the maximum rise of a buoyant plume in a stable atmosphere is then given by (Briggs. it cools as it entrains ambient air. If the plume kept on rising. 1 T w  ∆H 0 =  a s  β  Ts U p    1/ 2 Rs . 1/  3  Fm 3 1/ 3 ∆H m =  2  x . the plume rise must be limited. Both Briggs (1984) and Davidson (1989) recommend using β = 0. θ a ∂z (27) The characteristic frequency N is called the Brunt-Vaisalla frequency and it is the natural frequency of wave-like motions in the atmosphere. However. (26) The coefficient β is a measure of the rate at which ambient air is entrained by the plume. Up (25) and ∆H 0 is a constant which accounts for the initial size of the plume. (24) and (25) suggest that a buoyant plume will rise indefinitely. based on matching the model to observed plume trajectories. as a hot plume rises. Thus in a stable atmosphere. (23) In equation (23). 1989): 3 3 3 ∆H = ( ∆H m + ∆H b + ∆H 0 ) 1/ 3 − ∆H 0 . ∆H m is the rise component due to initial momentum. 1984): 13 .Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 The resulting plume rise trajectory in a neutrally stable atmosphere is then given by (Davidson. 2/3  β  Up 1/ 3 (24) ∆H b is the rise component due to initial buoyancy. The final rise is determined by the following atmospheric stability parameter S = N2 = g ∂θ a . The plume will eventually reach an elevation where its internal temperature is the same as the ambient air and cannot rise further because the potential temperature θa increases with height at the top of the mixing layer.3 Final Plume Rise Equations (23). 7. it would be cooler than the surrounding air and will experience negative buoyancy which pushes it down.6.

the plume velocity Up must be evaluated at the final rise height. there can be considerable error in applying the individual plume rise formulae.2  b U u  p *     3/5  F  H s + 1.2 b  . in the AERMOD model the following equation is used:  F ∆H = 1. Since this speed is unknown a priori. convective atmosphere with turbulence is generated by solar heating of the ground depends on the following heat flux parameter (Briggs. A factor of two error in 14 . In such a case the plume is eventually broken up by atmospheric eddies and a different formulation is necessary. N 2 → 0 and equation (28) over predicts the final plume rise. The final rise in an unstable. which ensures a conservative analysis – i. For example. it is correct to use equation (23) in all stability conditions. involving the friction velocity u*. 1975). (32) This result is based on the idea that the plume rise will terminate when the dissipation rate in the plume has decayed to that of the surrounding turbulent air. maximum ground level concentrations.66  b 2  .3  b  U   p 3/ 5 H *−2 / 5 . since they are based on an aggregate average of many observed plumes. In that case the plume rise is calculated from (Briggs. If the atmosphere is close to neutral.e. This is called the “plume break-up model” (Briggs. some sort of iteration is required.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 1/ 3  F  ∆H = 2. H* = κ LMO Ta (31) In this case the final rise is given by: F  ∆H = 4. Prior to reaching the final plume rise.   U pu*    (29) Equation (28) is also found to over predict final rise in a stable atmosphere with calm winds. 1984): F  ∆H = 4  b3  N  1/ 4 . 1975): 3 g w 'T ' u* = . Most simple regulatory models simply use the wind speed at stack exit Us. For a particular plume. In calculating the final plume rise using any of the above models. (30) AERMOD takes the final plume rise to be the minimum of equations (28)-(30).. U N   p  (28) Here Up and N2 are calculated at the final plume height.

