DOE TIC-11223 (Handbook On Atmospheric Diffusion)

DOE/TIC-11223
( DE82002045)
Handbook on
ATMOSPHERIC DIFFUSION
Steven R. Hanna
Gary A. Briggs
Rayford P. Hosker, Jr.
Atmospheric Turbulence and Diffusion Laboratory

National Oceanic and Atmospheric Administration
. .
Prepared f o r the
Office of Healtli a n d Environmental Research
Ofilce of Energy Research
U . S. Department of Ener,q\
1982
Jean S. Smith, Publication Editor

Editing, composition, proofreading, book design,
illustrations, and page makeup for this publication
were performed by staff members of the Technical
Information Center. This documait is
Published by
TECH N 1C A L IN F 0R M A TI 0 N CENTER
U. S. DEPARTMENT OF ENERGY
NOTICE
International Cop>-right. 0 1:. S. Department o f Energy. 1982, under the

provision, of rlic lInn,vr\al Copyright Convenrion. United States copyright
i s nor ,i\xrted iindrr r h c llnircd Sratrs Copyright L a w . 'I'irle 17, U n i r e d -
S t a t e s Cock.
Library o f Congress Cataloging in Publication Data
Hanna, Steven R.
Handbook on atmospheric diffusion.
Includes bibliographical references and index.

"DOE/TIC-1 1223."
1. Atmospheric diffusion-Handbooks, manuals, etc. 2. Smoke
plumes-Handbccks, manuals, etc. 3. Cooling towers-Climatic
factors-Handbooks, manuals, etc. I. Briggs, Gary A. ,
II. Hosker, Rayford P. Ill. United States.
Dept. o f Energy. Office o f Energy Research. IV. United States. Dept
of Energy. Office o f Health and Environmental Research. V. Title.
QC8 80.4.044 H36 551.5'153 81 -1 5 1 49
ISBN 0-87079-127-3 AACR2
Available as DE82002045 (DOE/TIC-11223) for $10.75 from
National Technical Information Service

U. S. Department of Commerce
Springfield, Virginia 22161
DOE Distribution Category UC-11
Printed i n the United States of America
1982
With the Clean Air Acts and increased environmental Basic meteorological concepts are covered first
consciousness, many engineers, consulting companies, and then plume rise, source effects, and diffusion
planners, and meteorologists find themselves pro- models. Chapters on cooling tower plumes and urban
pelled into the work of calculating atmospheric diffusion are included. Suggestions are made for
diffusion. Many of these people are not interested in calculating diffusion in special situations, such as for
knowing the detailed theoretical derivation of a instantaneous releases (puffs), over complex terrain,
formula and its complete set of references. All they over long distances (10 km t o global scales), and
want to know are the best current formulas for their during times when chemical reactions or dry or wet
problems plus a simple physical description of the deposition are important.
principles of analysis. This book should be helpful to This work was performed under an agreement
those who must make such problem-solving calcula- between the U. S. Department of Energy and the
tions of atmospheric diffusion. National Oceanic and Atmospheric Administration.
The book can be used also as a textbook for a
one-quarter course at either the upper undergraduate
or the graduate level. In fact, the basic outline [Editors Note: Dr. Steven Hanna and Dr. Gary
evolved from a graduate course on atmospheric Briggs were with National Oceanic: and Atmospheric
diffusion taught in the Environmental Engineering Administration, Atmospheric Turbulence and Diffu-
Department at the University of Tennessee. sion Laboratory, Oak Ridge, Tenn., during the
The number of pages was purposely limited to preparation of this book; at present Dr. Hanna is with
make the book more usable. A detailed index permits Environmental Research and Technology, Lexington,
quick location of subject areas, and a few problems Mass., and Dr. Briggs is with Environmental Pro-
are provided after each chapter. tection Agency, Research Triangle Park, N.C.]
Steven R. Hanna
Atmospheric Turbulence and Diffusion Laboratory
September 1981
iii
Preface . . . . . . . . . . . . . . . .
...
111 2-8 Multiple Sources . . . . . . . . 17
Problems . . . . . . . . . . . . . 17
1 Meteorology . . . . . . . . . . . . 1
1-1 Introduction . . . . . . . . . . 1 3 Source Effects . . . . . . . . . . . 19
1-2 General Circulation . . . . . . . 1 3-1 Overview . . . . . . . . . . . 19
1-3 Vertical Temperature Structure 3-2 Stack Aerodynamic Effect . . . . 19
and Stability . . . . . . . . . . 2 3-3 Structure of Flow Around
1-3.1 Adiabatic Temperature Gradient 2 Buildings . . . . . . . . . . . 19
1-3.2 Stability . . . . . . . . . 3 3-4 Diffusion Calculations
1-4 Structure of the Planetary Around Buildings . . . . . . . . 21
Boundary Layer . . . . . . . . 4 3-4.1 Isolated Sources Upwind
1-4.1 Turbulence Fluxes . . . . . 4 of Buildings . . . . . . . 22
1 - 4 2 Ekman Spiral . . . . . . . 5 3-4.2 Sources Close to Buildings . . 22
1-4.3 Similarity Theory Gives Wind Problems . . . . . . . . . . . . . 24
and Temperature Profiles in
Surface Layer . . . . . . . 6 4 Gaussian Plume Model for Continuous
1-4.4 Turbulence Parameters . . . 7 Sources . . . . . . . . . . . . . . 25
41 Why Use theGaussianModel? . . . .
..
1-5 Use of Spectra to Estimate 25
Turbulence Parameters . . . . . . 8 4 2 Formof theGaussianMode1 . . . . 25
1-6 Lagrangian Turbulence . . . . . . 9 4-3 Stability Classification Schemes . . 27
Problems . . . . . . . . . . . . . 10 4 4 Choice of oYand . . . . . . . 27
44.1 Stability ClassMethod . . . 27
2 PlumeRise . . . . . . . . . . . . . 11 4-4.2 Theoe ando. Method . . . 30
2-1 Introduction . . . . . . . . . . 11 4 5 Wind-Speed Variation with Height . 31
2-2 Top-Hat-ModelEquations . . . . . 11 4 6 Maximum Ground Concentration
2-2.1 Definitions . . . . . . . . 11 and Fumigation . . . . . . . . . 32
2-2.2 Set of Equations for 4 7 Averaging Times and Peak-to-Mean
Vertical Plume . . . . . . 12 Concentration Ratios . . . . . . 33
2-2.3 Set of Equations for 4-8 Sector Model for Long Sampling
Bent-Over Plume . . . . . 13 Times . . . . . . . . . . . . 34
2-3 Plume Trajectory Near Source . . . 13 Problems . . . . . . . . . . . . . 35
2-3.1 Vertical Plumes . . . . . . 13
2-3.2 Bent-Over Plumes . . . . . 13 5 Statistical Models of Diffusion from
2-4 Plume Rise Limited by Ambient Continuous-PointSources . . . . . . . 36
Stability . . . . . . . . . . . 14 5-1 Introduction . . . . . . . . . . 36
2 4 . 1 Vertical Plumes . . . . . . 14 5-2 Taylors Theorem . . . . . . . . 36
2-4.2 Bent-Over Plumes . . . . . 14 5-3 InfluenceofEddy S i z e o n o . . . . 38
2-5 Plume Penetration of Elevated 5-4 Lagrangian-Eulerian Relations . . . 39
Inversion . . . . . . . . . . . 14 5-5 Monte Carlo Particle Trajectory
2-6 Plume Rise Determined by Models of Diffusion . . . . . . . 40
Ambient Turbulence . . . . . . 15 Problems . . . . . . . . . . . . . 40
2-6.1 Nearly Neutral Conditions . 16
2-6.2 Convective Conditions . . 16 6 Puff Diffusion . . . . . . . . . . . 41
2-7 Maximum Ground Concentration 6-1 Introduction . . . . . . . . . . 41
withBreakupMode1 . . . . . . . 17 6-2 Statistical Approach . . . . . . . 41
V
-.--
vi
6-3 Similarity Approach . . . . . . . 42 9-5.2 Trajectory Models . . . . . 62

6-4 Applications . . . . . . . . . . 44 9-5.3 Grid Models with Winds
Problems . . . . . . . . . . . . . 44 Prescribed . . . . . . . . 62
9-6 Environmental Protection Agency
7 Similarity Modelsof Diffusion . . . . . 46 Models . . . . . . . . . . . . 62
7-1 Introduction . . . . . . . . . . 46 9-7 Model Evaluation . . . . . . . . 63
7-2 Diffusion of Continuous Plumes Problems . . . . . . . . . . . . . 66
in the Surface Layer . . . . . . . 46
7-2.1 Neutral Conditions . . . . . 46
7-2.2 Nonneutral or Adiabatic 10 Removal Mechanisms . . . . . . . . 67
Conditions . . . . . . . . 47 10-1 Introduction . . . . . . . . . 67
7-3 Diffusion in the Full Depth of 10-2 Dry Deposition . . . . . . . . 67
the Daytime Planetary 10-2.1 Gravitational Settling . . . 67
Boundary Layer . . . . . . . . 48 10-2.2 Deposition of Gases and of
Problems . . . . . . . . . . . . . 49 Particles with Radii Less
Than About 1 0 p m . . . . 68
8 Gradient Transport(K) Models . . . . . 50 10-3 Wet Deposition . . . . . . . . 71
8-1 The Basic Gradient Transport . 10-4 Chemical Removal . . . . . . . 72
Model . . . . . . . . . . . . 50 10-5 Removal Processes in the Box
8-2 Analytical Solutions . . . . . . . 50 Model . . . . . . . . . . . . 73
8-2.1 One-Dimensional Equation. Problems . . . . . . . . . . . . . 73
Time.Dependent. Constant K.
No Wind. Instantaneous
Area Source . . . . . . . 50 11 Cooling Tower Plumes and Drift
8-2.2 Three Dimensions. Time- Deposition . . . . . . . . . . . . . 74
Dependent. Constant K. 11-1 Introduction . . . . . . . . . 74
No Wind. Instantaneous 11-2 Plume Rise from Cooling
Point Source . . . . . . . 51 Towers . . . . . . . . . . . 74
8-2.3 Two.Dimensiona1. Time- 11-2.1 Visible Plume
Independent. Variable u Dimensions . . . . . . . 76
and K. Continuous 11-2.2 Numerical Approach for
Ground-Level Line Source . . 51 Deep Visible Plumes . . . . 77
8-2.4 Three.Dimensional. Time- 11-3 Drift Deposition . . . . . . . . 78
Independent. Constant u Problems . . . . . . . . . . . . . 80
and K. Continuous-Point
Sourceat GroundLevel . . . 52 12 Air-Pollution Meteorology in Complex
8-3 Numerical Solutions of the Terrain . . . . . . . . . . . . . . 81
Diffusion Equation . . . . . . . 53 12-1 Introduction . . . . . . . . . 81
8-3.1 Numerical Instabilities . . . . 53 12-2 Meteorology . . . . . . . . . 81
8-3.2 Specifying the Vertical 12-3 Diffusion Calculations . . . . . . 84
Diffusivity . . . . . . . . 54 Problems . . . . . . . . . . . . . 86
8-4 Higher Order Closure . . . . . . . 55
Problems . . . . . . . . . . . . . 56
13 Long-Range Transport and
9 Urban Diffusion Models . . . . . . . 57 Diffusion . . . . . . . . . . . . . 87
9-1 Importanceof Emissions . . . . . 57 13-1 Introduction . . . . . . . . . 87
9-2 Box Model . . . . . . . . . . . 57 13-2 Modeling Concepts . . . . ... 87
9-3 The Atmospheric Turbulence and 13-3 Application t o an Inert Tracer . . . 89
Diffusion Laboratory Model . . . . 59 Problems . . . . . . . . . . . . . 90
9-4 Street Canyon and Highway
Submodels . . . . . . . . . . .. 61
9-5 Computerized K Models for References . . . . . . . . . . . . . . 91
Urban Diffusion . . . . . . . . 62
9-5.1 An Urban Diffusion Author Index . . . . . . . . . . . . . 98
Model That Also Predicts
Windsand Temperatures . . 62 Subject Index . . . . . . . . . . . . . . 100
Meteorology
1-1 INTRODUCTION North of 30" N latitude, the direct Hadley sell

breaks down, and energy is transported by traveling
high and low pressure systems moving from west to
To set the stage for the remainder of the
east. Warm southerly winds and cold northerly winds
handbook, in this chapter we must briefly review
help accomplish the energy transport. Even more
several aspects of meteorology, including the general
energy is transported by latent heat processes, where,
circulation, vertical stability, and surface-layer struc-
for example, Gulf of Mexico water is evaporated,
tiire. Most students are eager to begin immediately
transported northeastward, and condensed. again as
the study of applications of the Gaussian plume
precipitation. For each gram of water involved in this
model; however, even the application of the Gaussian
process, 540 cal is transported toward the north.
plume model requires a knowledge of wind roses and
Between 60" N and the pole is another wind belt with
stability and an appreciation of the influence of wind
an easterly component a t the surface, but this
shear o n the range of usefulness of the model. Also,
circulation is not well defined.
the latest developments in plume-rise theories require
Upper-level winds strongly influence winds near
the ability to understand and estimate vertical pro-
the surface, where most diffusion problems occur. In
files of eddy dissipation rate. Therefore this chapter
will be a useful reference for the remainder of this general, the speed of the upper-level winds is propor-
handbook. tional to the slope of surface of constant pressure.
The atmospheric pressure -(p) typically varies by no
more than about 5% a t sea level over the earth's
surface. However, the temperature (T) could be
1-2 GENERAL CIRCULATION 300K at the equator and 240K a t the poles. The
equation of state for the atmosphere,
The sun is the source of nearly all energy received
by the earth's atmosphere, and the spherical shape of
the earth is responsible for the unequal absorption of
this energy by the earth's surface and the atmosphere. where R is the gas constant (0.287 x lo' ergs g-'
Without the transport of heat by the atmosphere and OK-'), tells us that the density ( p ) must b e less a t the
the oceans from the equator to the poles, tempera- equator than at the poles. The hydrostatic equation,
tures would be several tens of degrees colder a t the
poles and warmer at the equator. However, the fact is
that there is a strong poleward transport of heat that
is accomplished b y direct Hadley cells, traveling high
and low pressure systems, and major perturbations,
such as hurricanes. In the northern hemisphere where.2 is the height and g is the acceleration of
Hadley cell, air rises over the equator (causing much gravity (980 cm/sec2), then suggests that the pressure
rainfall), moves a t high elevations toward the north, decreases with height faster a t the poles than a t the
descends a t about 30" N latitude (causing dry desert equator. It follows that, if pressure p is constant (say
regions), and then moves as the Northeast (NE) trade 1000 mb, or lo6 dynes/cm2) a t sea level, then any
winds near the surface from 30" N toward the other constant pressure surface aloft (say 500 mb, or
equator. The NE trades are known as the most 0.5 x lo6 dynes/cm2) will slope downward from the
persistent general wind system on earth. equator to the pole, as in Fig. 1.1.
1
2 ATMOSPHERIC DIFFUSION
p = 0.5 x lo6
dynes/cm2
/LOW TEMPERATURE\
HIGH DENSITY
FORCE .
Fig. 1.1 Cross section of the earths atmosphere, showing how sloping pressure surfaces result at
mid-atmosphere. Westerly winds are caused by a balance between pressure forces and Coriolis forces.
The equation of motion says that air will first be Other hydrodynamic forces, which are beyond the
accelerated toward the poles along the upper constant scope of this chapter, frequently cause the westerly
pressure surface in the figure: flow to be compressed into narrow belts, called jet
streams, with speeds up to 200 km/hr.
Meteorological data are gathered from many
stations across the globe and are stored at the
National Climatic Center, National Oceanic and
where u = easterly component of wind speed Atmospheric Administration ( N O M ) , Asheville, N.C.
v = northerly component of wind speed Many statistical operations (e.g., annual wind roses or
y = northerly
. - coordinate axis frequency distributions of wind direction and speed)
subscript p = constant pressure surface have already been carried out and can be obtained
f = Coriolis parameter, which is equd to from the National Climatic Center at very reasonable
two times the earths rotation rate times prices. Surface weather summaries at larger National
the sine of the latitude Weather Service stations are collected into reports
called Local Climatological Data, which are mailed
The parameter f is of the order of sec- . The to subscribers monthly. The Climate Atlas of the
apparent Coriolis force arises as a result of the earths United Stutes (U. S. Department of Commerce, 1968)
rotation, which constantly displaces a Cartesian co- contains many data useful for diffusion calculations.
ordinate system fixed to the surface. An analogy is
given by rolling a marble from the edge o f a rotating 1-3 VERTICAL TEMPERATURE
record turntable toward the center. The marble will
STRUCTURE AND STABILITY
encounter regions with less angular momentum than
it has. To an observer fixed to the turntable, the
marble will always curve toward the right if the
1-3.1 Adiabatic Temperature Gradient
turntable is rotating counterclockwise. SiSlarly, in
the northern hemisphere the Coriolis force is t o the
right, and in the southern hemisphere it is to the left. If a parcel of dry air is moved vertically without
The poleward pressure force is thus balanced by a exchanging heat with its environment (i.e., adiabati-
Coriolis force toward the equator in both hemi- cally), the first law of thermodynamics becomes
spheres, which causes general westerly flow at mid-
levels in the atmosphere at mid-latitudes. The magni- 1
tude of the resulting geostrophic wind speed is 0 = cp dT - -dp
P
given by setting du/dt = 0 in Eq. 1.3, which yields the
formula
where cp is the specific heat of air a t constant
pressure (lo ergs g- K-l) and T must be in
degrees Kelvin (or absolute). Substituting from the
METEOROLOGY 3
hydrostatic equation (Eq. 1.2) yields the formula for temperature gradient is a function of temperature
the adiabatic temperature gradient: ranging from about -0.9"C/100 m in cold polar
climates to about -0.4"C/lOO rn in warm tropical
climates. The saturated water-vapor mixing ratio (ni,)
is a function of temperature and is presented graphi-
cally in Fig. 11.4 of Chap. 11. With each 10C rise in
or a temperature decrease of about 1C for each temperature, m, roughly doubles.
elevation increase of 100 m. The potential tempera-
ture (e) is an important parameter defined from
Eq. 1.5 by substituting for (l/p) with the use of the
equation of state (Eq. 1.1) and by integrating from 1-3.2 Stability
sea-level pressure ( l o 6 dynes/cm2) to the pressure p
a t any level. Meteorologists distinguish threc states of the
atmospheric surface layer: unstable, neutral, arid
stable. These adjectives refer to the reaction of a
parcel of air displaced adiabatically in the vertical
direction. Figure 1.2 shows the environmental l a p
rates that give rise to these stability classes. In each
where the ratio R/cp equals 0.286. In other words, example the parcel originates a t the height indicated
the potential temperature of a parcel of air a t by the circle in the figure; at this height the
temperature (T) and pressure (p) is the temperature temperature of the parcel is the same as that of its
that would result if the parcel were brought adia- environment. If the density of the parcel is less than
batically from a pressure p to a pressure of lo6 that of its environment (pp < pe or T, > Te), then
dynes/cm'. It follows thet the parcel is accelerated upward. If the density of the
parcel is mort: than that of its environment (pp > pe
or Tp < Te), then the parcel is accelerated downward.
If the density of the parcel is the same as that of its
environment (TP = Te), then the parcel continues a t
and its original speed. For the example of the unstable
layer, the parcel is continually accelerated away from
its origin. The example of the neutral layer shows
that the temperature of the parcel is always the same
where z is the height above mean sea level. The as that of its environment, and there is no force on it.
adiabatic potential temperature gradient (de/dz) is The sketches above the temperature profiles illustrate
zero. A wide range of temperature gradients is the gravitational analogy for a ball on top of a hill
observed in the atmosphere, but the average value in (unstable), on a flat plain (neutral), and in a valley
the troposphere (lowest 10 km) is -0.65"C/lOO m. (stable).
This represents a balance between vertical mixing We can formalize these stability criteria:
processes and radiztive heat exchanges.
When air is saturated with water vapor and is aT,
rising vertically, the adiabatic temperature decrease is Unstable:
'
-aZ
< -0.98"C/lOO m
less than that given by Eq. 1.6. As air cools, its
capacity for water vapor decreases, and liquid water is
condensed. This process releases heat to the air a t a Neutral: -
aTe - -0.98"C/lOO m (1.11)
a2
rate of about 540 cal/g (latent heat of vaporization,
L) of condensed water. Thus part of the internal aT
energy used in expansion is recovered from latent Stable: e>
aZ
-0.98"C/lOO m (inversion
heat release, and the moist adiabatic temperature condition)
gradient (dT/dz)m is given by
Typically, the criterion for instability is satisfied only
within about 100 m of the surface on a sunny day.
The atmosphere is neutral on a windy and cloudy day
or night and is stable near the surface a t night or at
where m, is the saturated mixing ratio (mass of water any time in an elevated inversion layer. During stable
vapor per mass of air a t saturation). The wet adiabatic conditions a parcel displaced from an equilibrium
A V
t t
- - b
TEMPERATURE TEMPERATURE TEMPERATURE
UNSTABLE NEUTRAL STABLE
Fig. 1.2 Illustration of unstable, neutral, and stable environmental temperature profiles (- - -). An
air parcel moved adiabatically cools as it rises vertically (-).
level, as in Fig. 1.2, will oscillate about the equi- the atmosphere, which.& in turn, directly related to
librium level with the Brunt-Vaisala frequency nBV diffusion.
(radians/sec):-
w 1-4 STRUCTURE OF THE PLANETARY
BOUNDARY LAYER
The earths surface exerts a drag on the atmo-

(1.12) sphere which influences wind speed up to a height of
about 1 km. Diurnal variations in temperature and
mixing ratio are also noticed up to the top of this
where 6 , is the potential temperature. The stability
layer, which is called the planetary boundary layer
parameter (s) in Eq. 1.12 will be important in the
(PBL). Since most diffusion problems, with the
calculation of plume rise in Chap. 2. For typical
exception of such problems as aircraft emissions and
stable temperature gradients (OC/lOO m and
high level bomb blasts, occur in this layer, it is
2C/100 m), the Brunt-Vaisala period (2n/ngv)! is
important to know the variations of winds, tempera-
355 sec and 200 sec, respectively.
tures, and turbulence parameters in the PBL.
The parameter s may also be thought of as being
proportional to the rate at which stability suppresses
the generation of turbulence. On the other hand, 1-4.1 Turbulence Fluxes
turbulence is being generated by mechanical shear
forces at a rate proportional to (au/az)2. The ratio of The so-called eddy diffusivity (K), eddy viscosity
these two processes is called the Richardson number (Km), and eddy conductivity (Kh) coefficients are
(Ri): derived by assuming that any variable A is the sum of
an average and a turbulent fluctuation A:
A = ~ + A (1.14)
Clearly the stability parameter Ri gives us more where A could represent such variables as tempera-
information than s about the state of turbulence in ture, absolute humidity, or pollutant concentration.
METEOROLOGY 5
The average is usually over a time period of about 1 If -4is the concentration of a pollutant, then
hr. Further, the Reynolds averaging procedure is Eq. 1.18 yields the diffusion equation:
used :
-
A=O and A=A (1.15)
Next, consider the continuity equation for A:
dA aA aA aA aA
-=- + u --+ BtS (1.16) where S can represent internal processes, such as
dt at ax v-
ay t w -aZ
=
chemical reactions. If A is the wind-speed component
u or v in the x or y direction, then Eq. 1.20 yields the
where B includes all external effects, S includes all equations of motion. For unaccelerated flow homo-
internal sources, and w is verticalspeed. Also, assume geneous in the x and y directions, these equations
that the atmosphere is incompressible: become
t - t -av= o aIv
-aU (1.17)
ax ay aZ
By substituting Eq. 1.14 into Eq. 1.16, using Eq. 1.17
(multiplied by A), and averaging according to
Eq. 1.15, we obtain
where K,, refers to the vertical component of the
diffusivity coefficient for momentum, or the eddy
a A - a -. 4t
-- aA
ii-tv--+a i w-a?l
- viscosity. There is a balance among pressure, Coriolis,
dt at ax ay aZ and frictional forces in Eqs. 1.21 and 1.22. From this
point on, the bar notation for averages has been
- -
= B + --(u
a
s ax .4)
-I
removed.
- a
aY
(v)-a
aZ (w)
(1.18) 1-4.2 Ekman Spiral
- A simple expression for the variation of the wind

The term A is the flux of A in the x direction due velocity through the whole depth of the PBL was
to turbulent fluctuations. - developed by Ekman in 1902. He assumed that the
Since such turbulent fluxes as uA can be eddy viscosity coefficient for momentum in the
measured only with fast-response instruments and are vertical (Kmz)was constant and that the geostrophic
difficult to treat theoretically by analogy with the windspeed approximation was valid:
molecular case, the turbulent flux is commonly
assumed to be proportional to the mean gradient: ug = -;Elp (1 2 3 )
(1 24)
where K is a diffusivity coefficient (in units of

m2/sec). The negative sign is included so that the flux The substitution of these equations into Eqs. 1.21
is down the gradient (i.e., from high values of A to and 1.22 yields the following forms of the equations
low values). This technique is also called first-order of motion:
closure. Second-order closure (e.g., Donaldson, 1973)
is a more recent scheme that goes one step further 0 = f(v - vg) + K,,
a2
- (1 2 5 )
and a p p r o x i m e such terms as uwA with mean az2
gradients of (uA). Closure can be extended to any
order, but the technique soon becomes unbearably
com ple x.
If we orient the x axis parallel to the geostrophic The equation of motion (Eq. 1.25) can be used to
wind vector and assume that ug and vg are constant show that 4 usually vanes by less than 20%in the
with height, the solution to these equations is surface layer, or the lowest 50 m of the atmosphere.
During neutral or adiabatic conditions, the hypothesis
u(z) = ug(l - e-az cos az) (1.27) K, a (scaling speed) x (scaling length) can be used,
where the surface value of friction velocity (u+) is the
v(z) = ug e-az sin az (1.28) scaling speed and the height (z) is the scaling length,
i.e.,
where a = (f/2Kmz) w.
This solution is shown in Fig. 1.3, which shows K=kkz (1.30)
that the predicted angle between the surface wind
von Klirmans constant (k) is measured to be 0.35
and the geostrophic wind is 45. The predicted angle
over very smooth terrain and 0.4 over most other
is higher than is usually observed because of the
terrain.
restricting assumptions made to get a solution.
If k is assumed to equal 0.4, then Eqf1.29
Typical observed angles between the surface wind and
becomes
the geostrophic or free stream wind are 5 to 10 in
unstable conditions, 15 to 20 in neutral conditions, aU= 0.4 U* z -
K- au- 2
and 30 to 50 in stable conditions. Wind-direction aZ aZ
- U*
shear is very important for diffusion at large dis-
tances, where the bottom and t o p of the plume can a--U - U+ (1.31)
move in directions differing as much as 40 or 50 and aZ Ti&
thus yield a much larger plume spread than that
possible as a result of turbulent diffusion alone.
The real atmosphere usually does not agree with u = %In
0.4
(6)
the Ekman spiral at the top of the mixed layer (zi)
because of the development of an inversion at that This is the well-known logarithmic wind profile,
-
height. On most days this capping inversion at zi where the integration constant (zo) is called the
results in strong discontinuities in such parameters as roughness length. In general, zo h/10, where h is
water vapor mixing ratio and eddy dissipation rate. the height of the roughness elements, such as build-
Holzworth (1972) has analyzed observed temperature ings or plant cover.
profiles and published maps and tables of zi for most During diabatic conditions [(aO/a_zlfO] we de-
regions of the United States. Typical afternoon fine a scaling temperature [T, = (wT/u,)] and
mixing depths (zi) are about 1000 to 2000 m. another scaling length
3
1-4.3 Similarity Theory Gives Wind and L= U*