suppose the P-G equation for calculating σz for a point source is given by 15 .   3. The initial rate of plume spreading is greater than that due to ambient turbulence alone.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 the final rise height is quite possible. so it is often best to choose the minimum final rise height from a variety of formulas to ensure a conservative estimate of the resulting ground-level concentration. A similar approach is used in ISC3 and AERMOD. 1989). For example. As shown in Figure 10. it entrains air and is spread out by turbulent eddies. In addition. In practice.6). 7. R p = R0 + β∆H . we are often interested in area or volume sources of pollution with finite initial lateral and vertical dimensions. In order to account for the enhancement of the plume spread due to the buoyant rise phase. which is related to the physical source diameter by (Davidson. 8. (34) and β is the usual air entrainment parameter (β ≈ 0. T w  R0 =  a s   Ts U s  1/ 2 Rs . (33) Here R0 is an effective source radius. it is possible to account for any initial source dimensions by projecting a hypothetical point source some distance upstream of the actual area/volume source. This is done by first calculating the effective initial plume width σy0 and plume height σz0 due to the area or volume source. The upstream virtual source distances xy0 and xz0 are then calculated using the P-G sigma formulas.5 Buoyancy-Induced Dispersion As a plume rises to its final rise height.5    1/ 2 (35a)  2  ∆H  2  σ ze =  σ z +    . the plume size can be shown to increase linearly with rise height.5     1/ 2 (35b) Here ∆H is the plume rise. the SCREEN3 model calculates effective plume spread parameters downwind of the stack according to:  2  ∆H  2  σ ye =  σ y +       3. Initial Source Size and Virtual Source Methods The tabulated P-G plume sigmas given in Appendix A are based on an ideal point source of pollutant. and σy and σz are the P-G sigmas calculated for a point source at the stack (x = 0). nearby buildings can enhance the initial growth of plumes and can produce an effective volume source in their wakes. During the rise phase. for example.

Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 (36) σ z = ax b . (38) This is equivalent to assuming that the plume originates as a point source a distance (x + xz0) upstream. Table III gives recommendation for initial spread parameters based on the dimensions of the actual volume sources. In AERMOD. These methods are used in the ISC3 dispersion model. Vertical dimension ÷ 4.3 Vertical dimension ÷ 2. This idea is shown schematically in Figure 10. Procedures for Estimating Initial Lateral Dimensions and Initial Vertical Dimensions for Volume and Line Sources (ISC3 Model) Source Type σy0 Calculation σz0 Calculation Surface-based volume source Elevated source Length of side ÷ 4. the initial variance of plume as given in Table III is added to the predicted plume variance for a point source without virtual displacements: 2 2 2 σ y = σ y 0 + σ y. (39) In this equation: σ y 0 = the initial plume size σ y . ps . Although ISC3 uses a virtual source method. 16 .3 if elevated line source Line source represented by Length of side ÷ 2. then the virtual source distance xz0 is given by: σ  xz 0 =  z 0  .15 if ground level line source.15 series of volume sources It can be shown that the plume sigmas in Table III are equivalent to choosing the Gaussian plume so that it will have a concentration of 10% of its peak value at the physical edges of the source.15 Vertical dimension ÷ 4.3 Length of side ÷ 4.3 Vertical dimension ÷ 2. AERMOD differs in its treatment of the initial source size.  a  1/ b (37) At later stages of plume development downstream of the source we would calculate σz from σ z = a ( x + x z 0 )b . ps = the plume size assuming an initial point source σ y = the resultant plume size (including effect of initial source size). Table III.

1 Area Sources and Urban Pollution Problems MME 474A Wind Engineering December 2003 Often one is interested in calculating the cumulative effect of numerous small sources (small industries. The concentration downwind can be calculated by applying equation (8) to each infinitesimal area source dy×dz and then integrating over the whole area: q C ( x. if there is a vertical limit to mixing due to an elevated inversion. we get the result given by Gifford and Hanna (1970) for concentrations due to urban area sources: x1−b . the area integration must be performed numerically. For most problems.0. C= π a (1 − b) 2 (42) Assuming b ≈ 0. then the concentration is given by the simple expression: C ( x. if the receptor is within the boundaries of the area source of interest. the source rate is best expressed as an average pollutant flux per unit area [kg/m2-s].) that are distributed over a large area. etc.75 for typical urban dispersion. It is a measure of the area under the Gaussian distribution function. For the simple case where σ z = ax . 17 . if we are interested only in ground level concentrations (z = 0). and a simple box model calculation yields (for σz > zi): 1 The error function erf(x) occurs frequently in probability theory and diffusion problems. However in a large city where Ly is large compared to σy. we can replace b Lz by the distance x from the upstream edge. the dependence of the concentration on x (the size of the city) is weak. (40) The resulting integral can be expressed in terms of the error function erf(x) 1. residential heating. In such cases. the error function is approximately one.0) = σ π∫ 0 2 Lz −1 z dx (41) In addition. Of course. In addition. z ) = πU = 2 q πU Lz L /2  y2    z2   y   ∫ σ yσ z exp  − 2σ z2   − L∫ / 2 exp  − 2σ y2  dy dx    y 0      1 Lz  z  ∫ σ z exp  − 2σ z2  erf   0 1 2  Ly   dx  2 2σ  y   . vehicles. Consider a rectangular area source that has crosswind dimension Ly and along-wind dimension Lz. because the lower layers of air over a city tend to be well mixed. and is defined as: erf ( x ) = 2 π ∫e 0 x −u 2 du .Modelling Air Emissions for Compliance 8.0. then there will be a buildup of concentration in the atmosphere.

2 Complex Terrain When the complex terrain option is selected in ISC3. the elevated simple terrain option is chosen when the terrain height at the receptor is above the stack base elevation. U p zi (43) This last result explains why air quality in the center of large cities is generally very poor on calm days with a low-level capping inversion that traps the pollutants near the ground.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 C≈ q x . the plume is assumed to be terrain following. it does not follow the terrain contours. 9.y) downwind depends on the local elevation z(x. This is shown in Figure 11. With the exception of the CALPUFF model. y ) where: Hp′ Hp zs z(x.. there is no mechanism for lateral deflection of the plume due to terrain in these standard regulatory air dispersion models. i. The effective plume stabilization height above the ground at a point (x. Hp = Hs + ∆H). Complex Terrain Algorithms Although the derivation of the Gaussian Plume model assumes ideal conditions such as an infinite.1 Elevated Simple Terrain In SCREEN3. the plume height above ground at a point receptor (x. homogeneous area. the Gaussian plume model is often used to predict concentrations at receptors in complex. SCREEN3 and ISC3 are both similar in their approach to terrain modelling. Instead. The modified plume height above the receptor is calculated as H p ' = H p + z s − z ( x. 9. 9. only the vertical distribution function is affected. In all these models. and the method applied depends somewhat on the atmospheric stability class and whether the “elevated simple terrain” or “complex terrain” option is selected in the model. elevated terrain.y). which includes the effect of plume rise (i. is then kept constant. but below the release height. The approach used in AERMOD is more sophisticated. and also by the local terrain elevation above the stack base. flat.y) is then given by: 18 .e.y) = the effective plume height above the receptor = uncorrected plume height above the base of the stack = elevation of stack base = elevation of terrain at the receptor location (44) If the receptor is located above the ground.e. then the effective plume height above the receptor is reduced by the receptor height above ground (the so-called “flagpole” height). There are several ways to account for these effects. The net plume height above sea level.

The plume axis is assumed to remain fixed at the plume stabilization height above mean sea level. The terrain adjustment factor is stability dependent: f t = 0. zr).y). leading to high concentrations. where: (46) CT Ch .s .zs).s + (1 − ft ) × Ct . then the plume can impact on the obstacle.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 (45) H p ' = H s + ∆H − (1 − ft ) H t . However. who found from the results of wind tunnel experiments and potential flow theory that the wind streamlines never fully follow the underlying terrain. the terrain features are first processed by AERMAP (the terrain preprocessor). This concept is based on a recommendation by Egan (1975). the plume height relative to the stack base is deflected upwards by an amount equal to half the terrain height as it passes over complex terrain. during unstable and neutral conditions. zr). is then calculated as a weighted sum of these two contributions.5). The concentration at a receptor C(xr.s ft = total weighted concentration = concentration due to the horizontal plume state = concentration due to the terrain following plume state (i. In order to calculate the concentrations at downwind receptors located on elevated terrain.5 for stability categories A-D f t = 0 for stability categories E. and 2) a plume that completely follows vertical terrain (terrain following state). yr. the plume stabilization height is adjusted to partially follow the terrain. 9. CT = ft × Ch . yr. so that its centerline height above the local terrain is equivalent to the plume height above flat terrain. If the plume encounters an obstacle such as a hill that is above this height.. s Ct .e. AERMOD performs two concentration calculations corresponding to two extreme plume states. F . In neutral and unstable conditions (categories A-D).(= z(x. Figure 12. which assigns an effective height of terrain hc to each receptor point (xr. These are: 1) a horizontal plume as under very stable conditions such that the flow remains horizontal (and may impact on a large hill). where: Hs ∆H Ht = stack height = plume rise assuming flat terrain = terrain height of the receptor location above the base of the stack. under stable conditions the plume height is not deflected by the terrain. As a result of the terrain adjustment factor (ft = 0. 19 .3 AERMOD Approach to Complex Terrain In the AERMOD approach to complex terrain. the plume rises with the terrain) = terrain weighting factor.