(1.32)
Temperature Profiles in Surface Layer 0.4 (g/T) (-wT)
- This is called the Monin-Obukhov length, after its
The momentum flux [- uwl or Km,(au/az)] is founders (Monin and Obukhov, 1953). L is positive
used to define the quantity u:, where u+ is called the for stable conditions (usually at night), negative for
friction velocity: unstable conditions (usually daytime), and ap-
- au proaches infinity for neutral conditions (dawn and
I
d = - u w 1
=K,,- (1.29)
aZ dusk transition periods and cloudy, windy condi-
\
Fii. 1.3 Ekman wind spiral. The \
windvelocity vectors at increasing
heights (z, , z2, and z3) approach the 2,
I
geostrophic wind-velocity vector at +,J d
the top of the mixedlayer (Zi). GEOSTROPHIC WIND SPEED

METEOROLOGY 7
tions). The absolute value of L can be thought of as (1.40)

the depth of the mechanically mixed layer near the
surface. The ratio z/L is another stability parameter
that has been found to approximate the Richardson The eddy diffusivity coefficient (K) for pollutants is
number (Ri) during unstable conditions and equal usually assumed to equal K h , and experimental
Ri/(l - 5 Ri) during stable conditions. We can now evidence tends to support this.
make the similarity predictions:
1-4.4 Turbulence Parameters
(1.33)
In convective conditions, i.e., when the surface is
warmer than the air above it, the important scaling
parameters are the mixing-layer height (zi) and the
scaling velocity ( w ~ )which
, is defined by
where subscript "0" refers to a surface value and fn (1.41)

and fn' are universal functions.
Consistent with the above, observations (Businger
e t al., 1971) show that Deardorff (1970) first made use of this parameter,
and Kaimal e t al. (1977) applied it in their analysis of
a field experiment. The quantity (g/T)w"is called
the surface buoyancy flux (H); i t is proportional to
the surface heat flux. The surface layer extends to a
height of about 0.1 zi, and above that height, during
z unstable conditions, wind speed is nearly constant
=lt5- (stable) (1.35)
L and wind direction turns slightly (5 to 10") to the
right (in the northern hemisphere). Wind profiles
above the surface layer in neutral and stable
conditions are more complicated; the interested
reader should consult the basic references (Wyngaard,
Cote, and Rao, 1974; Wyngaard, 1975).
= 0.74 + 5 L2 (stable) (1.36) The parameters uU> u,,, and ,u are the standard
deviations of turbulent ve!ocity fluctuations in the x,
where @m and & are the dimensionless wind and y, and z directions, respectively. Panofsky e t al.
temperature gradients, respectively. Paulson (1970) (1977) have recently studied data from several
has integrated Eq. 1.35 for unstable conditions to different sites and have determined formulas 1.42,
give the solution 1.48, and 1.51 for u ~ u,,,, and uu, respectively, for
unstable conditions. Irwin (1979b) developed the
power-law formulas given in Eqs. 1.43 to 1.45 for the
variation of U,/W* above the surface layer. Formulas
(1 k)] 2 tan-'
for neutral and stable conditions were obtained by
1 --}
fitting analytical formulas to curves presented by
-
(e$r
- In[: + + n (1.37)
Wyngaard e t al. (1975) and Wyngaard (1974).
@m 2
In stable conditions the following solution is easily

derived: 3
W* =
0.96 t (0 <<:$) (1.42)
(1.38)
= 0.763 (t)0.1 7 5
(2< -< 0.4)
z
zi
(1.43)
The eddy diffusivity for momentum (Km) and for

heat (Kh) can be defined by Eq. 1.9 by using A = u = 0.722 (1--5y'207
(0.40 <-<
z
and A = 8,respectively: zi
(1.39) = 0.37 (0.96 < ?< 1) (1.45)

zi
where q / z , is the height at which the first and second conditions, however, the data show that E = u:/O.4 z
formulas give equal values of uW/w*. up to heights of several hundred meters.
aw
u, = 1.3 exp (-2:) (neutral) (1.46)
1-5 USE OF SPECTRA TO ESTIMATE
TURBULENCE PARAMETERS
= 1.3 (1-+) (stable) (1.47)
The eddy energy spectrum [ S(n)] gives informa-
2=
u,
(12 - 0.5 z) k
(unstable) (1.48)
tion on the amount of energy carried by eddies of
different sizes; i t involves a Fourier transform of the
correlation coefficient R(T) :
= 1.3 exp (- 2 k) U*
(neutral) (1.49)
= 1.3 (1)-: (stable) (1.50)

where T is the time lag. The turbulence time scale is
h defined by
(Tu = (12 - 0.5:) (unstable) (1.51)
u,
= 2.0 exp (- 3 k) U*
(neutral) (1.52) The larger the eddies, the slower R drops off with
time and the larger the time scale T.
Energy spectra have been found to follow simi-
= 2.0 (1)-; (stable) (1.53) larity theory also, and universal equations for their
form can be written. (These equations will not be
reproduced here but can be found in Kaimal et al.,
In the daytime zi is usually marked by an inversion 1977). One importantscaling parameter is the wave-
capping the unstable well-mixed layer above the length A)(, at which the eddies are carrying the
ground surface. At night an inversion is present to maximum energy. Kaimal e t al. (1977) deduce the
some degree at all levels, and zi marks the height at following form for ,A during convective daytime
which surface-induced mechanical turbulence dies off conditions:
to zero. Turbulent intensity in the surface layer can
be estimated by dividing the appropriate values in the w , :,A =
Z
(2 < -L)
set of Eqs. 1.42 to 1.53 by the appropriate values in (0.55 + 0.38 z/L)
the set of Eqs. 1.37 and 1.38. The turbulence
intensities u,/u and ou/u decrease with height for all = 5.9 z (-L < z < 0.1 zi) (1.57)
stabilities, whereas uW/u increases with height in
unstable conditions and decreases with height in
neutral and stable conditions.
The eddy dissipation rate (E) gives the rate at U, v : ,A =, ,A = 1.5 zi (1.58)
which turbulence is being dissipated into heat at small
scales. This rate will be important in the calculation For neutral conditions, the wavelength A, of peak
of plume rise. In the surface layer, E is given by the energy is assumed to be equal for all three compo-
formula (from the energy equation). nents of turbulence. Several researchers have proved
the validity of the assumption that A
, is proportional
to height in the surface layer but asymptotically
approaches a constant at great heights:
Z
(1.54) ,A =5 (neutral) (1.59)
(1 + 15 fzlu,)
Experiments indicate that E e 0.5 H a t heights Caughey, Wyngaard, and Kaimal (1979) give observa-
above the surface layer at midday. During neutral tions of the variation with height of the wavelength
METEOROLOGY 9
h, for the three turbulence components during on a meteorological tower (see Fig. 1.4). The wind
stable conditions. The observed points can be fit by and turbulence are measured by an anemometer as
the following simple power laws: the air flows past. Another type of Eulerian measure-
ment is made by an aircraft, which flies through the
A,,=1.5 (/-)
z 0.5
(1.60)
turbulence along a nearly straight line. Also, the
measurement made by an anemometer moving with
zi
the mean wind speed through the flow is called a
-
zi - 0.7 (E)
hmv-
0 .s
(1.61) Eulerian measurement. In none of these cases does
the measuring instrument move with the air.
Measurements of an air molecule (1 or 2 in
Fig. 1.4) that has been tagged and followed as it
--
Amw - 1.0 (;)o-8 (1.62) moves through the turbulent field are called La-
zi
grangian measurements. Clearly the diffusion of
The values of, ,A given above can be used to pollutants is a Lagrangian process, which unfortu-
estimate K, by using the formula suggested by nately must usually be estimated by using Eulerian
Hanna (1968) : measurements, and some relationship between the
two systems should be established.
K m = A uw hmw (1.63) An air molecule will generally think a given
turbulent eddy has a lower frequency than that
where Pasquill (1974) has determined that the best measured by a fixed anemometer. This is crudely
value for the constant A is 0.15. The parameter uW illustrated by Fig. 1.5, in which a circular eddy with
can be estimated by using the suggestions in Sec. tangential speed (w) is immersed in a mean wind (ii).
1-4.4. The molecule travels once around the eddy in time
2nR/w, whereas the fixed anemometer sees the eddy
pass in time 2R/u. Therefore the ratio of Lagrangian
1-6 LAGRANGIAN TURBULENCE to Eulerian time scales (0) in this figure is given by
So far we have been discussing Eulerian turbu- (1.64)

lence, which is traditionally measured a t a fixed point
PARTICLE 1
PARTICLE 2
Fig. 1.4 Eulerian and Lagrangian windmeasuring systems. True Lagrangian wind measurements are
given by tagged air particles 1 and 2
Fig. 1.5 Large eddy of radius R approaches an anemometer on a tower. The eddy moves with mean
speed (ti) and has a mean tangential velocity (9.
where i is the turbulence intensity, usually called 2. Suppose that the 500-mb surface is 1000 m
u,/u. As turbulence intensity increases (stability higher over New Orleans, La., than it is over Chicago,
decreases), the ratio p decreases. Pasquill (1974) 111. What is the eastward component of the geo-
suggests that a good average value for 0 is 4. Reid strophic wind at that level?
(1979) finds that more accurate diffusion calculations 3. On a windy day the following wind observa-
can be made if tions are made:
. -
0.5 z, m 0.5 - 1.0 2.0 4.0 8.0
p =- (1.65)
u, m/sec 1.6 2.3 3.0 3.7 4.4
which indicates that the relation p a l / i is correct but Estimate the roughness length (zo) and the
that the crude model in Fig. 1.5 overestimates the friction velocity (u+).
proportionality constant.
4. Assume that zi = 1000 m, L = -50 m, and =
0.3 m/sec. At what height does uW reach a maxi-
Problems mum? What is its value? What is the value of u, at
that height? If u = 5 m/sec a t that height, what is the
1. Calculate the mass of the atmosphere. (The ratio of Lagrangian to Eulerian time scales (0)for the
easiest derivation will take less than one-half page.) y and z components?
Plume Rise
2-1 INTRODUCTION 2-2.1 Definitions
Figure 2.1 is a schematic drawing of a vertical

Plume rise is a very important factor in determin-
plume and a bent-over plume which illustrates many
ing maximum ground-level concentrations from most
of the variables and parameters important in calcu-
sources since it typically increases the effective stack lating plume rise. A plume is usually more or less
height by a factor of 2 to 10 times the actual release vertical if wind speed is less than about 1 m/sec.
height. Because maximum ground-level concentration Note the difference on the figure in definitions of the
is roughly proportional to the inverse square of the plume volume flux :
effective stack height, i t is clear that plume rise can
reduce ground-level concentration by a factor of as
much as 100. Most industrial pollutants are emitted V = wR2 (vertical)
with high velocity or temperature, and plume rise (2-1)
must be calculated. However, pollutants released V = uR2 (bent over)
from some building vents or motor vehicles have very
little plume rise. where w is plume vertical speed, u is ambient wind
A few areas of plume rise are well understood, speed, and R is plume radius in a plane perpendicular
such as the trajectory before final rise is reached and to the plume axis. In Eqs. 2.1 as well as in other
final rise in stable conditions. In both of these cases, equations in this chapter, the factor ?I is left out. The
the effect of ambient turbulence in the air outside the initial volume flux is defined by using the initial
plume is negligible. When ambient turbulence affects plume vertical speed at stack exit:
the plume, such as during final rise in neutral
conditions and during the last half of rise in con-
vective conditions, the models are less certain, and
more research is needed. Initial buoyancy flux (Fo) and momentum flux (Mo)
are defined by the following equations:
2-2 TOP-HAT-MODELEQUATIONS
The review article by Brigs (1975) provides

background material on available plume-rise models.
Most models are based on fundamental laws of fluid
mechanics: conservation of mass, potential density,
and momentum. The distribution of temperature,
speed, or other quantities across the plume is assumed where subscripts p, e, and 0 indicate plume, environ-
to have t o p h a t form (n that)is, a ;
variable ment, and initial values, respectively. For plumes
has a certain value inside the plume, another value whose molecular weight (m)differs much from that
outside the plume, and a discontinuity at the plume of air, Eq. 2.3 should be rewritten by replacing all
radius (R). Basically we are looking at integrated temperatures (T) with the ratio Tlmo (m,o can be
averages of variables in a plume cross section. assumed to equal 28.9). The values of the buoyancy
11
ENVIRONMENTAL STABILITY. s $(%+ 0 01 OC/rn)
BENT-OVER PLUME
VERTICAL PLUME VOLUME FLUX: V = uR2
VOLUME FLUX: V = wR2
U-
I N I T I A L BUOYANCY FLUX
9
Fo = -U p 0 - T ~ owORg
)
TPO
VERTICAL PLUME BENT-OVER PLUME
Fig. 2.1 Schematic diagram of vertical and bent-over plumes which illustrates some of the parameters
and variables important for plumerise calculations.
flux (F) and momentum flux (M) can change with independent variables. -An additional relationship,
height, where they are defined by the relations: called the closure assumption, is needed to solve
the equations. The closure used most often is the
Taylor entrainment assumption
F= (TP - T,) V (2.5)
TP
M=wV -_ - 2 Rv,
dz
Environmental stability (s) plays a prime role in v, = olw (vertical plumes)

slowing the plumes vertical motion:
v, = ow (bent-over plumes)
The entrainment velocity (ve) is the effective

speed at which environmental air is drawn into the
. The last factor is the adiabatic lapse rate, which is
plume through its boundaries. Taylor (1948) pro-
simplified to O.Ol0C/m from its value of 0.0098C/m
posed that ve is proportional to plume vertical speed.
in Eq. 1.6. Thus the last two terms could be replaced
The constants (Y and /3 are functionally similar, but p
by ae,/az. From Eq. 1.12 we see that s is the square
is much larger than a.
of the Brunt-Vaisala frequency nBV.
The ratio (S) of the effective area influenced by
the plume momentum to the cross-sectional area of 2-2.2 Set of Equations for Vertical
the so-called thermal plume (Briggs, 1975) has now Plume (from Briggs, 1975)
been recognized to be about 2.3 for bent-over
plumes. This factor permits the accurate assessment Buoyancy conservation
of both plume trajectory and final plume rise.
Because we neglect the details of turbulence, we
-d=F -sv
have one less conservation equation than we have dz
PLUME RISE 13
Momentum conservation 2-3 PLUME TRAJECTORY NEAR

SOURCE
_- - F
-dM
dz w
Sometimes the plume trajectory near the source
Closure must be calculated before ambient stability or ambi-
ent turbulence has much effect. Stability has little
-_ddzV - 2 a R w = 2 ~ M ' (2.10) influence at time periods less than s-%, which varies
between about 10 and 100 sec, and ambient turbu-
where (Y equals 0.08. lence is not important at distances less than about ten
stack heights at typical power plants.
2-2.3 Set of Equations for Bent-Over 2-3.1 Vertical Plumes

Plume (from Brigs, 1975)
For most plumes, early rise is dominated by
Buoyancy conservation
momentum. In this stage radius (R) = 0.16 z and
average vertical velocity (w) = 6.25 M%/z. Transition
dF-
-_
dz
_-SV
S (2.1 1) to buoyancy domination occurs a t t = M / F o (typi-
cally less than 10 secj, after which radius (R) = 0.15 z
and average vertical velocity (w) = 2.3 (Fo/z) %.
where S equals 2.3.
Momentum conservation
2-3.2 Bent-Over Plumes
(2.12) For short times, the buoyancy flux can be
assumed to be constant. The transition to buoyancy
Closure domination occurs at t = M/FO, the same as for a
vertical plume. This time has been found to be
typically about 5 sec, which corresponds to a travel
. --
dV-2pRu
dz distance. of only about 50 m. The plume trajectory in
this region is given by the equation
.>"
or, if u equals a constant,
(2.13)
z= ;( gx .t $7
Fo x- (2.14)
where P is equal to 0.6 for a "Joyant p a m e and This equation is compared with observations in
0 = 0.4+ 1.2 (u/wo) for a jet. Fg.2;2, where plume rise and downwind distance
R EFFLUX V E L O C I T Y / W I N D S P E E D
"2/3 LAW" FOR BUOYANT PLUMES
MOMENTUM FLUX ENHANCEMENT DUE TO BUOYANCY (Fox/Mu)
Fig. 2.2 Ohserved plume lr&ctories as a function of downwind distance (both nondirmensionalized),
showing transition from jet "% law" to buoyant plume law.""x
have been suitably nondimensionalized, and different For buoyant plumes, Briggs (1981) finds that
curves are plotted for values of wo/u equal to 4, 8,
and 16. Consistent with the recommendations, zeq = 5.3 Fis-i - 6 Ro (2.18)
flz0.6 is used with the Fo term in Eq. 2.14 and
fl = 0.4 + 1.2 (u/wo) is used with the M term. The correction term 6 & says that a virtual source
For buoyancy-dominated plumes, Eq. 2.14 be- exists a distance of six stack radii below the actual
comes stack height.
z = 1.6 F t u- xs (2.15)
2-4.2 Bent-Over Plumes
This is the famous g law, which has been shown to
agree with a great bulk of field and laboratory data. When the plume is bent over, the ratio of
The coefficient 1.6 can be expected to be accurate
maximum plume height to equilibrium plume height
within *40/0 with variations due to downwash or
is predicted to be 1.5 and 1.2 for a jet and a strongly
local terrain effects (Fay, Escudier, and Hoult, 1969;
buoyant plume, respectively. Formulas for the final
Briggs, 1981).
rise of bent-over jets in a stable environment are not
satisfactorily developed because of the lack of data.
However, the formula for the final rise of a buoyant
2-4 PLUME RISE LIMITED BY
plume is well known:
AMBIENT STABILITY
At night a deep (100 to 200m) stable layer

Ah = 2.6 ($7 (2.19)
usually forms near the ground with other layers of
varying stability above it. In the daytime there is The coefficient in this formula was developed from
usually a well-mixed convective boundary layer of comparisons with many observations, including the
depth 500 to 2000m with a stably stratified cap- TVA power-plant plume-rise data reproduced in
ping inversion at the top. Plumes will nearly always Fig. 2.3. For these data, the coefficient 2.6 is slightly
have to contend with stable air at some point in their conservative (underestimated plume rise). The wind
trajectory. speed (u) in this formula is an average value between
By recognizing that w = dz/dt, differentiating the heights h, and h, + Ah.
Eq. 2.9 or 2.12 with respect to t, and substituting
from Eq. 2.8 or 2.11, we obtain
2-5 PLUME PENETRATION OF
d2 M
- = - sM (2.16) ELEVATED INVERSION
dt2
This is the equation of a harmonic oscillator with If a plume can penetrate an elevated inversion,
period equal to the Brunt-Vaisala period 2ns-%.The ground-level concentrations may be dramatically re-
vertical velocity (w) must act like a damped harmonic . duced because the inversion then is a strong inhibitor
oscillator since .the volume flux (V) always increases of downward diffusion. This situation is depicted in
with time as a result of entrainment. Where vertical Fig. 2.4. If the plume does not penetrate the inver-
velocity first drops to zero, maximum plume rise is sion, the plume is trapped below it, and high
achieved. Plumes d o reach a maximum rise and then ground-level concentrations may result.
drop down to an equilibrium height. As a result of Let us arbitrarily define an eievated inversion to
wave drag, no more than one or two of the be a jump (AB,) in potential temperature at a height
oscillations predicted by Eq. 2.16 are visible.
(zel) above the stack. Define an inversion strength
(de,)AB? If a plume penetrates the inversion, its
2-4.1 Vertical Plumes buoyancy flux is reduced by (de,)ABi Vi, where Vi
is the plume volume flux a t height zel. We can
Harmonic oscillator solutions and field and labe conclude that the plume will penetrate this inversion
ratory experiments suggest that the equilibrium if F > (g/B,) ABi Vp Remember that the plume vol-
height (zeq) of a vertical jet in a stable environment is ume flux (V) has been defined as being equal to the
true volume flux divided by 1 ~ .
The volume flux for vertical buoyant plumes and
(2.17)
vertical jets is given by 0.07 F;zK and 0.16 M%,
PLUME RISE 15
/
/
/-- --. /
\ -
-------
e---
e-==- ---
'* -
/#- 4). e
=
--
/
.
a
//----
/ ///
4-
,
0
/ 0
---- PLUME TOP

PLUME CENTER L I N E
0 0 0 0 Ah = 1.6FV3u-' x2/3
0
0 1 2 3 4 5 . 6
NON D I M ENSIONA L DISTANCE DOWNWIND ( X/US-'~)
Fig. 2.3 Observed plume t ~ + ~ t o r i eass a function of downwind distance (both nondimensiodid),
showing maximum rise achieved by buoyant plumes.
respectively. Thus penetration of the inversion is This formula provided fair agreement with data from
forecast if the following conditions are met: the .Ravenswood power plant in New York City
Vertical buoyant plume reported by Simon and Proudfit (1967). If the final
plume rise (Ah) is within a factor of 2 of the
inversion height (q) above the stack, only a fraction
(P) of the plume can penetrate the inversion. In this
case Briggs (1975) suggests the (untested) formula
Vertical jet
Zel
P = 1.5 -- (2.23)
Ah
A fraction (1 - P) of the plume reflects off the

Bent-over jets will not be covered here because bottom of the inversion and diffuses downward.
they have little ability to penetrate inversions. The
volume flux (divided by n) of buoyant bent-over
plumes is given by V = 0 . l b uz2, and penetration will 2-6 PLUME RISE DETERMINED BY
occurif AMBIENT TURBULENCE
{
I
zi < 2.5 Fo (2.22) On cloudy, windy days or typical sunny summer

u(g/ea) AeiI
afternoons, the stability can be neutral or adiabatic in
1
Fig. 2.4 Diagram of buoyant plume penetrating an elevated inversion.