There is often a single. The flow field in figure 13 results when the wind blows normal to a face. the parameter ft is in the range 0. high velocity gradients. a strong elevated trailing vortex may be generated. The time-averaged flow field around a rectangular building in a turbulent shear flow is shown in Figure 13. the reader is referred to the model formulation document by Cimorelli et al. The dispersion patterns around isolated buildings are generated by several flow features and are a function of obstacle shape.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 The weighting factor in this case is a function of the fraction of plume mass that lies below the dividing streamline height Hc2. The regions of separated flow are characterized by low wind speed. then there is usually no significant influence of the building on the dispersion (Hanna et al. 3. if a nearby source is released from a height greater than two-and-one-half times the building height. the regions of flow separation. As a rule of thumb.. In this case. approach flow turbulence and wind direction. For sources nearer the building. 2. there are five fundamental regions to consider: 1.1 Separated Flow Separation occurs at the sharp upwind edges of bluff obstacles embedded in a turbulent flow. 10. where the incident wind is first influenced by the building A region of separated flow over the upstream edges (roof cavity zone) A ground-based “horseshoe” vortex system that wraps around the base of the building A wake cavity of recirculating flow behind the building. 10. (47) 2 The dividing streamline Hc is the theoretical streamline height which divides the flow into two layers. one which remains horizontal and one which rises over the terrain feature. There is always some contribution due to the horizontal plume state. (1998). When the wind approaches the building on a diagonal. A displacement zone.5 – 1. 5. a representative building length scale can be defined as (Wilson. which is very unsteady. 1982). 20 . and flow reversal. Most research on this problem had been done for releases on or nearby individual buildings. taken from Hosker (1979). To quantify the size of the separated flow regions around a building. For further details on this approach. which means that the plume never completely approaches the terrain-following state. 4. high turbulence. The Effect of Buildings and Building Wakes on Plume Dispersion Many industrial stacks are located on top of buildings or in plant sites where there are large buildings nearby. 1979): 2 1/ LB = BS / 3 BL 3 . trailing vorticity and enhanced turbulence as shown in Figure 14 will likely affect the plume. One of the major challenges in regulatory dispersion modeling is to account for the effects of buildings on the near-field dispersion of a plume. A slowly decaying wake with reduced mean velocity and enhanced turbulence. such as a nuclear reactor building. Within this flow. large structure that dominates the scene. This concept comes from experiments involving flow over hills in stratified flow.

The frontal eddy trails around the sides of the obstacle and creates a trailing vortex with longitudinally oriented vorticity (Figure 13).3 ≤ L ≤ 3. The exact dimensions of the recirculation region downwind of an obstacle depend somewhat on the intensity and scales of turbulence in the approaching flow. H c = 0. This vortex can induce a “downwash” flow that draws the plume from rooftop stacks downward. It is shorter for narrow buildings. longer for wide buildings (Peterka et al. 10.3 1. but with reduced mean speed and enhanced turbulence..9 LB . At acute angles to the wind. there is a tendency for a frontal eddy to form at the upwind face. 21 . Depending on the length of the building. extends to 10-20H downwind. LR/H ≈ 1.24(W / H ) 0. 1985). depending on the W/H ratio of the building.0 H (49) For a building in the shape of a cube. which is approximately bell-shaped. there is a downward flow induced on the lower 2/3 of the upwind obstacle face as seen in Figure 14.5.8(W / H ) . The crosswind velocity defect. the rooftop leading edges of the building can produce longitudinally oriented vortices that transport the higher velocity air from above down into the central portions of the wake. the mean length of the “wake bubble” can be calculated using the following formula (Fackrell. the low speed flow recirculates and the instantaneous wind velocities fluctuate randomly so that any pollutants are rapidly dispersed throughout the cavity region. the roof top cavity may or may not reattach before the rear edge of the building.2 Frontal Eddy and Trailing Vorticity When a bluff body is placed in a shear flow. There is a characteristic delay time before any entrained pollutant is released to the outer flow (the “residence time”).22 LB . 1 + 0. 1979): Lc = 0. (48) Therefore the roof cavity will reattach to the roof as in Figure 14 if Lc < L. The length and height of the rooftop recirculation cavity is estimated as (Wilson. However. This generally leads to a greater rate of dispersion for plumes that are entrained in the wake. Within the wake bubble. Downwind of the wake cavity the wind begins to return to the conditions of the approach flow. where L is the alongwind length of the building. Because the stagnation pressure increases with height above the ground. This effect can extend some 50-100H downwind of the building (Hosker. 1984): LR  L  =  H H −0.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 where: BS BL = smaller of upwind building face dimensions H or W = larger of upwind building face dimensions H or W. 1984). for a broad range of building dimensions.

1984). some multiple of the building height) then the plume is assumed to enter the wake cavity.g. the modified plume rise algorithm due to Schulman and Scire (1980) can be used to calculate the initial rise. and increased lateral spreading. the SCREEN3 model calculates a cavity concentration using the following formula (Hosker. K c Af U (50) = emission rate = building cross-sectional area normal to wind (H×W) = reference wind speed (typically at 10 m height) = non-dimensional concentration constant (= 1. A simple model recommended by Turner (1969) is: σ 'y0 = W H . Cc = where: Q Af U Kc Q . ignoring the effect of the building. σ 'z 0 = . 10. 4.. This building-enhanced dispersion can be accounted for by virtual source displacements: σ 'y0 = σ y ( x + xy0 ) σ ' z 0 = σ z ( x + xz 0 ) (51) The virtual distances xy0 and xz0 are chosen so that the initial value of the dispersion parameters at the rear face of the building (σ′y0 and σ′z0) are some fraction of the building width and height. it starts with an initial size approximately equal to the building cross-sectional area Af.3 Effect of Building Downwash and Wake Flow on Plume Dispersion There are several available models to account for the enhanced dispersion of effluent caused by buildings located near a stack. If a plume is entrained in a building wake.15 (52) These virtual source estimates are illustrated in Figure 10. If this plume height is less than some criteria (e. Most of these are empirical correlations based on the results of controlled wind tunnel experiments and more limited field experiments. Since the size of the wake cavity depends on the building orientation relative to the wind.5). 22 . In many simple models for rooftop stacks.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 and can contribute to downwash of plume material into the wake. When a plume is fully entrained into a building wake. the plume rise due to momentum is calculated at a distance of two building heights downwind.3 2. In cases where the source is right on or very near the building. respectively. the cavity dimensions should be calculated for at least two extreme building orientations in order to get a reasonable bound on the cavity concentration estimate.