. .
the lowest 500 to 2000 m. If the plume reaches no dissipation rate. For a buoyant, bent-over plume, z =
stable layers, i t may find its rise limited by ambient h, + 1.6 F3u-l X~ (Eq. 2-15), and, by definition,
turbulence, which eventually dilutes the remaining w = dzldt. With the use of dx = udt and E =u2/0.4 z
plume buoyancy. Prior to that time the internal evaluated a t z = h, + Ah (valid during nearly neutral
turbulence of the plume is significantly greater than conditions), it is possible to arrive at the following
the ambient turbulence. Briggs (1981) has developed simplified formula for plume rise Ah:
the breakup model for the cases when rise is
limited by ambient turbulence. In the breakup
model the plume rise is assumed to terminate when Ah = 1.54(3) h: (2.25)
uu2,
the internal plume eddy dissipation rate just equals
the ambient eddy dissipation rate. For a jet, n o verifying data exist, but a theoretical
The results in this section are highly dependent estimate of plume rise at breakup in neutral
on Chap. 1, which gives methods of estimating conditions is
boundary-layer profdes of eddy dissipation rate,
turbulent energy, and wind speed.
Ah = 3D ($ - 1) (2.26)
2-6.1 Nearly Neutral Conditions
where D is the stack diameter and wo is the initial
Final plume rise in the breakup model rise is plume speed.
assumed to occur when the following condition is
met:
w3 2-6.2 Convective Conditions
1.5 7= E (2.24)
5
.
During convective conditions the neutral value for

The left-hand ratio is the internal eddy dissipation the eddy dissipation rate, e = &0.4 z, must be
rate of the plume, and E is the ambient eddy replaced by an estimate made by Brigs for convec-
PLUME RISE 17
tive downdrafts, E = 0.25H, where H is the surface If there is some question about whether to use the
buoyancy flux (see Chap. 1, Sec. 1-4.4). The resulting neutral or unstable formulas, use those which give the
prediction for buoyant plume rise is highest (most conservative) MGC. In quantitative
comparisons of MGC data measured near power
plants by Moore (1974), Briggs (1974)finds that the
(2.27)
dividing line for the neutral and bnstable formulas is
u = 7 m/sec.
Please note that this is a tentative formula. Plume-rise
observations in unstable conditions are the least
satisfactory owing t o rapid dilution. 2-8 MULTIPLE SOURCES
Many sites have several stacks that are close

2-7 MAXIMUM GROUND enough to each other that visual observations prove
CONCENTRATION WITH BREAKUP that the plumes indeed merge. The reduced entrain-
MODEL ment and increased buoyancy of the merged plumes
may increase plume rise significantly. Examples of
So that the dependence of maximum ground multiple sources are mechanical-draft cooling towers
concentration (MGC) on effective plume rise or lines of power-plant stacks.
he = h, -+ Ah can quickly be seen, suppose concentra- Briggs (1974) developed the following empirical
tion in the plume was uniform. Thus MGC would be method, which is based on TVA plume-rise data
inversely proportional to the volume flux TU h: at (Carpenter, Thomas, and Gartrell, 1968). The en-
the moment the plume first strikes the ground. This hancement factor (EN) is defined as the ratio of the
result is consistent with that obtained from the model plume rise from N stacks to that from one stack. The
that assumes a Gaussian distribution of material buoyant, bent-over plume rise from one stack (Ahl)
across the plume (see Chap. 4,Sec. 4-4)with standard is assumed t o be given by Eq. 2.19 [Ahl =2.6
deviation uY and U, in the crosswind horizontal and (Fo/us)%]. The spacing between the stacks is Ax.
vertical directions, respectively. For u,/uY constant, Then a formula for E, that has the proper asymp-
the resulting MGC prediction is totic behavior is:
MGC = 0.234 5 Q (2.28)

(2.33)
4y uhe
where
Maximum ground concentration is low a t low
wind speeds because of high plume rise and is low at %
high wind speeds because of high dilution. As the (2.34)
wind speed increases, plume rise decreases but dilu-
tion increases. At a critical wind speed (uc), MGC There was no dependence of EN on wind direction
reaches a maximum. This occurs at a downwind relative to the line of stacks for these data, although
distance xc. For the neutral "breakup" model, if we Hanna (1974) has found that the plume rise from
assume that uz/uy = 0.7, the critical wind speed
occurs at Ahjh, = x: lines of mechanicaldraft cooling towers is greater for
wind directions parallel to the line of sources.
For vertical plumes, merger is assumed at a height
(2.29) (z,) where the radius (R) equals one-half the stack
spacing (0.5 Ax). The plumes are then treated as if all
buoyancy combines but comes from a virtual source a
xc = 0.043 -
(":)"-
F@
;
(2.30) distance (1 - N%)zm below the actual stack heights.
There are no data to test this model.
For the unstable model, the critical wind speed

occurs a t Ah/h, = 5: Problems
uc = 0.43 FoH-' h,' (2.31)
1. A stack has an inside diameter of 3 m. The
plume has an initial speed of 1 0 d s e c and a
-
xc = 0.015 QH ?!i (2.32) temperature of 473'K. Ambient wind speed is
Fob! 5 m/sec, temperature is 295"K, and vertical tempera-
ture gradient is O.Ol'K/m. Calculate VO,Mo , Fo , and rises into a calm, dry atmosphere with Teo = 280K
s. Calculate final plume rise Ah. and aT/az = 0. How high will the plume rise above
2. For the plume in Problem 1: (1) At what the ground?
downwind distance does the buoyant rise term equal 4. Surface friction velocity (u*) is 0.3 m/sec,
the momentum rise term? (2) At what downwind ambient temperature is 270"K, aT/az = 0, and wind
distance does the 'Y3 law" give a plume rise equal to speed at stack height is 3 m/sec. Initial plume
that calculated in problem l? parameters are Tpo = 400"K, wo = 10 m/sec, Ro =
3. A cooling-tower plume with an initial height of 0.5 m, and the stack is 50 m tall. Calculate the final
100 m, Ro = 10 m, wo = 5 m/sec, and Tpo = 300K plume rise Ah.
Source Effects
3-1 OVERVIEW
Diffusion calculations would be greatly simplified

where D is the internal stack diameter, as illustrated
if stacks and buildings did not obstruct the airflow.
in Fig. 3.1.
Most stacks are built near other industrial buildings.
On the positive side, however, some utilities build -U
I 1
huge 300-m stacks at fossil-fired power plants to
ensure that pollutants from the stacks will be emitted
high enough above the ground so that there will be no
possibility of pollutant interaction with their other
buildings. Other industries use short vents con-
structed on building roofs to release toxic materials
to the atmosphere. Stacks on most residential dwell-
ings and commercial structures are usually short and
unobtrusive because of aesthetic considerations.
The purpose of. this chapter is to provide guidance
on:
Fig. 3.1 Schematic diagram of stack and important
1. Calculation of concentrations in building com- physical parameters: w,, ,efflux velocity;u, wind sped,
plexes due to emissions from existing sources. D, internal stack diameter; h,, stack height; and hd,
downwash distance.
2. Design of stack placement for minimizing
air-pollution effects.
The reviews by Hosker (1980, 1981) provide back- 3-3 STRUCTURE OF FLOW AROUND
ground information on the relevant phenomena and BUILDINGS
methods of calculation.
We have a fairly good idea of the flow around a
simple building, as drawn schematically in Fig. 3.2, as
3-2 STACK AERODYNAMIC EFFECT
a result of many wind-tunnel and field experiments
that have been done. The upstream boundary layer is
Low pressure in the wake of the stack may cause assumed to be a turbulent shear flow. The face of the
the plume to be drawn downward behind the stack. upstream building is not shown in this figure, but we
Downwash can be effectively prevented by maintain- know that there is a stagnation point about two-
ing the efflux velocity (wo) a t a magnitude greater thirds the way up the face; downward-moving air is
than the crosswind velocity (u). The fact that beneaththis point. The important features of this
downwash will not occur for wo/u greater than 1.5 diagram are the separated recirculation zones on the
has been generally recognized. For w o / u less than 1.5, roof and sides, the turbulent wake cavity zone, and
Brigs (1973) suggests that the distance (hd) that the the turbulent wake.
plume downwashes below the top of the stack can be Let the building height be H, the crosswind width
obtained by the following formula: W, and the alongwind length L. The wind is assumed
19
INCIDENT WIND
PROFILE
Fig. 3.2 Model of flow near a sharp-edged building in a deep boundary layer (Hosker, 1979).
to be perpendicular to the building face; if it is not,

x,
il
- (3.3b)
the flow is drastically changed, and the following ^N
2
techniques d o not apply. For ratios L/H greater than
about one, reattachment of the streamlines to the This cavity is bounded above by a turbulent shear
roof and sides can be expected unless W/H is very layer (zone II in Fig. 3.3) and above that by the
large. Between the upwind roof edge and the line of turbulent roof wake (zone III). The upper edge of the
flow reattachment is a zone of recirculation (see shear layer, which begins near the maximum cavity
Fig. 3.3). Pollutants released into the region by low height, is given approximately by
velocity roof vents or short stacks may produce very
high concentrations within this zone. Let 5 be the ZII X
smaller of H and W and ,$ the larger; define the --x 0.27 - 0.1 - (3-44
characteristic length 5%5%. Wilson (1979) finds
R il
that the roof cavity extends a distance
and the roof wake boundary is approximated by
L, = 0.9 E
from the upwind edge and reaches a maximum
(3.2)
%R= 0 . 2 8 (-7 (3.4b)
height
In Eqs. 3.4, x is measured from the upwind roof edge
H, zz 0.22 a (3.3a) and is greater than x,. Wilson suggests that, if a
roof-mounted stack is tall enough that the lower
at a distance from the upwind edge of plume edge remains above zone 11, contamination of
SOURCE EFFECTS 21
UPWIND VELOCITY PROFILE

I u = U(Z) 7 z, RECIRCULATION CAVITY BOUNDARY
zI, HIGH TURBULENCE ZONE BOUNDARY
zll, ROOF WAKE BOUNDARY
YL-I
Fig. 3.3 Flow over Center of a long flat building roof for wind perpendicular to the upwind face.
[From D. J. Wilson, Flow Patterns Over Flat-Roofed Buildings and Application to Exhaust Stack
Design, ASHRAE Trans., 85(Part 2): 284-295 (1979).]
the roof is unlikely, and if the plume clears the wake

boundary as well, there is virtually no danger of B = - 0.15 + 0.305 (3.6a)
contamination. Wilson recommends that a line of
slope 0.2 be drawn in the upwind direction from a
whereas for L/H 3 1, where flow reattaches to the
point on the zone IT or zone I11 boundary (depending
roof and sides,
on the safety margin desired) immediately above the
air intake or other critical receptor that is closest to
the lee edge of the roof. The effective stack exit
A = 1.75
(3.6b)
should then be above this line.
B = 0.25
The wake cavity behind the structure is important
because plumes caught in the cavity can be quickly
Cavity.extent is measured from the lee building face.
mixed to the ground. Let xr,yr, and z, be the length,
For cubical buildings, these formulas predict that the
width, and height of the wake cavity, respectively.
wake cavity length (xr) is about 2.5 times as great as
Crosswind cavity dimensions yr and zrseldom exceed
the building height. The maximum wake cavity length
the building dimensions W and H by more than 50%.
(-1OH) is predicted for large W/H (i.e., two-
Hosker (1979, 1980) has developed the following
dimensional buildings).
empirical formulas for wake cavity length (xr):
(3.5)
3-4 DIFFUSION CALCULATIONS
AROUND BUILDINGS
where, for L/H I1 or so,
The literature on diffusion around buildings is full
\a
of rules of thumb that are based on practical
experience and wind-tunnel observations. In some
A = - 2.0 + 3.7 -
- cases these rules of thumb can be verified by physical
reasoning and theoretical derivations. The following sides, close to the ground, whereas a vent located
sections are arbitrarily split into sources upwind of a higher up will emit into flow streaming over the
building and sources near a building. building. If roof-vented vents are located in the
separated areas near the leading edge of the roof and
sides, as discussed above, ineffective ventilation can
3-4.1 Isolated Sources Upwind of lead to high concentrations and recirculations.
Buildings Halitsky (1963) and Wilson (1976) present wind-
tunnel observations of concentration patterns result-
If the source is a distance greater than 2 H ing from various types of vent positions, building
upstream of the building and its height (h,) is greater shapes, and efflux-to-wind-speed ratios (wo/u). An
than two-thirds the building height (n, > 2 H/3), the example is given in Fig. 3.4, where the plotted
plume will rise over the building face. Parts of the concentration K is made dimensionless in the follow-
plume at heights less than 2 H/3 will be caught in the ing way:
downwash in the frontal eddy over the lower part of
the building. Some of this material will be caught in CAu
Dimensionless K = - (3.7)
the horseshoe vortex trailing off along the edge of the Q
wake. If the flow reattaches to the building roof and
sides, high concentrations may result a t these points. where C = concentration (pg/m3)
If the building length (L) is small, the flow will not u = wind speed (m/sec)
reattach, and the plume is deflected above the cavity. Q = source strength (pg/sec)
Any pollution caught in the wake cavity will mix A = convenient characteristic area (e.g., WH)
thoroughly to the ground. The overall effect of the for the building being studied (m)
building is to increase the dispersion of the plume, The wind-tunnel experiments suggest that flush,
although locally high ground-level concentrations low-velocity, roof vents give maximum values of K
may result owing to aerodynamic effects of the ranging from 10 t o 100 a t the roof surfaces. In the
building. design of vent exhaust and intake systems, it may be
Wilson and Netterville (1978) conclude that in wise to use a conservative value of 100 for K.
this situation the most important effect of a building The effects of variations in efflux speed (wo) and
is to mix the plume concentrations between the vent stack height (ho) are shown in the experiments
ground and the roof. Thus the roof-level concentra- by Halitsky (1963) and Wilson (1976). Roof concen-
tion should be held within limits. In design studies trations can be dramatically decreased by increasing
the stack height in a model can be varied so that the
the upward efflux speed and by installing relatively
height of the maximum tolerable concentration iso-
small stacks. If the ratio wo/u is increased from 0.5 to
pleth, as calculated by a flat-terrain diffusion model,
1.0, maximum roof concentration is reduced by a
is higher than the maximum downwind building
factor of 2 to 3. Wilson (1976) found that the upper
height.
bound on the ratio of concentration at distance x
For building clusters or out-of-the-ordinary build-
from the vent to the vent exit concentration was a
ing shapes, wind-tunnel and/or field tests are neces- function of wind speed (u), distance (x), effluent
sary for a realistic assessment of a site. Cagnetti speed (wo), and vent area (AY):
(1975) shows how results for diffusion around a
reactor complex vary strongly when there is a minor
change in wind direction.
bound
3-4.2 Sources Close to Buildings (b) Stack Effective Height. A major concern is
to keep the plume away from the wake cavity, where
In most cases vents and stacks are located on or it would be brought to the ground and recirculated in
near the building. Primary concerns are whether a region with low ventilation rates. This problem
pollution is emitted directly into a cavity o r recircula- results if the plume is released too close to the
tion zone, the magnitude of the concentration in building and/or at too low a speed W O - The familiar
these zones; and the deflection of streamlines and c c21
/2 times rule is often applied because experience
increase in dispersion for plumes outside these zones.
has shown that there are few downwash problems if a
As mentioned earlier, a vent located
( a ) Vents. stack whose height is at least 2.5 times as great as the
on the lower two-thirds of the upwind face w
ill emit height of the building is built. From the previous
into the upwind eddy and be transported around the diagrams and discussion, it is clear that the wake
SOURCE EFFECTS 23
I\
Fig. 3.4 Dimensionless concentration (K = C A d Q ) contours for centered flush roof vent on building
where L/H= W/H= 3.0 and w,/u 1.0. [From J. Halitsky, Gas Diffusion Near Buildings, ASHRAE
Trans., 69: 476477 (1%3). J
cavity does not extend to this height, a fact obviously

recognized over the years by plume watchers. In
many cases the 272 times rule can be relaxed, and the
pollutant concentrations can still be held within
acceptable limits. Briggs (1973) has developed a
straightforward method of determining the effective
height of a plume near the wake cavity, as described
in the next paragraph.
First, stack downwash is calculated by using
Eq. 3.1. Th,e effective stack height after downwash is
h, which equals the stack height (b)minus the Fig. 3.5 Diagram of effective plume height in the case
downwash (hd) (see Fig. 3.5). Next, let 1be defined of downwash due to a nearby building.
as the smaller of W and H. Then, if h is greater than
H + 1.5 1, the plume is out of the wake, and the For h < H + 1.5 f and h > H,
effective plume height (he) equals h. Otherwise the
plume is affected by the building. he = 2h - (H + 1.5 f ) (3.10)
For h > H + 1.5 f , Forb < H,
he = h (3.9) he = h - 1.5 f (3.11)

Finally, the plume is assumed to be trapped in the C= Q cWH)u (3.14)

cavity if the effective source height (he) calculated ( T U ~ U ,f
above is less than 0.5 {. In this case the plume can be
treated as a ground-level source with initial area {. The constant c is intuitively estimated by Gifford
On the basis of field observations of plumes (1975) to be between 0.5 and 2, with the lower value
released from a large vent from a reactor building, agreeing fairly well with test results. The factor cWH
Johnson et al. (1975) developed the split h concept represents the effective crosswind area of the build-
for evaluating effective plume height. They found ing. Dispersion parameters uy and u, can be obtained
that, for a particular reactor sample, for speed ratios by methods given in Chap. 4, although measurements
wo/u between 1 and 5: the plume would alternate suggest that uy and uz in wakes may need to be
between being in and out of the cavity. The percent modified in the future. For example, the use of
of the time that the plume was in the cavity was given open-terrain uy and u, values leads to concentration
by decreasing in proportion to distance raised to the
-1.3 to -1.6 power, depending on stability. Meroney
E = 2.58 - 1.58 0
U
(1 < 0 < 1.5)
U
(3.12)
and Yang (1971) and Huber and Snyder (1976) have
found the power to be closer to -0.8 out to a
distance of 50 H, which is presumably due to the
dispersion inhihiting effects of a very persistent
E = 0.3 - 0.06 3 (1.5 < 3 < 5) (3.13) horseshoe vortex. In full-scale field studies (eg.,
U U
Dickson, Start, and Markee, 1969), the power has
been found to be about -1.3. The difference between
This result is probably highly site specific, however,
the field and laboratory results may be due to
depending on vent size, placement, and building
mesoscale wind-direction fluctuations observed in the
geometry. field but not present in the laboratory.
(c) Concentrations in and Downwind of Cav-

ity. The dimensionless concentration (K = CAu/Q) Problems
in a wake cavity is generally found to be between 0.2
and 2.0 if we assume that the plume is nearly 1. Assume that the vent in Fig. 3.4 is emitting
completely drawn into the cavity (he < 0.5 5 ) (Barry, toxic gases at a rate of 1 g/sec and that wo =
1964; Meroney, 1979). Concentrations are relatively u = 1.0 m/sec. Estimate the maximum concentration
uniform from the lee side of the building throughout on the building roof and on the downwind building
the cavity. Because of this rapid mixing, stack height face. The building is 10 m high and 30 m wide.
usually makes little difference for plumes caught in a 2. A stack 50 m high emitting 10 g of toxic gases
cavity. Because of this rapid mixing, release within per second is located just upwind of a 40-m cubical
the wake cavity of pollutants from ground-level or building. The stack diameter is 1 m and wind speed
from low stacks makes little difference in the and initial plume effluent speed are 1 m/sec. The
resulting concentrations. However, a taller stack of plume is nonbuoyant. Estimate the ground-level
about the same height as the wake cavity may result concentration on the plume axis at a downwind
in effluent being transported quite directly to the distance of 300 m.
ground by the downward curving shear layer bound- 3 . b . McGraw is cooking cabbage on the third
ing the cavity. This is one instance where a tall stack floor of a 30-m cubical apartment building. Her
may produce higher ground-level concentrations than kitchen exhaust fan emits 5 g of cabbage gas per
a short one. second to the outside, downwind side of the building.
For plumes not entrained into building cavities Wind speed is 2 m/sec. What concentration of cab-
(he> 5/2), the standard Gaussian plume formula (4.1) bage gas is smelled by Ms. Jones while she hangs up
can be used to estimate concentrations. Usually wind her wash a distance of 30 m from that same side of
speed (u) is defined at z equal to H. If a plume is the building?
entrained into the building cavity, modifications to 4.For the situation in problem 3, what is the
the Gaussian plume formula for the centerline con- concentration on the plume centerline a distance of
centration due to a ground source are used: 1000 m from the building? Assume sunny conditions.
Gaussian Plume Model
for Continuous Sources
4-1 WHY USE THE GAUSSIAN MODEL?
Other dispersion models, such as the K-model, the

statistical model, and the similarity model, will be
described in succeeding chapters. The Gaussian model
is discussed first because it is still the basic workhorse The coordinate y refers to horizontal direction at
for dispersion calculations, and it is the one most right angles to the plume axis with y equal to zero on
commonly used because the axis. The coordinate z is height above ground,
which for the time being is assumed to be flat and
1. It produces results that agree with experi- uniform. The parameters uy and uz are standard
mental data as well as any model. deviations of the distribution C in the y and z
2. It is fairly easy to perform mathematical directions, respectively. The purpose of the last term
operations on this equation. is to account for reflection of the plume at the
3. It is appealing conceptually. ground by assuming an image source at distance h
4.It is consistent with the random nature of beneath the ground surface. Figures 4.1 and 4.2
turbulence. contain diagrams of the Gaussian plume. All variables
5. It is a solution to the Fickian diffusion are assumed to be averaged over a period of about
equation for constants K and u. 10 min. Corrections for different averaging times are
6. Other so-called theoretical formulas contain given in Sec. 4.5. Averaging time should be chosen so
large amounts of empiricism in their final stages. that most of the turbulent eddies go through at least
7. As a result of the above, it has found its way one cycle during that averaging time. Unfortunately
into most government guidebooks, thus acquiring a averaging times in experiments are often dictated by
blesed status (Environmental Protection Agency, other operational considerations, such as the duration
1978). of release of a tracer.
Newcomers to this field often ask, What happens
in the Gaussian equation when wind speed (u) goes to
zero? The standard reply is, Calm winds are defined
4-2 FORM OF THE GAUSSIAN MODEL as u equal to 0.5 m/sec. The truth is that anemom-
eters near the surface may register u = 0, but the
The origin of the Gaussian model is found in winds in the planetary boundary layer very seldom
work by Sutton (1932), Pasquill (1961, 1974), and stop entirely. There is nearly always a slight drift, and
Gifford (1961, 1968). Consider a continuous source the seemingly facetious answer to the above question
of strength Q (in micrograms per second) at effective is based on considerable experience,
height (h) above the ground. Assume that the wind The effective height (h) can be estimated by using
speed (u) is uniform. The concentration C (in the techniques outlined in Chaps. 2 and 3. The
micrograms per cubic meter) is then given by the remaining problem, then, is specification of the
formula: dispersion parameters uY and 0., These are given as
25
SECTION A-A'
I I /
/
\ I /
Fig.41 Diagram of typical plume, illustrating concepts important in the Gaussian plume formula
Symbols: u is wind speed; h is effective source height; and ay and a, are standard deviations of
crosswind concentration distributions.
40 I 1 I I I I I I I 1
-40 I
I I I I I I I I I I I I
-60 -50 -40 -30 -20 -10 O 10 20 30 40 50 D
Y. m
Fg.4.2 Cross section through Gaussian plume with u y = a m , u z = lorn, and centerline
concentration of 1.0.
GAUSSIAN PLUME MODEL FOR CONTINUOUS SOURCES 27
functions of downwind distance and stability and are classes. This method does not explicitly consider the
based on a combination of experimental results and effects of wind speed on diffusion and consequently
theory. is not a sufficient method for all situations. It would
be better to use Richardson's number (Ri) or the
Monin-Obukhov length L to characterize stability
4-3 STABILITY CLASSIFICATION
since they are direct measures of atmospheric stabil-
SCHEMES
ity, which accounts for the effects of both mechan-
ical mixing and buoyancy forces. Golder (1972)
For an estimate of uy or uZ in the absence of analyzed diffusion data to determine the graphical
research-grade turbulence measurements, the stability relation among roughness 20, the Monin-Obukhov
class must first be determined, preferably by a simple length L, and the Pasquill class shown in Fig. 4.3. If
scheme based on inexpensive and easy measurements. two of these parameters are known, the other can be
The most widely used scheme was developed by estimated from the figure.
Pasquill (1961) and was modified slightly by Turner The following recommendation was made by the
(1967). Table 4.1 contains the criteria for Pasquill's American Meteorological Society Workshop on Sta-
six stability classes, which are based on five classes of bility Classification Schemes and Sigma Curves
surface wind speeds, three classes of daytime insola- (Hanna et al., 1977):
tion, and two classes of nighttime cloudiness. In . . . The following quantities are required to char-
general, stability classes A through C represent acterize uy and uz in the boundary layer:
unstable conditions, class D represents nearly neutral 1. Roughness length, zo, and friction velocity,
conditions, and classes E and F represent stable u,, as measures of the mechanical turbulence.
conditions. Some users have filled in the blank in 2.Mixing depth, zi, and Monin-Obukhov
length, L, or the heat flux H, as measures of the
Table 4.1 with a so-called "G" class, which they convective turbulence daytime.
assert applies during light wind, stable conditions. 3. Wind speed, u, and standard deviation of
If turbulence measurements are available, it is wind direction fluctuations, u,g (the wind vector is
preferable to estimate uY and U, by using 0 0 and Ue, needed to specify the transport wind, and u,g is
standard deviations of wind direction fluctuations in required to estimate uy, especially in stable condi-
tions).
the horizontal and vertical directions, respectively.
Early advocates of stability classification schemes On a 61-m tower, all quantities but mixing depth (zi)
based on 00 or ue were M. E. Smith (1951) and can be measured by standard instruments required by
Cramer (1957). The Brookhaven National Laboratory the Nuclear Regulatory Commission, and remote
(BNL) classes are defined by M. E. Smith (1951), sounders that can measure zi are available. Methods in
using wind direction 6 recorded over a 1-hr period, in Chap. 1 (Eqs. 1.37 to 1.53) can be used to extrapo-
the following manner: late U,g U,/U to plume height from a height of
A: Fluctuations (peak to peak) of 8 exceed 90". 10 m.
B1: Fluctuations of 6 from 40 to 90". Despite the strong recommendation to use turbu-
Bz : Fluctuations of 8 from 15 to 40". lence measurements to estimate diffusion, most
C: Fluctuations of 6 greater than 15" with strip people today still use the Pasquill letter classes
chart showing an unbroken solid core in the because they have produced satisfactory results in
trace. most cases and because they are easy to use.
D: Trace in a line, short-term fluctuations in 6 However, the user must beware if he applies letter
less than 15". classes to problems outside the area of their deriva-
Cramer's (1957) classes, which are based on observation (eg., complex terrain, distances greater than
tions of wind fluctuations at a height of 10 m, are 10 km, effective release heights of above 100 m, and
defined in Table 4.2. Note that he distinguishes many other problems). For these problems, direct
between two roughness types in the neutral class. The turbulence measurements or theoretical extrapola-
basic turbulence typing methods are compared in tions are necessary.
Table 4.3, as reproduced from a review article by
Gifford (1976). The fact that these divisions and
4-4 CHOICE OF ay AND uz
comparisons are arbitrary is important, and this
system should not be considered perfect.
To further confuse the reader, the Tennessee 44.1 Stability Class Method
Valley Authority (TVA) (Carpenter et al., 1971) and
the Nuclear Regulatory Commission (NRC) use verti- Most published uY and uz curves as a function of
cal temperature gradient (DT/Dz) to define stability downwind distance and stability are based on a few
T a b l e 4.1 Meteorological Conditions Defining

Pasquill T u r b u l e n c e Types"
A: Extremely unstable conditions D: Neutral conditions?
B: Moderately unstable conditions E: Slightly stable conditions
C: Slightly unstable conditions F: Moderately stable conditions
Nigbttime conditions$
Daytime insolation
Surface wind Thin overcast or <YE
speed, m/sec Strong Moderate Slight >YE low cloud cloudiness
<2 A A-B B
2 A-B B C E F
4 B B-C C D E
6 C C-D D D D
>6 C D D D ' D
*From F. A. Gifford, Turbulent Diffusion-Typing Schemes: A Review,

NucL Sa$, 17(1): 71 (1976).
?Applicable to heavy overcast day or night.
$The degree of cloudiness is defined as that fraction of the sky above
the local apparent horizon that i s covered by clouds.
Table 4.2 Cramer's T u r b u l e n c e Classes" T a b l e 4.3 Relations A m o n g T u r b u l e n c e

' T y p i n g Methods"
De, deg
Stability description og, deg (at 10 m) Stability 06, deg
description Pasquill Turner BNL (at10m)
Extremely unstable 30 10 ~ ~~~
Near neutral (rough surface; Very unstable A 1 A 25

trees, buildings) 15 5 Moderately unstable B 2 B, 20
Near neutral (very smooth grass) 6 2 Slightly unstable C 3 B, 15
Extremely stable 3 1 Neutral D - 4 C 10
. .
- Slightly stable E 6 5
*From H. E. Gamer, A Practical Method for Esti- Moderately stable F 7 D 2.5
mating t h e Dispersal of Atmospheric Contaminants, in
Proceedings of the First Natioml Conference on Applied *From F. A. Gifford, Turbulent Diffusion-Typing
Meteorology, pp. C33 t o C 5 5 , American Meteorological Schemes: A Review, Nucl. Saf., 17(1): 72 (1976).
Society, Hartford, Corn, 1957.
o.2k \
0.5 \ I /
20 - A
-
-
50- I I I I I
-0.12 -0.10-0.08 -0.06 -0.04-0.02 0 0.02 0.04 0.06 0.08
I I L . rn-'
Fg. 4 3 Curves showing Pasqidl's turbulenm types as a function of t h e Monin-Obukhov s t a b r a y

length a n d the aerodynamic roughness length. A, extremely unstable conditions; B, moderately
unstable conditions; C, slightly unstable conditions; D, neutral conditions (applicable to heavy
overcast day or night); E,slightly stable conditiom; F,moderately stable conditions. [From D. Colder,
Relations Among Stability Parameters in the Surface Layer, Boundary-Layer MeteoroL , 3 56 (1972).]
carefully performed diffusion experiments during the

1950s and 1960s. Project Prairie Grass (Haugen, Table 4.4 Brookhaven National Laboratory
1959) is probably the most frequently quoted diffu- Parameter Values in the Formulas
sion experiment. The terrain was uniform, releases uY = axb and uZ = cxd
were from near ground level, and concentration
measurements were at downwind distances less than Parameter
1km. These experiments resulted in Pasquills (1961) Type a b C d
curves, which were adapted by Gifford (1961, 1968,
1976) into the forms shown in Fig. 4.4. Note that, at B, 0.40 0.91 0.41 0.91
distances beyond 1 km, the lines are dashed (i.e., a B, 0.36 0.86 0.33 0.86
C 0.32 0.78 0.22 0.78
guess). They may work under certain ideal conditions D 0.31 0.71 0.06 0.71
at greater distances, but there is little basis in
observations.
l , 11111, , ,, il,ll, l
E
I-
-
---
----
-
- -
18. I I I.ililll I I I111111 I 1 I l l l ~
10-1 100 10 102 lo- 100 101 102
DISTANCE DOWNWIND. krn DISTANCE DOWNWIND, km
(a) (b)
Fig. 4 4 Curves of uy and uz for turbulence typea based on those reported by Pasquill (1961). [From
F. A. Gifford, Turbulent Diffusion-Typing Schemes: A Review, N w L Sac., 17(1): 71 (1976).]
Because calculators and computers are in such formulas given in Table 4.5. Initial plume spread a t all
widespread use a t present, most people would rather stabilities is proportional to x, the proportionality
have a formula than a graph or table. Several factor being 0 0 (in radians). At large distances, uy is
researchers have worked out analytical power-law proportional to xu, as predicted by the Fickian and
formulas for uY and uz. One of the early suggestions Taylor theories of diffusion. Note that uy and U, are
was by M. E. Smith (1968). He summarized the BNL independent of release height and roughness in these
formulas, which are based on hourly average measure- formulas. There are too few experimental data to
ments out to about 10 km of diffusion of a nonbuoy- support more complex formulations including these
ant plume released from a height of 108 m: two variables.
The Prairie Grass experiments were camed out
uy = axb u, = cxd (x in meters) (4.2) over terrain with roughness zo of 3 cm. F. B. Smith
(1972) and Pasquill (1975, p. 19) have found that U,
Values of the parameters a, b, c, and d are given in varies as zi,where p lies in the range 0.10 to 0.25. A
Table-4.4. technique for incorporating Smiths (1972) recom-
Briggs (1973) combined the Pasquill, BNL, and mendations into analytical forms for U, in each of the
TVA curves (observations out to x = 10 km), using P-G categories was given by Hosker (1973). The
theoretical concepts regarding asymptotic limits of larger values of the exponent p are applicable to
the formulas, to produce the widely used set of shorter distances and rougher surfaces. Over rough
urban surfaces, especially under the influence of the 44.2 The ag and O e Method
nighttime urban heat island, the increased roughness
should be taken into account. McElroy and Pooler's Much research concerning the best way to relate
(1968) diffusion experiment in St. Louis was used by uY and u, to 08 and 0, is being done. In perhaps 5
Briggs (1973) to develop the formulas given in Table years the subject will reach a stage where definitive
4.5. Other people make the assumption that urban uY conclusions can be drawn. The recommendations in
and u, should be moved up one stability class (e.g., C this section are based on the latest available research.
to B) to account for increased dispersion over urban The idea behind this research is to remove a layer of
areas. empiricism (the Pasquill-Gifford curves) from diffu-
Table 4.5 Formulas Recommended by Briggs (1973)

for u,(x) and uz(x) (lo2 < x < 104m)
Pasquill
type 0,. m
Open-Country Conditions
A 0.22x(1 t 0 . 0 0 0 1 x ) - ~ 0.20x
I3 0.16x(1 + O.OOOlx)-w 0.12~
C O.llx(1 t O.OOOlx)-% O.O8x(1 t O.O002x)-~
D O.O8x(1 + O.OOOlx)-% O.O6x(1 + O . O O l S x ) - ~
E O.O6x(l t O.OOOlx)-% O.O3x(1 t O.O003x)-'
F O.O4x(l t 0.0001~)-% O.O16x(l + O.O003x)-'
Urban Conditions
A-B 0.32x(l + O.O004x)-$ 0.24x(l + 0 . 0 0 1 ~ ) ~
C 0.22x(l + 0.0004x)-% 0.20~
D 0.16x(1 + 0.0004x)-% 0.14x(1 + 0 . 0 0 0 3 ~ ) - ~
E-F O.llx(1 + 0.0004x)-% O.O8x(l + O.O0015x)-~
Recent diffusion experiments under clear, nearly sion calculations by developing a technique that
calm nighttime conditions (so-called category G) sug- relates diffusion directly to turbulence. Taylor's
gest that horizontal diffusion is actually greater (1921) work suggests the formulas:
during these conditions than under conditions asso-
ciated with category F (Sagendorf and Dickson, 1974)
because the plume often meanders during G condi- uy = u,tf (L) (4.3)
TY
tions, which results in 08 equal to 20" or more and
relatively low hourly average ground concentrations
at a given point. Van der Hoven (1976) analyzed (4.4)
several category G diffusion experiments and found
that observed U, values corresponded to anything
where fy and f, are universal functions and T, and T,
between categories A and F. Thus diffusion in terms
are turbulence time scales in the y and z directions.
of tabulated dispersion parameters is indeterminate
The fact that averaging times for all variables are
when category G stability is found. Of course,
equal and that turbulence parameters are measured or
diffusion can still be estimated on the basis of actual
estimated near the release height is implicit. Since
measurements of ~ 8 .
diffusion calculations are generally made in practice
With the use of stability classes in complex terrain
for downwind distances (x) rather than times (t), the
situations, u s are also difficult to determine. The
following forms of writing Eqs. 4.3 and 4.4 are
diffusion experiments summarized in Chap. 12 gen-
desirable:
erally show that uy and u, in complex terrain are a
factor of 2 to 10 greater than that predicted from the
Pasquill curves. Again, measurements of Ue and Ue are
(4.5)
the best solution to this uncertainty.
u, is less than the source height; otherwise reflection

and absorption at the ground surface distort the
plume. Of course, no matter what the plume shape is,
where it is assumed that 0 8 = u,/u and ue = uW/u. u, can always be calculated as the standard deviation
Once the universal functions f, and f, have been of the observed concentration distribution. Draxler
determined, then uy and u, are completely deter- (1976) plotted observed values of f, = u,/(u,t) as a
mined by observations of 0 6 and ue. Presumably T, function of time after release for several diffusion
and T, are similarity functions of z/zg, zi/L, and experiments. He divided the diffusion data into
Lflu,. ground-level and elevated sources and suggested the
Pasquill (1976), Draxler (1976), Doran, Horst, following formula for ground-level sources for fy (all
and Nickola (1978), and Irwin (1979a) are among stabilities) and f, (stable and neutral):
those actively investigating the best form for the
universal functions f, and f,. Although Taylors fy = fz = [I + 0.4O(x/uTy)~]- (4.8)
(1921) statistical diffusion equation (see Chap. 5,
Sec. 5-2) can be used to derive a theoretical form for where Ti is assumed to equal 5 T,. Draxler presented
f, and f,, current data are scattered enough that a slightly different formula for elevated sources, but
simple empirical formulas are justified. For example, at all x the formulas agree within *20%, which is
Pasquill (1976) gets good results by removing the probably as good accuracy as we can hope for.
dependence of f, on the time scale T,. Irwin (1979a) For unstable f,, Irwin (1979a) plots observations
has derived the following approximations to Pasquills of uz/(uwt) = f, vs. t = w,t/zi for elevated and
empirical table: ground-level releases. These figures are reproduced
here as Figs. 4.6 and 4.7, where the best fit curves
f, = (1 + 0.031 X * ~ ~ ) - (x < lo4 m) are forced to obey the theoretical law u 0: t % a t large
times. Unfortunately the data do not seem compelled
= 33x-?4 (x > io4 m) (4.7) to follow the Same law. Values of the convective
velocity scale (w*) and the mixing depth (zi) can be
where x is in meters. This solution is diagrammed in obtained from Eq. 1.43 and observations, respec-
F i g 4.5. The approximation has the desirable prop- tively.
erty that u y a x at small distances. At distances
greater than about 10 km, mean wind direction shear
in the planetary boundary layer may become impor-
0 2; = 0.24
tant, and Pasquill (1976) suggests the rough rule: add 0 2 ; = 0.26
to the u; already derived the term 0.03 (ACY)~X, 8 Zi = 0.50
0 z- = 0.71
where Aa (in radians) is the change in wind direction v zj = 0.15
over the plume depth.
The universal function fz is more difficult to
determine since there are fewer data on the vertical
distribution of concentration. Furthermore, a Gauss-
ian distribution in the vertical is found only as long as
o. 1 I I I Illlll I I I I1111 I I I I Ill

0.01 0.1 1 10
0.8 - to= w.t/h
0.6 - - Fig. 4.6 Values of oJ(owt) vs. t* for elevated

>
r
releases (q) duriing unstable conditions. Release
0.4 - heights range from 0.15 to 0.71.
0.2 -
I I I I
01 I I I I11l11 I I 1 Ill1 1 i I ~ I I J
4 5 WIND-SPEED VARIATION WITH
Id 103 104 105
x. m
HEIGHT
Fq 4.5 fy = oy/(oex) according to Eq. 4.7, which is The fact that the wind speed (u) appearing in the
Irwins (197%) attempt to fit Pasquills (1979) table. basic Gaussian plume formula (Eq. 4.1) should be an
4-6 MAXIMUM GROUND CONCENTRA-

TION AND FUMIGATION
0 zi = 0.025
0 2. = 0.061
0 z; = 0.002
Because regulations are written in terms of
-- maximum ground concentrations, it is informative to
.
z31
0
N
differentiate the Gaussian plume equation with
respect to x and set the result equal to zero to
determine the distance to the maximum concentra-
tion. When uY mu,, this occurs at the distance
downwind where = h2, where h is effective
plume height (equal to hs -t Ah). The concentration
at that distance is given by the formula:
Fig, 4.7 Values of uz/(uwt) VL t* for nearsurface (4.10)

releases (zi) during unstable conditions. Release
heights range from 0.002 to 0.067.
Experience shows that this critical distance is a few
tens of stack heights (h,) downwind. There is a
average value over the plume depth is generally critical wind speed at which C,,, itself is a
recognized. In practice, the wind speed a t the maximum if there is any plume rise at all. This
effective release height (h) of the plume is used. phenomknon is called high wind fumigation and
Sometimes observations of this wind speed are can persist for hours. The term fumigation in this
available, but usually the wind speed must be context means a situation in which high concentra-
estimated by using observations near the surface. The tions are brought to the ground from an elevated
theoretical formulas (Eqs. 1.37 and 1.38) or a plume. Equations 2.29 and 2.30 can be used to
power-law formula can be used: estimate the critical wind speed (uc) and the distance
(xC) to the maximum concentration point. Equation
= );(
10
P
(4.9)
2.25 gives the plume rise (Ah) under those condi-
tidns. Concentration (C) can then be calculated from
Eq. 4.9.
where z is height in meters and u l o is the observed Limited-mixing fumigation occurs when up-
wind speed at a height of 10 m. This formula is used ward diffusion or penetration of the plume is
by several of the U. S. Environmental Protection restricted by an inversion, below which strong mixing
Agency (EPA) models, with values of the parameter p occurs. In Los Angeles, for example, a fairly persis-
estimated by Irwin (1979b) given in Table 4.6. The tent marine subsidence inversion exists at a height of
power law is less accurate than Eqs. 1.37 and 1.38 about 500 m. If the inversion height is zi and i t is
assumed that the vertical distribution of material in
the plume is uniform from the ground to zi, then the
Table 4.6 Estimates of the Power (p) in Eq. 4.9 ground-level concentration is given by
for Six Stabiiity Classes and Two Roughnesses*
C= Q (4.11)
Stability class (2n)%l zi uy
A B C D E F
Tennessee Valley Authority experience shows that
Urbanp 0.15 0.15 0.20 0.25 0.40 0.60 limited-mixing fumigation more frequently gives the
0.10 0.15 0.35 0.55
,
Rural p 0.07 0.07

highest ground concentrations a t their very tall stacks
*Table is based on information by Irwin (1979b). (hs> loo,), whereas high wind fumigation is more
frequently critical at their shorter stacks.
This can be seen as follows: consider the ratio of
the concentrations predicted by the two methods
and should not be used a t heights above 2 0 0 m
(assume x = u200 for z > 200 m), but it is an
easy-to-use solution to the problem of wind variation
(4.12)
with height.
ON OFF
If we assume that h2 = 2 U: at the maximum point,
then Eq. 4.12 becomes
CLM h
- X 2.4- (4.13) C
CHW zi
If mixing height (zi) is 500 m , then limited-mixing

fumigation will be more important than high-wind
fumigation when the effective plume height is greater
t
than about 200 m. (a) SAMPLING TIME (TS) IS LENGTH OF TOTAL DATA RECORD
4-7 AVERAGING TIMES AND PEAK-TO- - ACTUALDATA

MEAN CONCENTRATION RATIOS 0 INSTRUMENT READING
A measuring instrument must be turned on and

off, and it always has some inertia that prevents i t
from responding to very rapid fluctuations. Also, so
that the number of data points can be kept at a
manageable level, groups of data are usually averaged
together at certain interviils. The total period over (b) INSTRUMENT TIME LAG (Ti) DEFINED BY AC/AC, = e-(f%)ni
which the instrument operates is called the sampling
time, and the time imposed by instrument inertia
and/or averaging is called the averaging time. These
concepts are illustrated in the data records in Fig. 4.8.
The effect of-finite sampling and averaging times
is to remove very high and low frequency fluctuations
C
from the data record, which would thus reduce the
total variance. If the sampling time is T,, then eddies
with periods greater than T, will not contribute to
the calculated variance. Similarly, if the averaging
time is T,, then eddies with periods less than T, will 1
f
+
not contribute. Pasquill (1974) has expressed this (c) AVERAGING TIME (T,)
analytically:
Fg,4.8 Illustration of sampling time (T,) (a),
instrument time lag (Ti)@), and averaging time
(Tal (4-
cated. Furthermore, since the turbulent energy

spectra of the y and z components are not necessarily
the same, the increase of uY and uz with averaging
where uZ(T,,Ta) is the variance for sampling time T, time is not necessarily the same either. Gifford
and averaging time Ta and F(n) is the spectrum
(1975) suggests accounting for the effects of sampling
function at frequency n.
time through the empirical formula:
The diffusion parameters (uy and uz) are directly
related to the standard deviations of the turbulent (4.15)
velocity fluctuations (a, and ow). As sampling time
increases, intuition and Eq. 4.14 tell us that observed
values of U, and ,a, and hence uy and uzr increase. where d and e represent two different cases, and q is
The exact level of increase is clearly a function of the in the range 0.25 to 0.3 for 1hr < T,d < 100 h r and
spectrum F(n) and the times T, and Ta, and the equals approximately 0.2 for 3 min < T,d < 1 hr. The
analytical equation quickly becomes very compli- standard Pasquill-Gifford curves represent a sampling
-
time T,, of about 10 min. Thus uy for a sampling
time of 1 hr equals 6.2, or 1.43, times the uy found
P
-=-
L
v2 + D 2
,
f 7
in Fig. 4.4. This is very close to an increase of one y2 ex IS(++ D2)
1
stability class interval.
An air-pollution concentration observation (Ca) (z - h)
involves an average over a period Ta. If the concentra-
tion record (as in Fig. 4 . 8 ~ has
) significant energy
+ 2(y2- +D2) (4.17)
a t periods greater than Ta, then the individual Cs are

not very representative of the true mean (c). Physi- Unfortunately little is known about the behavior of
cally, this situation is caused by the plume meander- either the large-scale component (D2)or the small-
ing about, as in Fig. 4.9. As the narrow plume passes scale component (yz ).
over a given point, the concentration is high for a
brief time but then drops to a low value again as the
plume meanders over to a new position. Gifford
4-8 SECTOR MODEL FOR LONG
(1959a) stated that the ratio of peak-to-mean concen-
SAMPLING TIMES
trations (P/M) is in the range from 1 to 5 for source
and receptor at the same elevation. As effective
plume height increases while the receptor remains at Over a long period of time, such as days or
the ground, P/M increases, with values of 50 to 100 months, the wind direction is likely to touch on all
or greater near a 100-m stack. The ratio P/M points of the compass. The horizontal crosswind
decreases as distance from the source increases. distribution in the plume, usually represented by u,,,
Cifford suggested that the P/M problem could be will no longer be important since the wind-direction
analyzed by assuming that there are t w o separate frequency distribution will be fairly smooth as we go
scales of diffusion, where the total diffusion c o v e r a around the compass. In this case the sector model can
long period equals the sum of the diffusion y2 of the be used, where it is assumed that there is no
instantaneous plume due to small scales of turbulence horizontal crosswind. variation in concentration
plus a contribution D2 due to meandering of the within an angular sector equal to the resolution of the
plume: wind-direction data. For example, wind direction is
commonly reported as N, NNE, NE, etc., giving rise
(4.16) to 16 sectors with width 22%. The formula for
ground-level concentration from a continuous-point
This assumption leads to .the following formula for source within sectors of arbitrary angular width 2n/n
the P/M: (in radians) is as follows:
1-HR AVERAGE PLUME POSITION

--- 1-MIN AVERAGE PLUME POSITION
Fig. 4.9 Comparison of short-and long-term average plume position.

2. Suppose that you are burning your trash after

dark in your backyard. The fire has died down so that
there is no plume rise, and assume no downwash or
where f is the fraction of the time the wind blows source effects. Source height is 1 m, wind speed a t
toward that sector. Thus, given an annual wind rose, that height is 2 m/sec, stability class is F, and source
the distribution of annual average concentrations strength of suspended particles is 1 g/sec. What
around a continuous source can be estimated. concentration is your neighbor, sitting in his lawn
chair on the plume centerline 100 m downwind,
exposed to? How far must he move his chair sideways
Problems before the concentration decreases to 10 mg/m3 ?
3. A big power plant releases 1000 d s e c of SOZ.
Effective source height is 200 m, wind speed a t a
1. Assume that the effective plume height is height of 10 m is 6 m/sec, and it is cloudy. Plot the
100 m and that the continuous-source release rate is ground-level plume centerline concentration as a
10 g/sec. Stability is neutral, wind speed is 5 m/sec, function of distance from x = 0 to x = 20 km.
ue is lo, and U, is 5. Calculate ground-level plume 4. Calculate maximum ground-level concentration
centerline concentration at a distance of 1 krn from and the downwind distance a t which it occurs using
the source for the following u methods: BNL, Briggs both the high-wind and limited mixing formulas
open country, Pasquill-Gifford-Turner figures, and for the following conditions: Q = 1000 g/sec,
Draxler (Eq. 4.8). Give the deviation of each answer u(10 m ) = 4 m/sec, sunny, h = 2 5 0 m , and zi=
from the average of the four. 1000 m.
Statistical Models of Diffusion
from Continuous-Point Sources
5 - 1 INTRODUCTION
The statistical, similarity, and K theories of

diffusion provide alternatives to the Gaussian plume
model discussed in Chap. 4. These theories are dis-
cussed in Chaps. 5 to 8. The definition of statistical
models used in this chapter is based on the fact that
diffusive motions have a certain random or stochastic
nature. Thus the path of an individual particle can be
described by a statistical function. If particles are Fig. 5.1 Orientation of steps m and n in Monte Carlo
assumed to have no memory of their previous problem.
motions and if their chances of going left or right at
any time are equal, then they will follow a
drunkards walk or Monte Carlo path. This not valid for atmospheric diffusion. In the boundary
simple discrete-step stochastic diffusion model is valid layer of the atmosphere at heights of 10 to 100 m
for molecular diffusion. Let n be steps downstream during the daytime, the turbulent speed v(t) will be
from the source and m be steps perpendicular to the correlated with the speed v(t + At) for time lags At
axis (see Fig. 5.1). The particles flow downstream in as great as several minutes. For very short time lags,
the n direction at a constant rate and can diffuse only on the order of 1 sec, the speeds v(t) and v(t f At)
in the crossstream direction. The probability of are nearly identical. An autocorrelation coefficient
finding the particle at steps n,m is given by the R(At) is defined by the formula
formula
v(t) v(t f At)
R(At) =
o, (5.2)
where the bar indicates a time average. For small time

As n increases this solution approaches the familiar lags, R + 1,and for large time lags, R -+ 0.
Gaussian or bell-shaped curve discussed in Chap. 4, The speeds (v) discussed in this section refer to
with u2 = n. Since n is proportional to time, the those felt by the particle or parcel as i t moves. This
result of a pure Monte Carlo diffusion model is system is called a Lagrangian system of motion. In
u a t%. contrast, the speeds felt by an anemometer fixed to a
mast are called Eulerian. Relations between these two
systems were discussed in Chap. 1, Sec. 1-7, and are
5-2 TAYLORS THEOREM further discussed in Sec. 5-5.
Taylors (1921) theorem of diffusion from a
The physical basis of the drunkards walk or continuous source begins with the assumption that y
Monte Carlo method, i.e., noncorrelated motions, is is the crosswind deviation from a fixed axis of a
36
STATISTICAL MODELS O F DIFFUSION FROM CONTINUOUS-POINT SOURCES 37
typical particle due to eddy speed vr after a time t. initial motions, the problem reduces to a Monte Carlo
The symbol 7 (equal to u;) indicates the mean problem, and uy 0: tk
square of a large number of values of y. Particles have A simple exponential form
(-3
been assumed to be released from a single point and
travel paths illustrated in Fig. 5.2. Each particle is
R(t) = exp (5.7)
inertialess and follows the airflow exactly. Then the
rate of change with time of 4
is given by
has often proved to be a useful approximation to the
autocorrelogram. The integration of Eq. 5.4 with the
lt
use of this form for R(t) results in the solution
=2 vr(t) v(t + t) dt (5.3)
If the turbulence is homogeneous (does not vary in

space) and stationary (does not vary in time), then
PARTICLE 1 which is plotted in Fig. 5.3. It is interesting t o note

that the asymptotic lines oy = &t2 and = 2 4Tt
meet roughly at a time equal to twice the time scale
T.
\
PARTICLE 2
Fig. 5.2 Traijectories followed by two typical

particles.
Eq. 5.2 can be substituted into Eq. 5.3 to yield, upon

integration,
u; = 2 .fi JiR(t) dt dt (5.4)
This is usually referred to as Taylors equation.

We can use simple approximations to determine
the behavior of 6 at small and large times.
AstO,
lo-+ / 4
R(t)+ 1 and 6%&?
(5.5)
or uy a t
J; Yt) dt = T
L
where T is a constant known as the time scale, and t I I
1
Thus, as travel time for a continuous plume increases,
the rate of diffusion decreases. Particle motions are a t
Fig. 5.3 Analytical solution (Eq. 5.8) to Taylors
first linear because the particles remember their equation with the assumption that R(t) = exp (-t/T).
initial velocity. However, when such travel times are Asymptotic solutions for andl and large times are also
reached that the particles no longer remember their ShOWn.
Equation 5.8 can be used to derive a theoretical

form for the function fy discussed in Sec. 4-4.2: 4' 2u"
(1 - n)(2 - n)&
(uz,*t)2-"
(5.11)
The factor n is determined from
(5.12)
This formula for fy (in which it is assumed that t = x/u
and u = 5 m/sec and T = 50 sec) is compared with
Pasquill's (1976) curve in Fig. 5.4. There is good This method was used extensively in the 1950's but
agreement in the midrange, where f y is approximately eventually got so complicated by empirical factors
0.5 and x is approximately 1 km, but formula 5.9 that i t was replaced by the simpler Pasquill-Gifford
slightly overpredicts at small x and slightly under- method.
predicts at large x when compared with Pasquill's
curve.
5-3 INFLUENCE OF EDDY SIZE ON 0
According to the results of Taylor's theory