Using extensive wind tunnel and field data. For a very tall building (H >>W). the dispersion parameters depend on the building dimensions. This is done in SCREEN3 and ISC3 models. Neither of these models will predict concentrations directly in the wake cavity (x < 3H).W). otherwise unrealistically high vertical dispersion is predicted. The Huber-Snyder model has different expressions for the near wake (3 < x/H < 10) and far wake (10 < x/H) regions. ISC3 and AERMOD use the HuberSnyder (1982) model to calculate the enhanced dispersion for sources entrained in the wake. evaluated at two building heights downstream. This model is described in some detail in the original reference and in the model formulation documents for ISC3 and AERMOD. a more sophisticated algorithm due to Schulman and Scire (1980) is used to calculate the downwash effect because the initial rise is reduced by the building-enhanced dilution. then the latter are used. the plume variances are added: 2 2 2 σ yT = σ y + σ y '2 and σ zT = σ z2 + σ z '2 .35W + ( x − 3H ) /15 σ z ' = 0.H) is the lesser of the building frontal dimensions.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 For building-affected dispersion calculations in the wake.H). is less than H + 1. (54) where: 2 σ yT 2 σy = total variance of the plume = Pasquill-Gifford variance calculated for a point source σ y '2 = Huber-Snyder variance evaluated at x = 10H Special cases may arise when applying equation (53).2 building heights. if the stack height is less than H + 0. In the near wake. the P-G sigmas are used with virtual source displacements xy0 and xz0 chosen so as to match the plume dimensions at x = 10H. Further details on incorporating the H-S model are described in the ISC3 Model documentation. If W >>H (a squat building) then H replaces the value of W in calculating σy′. The overall effect of enhanced dispersion in building wakes is to decrease the maximum groundlevel concentrations for low level releases or to increase the ground level concentrations for elevated sources. W replaces H in the σz′ equation. The ISC3 model modifies only σz if the plume height is greater than 1. In ISC3 and AERMOD. If the building-affected sigmas from Equation (53) are less than the P-G sigmas at the same distance downwind. In AERMOD. The Huber-Snyder enhanced dispersion model is only applied if the plume height. Fackrell (1984) found that the differences between the measured and predicted concentrations using the Huber-Snyder and other simple building-affected dispersion models was often less than a factor of two or three. 23 . which uses a more complicated method of calculating the plume sigmas. with: σ y ' = 0.5×min(W.7 H + ( x − 3H ) /15 (53) In the far wake. where min(W.5×min(H. however SCREEN3 can be used for that.

selection of model options and the interpretation of results. ISC3. However.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 This is within the expected limits of accuracy for most Gaussian plume models (Beychok. Summary Although the Gaussian plume model (GPM) is based upon many simplifying assumptions about the dispersion process. sensitivity analyses.S. such as SCREEN3. 1995). non-uniform wind profiles. 11. To become an “expert” user. dry or wet deposition. buoyancyinduced dispersion. 24 . In this treatment of buildings. complex terrain and the influence of buildings. stack-tip downwash. These include the influence of atmospheric stability. EPA models make use of several special algorithms or semi-empirical corrections to account for the various effects. 1994). have been run through extensive physical audits. it is applied to a wide array of dispersion scenarios and some form of this model is adopted in most regulatory air pollution models for continuous releases. the resulting model codes are quite robust and can be used in a wide variety of situations combining many separate effects. The net concentrations downwind are found by adding the two concentration distributions due to the wake and elevated source. finite initial source dimensions. none of the models seems to cope well with acute wind angles. and quality assurance studies using benchmark data in order to justify their use in environmental assessment. This part is then accounted for by locating a virtual source upwind (Figure 15).. The modifications to the basic GPM make extensive use of wind tunnel and measured field data.5H). it is essential that one read the model formulation documentation and some of the references in this document. a rooftop plume is partitioned so that only a fraction of the plume is entrained in the wake. Because of strong peer reviews and model validations studies. plume trapping below elevated inversions. the U. An understanding of the fundamental concepts used in these models is important for the most intelligent use to be made of the models. The part of the plume that is not entrained is treated as an elevated point source. All of the EPA codes. A brief summary of the algorithms used to incorporate these various features has been provided in this review. One promising approach is that used by the ADMS model (Carruthers et al. In order to extend the applicability of the GPM to realistic scenarios. This background knowledge is required to ensure that the most sensible choices are made in all aspects of the data input stages. where the trailing vortices can induce severe downwash and increased ground level concentrations even for more elevated sources (Hs > 2. many of the algorithms in the advanced EPA models are based on simplified physical models of the various dispersion processes. combined with empirical data. As shown in this review. fumigation. AERMOD and CALPUFF.