11.o . 2 derived above, the
~ rate of change of uy with time is a
decreasing function of time of travel for continuous
plumes. We can guess intuitively that the influence of
small eddies decreases as the plume grows since small
oscillations of particles deep within the plume do not
contribute to "diffusion" of the plume as a whole.
Instead, eddies with diameters roughly equal to and
larger than u cause most of the diffusion. This result
can be shown analytically since the autocorrelation
x. km coefficient [ R(t)] and the energy spectrum [ F(n)]
are known to be Fourier transforms of each other.
Fig.5.4 Comparison of Pasquill's empirical sugges- The variable nF(n) is proportional to the amount of
tion for f(x) with the analytical solution (Eq. 5.9) to turbulent energy camed by eddies of frequency n.
Taylor's equation with the assumption that t = x / u (The frequency n of an eddy equals the inverse of its
and u = 5 m/sec and T = 50 sec. period; n increases as eddy size decreases.)
Without going through the details of the mathe:
matics, we can rewrite Taylor's equation (5.4) as:
Doran, Horst, and Nickola (1978) and Draxler
(1976) define a different time scale (Ti) as the time at
which the function fy drops to 0.5. Equation 5.9 can
be used to show that T i z 7 T , where T is the
Lagrangian time scale. Of course, any user of Taylor's Sampling time (T,) is assumed to be much greater
equation must realize the strict physical assumptions than travel time (t); thus the plume is assured of
made in its derivation. Since turbulence is assumed to being continuous instead of instantaneous. The term
be stationary and homogeneous, the theory cannot be sin2(nnt)/(nnt)' is unity for large eddies with fre-
expected to be strictly valid in shear zones, such as quencies less than roughly l/t and drops to zero for
the near-surface layer of the atmosphere. small eddies with frequencies greater than roughly
Sutton (1953) solved Taylor's equation (5.4) by l/t. In other words, this term is a filter that passes
using the following assumed form for R: only low frequency eddies, and, as travel time
increases, fewer and fewer eddies are passed by this
filter; hence oy a t for small t and uy 0: t i for large t.
(0 < n < 1) (5.10) There is a low frequency cutoff imposed by the
finite sampling time, T,. This cutoff is accounted for
where v is viscosity. It develops that by multiplying inside the integral (5.13) by [l - sin2
STATISTICAL MODELS OF DIFFUSION FROM CONTINUOUS-POINT SOURCES 39
(nnTs)/(nnTs)2] (see Eq. 4.14). We can conclude that to calculate RL(t) and RE(t) by using concurrent
continuous-source diffusion is influenced only by observations from free-floating balloons and fixed
eddies with periods ranging roughly between the anemometers. Gifford (1955) and Hay and Pasquill
travel time and the sampling time. Pasquill (1974) (1959) suggested that Lagrangian and Eulerian
points out that these results are equivalent to the spectra and autocorrelograms were similar in shape
simple equation: but displaced from each other by a factorp.
Figure 5.5 illustrates this similarity concept, which
can be written analytically;
where p is the ratio of Lagrangian to Eulerian time

scales, which has been observed to average about
four. The terms in the parentheses imply that o', is
sampled by a fixed anemometer over a total time
TAP, where consecutive averages over a time t/P are where 0 can be formally defined as the ratio of the
taken within that sampling period before & is Lagrangian to the Eulerian time scale:
computed. The concept of sampling and averaging
time has been illustrated in Fig. 4.8. Thus for p =T L (5.17)
constant T, the effective turbulent energy U: will TE
decrease as travel time increases. With Eq. 5.14,
diffusion can be calculated by using basic wind-speed A simple method of evaluating 0is to consider the
observations, which makes this equation a very circular eddy in Chap. 1, Sec. 1-6, which gave the
practical but so far underused model. result that P was inversely proportional to turbulence
Doran, Horst, and Nickola (1978) assume a intensity [ P = 0.5/i = O.~/(O,/U)]. Pasquill (1974) sug-
standard form for the spectrum F(n) for stable and gests that, on the average, in the planetary boundary
unstable conditions and integrate Eqs. 4.14 and 5.13, layer, P equals 4 for neutral conditions, + 10 as
which thus accounts for the effects of averaging
timet and sampling time T, on 6. They find that
larger values of gy/(aex) are associated with longer
averaging time and sampling time. However, because
of the scatter in the results of diffusion experiments,
the improvement over simpler models for uy/(oex)
(e.g., Fig. 5.4) is difficult to see.
\
5 4 LAGRANGIAN-EULERIAN 0 I I I I
RELATIONS
Small elements of pollutant gases or inertialess

particles follow the airflow exactly. Intuitively,
velocities measured by parcels following the airflow
(Lagrangian) are more slowly varying than those
measured by a fixed anemometer (Eulerian). The
problem is that we must use fixed anemometer
observations to estimate diffusion, which requires
that relations between Eulerian and Lagrangian
turbulence be developed. The spectrum F(n) in
Eq. 5.13 and the autocorrelation R(t) in Taylor's
equation (5.4) are Lagrangian parameters.
If08 is known, the nRL(t) a: can be estimated
(see Eq. 5.4) from d20;/d$. However, second
Fig.5.5 Eulerian (RE) and Lagrangian (RL)auto-
derivatives of field observations are very difficult to correlograms (a) and spectra (b), which are assumed to
compute accurately. A more common method of have similar shapes but time scales that are different
comparison of Lagrangian and Eulerian turbulence is by a factor p .
stability increases, and -+ 1 as stability decreases. comes from the fact that the spectrum for an
There is much scatter in observations of 0. assumed autocorrelogram of the form exp (-At/TE)
has peak energy a t a period of about ~ T EThen,
. if we
substitute Eq. 1.65 (with i = uUr/u) in Eq. 5.22, the
5-5 MONTE CARLO PARTICLE time scale (TL) becomes
TRAJECTORY MODELS OF
DIFFUSION
(5.23)
Thousands of particle trajectories can be easily

Typical time scales (TL) for a convective boundary
calculated by todays high-speed computers, and the
layer of depth 1000 m and sampling times of about
statistics (e.g., cry) of the particle distribution after a
1 hr are on the order of 100 sec.
certain time can be estimated. This is potentially a
powerful technique for the evaluation of diffusion in The advantage of this technique is that diffusion
nonuniform and nonstationary wind and turbulence calculations are related directly to basic turbulence
fields. In practice, time steps At of a few seconds are characteristics. The calculated u,, values agree exactly
used in the computer model, and the following with the analytical solution to Taylors equation
equation is assumed: (see Eq. 5.8) when mean winds and turbulence are
assumed to be stationary and homogeneous. In these
x(t) = x(t - At) + u At (5.18) applications the particles are released from the same
point, and the initial turbulent velocity is chosen
where the total speed, u, is the sum of a mean and a randomly from a Gaussian distribution with zero
turbulent component, mean and variance 41.
Reid (1979) applied this technique t o estimate
u=ii+u (5.19) vertical dispersion from a ground-level source and was
able to satisfactorily simulate observed distributions
The turbulent component is the sum of a correlated in the Prairie Grass diffusion experiments. In that
component and a random or Monte Carlo com- case strong vertical shears of mean wind speed were
ponent: present. This Monte-Carlo-type method will be most
applicable to difficult situations, such as sea breezes
u(t) = u(t - At) R(At) + u (5.20) or complex terrain, in which the Gaussian plume
model does not apply.
where the random component u is assumed t o have
a Gaussian distribution with zero mean and variance
given by
c(,ip
Problems
41, 1 [l - R2(At)]
=4 (5.21)
1. Assume that R(t) = 1 - It/TI for It/TI < 1
This relation is necessary to conserve energy 41from and R(t) = 0 for It/T1 > 1. Find an analytic formula
one time step t o the next. The autocorrelation for 0y with Taylors equation.
coefficient R(At) is a Lagrangian variable. 2. A continuous release of neutrally buoyant
The turbulent energy C&I can be estimated from material is made at a rate of 1 g/sec from a height of
boundary-layer equations, such as 1.42 to 1.53. Any 100m. Lagrangian time scales T in the y and z
form for R(At) can be used, b u t the simplest is directions are 50 and 10 sec, respectively. Wind speed
R(At)=exp (-At/TL). Thus the Lagrangian time is 5 m/sec. Calculate the centerline concentration
scale must be estimated before Eq. 5.20 can be used. after a travel time of 30 sec.
We use the formula 3. In the daytime P FZ 2. A plume is sampled over
a period of 10 min, and averages of 10 sec are made
within that period. What sampling and averaging
(5.22)
times should be used in calculating 4 so that the
equation 4 4 = t2 is valid?
where h , ~is the wavelength of peak energy in the 4. Show that Eq. 5.22 follows from Eq. 5.21.
Eulerian spectrum, as given by Eqs. 1.57 and 1.58 for Hint: square both sides of Eq. 5.21 and average by
the three velocity components. The fraction (&) using Reynolds averaging conventions.
Puff Diffusion
6-1 INTRODUCTION Puff or relative diffusion parameters (a) appropri-

ate for the instantaneous plumes in Fig. 6.l(c) have
been measured in very few field experiments. Conse-
Possibly the most confusing aspect of atmo- quently questionable situations arise in which some
spheric diffusion is the difference between plume and model developers use Pasquill-Gifford continuous
puff diffusion. Plume diffusion formulas apply to plume u's for puffs. In this chapter several theoretical
so-called continuous plumes, where the release and approaches for estimating puff 0's are outlined, and a
sampling times are long compared with the travel few comparisons with data are made.
time from source to receptor. On the other hand,
puff or relative diffusion formulas apply to so-called
instantaneous sources, where the release time or 6-2 STATISTICAL APPROACH
sampling time is short compared with the travel time.
These definitions lead to a dilemma when the release In the Taylor statistical approach to diffusion
time is roughly equal to the sampling and travel from continuous plumes covered in Chap. 5 , the
times, as shown in Fig. 6.1. In this case a combination trajectories of single particles relative to a fixed axis
of the two techniques may be necessary. were discussed. However, for puff diffusion, no fixed
axis can be defined, and the motion of one particle
relative to another must be studied. For this reason
puff diffusion is often called relative diffusion.
Batchelor (1950) writes an equation analogous to
Taylor's.equation:
- -
la) y2 = y i +2 I,T J,t' 6v(t) 6v(t + t l ) dt, dt' (6.1)
where yo is the initial separation of two particles and

6v is their relative velocity (i.e., the difference
T, = t
between the velocities of the particles, 6v = v2 - vI).
0
However, very little is known about the correlation
term h ( t ) 6v(t + tl).
(b)
Smith and Hay (1961) expand on Batchelor's
analysis and assume an exponential correlogram with
length scale 1 to derive an equation for the growth of
a puff
s, [w]
0
2
dx = 2Pi2 - dn (6.2)
where i = uv/u is the intensity of turbulence, r = 011,

Fg.6.1 Plume shapes relative to sampling time (T,)
and travel time (t). (a)Continuoussource plume and n = K I (K is wave number). The weighting
(T, > t). (b) Continuous or instantaneous plume function (in brackets) is, in effect, a filter function
(T, = t). (e) Instantaneous plume (T, t). analogous to that used in the statistical theory for
41
continuous plumes. This filter function tells us that 6-3 SIMILARITY APPROACH
eddies with sizes roughly equal to the size of the puff
or cluster are most important for the growth of the Batchelor (1952) isolated the basic physical
pulf. In contrast, eddies with sizes much less or much parameters involved in puff diffusion and used them
greater than the size of the puff contribute little to t o derive similarity formulas. For example, he rea-
the puff diffusion. The filter can be thought of as a soned that at short times, when the puff dimensions
window that passes eddy sizes between roughly 012 were a t scales within the inertial subrange, the rate of
and 50. The argument for the lack of contribution diffusion of the puff (dc?/dt) was a function of the
from small eddies is the same for plume and puff eddy dissipation rate (E), the time after release (t),
diffusion: small eddies move particles inside the and the initial size of the puff (uo):
plume or puff with little influence on diffusion. Large
eddies merely transport a puff bodily. On the other
hand, large eddies can contribute to the diffusion of a
plume from a fixed axis. This situation is simply
summarized in Table 6.1. or
Table 6.1 Influence of Eddy Sizes on

Puff and Plume Diffusion
In practice, this diffusion law applies for only a few
Diffusion seconds. At longer times, when the puff has forgotten
type Size range of eddies contributing to diffusion its initial size, the following similarity law is valid:
Plume Only eddies with sizes greater than u times travel
time and less than u times sampling time da2'
dt
a Et2
Puff Only eddies with sizes close to the puff size
(+ factor of 3)
or
Once the standard deviation (u) of the distribution of

material in a. puff is known, the concentration (C) of
material can be calculated by the Gaussian formula: The constant c2 has been evaluated from a few
experiments and is generally thought to be of order
c =- QP -r2 120'
unity. Most experiments yield the relation
(2n)%0 3 e 0.5 < c2 < 2.
At still longer times, when the puff dimensions
where QP represents emissions (in mass per second) are at scales larger than the eddy sizes in the inertial
and diffusion is assumed to be isotropic (i.e., the subrange, the rate of puff diffusion becomes equal to
same in all directions). The variable r is the radial the rate of plume diffusion.
distance from the puff center. Richardson (1926)
recognized the dependence of puff diffusion on puff
dimension when he used empirical data a t many
scales to derive an expression for the effective eddy Using these formulas, we can summarize the differ-
diffusivity for puffs: ences between puff and plume diffusion (Table 6.2).
K = 0.20' (6.4) Table 6.2 Comparison of Puff and Plume Diffusion
where K has units of square meters per second and u Short times Long times
has units of meters.
Puff oat (veryshortt) u a tH (t>TL)
The main result of the difference between plume
and puff diffusion is that there is a region in which o a t' (intermediate)
puff growth is greater than plume growth. This occurs
Plume oat (t<TL) oatH (t>TL)
when the puff is growing through sizes within the
inertial subrange of the energy spectrum, where the
spectral energy is rapidly increasing as eddy size The u a ts regime does show up in puff diffusion
increases. A' precise formulation of this effect is observations, as shown in Figs. 6.2 and 6.3 (Gifford,
better given by similarity theory. 1977). These data include radioactive clouds, tetroon
PUFF DIFFUSION 43
-
E
-
D
lot
I
:
I-
s
3
100
Fx
10-2 100 io1 io2 lo3 io4 105 106
100 io1 102 103 io4 io5 io6 TRAVEL TIME (t), sec
TRAVEL TIME (t), sec
Data Key
Data Key
0 Angel1 et al. (1971) X Edinger (1952)
CI Frenkiel and Katz (1956) + Crozier and Seely (1955)
Q Peterson (1968) Outbound 2 Kao and Wendell (1968)
Q Seneca (1955) Z Braham, Seely, and Crozier (1952) A Peterson (1968) Inbound I Roberts (1923)
*
A Smith and Hay (1961) X Machta et al. (1957) + Byzova et al. (1970) X Kazanski and Monin (1957)
+ Hogstrom ( 1964)- e Randerson (1972)
*
k Heffter (1965) Smith and Hefferman (1956)
e Pack and Angel1 (1963) Crzwford (1966) + Davies (1959) 8 Hanna (1975)
Fig.6.2 Ten tropospheric experiments on relative Fig. 6.3 Twelve tropospheric experiments on relative
diffusion. [From F. A. Gifford, Tropospheric Relative diffusion. [From F. A. Gifford, Tropospheric Relative
Diffusion Observations, J. Appl. MeteoroL, 16: 312 Diffusion Observations, J. Appl. MeteoroL, 16: 312
(1977).] (1977).]
pairs, photographs of plumes, artillery bursts, and

soap bubbles. Typically a t Klaw is valid over part of
-?)
the range of most experiments, although a t. law
seems to give the best agreement over the largest part
of the figure. The figures show that the o 0: t H region -0.5 (1 (1 - e-t/TLv)2 (6.11)
is not often present at large times owing to the
presence of mesoscale and synoptic scale eddies (high where the eddy diffusivity (Ky) applies to the entire
and low pressure systems). flow, i.e., it is strictly a large-scale quantity, and vo is
In an attempt to develop a theory for uy that fits the initial velocity at the source. If Eq. 6.11 is
both puff and plume observations, Gifford (1981) averaged over all possible initial velocities (vo), then
began with the statistical equation (5.20), v(t)= Taylors solution (Eq. 5.8) for diffusion from a
v(t - At) R(At) + v. Here the lateral component (v) continuous source is obtained. For fixed vo, however,
is used instead of the component u. He recognized the above solution applies to instantaneous (puff)
that this is a form of Langevins equation: diffusion at small times. As time increases, the
solution approaches the. Fickian equation,
dv
-+ pv(t) = v(t) (6.10)
$y = 2 K,t.
dt Equation 6.11 was compared by Gifford (1981)
with the field observations in Figs. 6.2 and 6.3,
where is a random acceleration and P- equals the assuming that K = 5 x lo4 m2/sec, vo = 0.15 m/sec,
Lagrangian turbulence time scale T L ~ .A solution to and TL = lo4 sec. The resulting curve provided a
Eq. 6.10 is as follows: good fit to the observations for all travel times.
106 I I I I I I I I I CONTINUOUS INSTANTANEOUS

PLUME PLUME OR PUFF
A Time-averaged or
-:..-,.. ..^^:^I^
s!,lylr-IJd'LlLlr
.J:u..-:-..
" I 4 IUIIVII
t < T, T, Z= t t > T,
CI
L
Instantaneous or hr
relative diffusion
104 50
N L km
E
N0
, --+ i = 5 rn/sec
Fe. 6.5 Diagram of plume sampled by an aircraft for,

1 hr (T, = 1 hr) at the downwind distances indicated.
When travel time (t) is less than T,, the observed
distribution indicates a continuous plume. When t is
greater than T,, the observed distribution indicates an
instantaneous plume or puff.
photographs taken by a U-2 aircraft of a large oil fog

10 20 50 100 200 500 1000 plume are first analyzed individually and then used to
TIME, sec form a composite plume. As shown, u is proportional
Fig. 6.4 Chart of u; vs. time for instantaneous and
to t%for the individual photographs (puff diffusion),
time-averaged plumes. (From C. J. Nappo, Relative and as predicted by our theory.
Single Particle Diffusion Estimates Determined from Figure 6.5 illustrates another case in which the
Smoke Plume Photographs, in Proceedings of the results of a single experiment can provide information
Fourth Symposium on Turbulence, Diffusion, and Air on puff diffusion, plume diffusion, and a combina-
Pollution, Jan. 15-19, 1979, American Meteorological
Society, 1979.)
tion of the two. Suppose a long power-plant plume is
sampled by several aircraft over a period of 1 hr. Each
aircraft flies through cross sections of the plume a t a
given distance downwind for the whole hour. The
6-4 APPLICATIONS
wind speed is 5 m/sec. -During a travel time that is
equal to the sampling time, the plume covers a
We have a few theories for puff diffusion and a
distance of 18 km. Thus, for downwind distances (x)
data set that is several orders of magnitude smaller
less than about 10 km, the aircraft observations yield
than the data set for plume diffusion. There is no
information about continuous plume diffusion. For x
ordering of the u curves by stability in Figs. 6.2 and
greater than about 30 km, the aircraft observations
6.3. It is no wonder, then, that many applied models
are related more to puff diffusion. Most photographs
for puff diffusion slip back and use plume U'S.
and aircraft observations of plumes at great distances
We recommend using Batchelor's formula,
(x > 50 or 100 km) are useful only in that they give
us information on puff diffusion.
$y = Et3
for puff travel times that are less than lo4 sec. The Problems
eddy dissipation rate should be evaluated locally a t
first and then at a height midway in the boundary
layer as u approaches 0.3zi. Equation 1.42 and the 1. The following table contains a variety of travel
equations in Chap. 1, Sec. 1.5 can be used to times and sampling times for a continuous plume. In
calculate e. For travel times greater than lo4 sec, the
constant c3 in the formula uy = c3t should be fitted Travel time 1
so that c3 = e H (lo4 secjh; i.e., the two curves
intersect at t = lo4 sec. A similar procedure is used
for uz except that u, = 0.32 for all times after uz first
attains that value.
A photograph of a continuous plume yields 30 rnin
information about puff diffusion since the sampling 2 hr
time is essentially zero. The results of an experiment
2 days
by Nappo (1979) are shown in Fig. 6.4. Four
PUFF DIFFUSION 45
the boxes write either puff or plume, depending 4. Assume you are conducting a field experiment
on which kind of diffusion analysis is appropriate. in which a long power-plant plume is being sampied
2. With the use of Fig. 6.2 and Richardsons by lidar on an aircraft. You take only a 5-sec traverse
formula, plot a graph of K, vs. travel time over the at each downwind distance, with distances greater
range of travel times from 1 sec t o io7 sec. than 1 km from the power plant. Wind speed is
3. Suppose a puff of 100 dandelion seeds is 3 m/sec. .4re you measuring plume or puff diffusion?
suddenly released from the top of a 100-m-high Suppose you take several traverses over a time period
tower. Assume that E = u%/O.4z and us = 0.2 m/sec. of 20 min a t each downwind distance and average the
Calculate the density of dandelion seeds in the center results. At what distance from the power plant does
of the puff after a travel time of 100sec. Assume the meaning of your measurement change from
c2 = 1.0. plume t o puff?
Similaritv Models of Diffusion
7-1 INTRODUCTION In Chap. 6 we showed how similarity theory
could be satisfactorily applied to the diffusion of a
puff in the inertial subrange, where the important
Useful models of diffusion are often possible to
derive if the important variables and governing variables and parameters are u, e, and t.
In this chapter the discussion is restricted to the
parameters for a problem are known. Such models are
diffusion of continuous plumes released near the
called similarity models because they imply similar
ground in the surface boundary layer. Detailed
behavior of the atmosphere from one place or time to
derivations of this theory are given by Gifford (1968,
another if we assume that certain parameters, such as
pp. 88-90), Pasquill (1974, pp. 116-123), Monin and
zi/L or z/zo, are held constant. The techniques that
Yaglom (1971, Chap. 5), and Horst (1979).
lead to similarity models are sometimes called dimen-
sional analysis.
The first step in developing a similarity model is
to isolate all the variables and parameters that apply
7-2 DIFFUSION OF CONTINUOUS
to a problem. In general, if there are n variables and
parameters (al , a z , . . . a,) of m different units (e.g., PLUMES IN THE SURFACE LAYER
grams and seconds), then n - m dimensionless inde-
pendent numbers can be formed. For example, 7-2.1 Neutral Conditions
consider the wind speed in an unstable surface layer.
We have the parameters u (in meters per second), u*
In the first example neutral conditions are as-
(in meters per second), z (in meters), zo (in meters),
sumed. As a result, the stability parameter z/L does
and L (in meters) and can conclude that 5 - 2 = 3
not enter the problem. In our search for important
independent dimensionless numbers can be formed:
governing parameters, we look for parameters that are
constant in the surface layer (0 < z 6 5 0 m). The
-U
= f(z/zo ,z/L) friction velocity u* is one such parameter, and i t is
C u*
known to be an important scaling velocity for the
which we know from Chap. 1 is a solution to the wind profile. Several people independently postulated
problem. the equation
If n - m equals 1, then we know the solution to
the problem to within a constant factor. This is how
Briggs arrived a t some of his plume-rise predictions in
Chap. 2. For example, plume rise Ah(m) in a calm,
stable environment is a function of initial plume where is the mean height at time t after release of
buoyancy flux F0(m4/see3) and environmental stabil- particles released continuously from some point near
ity s (see-). Therefore n - m = 3 - 2 = 1, and we the ground. This equation can be converted from
can write the universal similarity formula Ah = C1 F? dT,,dt to dZ/djl by dividing by the wind speed.
s-%. The powers and -% are necessary to make However, the wind speed transporting the plume is
the units correct, and the constant C1 can be effective at height .:a Pasquill recommended, on
determined by plotting observations of Ah vs. Fo b
theoretical grounds, that c = 0.4 and a = 0.56. Horst
s - ~ . This formula is found to be valid for plume rise9 (1979) recommends a = 0.63 in neutral conditions,
Ah ranging from centimeters to kilometers. with a between 0.5 and 0.6 in unstable conditions
SIMILARITY MODELS O F DIFFUSION 47
and 0.8 to 0.9 in stable conditions. With the use of is less than 1.5 for unstable conditions. The concen-
u = (u,/0.4) In (z/zo), Eq. 7.1 can be written in the tration is not very sensitive to the value of a used in
fo m Eq. 7.2.
One final equation is needed to obtain concentra-
tion predictions with this technique. We are dealing
here with crosswind-integrated concentrations:
C(x:z) = JY- C(x,y,z) dy. Then the mass continuity
which can be integrated, by using T = zo a t T = 0, condition is given by the equation
with a = 0.56, to give the solution
j i u(z) C(x,z) dz = Q
(-z - (7.5)
- -
X z z
-= 6.25 In- - 1 . 5 8 5t 1.58) (7.3)
ZO ZO ZO
where Q is the continuous-source strength. By substi-
tuting Eq. 7.4 into Eq. 7.5, we get the solution under
This solution is plotted in Fig. 7.1.
neutral conditions for the variation of crosswind-
integrated ground-level concentration with downwind
distance,
.
which must be solved numerically. (An analytical
0
IN
solution for the integral is available if r = 2.) Horst
(1979) gives a graphical solution that can be approxi-
2kL.d-J
mated by the simple power-law formula:
10102 2 . 5 lo3 2 5 lo4 2 5
x /zo This formula is within 10% of his curve for dimen-

sionless distances x/zo ranging from lo2 to lo.
Fig. 7.1 Solution to Eq. 7.3 for the variation of mean
cloud height C;i/z, ) with distance downwind (Y/z,) for Before these equations are applied out to X / Z O
ground-level sources in a neutral boundary layer. equal to lo, they should be checked to determine if
the mean plume height (Z) is still within the surface
layer., Remember that we are assuming that friction
The only information this analysis gives is the velocity (u+) is constant and that the wind profile is
mean height of the plume. It does not tell anything logarithmic, assumptions that are valid only for Z less
about the vertical distribution of material in the than about 50 or 100 m. For typical zo values, this
plume, which is not necessarily Gaussian because of will limit downwind distance X to just a few
the nearness of the ground and thestrong wind shear. kilometers. Also, a ground-level source has been
Several observations of crosswind-integrated concen- assumed.
tration C in the surface layer due to continuous
releases are summarized by Horst (1979), who found
7-2.2 Nonneutral or Adiabatic Conditions
that the following equation is valid:
--
C(?,Z)
C(X,O)
- exp [43] (7.4)
In nonneutral or adiabatic conditions, Eq. 7.1
must be rewritten to include a dependence on the
stability parameter (z/L):
The factor b equals r(l/r)/r(2/r), where r is the dz

-= cu* f(z/L)
gamma function (which is partially tabulated in dt
Table 8.1). The Gaussian distribution value r = 2
(which gives b = 1.77) is an overestimate except for Chaudhry and Meroney (1973) assume that
fairly stable conditions, the value r = 1.5 (which gives
b = 1.52) is valid for nearly neutral conditions, and r f(z/L) = @h*(Z/L) (7.9)
where r$h is the dimensionless temperature gradient vertical distribution parameter (r) equaled 1.5 for all
given by Eq. 1.36: stabilities. The following simple empirical formulas
provide a satisfactory fit (ffactor of 2) to the most
-w stable and unstable curves:
@h(z/L)= 0.74 (1 - 9 3 (unstable)
At zO/L = lo-
z
= 0.74 + 5 (stable)
(-)
-0.69
u*zo C(x,O)
0.4Q = 0.75 (7.11)
This result follows from their assumptions that f(z/L)
is similar to the stability-dependent term in the eddy At zO/L = -lo-
diffusivity formula (K = 0.4u,z/@) and that the eddy
diffusivity for pollutant equals the eddy conductivity
for heat.
As in Sec. 7-2.1, the dependence on time in
u*zo C(X,O)
0.4Q
= 35 );( -1.54
(7.12)
Eq. 7.8 is converted to a dependence on downwind

distance x through the relation: dF/dt = .(a?). How- Horst tested his predictions by using observations of
ever, the wind-speed function, as given by Eq. 1.37, is crosswind-integrated ground concentration from the
more complicated than in the neutral calculation. The Prairie Grass diffusion experiment (Barad, 1958) and
procedures used in the last section give the following an experiment at Idaho Falls (Islitzer and Dumbauld,
equation for the crosswind-integrated surface concen- 1963). He found good agreement for all stabilities at
tration: distances x/zo out to 2 x l o 5 .
The concentrations (C) calculated above refer to
Q/zo crosswind-integrated values. The actual point concen-
C(x,O) =
Jy u(z/z~,z/L)exp [-(z/bgr] d(z/zo) trations at the ground [C(x,y)] a t a crosswind distance
y from the plume axis can be calculated from Eq. 7.7
(7.10) or 7.10 or from Fig. 7.2 by the following conversion:
Fortunately Horst (1979) has already solved this

equation on the computer for 15 values of zo/L
ranging frorn--lO- to lo-, yielding the graph of
C(X,O)
-
C(X,y,O) = ( 2 I p uy exp (-6) (7.13)
normalized C given in Fig. 7.2. He assumed that the The dispersion parameter (uy) can be obtained
through standard techniques outlined in Chap. 4. In
fact, the published Pasquill-Gifford uy curves were
based on concentration measurements taken at
ground level for emissions at ground level.
7-3 DIFFUSION IN THE FULL DEPTH

OF THE DAYTIME PLANETARY
BOUNDARY LAYER
Other basic theories of diffusion owe their exis-

tence to the similarity approach. A good example of
this is the problem of diffusion in the full depth of
the daytime planetary boundary layer. Recent experi-
ments in Minnesota and England show that important
scaling parameters for this problem are the mixing
IU
102 103 104 lo5 lo6 10 depth (zi) and the convective velocity scale (w,),
NORMALIZED DOWNWIND DISTANCE (x/z,) which is defined by Eq. 1.41:
Fig. 7.2 Predicted crosswind-integratedconcentration

at ground level as a function of downwind distance for
various stability conditions. [From T. W. Horst,
Lagrangian Similarity Modeling of Vertical Diffusion
from a Ground Level Source, J. AppZ. MeteoroL, 18:
734 (1979).] We can postulate the similarity formula
SIMILARITY MODELS OF DIFFUSION 49
The result in Eq. 7.18 implies that C,, 0: z r ' ,

(7.14)
which is in disagreement with the prediction of the
Gaussian model (given by Eq. 4.9) that C m a x a z F z .
and find that this formula is indeed valid, with Observation of fumigation in daytime conditions is
Cz ~ 0 . 5 6 ,for times less than the Lagrangian time not detailed enough to permit selection of one model
scale. The Lagrangian time scale ( T L ~ )is given by over another.
another similarity formula:
zi
TL" = 0.15 - (7.15)
Problems
UV
1. With the use of the same coefficient C1 as that
where uV is a function of zi/L and w,. In general, we derived on Earth, pould you expect the plume-rise
find that any distance variable ( I ) and speed variable equation Ah = C1 F!S-~~ to be valid on Mars? Why?
(s) in the daytime planetary boundary layer are 2. What is the ground-level crosswind-integrated
described by the similarity formulas concentration during neutral conditions a t a distance
of 200 m from the source for roughness (zo) equal to
1 cm, friction velocity (u*) equal to 0.5 m/sec, and
surface source strength (Q) equal to 10 g/sec? Calcu-
-late the concentration on the plume axis assuming
uo = 0.1 radian.
3. Assume that the roughness length equals 0.3
cm. There is a continuous source of strength of 5
where f and g are universal functions. g/sec. What is the ratio of the ground-level crosswind-
Lamb (1979) has used the turbulence fields from integrated concentration for L equal to 10 m to the
Deardorff's (1974) numerical model of the daytime ground-level crosswind-integrated concentration for L
planetary boundary layer to calculate the maximum equal to -10 m a t a downwind distance of 300 m?
concentration expected from effective release heights How d o plume axis concentrations compare with
(q) greater than about 0.025zi. The calculated those made by using the standard Gaussian formula?
maximum ground-level concentration (&ax) and the (Assume same uy for both methods.)
distance (xmax) a t which it occurs are given by the 4. The mean height 0 can b e calculated for the
formulas standard Pasquill-Gifford-Turner Gaussian plume
model (half of area of Gaussian curve is within 0.670)
Q
cm,,(x,o,o) = 1.2-Z,ZiU (7.18)
for a surface-level source with full reflection. Calcu-
late the variation with x of Z for neutral conditions
for the Gaussian model and compare it with Fig. 7.1
for zo = 1 cm. How far apart are the two i estimates
(7.19)
a t x = 100 m?
Gradient Transport (K) Models
8-1 THE BASIC GRADIENT TRANSPORT sure to have scales less than the thickness of the
MODEL plume. For greater point release heights (e.g., tall
stacks), the diffusion equation should not be used
until the pollutant of interest is spread out over
Chapter 1, Sec. 1-4, gives a derivation of the
several hundred meters.
continuity equation for a substance (C), where
turbulent fluxes of C are assumed to be proportional
to the mean gradient of C:
8-2 ANALYTICAL SOLUTIONS
-wC =KZ-
ac
aZ Equation 8.2 is impossible to solve analytically
for completely general functional forms for the
The basic gradient transport model can be written: diffusivities K and wind speeds u, v, and w. Before
the days of computers, it was a popular exercise to
ac t u -
- - t w -ac
ac t v ac = a
S t -K, ac
- solve this equation for specific forms for K and u.
at ax ay aZ ax ax Carslaw and Jaeger (1959), Pasquill (1974, pp. 108-
116), and Sutton (1953) give further details. Many of
+-Ka -ac
t - K , a- ac (8.2) these cases are highly instructive.
ay y ay aZ aZ
So-called cross-diagonal terms, such as a/ax (Kxy 8-2.1 One-Dimensional Equation, Time-
aC/ay), are not included here, because they are Dependent, Constant K, No Wind,
usually insignificant. Instantaneous Area Source
It is important to point out that certain time and
space scales are implicit in the diffusion equation. Consider a simplified form of Eq. 8.2:
The mean wind components (u, v, and w) and mean
concentration (C) represent averages over a time scale ac
-= K-
a2c
(T,) and space scale (xa). Velocity 5uctuations with at ax2 (8.3)
time and space scales less than these values are
considered turbulence and are implicitly included in That concentration (C) varies only with time and
the K coefficients. However, as shown in Chap. 5, the distance (x) and that diffusivity (K) is a constant are
rate of diffusion of a plume depends on the plume assumed. Boundary conditions are
size. This statement contradicts the diffusion equa-
tion, which uses constant Ks. We can conclude that C + 0 as t + 00, all x
the diffusion equation is valid only if the size of the
plume is greater than the size of the dominant C + 0 as t + 0, all x except x = 0
turbulent eddies so that all of the turbulence implicit
in K is taking part in the diffusion. 1
: C dx = Q
Point sources and the diffusion equation are
therefore compatible for vertical diffusion when the where Q is the instantaneous area source strength
source is near the ground, where turbulent eddies are (mass per unit area). Thus this problem simulates an
50
GRADIENT TRANSPORT (K) MODELS 51
instantaneous emission from a very large plane The solution is
surface. The solution is
Qip
C=
(47rt)%(KxKyKz)%
which has Gaussian form with the standard deviation

of the distribution (0)given by the equation
Xexp
[ ;(
-- -+-+-
it cy $1 (8.8)
This solution, like Eq. 8.4, is Gaussian with standard

deviations
This type of diffusion (constant K) is called Fickian ox = (2Kxt)H

diffusion. The implied relation between u and K is
often used by researchers to estimate the diffusivity uy = (2KYtfh (8.9)
(K) from the observed diffusion parameter (0).
However, a quick look at the assumptions made to uz = (2KZt)%
derive this result tells us that it is likely to be
This type of diffusion is also known as Fickian
questionable because diffusivity (K) is seldom a diffusion.
constant in space and time.
Note that the u 0: tH dependence in Eq. 8.5 is the
same as the functional dependence given by Eq. 5.6, 8- 2.3 Two-Dimensional, Time -Inde pendent ,
the solution to Taylors statistical diffusion equation Variable u and K, Continuous Ground-
at large times. An assumption important for both Level Line Source
solutions is that the cloud of pollutants is larger than
Diffusion from an infinite crosswind continuous-
the space scales of the turbulent eddies. The follow-
line source is described by the following simplifica-
ing relation can be obtained by equating Eqs. 5.6 and
tion of Eq. 8.2:
8.5
K=$,T (8.6)
. .
(8.10)
where T is the Lagrangian time scale. This equation,
like Eq. 8.5, has often been used to estimate the with boundary conditions
Lagrangian time scale. Again, caution is advisable in
applying any of these formulas which are derived C + 0 as x, z + 00
under highly restrictive assumptions.
C + m a s x = z -+ 0
8-2.2 Three Dimensions, Time-Dependent, ac

-+Oasz+Oandx>O
Constant K, No Wind, Instantaneous Kz aZ
Point Source
The solution in See. 82.1 is simple to expand to
$0 UC d z = QI, x > 0
three dimensions, which implies diffusion of a puff where QI (in mass per unit length divided by unit
from an instantaneous point source[Qip(mass)] in an time) is the continuous-line source strength. The third
environment with no mean wind. The diffusivity is boundary condition ensures that there is no flux of
assumed constant in any given direction but can be material into the lower boundary (i.e., ground).
different for different directions. The basic equation Much work has been done on obtaining analytical
and boundary conditions are soh tions to this equation for specified functional
forms for K, and u. These solutions should be applied
to ground-level sources only, for which eddy sizes are
generally less than plume size. Roberts (1923) gave
the correct solution for the conditions
C + 0 as t + m,all x, y, z
(8.11)
C + 0 as t + 0, all x, y , z except x = 0, y = 0, z = 0
u = u1 );( (8.12)
The generalized solution is Pasquill (1974) gives references for further ana-
3'
lytical solutions under the conditions u = constant
Qlz?Ym - n + 2 ) [( U1 and
C(x,z) = 2u 1 r(s) m-n+2) 2
z1m-n K l x K a (h- z)" (0 < z < h)
1
u1 Zm-n+2 K az(h - z) (0 < z < h)
(8.13)
exp[ z t - " (m - n + 2)2 K l x
Kaz (0 < z < h/2)
where s = (m + l)/(m - n + 2) and l' is the gamma Ka(h-z) (h/2< z < h)
function (e.g., Abramowitz and Stegun, 1964,
Chap. 6). Representative values of m, n, s, and r(s)
are listed in Table 8.1. The first nine values of r(s) in 8-2.4 Three-Dimensional, Time-Independent,
Constant u and K, Continuous:Point
Source at Ground Level
Table 8.1 Gamma Function for Typical Values of s*
For continuous-point sources at ground level, we
m n 6 I'(4 can assume that u aC/ax P a(K, aC/ax)/ax; i.e.,
0.9 0.1 0.679 1.33 advection dominates diffusion in the downwind
0.8 0.2 0.692 1.31 direction. The basic equation is then
0.7 0.3 0.708 1.29
0.6 0.4 0.727 1.26
0.5 0.5 0.750 1.23
0.4 0.6 0.778 1.19
0.3 0.7 0.813 1.15 C-+ 0 as x, y, z + 00
0.2 0.8 0.857 1.11
0.1 0.9 0.917 1.06
0 0 0.5 1.77 C+masx,y,z+O
"First nine rows assume m = 1 - n; last row assumes con- ac

K,-+Oasz+Oandx,y >0
stant K and u. aZ
UC dy dz = Q, z > 0
Table 8.1 are calculated by using the conjugate power
law, m = 1 - n, which arises from the surface- where Q is continuous-point-source strength (mass
boundary-layer relation per time). The approximate solution for u and
Ky = K, equal to constants is
u:=K-
aU (8.14)
aZ
(8.18)
Since the stress uf is constant in the surface
boundary layer, it is required that m = 1 - n if
Kaznanduazm. The problem with this solution is that axial concen-
For the simple case of constant u and K tration drops off as x-', whereas observations show
(n = m = 0; the last column in Table 8.1), the solu- that x - ' . ' ~ is more accurate. The reason for this
tion is Gaussian: error is the assumption of constant K, which results
in a failure of the diffusion equation to account for
rapid 'diffusion at small times, where we know that u
should be proportional to x. However, the constant K
diffusion equation can give us only u a x'.
The only situations for which the solutions
(Fr
Note that, if we assume t E x x / u l , the standard
deviation of the distribution is described above are accurate are a continuous
ground-level line source and a continuous or instan-
taneous ground-level area source. Here the diffusion
u= = (2Kt)' (8.16)
scale is always greater than the turbulence scale.
These methods are also valid a t times or distances
which is the same result as that obtained for the downwind beyond which the diffusion scale is larger
instantaneous area source in Sec. 8-2.1. than the Lagrangian time or distance scale. However,
. . . .
these techniques should not be used for such situa- 0
tions as local diffusion of a neutrally buoyant
.
effluent from a tall smokestack.
n - 1 n n + l
a a a
i
17
8-3 NUMERICAL SOLUTIONS OF THE m+l m + l m+l
.
DIFFUSION EQUATION
E n-1 n n + l
Analytical solutions t o the diffusion equation are 0 0 a
a
E
m m m
interesting but are limited in applicability by restric- UI
.
a
.
V
tions on K and u. Furthermore, because of physical 0
v)
limitations, the models should not be applied to n - 1 n n + l

elevated point sources. As a result, analytical solu- a a 0
rn-1 m - 1 m - 1
tions are no longer widely used. This is in great
. .
contrast to computer solutions of this equation,
which are very common. Time and space variability in
.-At-
K and u can be handled by carrying a large number of 0
grid points in the computer. Major problems are -TIME STEPS

numerical instabilities and the fact that our knowl-
edge of the distribution of K and u is not up to the
potential of the computer. Fig. 8.1 Illustration of numbering system for finite
differencinggrid.
8-3.1 Numerical Instabilities

used in this problem, the numerical solution would
Computer solutions are obtained by stepping blow up.
forward in time or in space. If a random error This restriction on At can be somewhat relaxed
introduced a t some point will amplify indefinitely by using more accurate finite difference schemes.
with each succeeding time or space step, then the Even if the solution is stable, however, there still is
solution is said to be numerically unstable. An some undesirable diffusion that appears solely due
analysis of Eq. 8.3 shows this: to the finite difference method. The advection term
(u aC/ax) in Eq. 8.2 is especially noted for its
ac - a2c contribution to numerical diffusion. A standard
TKS technique for studying the diffusive characteristics of
a finite difference scheme is to begin with a cosine-
An explicit finite difference approximation to this
shaped hill and then advect it around the edge of a
equation is
circle by using the equation
ac
-t u -ac
tv ac
-=o (8.21)
at ax ay
The solutions for six different finite difference
techniques studied by Long and Pepper (1976) are
where i is the current time position and j is the plotted in Fig. 8.2. In each case the hill is advected
current space position (see Fig. 8.1). By writing C in once around the circle. The second moment, cubic
complex form, it can be shown that this solution is spline, and Chapeau function do not distort the hill
numerically stable only if the following condition is too much. However, numerical diffusion significantly
met: distorts the hill in the donor cell, fully implicit, and
At 1 Crank-Nicolson schemes. If these schemes were used
K-<- (8.20)
Ax2 2 to solve the diffusion equahon, it would be difficult
to know if the calculated diffusion was real or
Thus, if the diffusivity (K) is 10 m2/sec (typical numerical.
daytime boundary layer) and the grid distance (Ax) is Other techniques can be used to get around the
100111, then the time step (At) must be less than problem of numerical diffusion. The second-moment
500 sec, or about 8 min. If a time step of 1 hr were scheme devised by Egan and Mahoney (1972) belongs
54 ATMOSPHERIC DIFFUSION ,
DONOR CELL FULLY CRANK- SECOND CUBIC SPLINE CHAPEAU
(Upwind Differencing) IMPLICIT NICOLSON MOMENT (Quasi-Lagrangian)
. . FUNCTION
INITIAL
DISTRIBUTION
SLIGHTLY MORE THAN

% REVOLUTION
SLIGHTLY MORE THAN

%REVOLUTION
ALMOST
1 REVOLUTION
THREE-DIMENSIONAL CONTOURS
Fig. 8.2 Application of six types of numerical schemes to solve the advection equation (8.21). The
initial cosine hill is at the top of each series. The next three figures are the distributions resulting from
applications of the numerical schemes to advection in a circle around the plane. (Adapted from Long
and Pepper, 1976.)
to this class and was evaluated in the paper by Long heights above about 0 . l Z i . The old adage that you
and Pepper (1976). Spectral approximations sug- cannot get something for nothing applies to the
gested by Prahm and Christensen (1977) are useful diffusion modeling business also. The gradient trans-
but are too complicated to describe here. The port model may be physically attractive, but it
Particle-in-Cell (PIC) method (Lange, 1978) is some- encounters problems when the time comes to specify
times used, where an effective velocity equal to the K,.
actual velocity minus (K/C) aC/ay is used to trans- The eddy diffusivity coefficient can be assumed
port particles. to equal the eddy conductivity coefficient (Kh). In
It is sobering to think that none of the very the lowest 50 to 100 m of the atmosphere (z sO.lzj),
extensive work described in this section has anything formulas by Businger et al. (1971) (Eqs. 1.36 and
to do with the physics of the diffusion problem. The 1.39) can be used:
purpose of this work is solely to speed up the
computer calculations and to make them more true
to the physical equation.
8-3.2 Specifying the Vertical Diffusivity %(z/L) = 0.74 (1- 9 $)- (unstable)
In most applications of the gradient transport or

diffusion equation at scales less than 10 km, the
= 0.74 -+ 5 7
Z
(stable)
horizontal diffusivity (Ky) is neglected, but the
variation of the vertical diffusivity (K,) must be Note that Businger et al. prefer to use 0.35 for von
known. In addition, the vertical variation of the wind Kkmalns constant (k). During daytime conditions
speed must be input to the model. As a result, in the and a t heights between about 0 . l Z i and O.?zi, K, can
comparison of the Gaussian and the gradient trans- be assumed constant, with a value equal to its value at
port models, the question reduces to the relative 0 . l Z i . Above O.?Z~, K, probably decreases linearly to
accuracy of two empirical parameters, U, and K,. At a small value a t the mixing height zi. This type of
this point, K, is probably less well known than u, at behavior was reported by Crane, Panofsky, and
Zeman (1977), who calculated K, from observed
verfical pollutant fluxes w'C' and gradients aC/az
over Los Angeles during the daytime. At night K, is
quite uncertain above heights of about 50 to 100 m
but probably has some small residual value, say
0.1 m2/sec.
The coefficient K, used by Smith (1972) is based
on a suggestion by Hanna (1968) (Eq. 1.63):
K, = 0.150, ,A
where ,A is the wavelength of peak energy in the w

spectrum. The parameters U, and ,A can be calcu-
lated for all stability conditions by using Eqs. 1.42 to
1.47, 1.57, 1.59, and 1.82.
In Smith's analysis, he solves the diffusion equa-
tion and then extracts u, from the calculated
concentration distributions. The goal of this work is
to devise a revision to the Pasquill-Gifford CJ, curves,
where CJ, is dependent on surface roughness as well as
on stability. 0.01 0.02 0.03 0.04
Other researchers use different formulas for K,, fK&
but common characteristics of these formulas are a
linear variation near the ground, a constant value at
mid-mixing depth, and a decrease as the top of the Fig. 8.3 Dimensionless plot of Sht's (1973) formula-
mixing layer is approached. For example, Shir (1973) tion for &.
recommends the formula
K, = O.~U,Z~-*~~/"* (8.22) known. This method of solution is known as "first

order closure." However, this closure assumption is
which is based on a theoretical analysis of the neutral not always valid; for example, at the top of the
boundary layer. The depth of the neutral boundary mixing layer is an upward flux of heat against a
layer is about 0.25 u,/f, where f is the Coriolis positive temperature gradient. Also, as just shown,
parameter. Equation 8.22 is plotted in dimensionless the diffusivity K, is highly variable and is not well
form in Fig 8.3.
known in the upper half of the mixing layer.
We can eliminate some of these problems (and
generate some new problems) by invoking a higher
8-4 HIGHER ORDER CLOSURE order closure. Consider the z component of Eq. 1.16,
The derivation of the gradient transport equation

in Chap. 1, Sec. 1-4, used assumptions of the form ac
-+ ac B + S
w-= (8.23)
at aZ
C=CtC'
Substitute w = W + w' and C = t C' into this
This results in the -
appearance of second-order un- equation, and use the continuity equation to convert
known terms like ~ ' C ' a n d X in the continuity
w ac/az to awC/az. Then, multiplying the equation
equation. So that the system can be closed (i.e.,
by w' and averaging yields
reduce the number of unknowns to equal the number
of equations), these terms are eliminated by making
hypotheses like -a-
wfCft-
at
ac-+a-
aZ
w w
aZ w w c a-
aZ
t-W'W'C'
= 6+? (8.24)
Now the mean value C is the only unknown in the

equation since the diffusivity K, is assumed to be
7-4
We now have an extra e uation, for w'C', and a new
third-order term: w w C . At this step second-order
closure is introduced to reduce the number of
knowns to the same value as the number of equa-
Problems
tions.
1. The wind is blowing from the west a t 5.0 m/
(8.25) sec at a height of 5 m perpendicular to a highway
with carbon monoxide emissions of 1.0 g m-l sec- .
The power law parameter for the wind profile is 0.1,
In this expression A is an (unknown) length scale. and the diffusivity K, has a value of 1.0 m*/sec at a
This same procedure could be followed to generate height bf 5 m. Plot the variation of ground-level
methods of third-order, fourth-order, or whatever- concentration with downwind distance from the
order closure you desire. With each new order a new highway for x = 5 m to x = 200 m. At a height of
unrestricted term on the plus side is gained, but there 20 m, at what downwind distance does the maximum
is a longer, more-complicated, string of governing concentration occur?
equations on the negative side. 2. In a numerical model for bomb-debris trans-
At first sight the introduction of the new length port in the upper atmosphere, K, equals lo6 m/sec,
scale A appears to be a problem. It turns out, and the grid distance Ax is 200 km. What is the range
however, that A in the second-order scheme is not as of time steps that can be used to ensure numerical
variable as K in the first-order scheme. Donaldsons stability? Assume that a simple explicit finite dif-
(1973) review of second-order closure models gives ference scheme is being used.
methods of estimating A. He shows that these models 3. Plot the variation of K, with height up to
give predictions of boundary-layer winds and turbu- 1 0 0 m by using (1) Busingers formula and (2)
lence which are in good agreement with observations. Smiths formula. Assume that L = -100 m and
However, a t this point second-order closure models zi = 1000 m.
are useful for basic research only and have not made 4. Plot the variation of K, with height up to
their way into the field of applied diffusion modeling. 1000 m by using (1) Smiths formula and (2) Shirs
They usually consume large amounts of computer formula. Assume that L = 50 m, zi = 1200 m,
core and time. u, = 0.5 m/sec, and f = 10-4s-1.
Urban Diffusion Models
9-1 IMPORTANCE OF EMISSIONS from one edge of the urban area to the other, is often
useful. The pollutant is then assumed to be uniformly
mixed in a layer of depth zi between the ground and
An urban area contains thousands, or even mil-
the mixing height. The wind speed (u) is assumed
lions, of individual sources that range from small
constant within the layer (see Fig. 9.2). An additional
sources, such as incinerators, to large sources, such as
assumption can be made that the mixing depth is
power plants The application of a diffusion model to
increasing with time (azilat), as it does in t h e
each of these sources is impractical, even if the
morning. The concentrations upwind of the city and
assumption can usually be made that the contribu-
above the mixing height are Cb and &, respectively.
tions of individual sources to the total concentration
Then the continuity equation for this volume is:
at a point are additive. Consequently we combine
most of the small sources into larger area sources of ac
AX zi - = AX Qa f UZj (Cb - C) f AX
azi (Ca - C)
at
strength, Qa (mass per unit time per unit area), and at
assume that emissions from t h e ground surface are
uniform over that particular area. An example of (9-1)
average annual area source emissions for SOz in Change in = source + change due + &awe due to
C with to horizontal mixing layer
Frankfurt, West Germany, is given in Fig. 9.1, where time advection growth and
a square grid (4 by 4 km) is used. Most area vertical
source-emission inventories are given on square grids, advection
although often the grid size may vary over the urban
area. Some simplifications are possible. If conditions are
Diffusion from the largest point sources can be steady state (aC/at = azi/at = 0) and the background
calculated individually, and the resulting concentra- concentration ( c b ) is zero, then the solution is simply
tions a t a receptor point can be added to the
contribution from area sources. The number of point
sources treated this way is usually between 10' and
100. Some pollutants, such as carbon monoxide, have
very few point sources, whereas others, such as SOZ, This is the well-known box model solution.
are. mostly emitted from point sources. An accurate Lettau (1970) defines the equilibrium box model
emissions inventory is essential for successful urban concentration given by Eq. 9.2 as C" and defines a
diffusion modeling. This inventory includes a knowl- scalirg time Ax/u as the flushing time required for
edge of the seasonal, weekly, and diurnal variations of the air to pass completely over the urban area. Define
emissions. If emissions are not known within a factor a nondimensional time t" = tu/Ax. Then, if Cb and
of 2, then the diffusion model has that error imposed Ca can be neglected, Lettau's (1970) simplification of
on it even before it starts working. Eq. 9.1 can be written:
ac - C"
-- - c (9.3)
9-2 BOXMODEL at*
which has the solution:

The assumption that emissions in an urban area
are constant over a distance Ax, which runs roughly
57
45
0 0 0 0
41
37
33
E
Y
z
29
1.99
25
21
1
17
13 17 21 25 29 33 37 41 45 49
x, km
Fig. 9.1 Average annual SO, area source emissions (in milligrams per square meter per second) for
Frankfurt, W. Germany (Hanna and Gifford, 1977). Emissions are printed in the center of the 4-by
4-km grid squares, and isopleths based on these values are drawn. Monitoring station locations are
circled.
. .
C,(CONCENTRATION ABOVE MIXING DEPTH) and resulting uniform concentration in the box
az,/at (CHANGE OF MIXING DEPTH WITH TIME) model. Realistically, however, as clean air enters the
t
urban area, the pollutant is swept out; so, after Ax/u,
-b u (wind speed) there should be virtually no pollutant material left in
I F the box. Venkatram suggests instead a slug model
Cb
(UPWIND
C (UNIFORM CONCENTRATION) Ib
.- 0
I-
that gives the following solution for the situation
where emissions are suddenly shut off:
BACKGROUND (3
CONCENTRATION)
C = (1 - t) (9.5)
Q,(AREA SOURCE)
However, it should be recognized that a fundamental

assumption of the box model is that there are no
(BOX WIDTH)
extreme changes occurring, such as the situation
suggested by Venkatram.
Fig. 9.2 Parameters in box model
The box model is often used as a screening model,
where, for example, a government agency might wish
where Co is the initial value of concentration. As time to identify a few substances in a long list of toxic
(t*) increases, the concentration (C) in this equation chemical emissions that should be singled out for
approaches the equilibrium concentration (C*) given special attention. In many cases, though, the box
by Eq. 9.2. model has been used as the basic workhorse in a
Venkatram (1978) points out that, if the source diffusion study, and it has been found to perform
term Qa suddenly drops to zero in Eq. 9.4, the quite well. An example of this is the application of a
concentration appears to decrease exponentially, photochemical box model to ozone air quality in
dropping to 0.37 Co after one flushing time Ax/u has Houston, Tex., by Demerjian and Schere (1979).
elapsed, because of the assumption of rapid mixing They use Eq. 9.1 and retain the final term relating to
URBAN DIFFUSION MODELS 59
changes due to vertical advection. Ozone is often Qd