EPA (1995). Dept. Aero. Report AFCRC-RT-58-235. Cimorelli. U. EPA (1995). Briggs. Tennessee. Vols. Volume II – Description of Model Algorithms.J. Paine.C. Robins. (1975).Modelling Air Emissions for Compliance 12. US EPA Report EPA-454/B-95-004.G. Ind. pp.B. G.gov/scram001. USAF Cambridge Research Laboratory. Haugen. Beychok. A. The Danish Gaussian Air Pollution Model (OML): Description. MA.C. Publication no.R. Olesen and U.F. (1976). 99-105.J. Soc. SCREEN3 User’s Guide.J. Washington. D. J.J. US EPA Report EPA-454/B-95-003B. Amer. Mass. Air Dispersion Modeling: Keep Predicted Values in Perspective. Wilson. M. In Lectures on Air Pollution and Environmental Impact Analysis. ed. In Air Pollution Modelling and its Application V (ed. Briggs. R. Plume Rise Predictions. (1958). Diffusion Estimations for Small Emissions.A. Plume Rise and Buoyancy Effects. Turbulent Diffusion in Complex Terrain. Project Prairie Grass: A Field Program in Diffusion. Available at EPA Support Centre for Regulatory Air Models: www. Hunt. D. Environmental Engineering World.B. (1995). Randerson).A.J. J. Lee and W.R. I and II.. G. Boston. De Wispelaere. Apsley.. Atmospheric Egan. 139-153. Mass. Briggs. American Meteorological Society. R. Atmospheric Environment 23. Environment 10.epa. R. Draxler.G. 16-18. Schiermeier and Gillani).R. 341-349. Determination of Atmospheric Diffusion Parameters. US Atomic Energy Commission Report ATDL-106. Davidson. of Energy. References MME 474A Wind Engineering December 2003 Barad. Meteor. Smith (1994). Chapter 8 in Atmospheric Science and Power Production (ed. UK-ADMS: A New Approach to Modelling Dispersion in the Earth’s Atmospheric Boundary Layer. Test and Sensitivity Analysis in View of Regulatory Applications.D. Berkowicz. Thomson and F.A. M. A. Weil. 59. Weng. Oak Ridge.L.C. Boston. Holroyd. R. A.D. (1974). AERMOD: Description of Model Formulation. US Atomic Energy Commission.-Dec.. G.R. (1975)..A. User’s Guide for the Industrial Source Complex (ISC3) Dispersion Model. Simultaneous Trajectory and Dilution Predictions from a Simple Integral Plume Model. B.S. H. 52. 113 pp. Wind Eng.S. 51 pp. D. Peters (1998). Nov.. W.A. Geophysical Research Papers No. G. R. in Lectures on Air Pollution and Environmental Impact Analysis (D. D. J. D. DOE/TIC-27601. Carruthers. Venkatram. (1984).). S. (1989). R. 25 . Perry. Torp (1987).

26 . Reno Nevada.J. of Int. Flow and Diffusion Near Obstacles. In Fourth Symposium on Turbulence. Publication no. Wind Eng. NY 10017. J. J. Aero.P. and S. Public Health Service. Guidelines for Use of Vapor Cloud Dispersion Models.C. In Engineering Meteorology (ed. McElroy.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 Fackrell. J. F. S. Huber. Hosker. S. pp. 89-98. G. 1385-1398. Hosker. Amer. S. Gifford. Huber. 97-118. 481-525. New York. Randerson).C.R. Hanna (1970). Atmospheric Environment 25A.C. An Examination of Simple Models for Building Influenced Dispersion.R. Prevention Assocs. Pooler (1968). (1984). U. 102 pp.. P. and F. Urban Air Pollution Modelling. E. J. Hanna. 603-609. Washington. Elsevier Publishing Co. Atmospheric Environment 13. U. Diffusion and Air Pollution. 1237-1249.P. and W. Boston. D.H. Chang (1996). Irwin.H. The St. Hanna. Empirical Estimation of Wake Cavity Size Behind Block-Type Structures. 16. Ind. (1960). Publication no. D. 15-18.E. 345 East 47th St.A. pp. 191-194. Union of Air Poll. Washington. Chapter 7 in Atmospheric Science and Power Production (ed. F.. (1979). Drivas and J.R.E.A.S. Briggs and R.S.. Hanna. 285 pp. Plate). Handbook on Atmospheric Dispersion. Atmospheric Dispersion Calculations Using the Generalized Gaussian Plume Model. A Theoretical Variation of the Wind Profile Power-Law Exponent as a Function of Surface Roughness and Stability. Snyder (1982). (1984). Confidence Limits for Air Quality Model Evaluations as Estimated by Bootstrap and Jacknife Resampling Methods. R. Presented at 1970 Int. DOE/TIC-11223.L. J. Dept. Nuclear Safety 2. Turbulent Diffusion Near Buildings. Evaluation of a Method for Estimating Pollution Concentrations Downwind of Influencing Buildings. pp. Soc. Atmospheric Environment 16. R.. Fackrell. Wind Tunnel Investigation of the Effects of a Rectangular-Shaped Building on Dispersion of Effluent from Short Adjacent Stacks. Parameters Characterising Dispersion in the Near Wake of Buildings. A. Atmospheric Environment 23. of Energy. AIChE/CCPS. R.S. (1979).P. D. Mass. 241-326. Louis Dispersion Study. Hosker (1982).N. (1982). pp. 56-59. Gifford. D. Jan. (1984). NAPCA Report AP-53. U. Amsterdam.J. Atmospheric Environment 18. (1984).. Air Pollution Conf.S. Dept. Washington.H. Meteor. Meroney. 2837-2848.R.A. DOE/TIC-27601.. of Energy. A.C. (1989).

2nd Edition. Aero. p. D. (1980). H. Turner. Department of Health.. (1974).B. A104. 33-49. Proc. (1976). Sutton. Schulman. and J. Flow Patterns Over Flat Roofed Buildings and Application to Exhaust Stack Design. 502-519. Academic Press. (1979). ASHRAE .T. Dutton (1984). Air Pollution: Volume I . Workbook of Atmospheric Dispersion Estimates: An Introduction to Dispersion Modeling. New York. Publication no. J. Peterka.Air Pollutants. Ind.C.E. O. Atmospheric Environment 12. Stern. Meteorol. D. London.B. Environmental Research and Technology Inc. D. 640. (1923).F. A. 1791 Tullie Circle. Roy. Document P-7304.Modelling Air Emissions for Compliance MME 474A Wind Engineering December 2003 Panofsky. Atmospheric Diffusion. Wilson. CRC Press. Soc. (1969).M. Pasquill. Workbook of Atmospheric Dispersion Estimates. (1997). ASHRAE Transactions 85. Turner. (1953). J.B. Roberts.S. ISBN 1 883413 45 1 27 .J. Chichester. (1994). 21. Mag.S. R. ISBN 0 85312 015 3. van Ulden. Kothori (1985). 333 pp.A. Education and Welfare. Turner. 284-295. (1978).L. U. F. (1979). 29. L. London.A. Wilson. and K.R. ISBN 0471 057142. D. ISBN 0 12 666601 6. J.A. Micrometeorology. (London). 999-AP26. F. Pasquill. Atmospheric Turbulence: Models and Methods for Engineering Applications.J. Chapter 15 in ASHRAE Fundamentals. Wind Flow Patterns about Buildings. 21-38. 90. ISBN 1 56670 023 X. Air Poll. N. Meroney. A.N. Buoyant Line and Point Source (BLP) Dispersion Model User’s Guide. O. Airflow Around Buildings. Scire (1980). 2125-2129. Simple Estimates for Vertical Diffusion from Sources Near the Ground. Atlanta. (1961). New York.G.. Wiley. Control Assoc. Wind Eng. and J. GA 30329. Atmospheric Dispersion Modeling: A Critical Review. Ellis Horwood Limited. The Estimation of the Dispersion of Windborne Material. D. Their Transformation and Transport. McGraw-Hill. 2nd Ed.

Sign up to vote on this title
UsefulNot useful