Qa2 \
trapped above the inversion layer over urban areas at Qa1 \
\
\ \
L X P , T
night and is mixed down to the surface by the V \
\ \
growing mixing layer the nextmorning. Their applica-
tion requires a 3 2 k m grid size. The most complicated .:
part of the model is the 36-step chemical kinetic
mechanism. They find that predicted hourly concen- RECEPTOR
:\ I\
trations of hydrocarbons, carbon monoxide, NO,, \ \ \
and ozone are within a factor of 2 of observed \ \ \
\ Qa3
concentrations. They had an advantage here that \
might not be available at other locations in that Qa2
Qa1
detailed emission estimates were made as part of an
Environmental Protection Agency study in Houston. ----- a,w
APPROXIMATION SO THAT
Qa IS A FUNCTION ONLY OF x
9-3 THE ATMOSPHERIC TURBULENCE --- Qalx,V)
AND DIFFUSION LABORATORY ISOPLETHS OF AREA
SOURCE EMISSIONS
MODEL
Fig. 9.3 Illustration of MRDW plume hypothesis For
The Atmospheric Turbulence and Diffusion Labo- most purposes, emissions can be considezed a function
ratory (ATDL) urban diffusion model described by only of distance x.
Hanna (1971,1973) and Gifford and Hanna (1973) is
essentially a box model with the height of the top lid Next, the solution must be written in a form
proportional to the vertical dispersion parameter consistent with the typical square grid that is used to
[o,(Ax)] rather than to the mixing depth (q). For present urban emissions data. Consider Fig. 9.4, in
grid distances (Ax) less than about 10 km, u,(Ax) is which the receptor point is located in the center of
significantly less than zi, which is typically 500 to grid square 0 and upwind grid squares are denoted
1OOOm. This model uses the integral form of the by subscripts 1, 2 , 3, . . . N. The grid size is Ax. The
Gaussian plume model and treats an area source as an solution is obtained by piecewise integration:
infinite array of infinitesimal point sources of
strength Qa(x,y). The concentration C a t point x = 0,
y = 0, z = 0 is given by an integration over the upwind
half plane (Gifford, 1970):
We assume. that uz has the form

where sources are assumed to be at ground level.
Giffords (1959b) narrow plume hypothesis permits (52 = ax b (9.10)
the elimination of the y dependence in Eq. 9.6. As
shown in Fig. 9.3, a typical plume that subtends an Parameters a and b, suggested by Smith (1968), are
angle of only 10 to 20 crosses lines of constant Q, listed in Table 9.1. The source strength within each
in such a way that little accuracy is lost by assuming grid square is assumed to be constant.
that Q, is a function only of x. In this case Eq. 9.6
becomes
(9.?>
This can be written in the form
-=0
c =(2/n)%Ax
u (Jz
which is equivalent to the equilibrium solution Ftg.9.4 Area source grid pattern asumed in the
(Eq.9.2) to the box model. derivation of Eq. 9.11.
Table 9.1 Parameters a and b in Eq. 9.10 NE

WIND
Parameter
Meteorological
conditions a b
Very unstable 0.40 0.91

Unstable 0.33 0.86
NE
Neutral 0.22 0.80 WIND
Estimated Pasquill D 0.15 0.75
Stable 0.06 0.71
Smith's (1968) parameters are assumed to apply

to both urban and rural regions. The solution is then:
E
WIND
C = (2/n)'(Ax/2)'
ua(1 - b) -b {Qao + i Q a i [(2i + 1)' -b
k i=1 RECEPTOR
Fig. 9.5 Method of accounting for area source emis-

Values of the coefficients in the summation term for sions when' the wind direction is at an angle to the grid
several values of b and i are given in Table 9.2. squares.
Table 9.2 Values of Summation Term [ (2i + 1)' -b - Hanna (1973). The Gaussian plume model is used for
(2i - l)'-b]in Eq. 9.11 for Various Stabilities point sources in this program.
and i Values After this technique had been applied to several
urban areas, it was noticed that the calculated
Stability concentration (C) at any receptor was usually propor-
Very . Pasquill's tional to the emissions Qao in the grid square in
i unstable Unstable Neutral D Stable which the receptor was located. The reason for this is
that the distribution of emissions is usually quite
1 0.10 0.17 0.2 5 0.32 ' 0.38 smooth and the coefficients of the upwind Qai terms
2 0.05 0.09 0.14 0.18 0.22
3 0.04 0.06 0.10
in Eq. 9.11 are quite small. For most applications, it
0.13 0.16
4 0.03 0.05 0.08 0.1 1 0.13 is sufficient to use the simple ATDL model, which is
5 0.02 0.04 0.06 0.09 0.1 1 obtained by approximating the various source
6 0.01 0.03 0.05 0.08 0.1 0 strengths Qai by Qao in Eq. 9.11:
C = A -Qa o
U
Generally, N is the number of grid blocks necessary
to reach the upwind edge of the urban area. Equation
9.11 is most valid for time periods of about 1 hr, for
which the uz values in Table 9.1 apply, and the wind
direction and speed are fairly uniform (-+20", -+30%)). The dimensionless parameter A is evaluated in
Extension to longer averaging times is made by Table 9.3 for several values of Ax(2N + 1)/2 (distance
solving Eq. 9.11 for a variety of wind directions and to edge of city) and stabilities. Approximate values
then weighting each result by the frequency with for A for unstable, average annual (class D), and
which the wind blows from that direction. When the stable conditions are 60, 200, and 600, respectively.
wind direction croses emissions grid squares Shghtly different estimates of A might be made if
obliquely, as in Fig. 9.5, an arbitrary scheme must be Briggs's proposed formulas for uy and uz in an urban
devised for including the squares in Eq. 9.11. Also, region (Table 4.8) were used in the derivation follow-
Ax must be multiplied by l/cos At9, where .At9 is the ing Eq. 9.9. Gifford and Hanna (1973) verified the
departure of the wind direction from north to south annual value of 200, using suspended particle data
or east to-west directions (0 < At9 < 45'). A com- from several U. S. cities, and Hanna (1978) verified
puter program for these computations, which includes the variation with stability, using carbon monoxide
provisions for point sources, is given in a report by data from several cities. SO2 observations are best
Table 9.3 Evaluation of A in Eq. 9.12 for Various City where AC = carbon monoxide concentration (ppm)
Sizes and Stabilities N = traffic flow (vehicles/hr)
Stability
S = average vehicle speed (miles/hr)
Cityradius Very
u = wind speed at roof level (m/sec)
W = street width (m)
PasqUiUs
&(2N+ 1)/2,m unstable Unstable Neutral D Stable
x and y = horizontal distance and height (both m)
5000 48 57 100 180 545
10000 51 63 115 213 667 of the receptor point relative to the
20000 54 69 132 258 814 traffic lane, respectively
K = dimensionless best fit constant
simulated if the parameter .4 predicted above is The data suggest that I<-7. For wind directions
divided by four, presumably to account for the fact nearly parallel to the street,
that most SO2 sources are elevated. On the other
hand, carbon monoxide observations in most cities 1
AC, = [AC1 (windward) + ACl (lee)] (9.16)
are best simulated if the parameter A is multiplied by
three, presumably to account for the placement of
carbon monoxide monitors near busy streets. How- Equations 9.14 and 9.15 are subject to some revision
ever, A is calculated to equal about 200 (its value in since the average emission rate (e.g., grams of carbon
Table 9.3) from data from the St. Louis Regional Air monoxide per vehicle mile) changes with the vehicle
Pollution Study where monitoring stations are placed vintage and type of mixture, which would thus
away from busy streets in schoolyards and parks. In modify the numerator (source term) of both expres-
any case, the model described in this section ideally sions.
predicts concentrations over a broad area. The excess concentration AC1 contributed by a
major highway in an urban area is important for
perhaps 200 or 3001x1 downwind of the highway.
Many highway diffusion models are referenced by
9-4 STREET CANYON AND HIGHWAY Johnson et al. (1976), but most are based on the
SUBMODELS Gaussian model for an infinite line source:
The box modeland the ATDL model can give the (9.17)
average carbon monoxide concentration over a broad
area (say 10 by 10 km). In a street canyon or
adjacent to a highway in an urban area, there is an where Ql is line source strength (in mass per unit time
additional contribution to the concentration from per unit length), H is the effective height of emissions
local sources In this case the total concentration C is (probably 2 or 3 m), and 4 is the angle between the
the sum of a spatially averaged Ca and a local AC1
component:
MEAN BACKGROUND
C = C, + AC1 (9.13)
Johnson et al. (1976) outline methods of estimating

AC,. Consider the street canyon in Fig. 9.6, where
the important variables are defined. If the wind is
more or less normal to the street, the equations for
the concentration ACl in the street canyon are:
Lee side,
AC1 =
0.1 KNS- . (9.14)
LANE
(u + 0.5)[(x2 + z2)$ -t 21 -l -w
Fe.9.6 Schematic of crossstreet air circulation in a
Windward side, street canyon. [From W. B. Johnson, R. C. Sklarew,
and D.B. Turner, Urban Air Quality Simulation
Modeling, in Air Pollution, Vol. 1, 3rd ed., Chap. 10,
0 . 1 ~ ~ s - 05 - 7 p. 530, A. G. Stern (Ed.),Academic Press, New York,
(9.15)
AC1 = W(u + 0.5) 1977.1
wind direction and the highway. An initial value of U, which picks up pollutants emitted by the areas it
(e.g., 2 or 3 m) is assumed to account for the passes over. Vertical diffusion takes place by means
turbulent wake behind moving vehicles. If d is the of a vertical K, coefficient. Eschenroeder, Martinez,
perpendicular distance from the highway, then the and Nordsieck (1972) developed a model of this type
distance x to be used in evaluating uZ is d/sin @. for application to photochemical pollution in Los
krgeles. The advantage of the trajectory model is
that calculations must be made for only those few
9-5 COMPUTERIZED K MODELS FOR trajectories which end at the monitoring stations.
URBAN DIFFUSION
9-5.3 Grid Models with Winds Prescribed
In Chap. 8, K diffusion models were discussed.
The basic equation was given as well as various Several models use a fixed grid and assume that
techniques for estimating K. Most of the applications all meteorological parameters are known. Generally
of K models to urban diffusion described below took observed wind speeds at specific stations are inter-
place in the mid-1970s. Anyone interested in using polated to the grid points by means of a l / r 2
any of these models should consult the original weighting scheme. Only a few models (e.g.,
reference since the models and their input and output MacCracken and Grant, 1976) adjust the wind field
are too complex to cover fully in this brief summary. so that mass continuity is preserved.
Vertical K, profiles are usually linear up to a
height of about 100 m in these models. Above this
9-5.1 An Urban Diffusion Model That Also height KZ is assumed constant and sometimes de-
Predicts Winds and Temperatures creases near the top of the mixed layer (see Fig. 8.3).
Models by Reynolds and Roth (1973) and by Shir
Pandolfo and Jacobs (1973) applied their K and Shieh (1974) are typical of this group. Horizontal
model to estimate carbon monoxide concentrations diffusion can often be neglected for area sources, but
in Los Angeles. The model was originally developed in some models it is handled by specifying a constant
for studying the dynamics of the three-dimensional KY -
planetary boundary layer and includes radiative ex- Plumes from point sources cannot be resolved at
change processes. Basic dependent variables are wind scales less than the ghd distance and are usually
velocity and temperature. The diffusion equation was arbitrarily assigned to various grid squares on the
easy to add to the list of governing equations since basis of the length of the trajectory over the grid
the pollutant concentration has little feedback to the square. Pollutants whose major source is tall stacks
other equations. The model accounts for sloping cannot be accurately modeled at small scales by grid
terrain through the continuity equation. models.
With all the sophistication of this model, it is
surprising that the carbon monoxide predictions 9-6 ENVIR ONMENTAL PR 0TECTI0N
made by this model are no better correlated with AGENCY MODELS
observations (R = 0.6 to 0.8) than predictions made
by other models that did not include weather
A set of diffusion models recommended by the
variables. The explanation is apparently that the
U. S. Environmental Protection Agency is available
other models used the most recent observed winds on magnetic tape from the National Technical Infor-
and temperatures a t each stage, which are more mation Service, U. S. Department of Commerce,
accurate than the weather predictions of the Pandolfo Springfield, VA 22161. Eleven models are in this
and Jacobs (1973) model. There seems to be a UNAMAP system; all are based on the Gaussian
paradox: to run a complex model, we need good formula described in Chap. 4. The latest urban model
observations of initial and boundary data from a (RAM) uses the ATDL formulations for area sources
network of instruments. However, if these instru- derived in Sec. 9-3. The cast of models in this system
ments are in place and operating, there is often no is continually being updated as new models are
longer a need for the complex model. developed and evaluated. For example, plans have
been made to include a model for urban reactive
9-5.2 Trajectory Models pollutants in the near future. Turner (1979) gives a
detailed review of the status of the UNAMAP models
In a trajectory model, a box with dimensions of 1 as of May 1979. The appendix to his paper is
to 5 km is allowed to move with the observed wind, reproduced in this chapter as Table 9.4, which very
briefly descries each model and gives references to

users guides. The models are all state of the art and
are quite satisfactory. Some are conservative; i.e., the
predicted concentrations are probably higher than the
actual concentrations. 0
x x
1
1- x X . -
9-7 MODEL EVALUATION X
Most urban diffusion models yield correlations

between hourly values of observed and predicted
concentrations at a given station of about 0.6 to 0.8.
This result seems to be independent of the number of
statements in the computer program. Good results
. .
depend mainly on good knowledge of emissions and
.
wind velocities.
i
Hayes (1979) and Nappo (1974) discussed several
X
methods that can be used to evaluate models. These x I

include:
Bias evaluation: Ratio of mean predicted concen-
tration to mean observed concentration.
0 1 2 3 4 5 6 7 8 9 10
Error analysis: The mean square of the dif- TIME, hr
ferences between predicted and observed concentra- 0, Observed, X, Predicted.
The rime correlation R is 0.15, and bias is -0.28
tions is calculated.
Time correlation: Correlation between observed
..
and predicted concentration distributions with time
at a given station.
Space correlation: Correlation between observed
and predicted concentration distributions across a
monitoring network at a given time.
>
15
. . 4
4
Peak analysis: Comparisons of magnitudes and 13
0
locations of peak observed and predicted concentra- 0 1 2 3 4 5
tions from point sources are made. X
0, Urban area monitoring stations.
Distribution functions: Observed and predicted Number above is Observed C,
and number below is predicted C.
cumulative distribution functions are compared to see The space correlation R i s 0.76,
if they are significantly different. and bias i s 0
Fig. 9.7 IUustration of (a) bias, (b) time correlation,

and (e) space correlation.
The types of results obtained from some of these
techniques are illustrated in Fig. 9.7. A small bias in
an urban diffusion model can be corrected merely by
adjusting the model by using monitoring data. Of nine stations and 105 mg/m3 a t the final station, the
course, if there is a large bias, the modeler should correlation R will be a perfect 1.0. Consider an urban
look carefully for fundamental errors in his physical area with ten observing stations,all reporting concen-
assumptions or computer program. trations between 80 and 120 mg/m3. The model may
Time and space correlations are useful, but we d o very well predicting the mean concentrations, with
should realize that correlation coefficients can mask values between 80 and 120 mg/m3, also. However,
many strange variations in the data. For example, there may be a very low correlation coefficient. In
assume that the observed concentrations at nine this case the model is unfairly given a low rating by
monitoring stations are all 100 mg/m3, and the the spatial correlation method. A combination of
observed concentration at a tenth station is 1000 evaluation methods is best, including a subjective
mg/m3. If the model predicts 100 mg/m3 at the first judgment by an experienced modeler.
I
s
3
8
-3
3
E
n
E!
-
-
Table 9.4 Abstracts of Models i n UNAMAP (Version 3) C
z
APRAC: Stanford Research Institutes urban carbon monoxide model. Computes hourly averages for any urban location. Requires an extensive traffic inventory for the city of
interest. Requirements and technical details are documented in:
Users Manual for the APRAC-1A Urban Diffusion Model Computer Program (NTISP accession number PB-213-091).
Additional information is available on APRAC from:
A Practical, Multipurpose Urban Diffusion Model for Carbon Monoxide (NTIS accession number PB-196-003).
Field Study for Initial Eualuation of an Urban Diffusion Model for Carbon Monoxide (NTIS accession number PB-203469).
Evaluation o f the APRAC-1A Urban Diffusion Model for Carbon Monoxide (NTIS accession number PB-210-813).
Dabberdt, Walter F., F. L. Ludwig, and Warren B. Johnson, Jr., 1973, Validation and Applications of A n Urban Diffusion Model f o r Vehicular Pollutants,
Atmos. Enuiron., 7: 603-618.
Johnson, W. B., F. L. Ludwig, W. F. Dabberdt, and R. J. Allen, 1973, A n Urban Diffusion Simulation Model for Carbon Monoxide, J. Air Pollut. Control
Assoc., 2 3 (6): 490-498.
CDM: The Climatological Dispersion Model determines long-term (seasonal or annual) quasi-stable pollutant Concentrations at any ground-level receptor using average
emission rates from point and area SOUIC~Sand a joint frequency distribution of wind direction, wind speed, and stability for the same period.
Busse, A. D., and J. R. Zimmerman, 1973, Users Guide for the Climatological Dispersion Model, U. S. Environmental Protection Agency, Research Triangle Park,
NC, Environmental Monitoring Series, EPA-R4-73-024, 131 pp. (NTIS accession number PB-227-346).
HIWAY: Computes the hourly concentrations of non-reactive pollutants downwind of roadways. It is applicable for uniform wind conditions and level terrain. Although best
suited for at-grade highways, it can also be applied t o depressed highways (cut sections).
Zimmerman, J. R., and R. S. Thompson, 1975, Users Guide for HIWAY: A Highway Air Pollution Model, U. S. Environmental Protection Agency, Research
Triangle Park, NC, Environmental Monitoring Series, EPA-650/4-74-008, 5 9 pp. (NTIS accession number PB-239-944).
Three Point
Source Models: The three following point source models use Briggs plume rise methods and Pasquill-Gifford dispersion methods as given in EPAs AP-26, Workbook of Atmospheric
Dispersion Estimates, t o estimate hourly concentrations for stable pollutants.
PTMAX: Performs an analysis of the maximum short-term concentrations from a single point source as a function of stability and wind speed. The final plume height is used
for each computation.
PTDIS: Estimates short-term concentrations directly downwind of a point Source at distances specified by the user. The effect of limiting vertical dispersion by a mixing
height can be included, and gradual plume rise to the point of f i a l rise is also considered. An option allows thr calculation of isopleth half-widths for specific
concentrations a t each downwind distance.
PTMTP: Estimates for a number of arbitrarily located receptor points at or above ground level, the concentration from a number of point s o ~ r c e s .Plume rise is determined
for each source. Downwind and crosswind distances are determined for each source-receptor pair. Concentratiotls at a receptor from various sources are assumed
additive. Hourly meteorological data are used; both hourly concentrations and averages over any averaging time from 1 to 24 hours can be obtained.
Turner, D. B., and A. D. Busse, 1973, Users Guide to the Interactive Versions of Three Point Source Dispersion Programs: PTMAX, PTDIS, and PTMTP,
Preliminary Draft, Meteorology Laboratory, U. S. Environmental Protection Agency, Research Triangle Park, NC.
CDMQC: This algorithm is the Climatological Dispersion Model (CDM) altered t o provide implementation: of calibration, of individual point and area source contribution lists,
and of averaging time transformations. The basic algorithms t o calculate pollutant concentrations used in the CDM have. not been modified, and results obtained using
CDM may be reproduced using the CDMQC.
Brubaker, Kenneth L, Polly Brown, and Richard R. Cirillo, 1977, Addendum t o Users Guide for Climatological Dispersion Model, prepared by Argonnc National
Laboratory for the U. S. Environmental Protection Agency, Research Triangle Park, NC, EPA-450/3-77-015 (NTIS accession number PB-274-040).
CRSTER: This algorithm estimates ground-level concentrations resulting from u p t o 19 colocated elevated stack emissions for an entire year and prints out the highest and
second-highest 1-hour, 3-hour, and 24-hour concentrations as well as the annual mean concentrations at a set of 180 receptors ( 5 distances by 36 azimuths). The
algorithm is based o n a modified form of the steady-state Gaussian plumeequation which uses empirical dispersion coefficients and includes adjustments for plume rise
and limited mixing. Terrain adjustments are made as long as the surrounding terrain is physically lower than the lowest stack height input. Pollutant concentrations for
each averaging time arc computed for discrete, non-overlapping time periods ( n o running averages arc computed) using measured hourly values of wind speed and
direction, and estimated hourly values of atmospheric stability and mixing height.
U. S. Environmental Protection Agency, Research Triangle Park, NC, Monitoring and Data Analysis Division, 1977, Users Manual for Single-Source (CRSTER)
Model, EPA-450/2-77-013 (NTIS accession number PB-271-360).
PAL: Point, Area, Line source algorithm. This short-term Gaussian steady-state algorithm estimates concentrations of stable pollutants from point, area, and line sources.
Computations from area sources include effects of the edge of the source. Line source computations can include effects from a variable emission rate along the source.
The algorithm is not intended for application t o entire urban areas but for smaller scale analysis of such sources as shopping centers, airports, and single plants. Hourly
concentrations are estimated, and average concentrations from 1hour to 2 4 hours can be obtained.
Petersen, William B., 1978, Users G&de for PAL-A Gaussian-Plume Algorithm for Point, Area, and Line Sources, U. S. Environmental Protection Agcncy,
Research Triangle Park, NC, Environmental Monitoring Series EPA-600/4-78-013 (NTIS accession number PB-281-306).
VALLEY: This algorithm is a steady-state, univariate Gaussian plume dispersion algorithm designed f o r estimating either 24-hour or annual concentrations resulting from
emissions from up t o 50 (total) point and area sources. Calculations of ground-level pollutant concentrations are made for each frequency designed in an array defined
by six stabilities, 16 wind directions, and six wind speeds for 1 1 2 programdesigned receptor sites o n a radial grid of variable scale. Empirical dispersion coefficients are
used and includes adjustments for plume rise and limited mixing. Plume height is adjusted according t o terrain elevations and stability classes.
Burt, Edward W., 1977, VALLEY Model Users Guide, U. S. Environmental Protection Agency, Research Triangle Park, NC, EPA-450/2-77-018 (NTIS accession
number PB-274-054).
RAM: Gaussian-Plume Multiple-Source Air Quality Algorithm. This short-term Gaussian steady-state algorithm estimates concentrations of stable pollutants from urban point
and area sourcea. Hourly meteorological data are used. Hourly concentrations and averages over a number o f hours can be estimated. Biiggs plume rise is u s e d
Pasquill-Gifford dispersion equations with dispersion parameters thought t o be valid for urban areas are used. Concentrations from area sources are determined using
the method of Hanna; that is, sources directly upwind are considered representative of area source emissions affecting the receptor. Special features include
determination of receptor locations downwind of significant sources and determination of locations of uniformly spaced receptors t o ensure good area covcragc with a
minimum number of receptors.
Turner, D. Bruce, and Joan Hrenko Novak, 1978, Users Guidefor RAM. Vol. I, Algorithm Description and Use, EPA-60018-78-016a (NTIS accession number
PB-294-791). Vol. 11, Data Preparation and Listings, EPA600/8-78-016b (NTIS accession number PB-294-792), U. S. Environmental Protection Agency,
Research Triangle Park, NC.
~ ~~
*From D. B. Turner, Atmospheric Dispersion Modeling: A Critical Review, J. Air Pollut. Control Assoc., 29: 518-519(1979).
t NTIS, National Technical Information Service, U. S. Department of Commerce, SpringTield, VA 22161. 0
Z
Problems
1. If the initial concentration in an urban area is (3) Compare the solutions in (1) with the
zero and the sources are suddenly turned on, how answers you get from Eq. 9.12. Assume
many flushing times will pass before the concentra- that there are no sources beyond square 3.
tion equals 95% of its equilibrium value. 3. What is the concentration at receptor 5 in
2. What is the concentration in the receptor block Fig. 9.1 for east winds with a speed of 3 m/sec and D
of Fig. 9.4 if Pasquills D stability class is valid, wind stability class?
speed (u) is 2 m/sec, and grid size (Ax)is 5 km? 4. Motown, USA, has a street 20 m wide with 500
vehicles/hr driving on it at a speed of 20 miles/hr.
Wind speed at roof top is 5 m/sec and wind direction
is perpendicular t o the street. Background or area
source concentration of carbon monoxide is 4 ppm.
What is the added concentration on the windward
side of the street?
Removal Mechanisms
10-1 INTRODUCTION by Van der Hoven (1968) for spherical particles with
densities of 5 g/cm3 is plotted for sea-level situations
in Fig. 10.1. Settling speeds for particles with dif-
Most air pollution is eventually removed from the ferent densities (p,) can be approximated by multi-
atmosphere, either by transport to vegetation, soil, or plying the speed in the figure by [pp/(5 g/cm3)]. The
water or by chemical transformation to another settling speed of nonspherical particles can be calcu-
compound. For example, SO2 is removed by all these lated by dividing the speed of the equivalent spherical
mechanisms and has a half-life in the atmosphere of a particle [equivalent radius re = (3Vp/4n)%,V, = par-
few hours or days, depending on rain intensity and ticle volume] by a dynamical shape factor a. Typical
relative humidity, other chemicals in the air, and shape factors are shown in Table 10.1 (Chamberlain,
surface characteristics. On the other hand, about half 1975).
of all the COz we put into the atmosphere remains Van der Hoven (1968) suggests that, when v t is
there. The difference between the atmospheric half- greater than 100 cm/sec (radius r 7 100pm), the
lives of SO2 and COz is due to their reactivity with particles are falling through the turbulence so fast
substances in the air and at the ground surface. that diffusion is no longer important. In this case
Methods of transport to vegetation, soil, or water particle trajectories are calculated by a straight-
inc!ude dry deposition and precipitation scavenging.
Despite extensive field and laboratory experiments
and detailed theoretical calculations, there is much
uncertainty connected with fundamental parameters,
such as the dry deposition velocity.
10-2 DRY DEPOSITION
10-2.1 Gravitational Settling
T /
/
Particles with radii greater than about 5 pm have
significant gravitational settling speeds. Stokes' law
for the terminal settling speed (vt) is valid for ' 5O E
particles with radii less than 10 to 30 Pm, depending
on particle density:
(10.1)
where pp is particle density and p is the dynamic

viscosity of air (1.8 x g sec-' cm-'). For Fig. 10.1 Gravitational settling speeds for particles
particles with larger radii, Stokes' law must be with densities of 5g/cm3 near the earth'ssurface
modified somewhat. and the maDhicd solution kven (Adapted from Van der Hoven, 1968, p. 203.)
67
Table 10.1 Dynamical Shape Factor 01 ground. Thus, for h equal to l o o m , u equal to
5 m/sec, and 20-pm-diameter particles with densities
of 5 g/cm3 (vt = 6 cm/sec), about half of the material
is deposited within a distance of 8.3 km.
Ellipsoid 1.28
Cylinder 1.06
Cylinder 1.14
10-2.2 Deposition of Gases and of Particles
Cylinder 1.24
Cylinder 1.32 with Radii Less Than About 10 pm
Two spheres touching 1.10
Two spheres touching 1.17 Very small particles and gases are also deposited
Three spheres touching on surfaces as a result of turbulent diffusion and
As triangle 1.20 Brownian motion. Chemical absorption, impaction,
In line 3 1.34 photosynthesis, and other biological, chemical, and
In line 3 1.40 physical processes cause the material to be retained at
Four spheres touching
In line 4 1.58 the surface. In this case a deposition velocity (vd) can
In line 4 1.56 be defined as an empirical function of the observed
deposition rate (0) and concentration near the
From A. C. Chamberlain, The Movement of surface (Co):
Particles in Plmt Communities, in Vegetation and
the Atmosphere, Vol. 1, Chap. 5, p. 157, J. L. w
Monteith (Ed.), Academic Press, London, 1975. Vd = - (10.3)
t I n all cases preferential motion is perpen- co
dicular to long axes.
The height at which CO is measured is typically about
$Ratio of terminal velocity of equivalent
sphere to that of particle. 1 m. Once vd is known for a given set of conditions,
the formula o = v d C o can be used to predict dry
deposition of gases and small particles, where Co
forward ballistics approach that is based on the wind would be obtained from some appropriate diffusion
speed and the gravitational settling speed. model.
For a vt of less than 100 cm/sec (radius Many methods of incorporating dry deposition
r? l o o p m ) ; the particles are assumed to be dis- into existing effluent dispersion models are available
persed by turbulence in the same way as particles hav- [see Hosker (1980) for a survey]. Even the relatively
ing no inertia. A plume model is used, but h, = xvg/u is simple Gaussian plume model has been adjusted for
substituted for effective plume height (h) to account dry removal by at least four techniques. The most
for gravitational settling. This model, called the tilted common, and one of the easiest to use, is the
plume model, is illustrated in Fig. 10.2. Deposition scwxdled source depletion model, in which the
of particles with 5 cm/sec 5 vt 5 100 cm/sec at the apparent strength of the source is allowed to vary
ground at any position x, y is given by the expression with downwind distance to account for the diminish-
ing amount oi material remaining aloft. The rate of
change of Q with distance is
where w is the deposition rate in mass per unit area

per unit time. In a Gaussian model of this situation,
roughly half of the material is deposited by a distance
hu/vt, where the particle plume centerline strikes the
which leads to
Graphical solutions to this equation with the use of

Fig. 10.2 Tilted plume model. The plume axis drops standard Pasquill-Gifford curves for U, are given by
With speed vg. Van der Hoven (1968). The resulting Q(x) is then
REMOVAL MECHANISMS 69
sed instead of the usual fixed-source strength (Q) in 2Vd

o.(x)= 1 - (10.8)
le Gaussian plume equation (Eq. 4.1). Table 10.2 v t + vd t (uh - vtx) 0; (da,/dx)
ses Van der Hovens graphs to estimate the down-
iind distance at which 50% of the plume is depleted In this model dry deposition removes material from
or a wind speed (u) of 1.0 m/sec and a deposition the lower portions of the plume, which therefore
;peed(Vd) of 0.01 m/sec. Because of the exp ( h 2 / 2 4 ) begins to show a non-Gaussian vertical concentration
;erm in Eq. 10.5, the distance for 50% depletion in distribution. In unstable conditions, because of rapid
rable 10.2 is not always a continuously decreasing mixing within the plume, the results from Eqs. 10.5
function of stability for any given source height. This and 10.6 are not very different; in stable conditions,
model implicitly assumes that depletion occurs over however, the flux to the surface estimated by using
the whole depth of the plume rather than a t the Eq. 10.6 will be significantly smaller than Eq. 10.5
surface; the plumes vertical profile is therefore would suggest, especially far downwind.
invariant with distance. Horst (1979) has compared the results of various
deposition models. He concluded that, close to the
source, the partial reflection model is the easiest to
Table 10.2 Distance in Kilometers at Which 50%
of Q Is Depleted by Dry Deposition as a use and is fairly accurate. Far downwind, however,
Function of Pasquills Stability Class the source depletion model performed better than the
and Source Height partial reflection model. He then suggested a modi-
fied version of the source depletion model which
Source heiht. m permits the plume vertical profile to vary with
Pasauills class 0 10 50 100 distance. Field data to validate any dry deposition
model are largely unavailable or inadequate (Sehmel,
AandB >10 1980). The model is selected mostly on the basis of
C 1.8 18 43 60 the physical plausibility of the models assumptions
D 0.4 3.5 8.6 19
E 0.15 2.2 8.3 17 and predictions.
F 0.10 2.0 10.0 28 A knowledge of the dry deposition velocity (vd)
is necessary to operate any deposition model. Both
*u = 1.0 m/sec; vd = 0.01 mtsec. measured and theoretical estimates of vd are com-
monly used (Sehmel, 1980). Some theoretical models
of depositioil assume the deposition process is anale
The partial reflection model, summarized by
gous to electrical current flow across resistances. In
Overcamp (1976), is a somewhat different approach. effect, the deposition speed (vd) is inversely propor-
In this instance the image term (involving z + h) in tional to the sum of an aerodynamic resistance to
Eq. 4.1 is preceded by a reflection coefficient (CY),
turbulent mass transfer (ra), a resistance to transfer
which is thus a fraction of the strength of the real
across the surface boundary layer (q,),and a re-
source. This coefficient is determined by setting the
sistance to transfer into the surface (rs). In principle,
deposition flux equal to the difference in fluxes from then, deposition speed should be a function of
the real and image terms. The plume is also allowed
roughness length, friction velocity, and many other
to tilt to incorporate gravitational settling of large
surface parameters. Wind-tunnel data show clearly the
particles at terminal speed (vt): dependence on roughness length and friction velocity.
However, field measurements of deposition speeds
contain much scatter and d o not always agree with
theories or with measurements in controlled labora-
tory environments. The values of vd recommended in
this chapter are based on reviews of field data by
McMahon and Denison (1979) and Sehmel (1980).
Original references are given in their reviews. We
arbitrarily assume that the field measurements are
The reflection coefficient [ ~ ( x G ) ]can be computed more reliable than theoretical estimates.
by solving an implicit relation for XG Figure 10.3 shows measured deposition speeds for
particles to grass. For particles larger than about
1Pm, turbulent diffusion and gravitational settling
are the dominant processes, whereas, for particles
smaller than 0.1 jm, Brownian diffusion becomes
and the following equation for CY(X): increasingly important. Theory and laboratory data
10
0 Laboratory measurements 10
0 Field measurements
Y 1oc
5
>- \- .
I
0
t E 1
\ -
s -P
0
. A
.--
W
> 10-1 0 t
z
0
\ -
I-
0
0
k
v) 0
i>i 10-1
P
W z
0 0
10-2 Terminal
settling
t
v)
velocity 2
w
n
10-2
10-2
10-2 lo- 100 10
PARTICLE DIAMETER, pm
Fig. 10.3 Laboratory and field measurements of

1o -~
deposition speeds of particles to grass. [From T.A.
McMahon and P. J. Denison, Empirical Atmospheric lo- 1 10 102
Deposition Parameters-A Survey, Atmos. Enuiron., PARTICLE DIAMETER, pm
13: 1000 (1979); by permission of Pergamon Press,
Ltd.] Fig. 10.4 Predicted deposition velocities at 1 m for
u+ = 50 cm/sec and particle densities of 1,4, and 11.5
g/cm3. [From G. A. Sehmel, Particle and Gas Dry
indicate a minimum in vd for particles between 0.1 Deposition: A Review, Atmos. Enuiron., 14: 1002
and 1 p m in- diameter; the exact value for vd is (1980); by permission of Pergamon Press, Ltd.]
predicted to depend on such things as particle
density, friction velocity (u+), and surface roughness carbon monoxide, typically have nearly negligible
(zo). Figure 10.4 shows typical calculated values. deposition speeds, to cm/sec. If the gas is
However, the scatter and sparsity of the field data biologically or chemically active, however, the mea-
preclude a vigorous validation of the expected be- sured deposition speed is likely to be on the order of
havior of vd, although the general trends seem to be 0.5 to 3 cm/sec. For example, SO2 deposition speeds
verified. The inaccuracies of the data suggest that, in for various surface types are summarized in
any particular model calculation, a range (often Table 10.3 [McMahon and Denison (1979)l. The vd
orders of magnitude) of possible Vd values should be for SOz in forests is not significantly enhanced over
used rather than a single estimate, such as the average. that for grasses or crops. A dependence on wind
This will place upper and lower bounds on the speed cannot be discerned from the available data.
computed deposition results, which is really all the During the night or the winter, when stomata are
accuracy that can be expected with the present state closed, vd is about a factor of 2 to 5 less than during
of the art. A few special cases in which the deposition summer days.
rate is dramatically enhanced are known. For The deposition speed of other reactive gases is
example, the deposition of particles to a forest similar to that of S 0 2 . 12 is deposited at an average
canopy is about 2 to 16 times as great as that over speed (vd) of 1.5 cm/sec, and it has relatively little
open terrain, whereas mosses give rise to an order-of- scatter. Ozone is deposited at a speed (vd) ranging
magnitude increase in vd. Special caution is warranted from 0.02 to 1.4 cm/sec, with an average speed of 0.5
for unusual types of ground cover. cm/sec. The lowest ozone deposition speeds were
Many measurements of Vd for gases have been observed over water and snow (vd e 0.1 cm/sec).
reported by McMahon and Denison (1979) and Also, the ozone deposition speed for neutral condi-
Sehmel (1980). Biological and chemical activity play tions appears to be twice as large as that for unstable
key roles in gaseous deposition and may, in fact, conditions. Many other experimental factors in-
provide the rate-limiting steps for deposition and fluenced the results, such as reference height, rough-
uptake. For example, relatively inert gases, such as ness, surface conditions, etc., but the data d o not
Table 10.3 SO2 Deposition Rates vary with position to account for changes in precipi-
tation (and hence in scavenging rate) over the region
Surface V d , Cm/seC Comment of interest. The scavenging coefficient method cannot
Short grass 0.5 0.1 m in height be applied directly to aerosols with a range of
Medium crop 0.7 1.0 m in height diameters (polydisperse) unless an empirical value of
Calcareous soil 0.8 Wet or dry A is available for particles of that type and size. In
Acid soil 0.4 Dry particular, a scavenging coefficient cannot be applied
Acid soil 0.6 Wet
for an average particle whose size is equal to the
Dry snow 0.1 If wet, behaves like water
Water 0.7 geometric mean of the particle size range; the
Countryside I 0.8 scavenging rate will be too small by more than an
Cities 0.7 Based o n London data only order of magnitude. Methods of estimating scavenging
From T. A. McMahon and P. J. Denison, Empirical coefficients for polydisperse aerosols have been sum-
Atmospheric Deposition Parameters-A Survey, Atmos. marized by Dana and Hales (1976). The scavenging
Enuiron., 13: 575 (1979); by permission of Pergamon Press, coefficient method is also not applicable to gases
Ltd. which are not highly reactive or which are merely
soluble in water since it is essential in such cases to
permit any quantitative conclusions regarding these account for the possible desorption of gases from
factors. droplets as they fall from regions of high effluent
concentration (e.,.., an elevated plume) toward the
ground (Hales, 1972). Soluble gases are discussed in
10-3 WET DEPOSITION
reports by Hales et al. (1973) and Slinn (1974). Slinn
has found that close to the source the shape of a
The theoretical treatment of wet deposition is
Gaussian plume is unaltered by the wet removal
often divided into rainout (within cloud scavenging)
process but the concentration diminishes exponen-
and washout (below cloud scavenging). In practical
tially with distance. Far from the source, the plume is
applications the two processes are generally lumped
washed down or tilted from its initial height by
together since they can be modeled similarly [see the
an amount dependent on the precipitation rate,
reviews by Hosker (1980) or Slinn (198l)l.
droplet size, gas chemistry, wind speed, and distance
There are also. two methods of modeling this
downwind. The vertical dispersion parameter is also
problem. In the first the concentration (C) is assumed
enhanced by an amount strongly dependent on
to decrease exponentially with time:
droplet size.
Other methods of modeling wet removal use the
C(t) = C(O) e-*t (10.9) so-called washouts ratio (Wr). Let ko and Co be the
concentration of effluent in the precipitation (e.g.,
where A is the so-called scavenging coefficient raindrops) and in the air, respectively, at some
(time-) and t is time since precipitation began. The reference height; both quantities are measured in unit
precipitation-induced flux of effluent to the ground is of mass per volume. Then
given-by
Fwet = $tw AC dz (10.1Oa)

(10.1l a )
Other definitions are widely used (e.g., McMahon and

where Z, is the depth of the wetted plume layer. For Denison, 1979); let kb be the effluent concentration
rain falling completely through a Gaussian plume, in the precipitation in terms of mass (e.g., micrograms
per gram of HzO). In this case the washout ratio is
Fwet =
(2n)Auyu
exp (-2)
20;
(10.10b)
written as
(10.11b)
The method is, strictly speaking, applicable only to
particles of a single size (monodisperse) and to
highly reactive gases, which are irreversibly captured where pa is the density of air (-1.2 x g/cm3).
by the precipitation. Furthermore, the derivation of The dimensionless ratios are related by
Eq. 10.9 involves the assumption that the scavenging
coefficient is independent of both space and time. In
modeling practice, however, A is often allowed to
w =-w,
I Pa
Pw (10.11c)
where pw is the density of water (1 g/cm3). Thus Wr theoretical estimates of A must be used. The pro-
is almost 1000 times as large as W; since both cedure is most useful for single episodic events, where
definitions (Eqs. 10.11a and 10.11b) are common in the available theory does permit some adjustment for
the literature, the user must be careful. In the the peculiarities of individual storms.
following discussion, W, is used for convenience. There are many field observations of washout
The flux of effluent to the surface as a result of ratio (W,) which can be determined by measuring
the precipitation is just pollutant concentrations in air and in rainwater. The
washout ratio has been observed to decrease with
Fwet = ~OJO (10.12) precipitation amount during any given experiment,
presumably because the pollutant cloud becomes
where Jo is the equivalent rainfall rate in, for
more dilute. On the average, W, decreases by a factor
example, millimeters per hour. If W, is known and
of 2 for each order-of-magnitude increase in rainfall.
the concentration in air (C,) can be measured or
Over half of the washout ratios (W,) reported by
estimated from a plume model, then
McMahon and Denison (1979) are in the range Gom
(10.13) 3 x 10 to lo6 with a median of about 6 x 10.
There may be some variation with pollutant type,
The washout ratio can also be used to define a wet de- although the ranges in observed W, for different
position velocity by analogy to the dry: pollutants generally overlap each other and no statis-
tically significant differences can be proven. Washout
(10.14) ratios are probably best suited to long-term estimates,
in which the variability induced by single storm
This can then be used like Vd to develop models for events is integrated out.
the wet scavenging process. An analysis (summarized
by Hosker, 1980) of the microphysics of the wet
scavenging process yields an approximate relation 10-4 CHEMICAL REMOVAL
between the scavenging rate and the washout ratio:
Primary pollutants are those which are emitted
A x -W J o (10.15) directly into the atmosphere, such as SO?, CO, and
ZW NOZ. Secondary pollutants are those which are
where Zw is the depth of the wetted plume layer. created through chemical reactions involving the
Hence the washout ratio can also be used in ex- primary pollutants. For example, sulfates form when
ponential decay models of wet plume depletion SO2 is oxidized, or ozone is formed when a mixture
(Eq. 10.9). consisting of NO2, NO, and reactive hydrocarbons is
subjected to sunlight.
The scavenging coefficient (A) is theoretically a
Often an exponential chemical decay rate is
function of droplet size spectrum, physical and
assumed with time constant T,:
chemical characteristics of the particle or gas, and
precipitation rate. McMahon and Denison (1979)
report 20 field experiments in which A was measured (10.16)
for particles which gave A a median value of 1.5 x
sec- and a range from 0.4 x lo- sec- to
3x sec-. They found washout and rainout The conversion from SO2 to sulfate is often treated
coefficients to be nearly equal but did not find as an exponential process, but there is much debate
systematic differences in A as particie size or rain over the proper value for the time constant (T,). The
characteristics changed. Their median value of A for reaction rate depends on humidity and the presence
SO2 is about 2 x lo- sec-, and one laboratory of catalysts, and thus T, has been measured to range
experiment showed that A = 17 x lo- JOOS6, where from an hour to several days. A typical value of T,
Jo is rainfall (in millimeters per hour). A value of A used in long-range transport models is about 4 days.
ranging from lo- to sec- implies a half-life Chemical removal can also be studied by using the
for wet removal processes ranging from about 2 hr to kinetic equations. For example, assume that the
1 day. The use of scavenging coefficients for wet following two chemical kinetic equations are valid:
removal modeling is probably best regarded as an
kl
order-of-magnitude estimation procedure. This is A+B-D (10.17)
particularly true if empirical values of A are unavail-
k*
able or inappropriate for the conditions a t hand; thus D+E- A (10.18)
where k l and kz are rate constants (concentration-' If a steady state and no upwind background are
time-'). Then the rate of change of concentration of assumed, the continuity equation for this box be-
substance D due to chemical reactions is given by the comes
equation
AX Q, - U Z ~-
C V d c AX
Source Advec- Dry de-
tion position
loss loss
This term should be added to the right-hand side of - AC AX zi - (C/Tc) AX zi = 0 (10.20)

Precipita- Chemical loss
the continuity equation (8.2) for substance D. Equa-
tion loss
tion 10.19 is often used in the analysis of photo-
chemical smog. The set of chemical kinetic equations The solution is:
for photochemical smog is in a continual state of
development; a recent model (Falls and Seiiifeld,
1978) uses 55 equations. A rough schematic diagram
of smog formation is given in Fig. 10.5. Primary The numerator of Eq. 10.21 (Q, Ax/uzi) is tht:
emissions consist of hydrocarbons, NO, and some steady-state box-model solution in the absence of
NOz. Sunlight (hv, i.e., energy of a photon from removal processes. The terms in the denominator all
sunlight) results in the formation of free radicals and act to decrease the concentration. Realizing that
the ultimate production of the problem pollutants AX/U is the flushing time for the box, we can then
ozone ( 0 3 ) and peroxyacetylnitrate (PAN). think of all these correction terms as the ratio of the
flushing time to the characteristic removal time scale
for that process. The smaller the removal time scale,
10-5 REMOVAL PROCESSES IN the greater the ratio and the smaller the concentra-
THE BOX MODEL tion (C). Typical values of the parameters in this
equation are
Chapter 9, Sec. 9.2, presented the box model, in
which concentration (C) is assumed to be uniform vd = 0.01 m/sec A= sec-' Tc = lo4 sec
across a region of along-wind width (AX) and depth
(zi). Removal processes in this model can be ac-
AX =2x io4 m u = 5 m/sec zi = 500 m
counted for by using the models developed in These parameters might be used for urban SO2
previous sections: pollution on a rainy day. The denominator in the
equation for this set of parameters is thus 1.88, which
Dry deposition removal = -vdc Ax implies that the concentration is almost 50% less than
Precipitation scavenging = -AC Ax zi that given by the simple box model with no removal
processes.
Chemical removal = -(C/T,) AX zi
Problems
1. Particles with radii of 50 p n and densities of 5
g/crn3 are released at a rate of 1 g/sec in a
nonbuoyant plume from a stack 100 m high. What is
the deposition rate on the plume axis at a distance of
100 m from the stack for D stability and wind speed
of 5 m/sec?
2. For the situation in problem 1, what fraction
of the original effluent is still remaining airborne?
3. The area source strength of SO2 in an urban
region of width 30 km is 10 mg m-? sec-' . Mixing
depth is 1 km, and wind speed is 3 ni/sec. What
would the concentration be (a) in the absence of any
removal process, (b) assuming only dry deposition,
Fq.10.5 Simplified d i q a m of photochemical smog (c) assuming only wet deposition, (d) assuming only
reactions. hu is the energy of a photon (from sunlight), chemical removal, and (e) assuming that all three
and R is a free radical. removal mechanisms are acting?
Cooling Tower Plumes
and Drift Deposition
11-1 INTRODUCTION only if water vapor is condensed. The important
parameters are listed in Fig. 11.3, which is a sche-
More and more cooling towers are being con- matic drawing of a cooling tower plume. The in-
structed today to conserve water and prevent the fluence of latent heat on plume rise is shown by the
discharge of heated water to streams, lakes, and psychrometric chart of Fig. 11.4. The curved line is a
estuaries. Hot water from the industrial process plot of saturated specific humidity (qs) vs. tempera-
drips over wooden or plastic barriers in a cooling ture. Initial plume and environmental temperature
tower and evaporates into the air that passes through (T) and specific humidity (9) are indicated by points
the tower. As a result, about a 0 calories of heat are p and e, respectively. By simple mixing of plume and
lost for each gram of water evaporated. Cooling environmental air, the mixture T and q would follow
towers can be tall (-150 m tall and 30 m in radius) the straight line to any point 1. However, the plume
natural-draft towers (Fig. l l . l ) , in which vertical would then be supersaturated and would have to
motions are induced by density differences, or short condense water and warm itself to point 1'. At this
(-20 m tall and 5 m in radius) mechanical-draft point the amount of latent heat released is propor-
towers (Fig. 11.2), in which vertical motions are tional to the liquid water in the plume (qL), or the
forced by large fans. Vertical velocities of about 5 difference on the diagram between ql and qll (about
m/sec are observed in natural-draft towers and about 0.002 g/g in this case).
10 m/sec in mechanical-draft towers. Temperature
~
The ratio of latent heat released to initial latent

and moisture differences between the plume and the heat flux is
environment are about the same in both types of
towers, about 20C and 0.03 g/g, respectively. The
plume is saturated when it leaves the tower, and
liquid water concentrations are about 0.001 g/g.
Heat and moisture fluxes from cooling towers at where V and Vo are the current and initial volume
large power plants can cause fog or cloud formation fluxes, respectively. The ratio of volume fluxes
and at times can induce additional precipitation. (V/V,) can be approximated by (Tpo - Te)/
Another potential problem is drift deposition, in (TI# - T,,). Therefore, at point l',the fraction r is
which circulating cooling water with drop sizes
ranging from 50 to 10OO/.m1 is carried out of the
tower and may be deposited on nearby structures and
vegetation. These drops generally contain salts,
fungicides, and pesticides, which may harm the
surfaces they strike. A comprehensive review of Only 15%of the initial latent heat flux is released at
atmospheric effects of cooling tower plumes is given point 1. If this exercise were carried out a t the other
by Hanna (1981). points along the curved line between points e and p in
the figure, r would have a maximum of about 20%.
11-2 PLUME RISE FROM COOLING At a typical cooling tower with 80%latent heat and
TOWERS 20% sensible heat, the initial sensible heat flux (Fo)
would be no more than doubled owing to latent heat
H
About 80% of the total energy leaving a cooling release. Since plume rise is proportional to Fo, the
tower is latent heat, which can influence plume rise release of latent heat would increase plume rise by no
74
COOLING TOWER PLUMES AND DRIFT DEPOSITION 75
fig. 11.1 Typical natural-draft cooling towers. (Photograph courtesy of M. Kramer, Meteorological
Evaluation Services, Inc.)
Fig. 11.2 Typical mechanicaldraftcooling tower. (Photograph courtesy of J. Holmberg, The Marley
coding Tower Company.)
Fig. 11.3 Parameters important for coding plume analysis.
0.05 more than about 30%. Since this increase is probably

I 1 1 I I I/
I
within the errors of measurement, the formulas for
dry plume rise in Chap. 2 are recommended for
calculating cooling tower plume rise. At any rate the
dry formulas would always predict less plume rise and
therefore would be conservative. As shown by
experience, the bent-over plume-rise formula should
be used for wind speeds greater than 1m/sec, and the
calm plume-rise formula should be used for wind
speeds less than 1 m/sec.
11-2.1 Visible Plume Dimensions
The plume trajectory near the source can be calcu-

/1 d A f t e r condensation
)
lated by using the formulas in Chap. 2, Sec. 2-3. The
initial flux of moisture [Vo(q,o + qLo)] and the
variation with height of the saturation deficit flux
[V(q, - qe)] must be known to predict the dimen-
sions of the visible plume. The term qLo is the initial
specific humidity of liquid water in the plume and is
O 0 O L I I I I I I
often assumed to equal 0.001 g/g. The parameter qs is
-20 -10 0 10 20 30 40 50
r, c the saturation specific humidity of the environment.
The plume will be condensed when the following
condition is satisfied.
Fig.11.4 Psychrometric chart. Curved line is the
variation of saturation specific humidity with tem-
perature. As an illustration, equal parts of plume (p) Vo(qp0 + 9LO)
and environment (e) air are assumed to mix. >, V(q, - qe) (condensation) (11.3)
Chapter 2, Sec. 22.1, shows that the effective radius Entrainment assump tions :
(R,) of the momentum plume is about 1.5 times the
effective radius (R,) of the thermal plume. Nleyer et aR/az = . . . (11.9)
al. (1974) have found that observed visible plume
lengths are best simulated if the effective moisture Equation of motion (w, f possibly u):
plume radius (R,) is assumed to be about 0.71 times
the effecrive thermal plume radius. Thus the m e
_-
a w - buoyancy'force - entrainment
mentum flux ratio (V/Vo) at height z can be esti- aZ
mated by the formulas (Hanna, 1976): - drag due to water drops (11.10)
-=
VO
[ito.28- io
- (:TI2 (windy) (11.4)
aTP
--
First law of thermodynamics:
- dry adiabatic change + latent heat

-=
TO
iJ
(1 t-0.11- (calm) (11.5)
aZ
release - entrainment (11.11)
For calm plumes, Eqs. 11.3 and 11.5 can be Equation for saturated qps as a function of
combined to give the height (q)of the visible plume: temperature and pressure:
Clausius - Clapeyron or any one of a number

of empirical equations (11.12)
For bent-over plumes, E q s 11.3 and 11.4 and the Equation for variation in water vapor:
equation z = 1.6F&-'xKcan be used to predict the
height (2)) and length (xl) of the visible plume : --
%P - - entrainment (if unsaturated)
aZ
- 1'
(windy) (11.7)
---a ' ~(if ~saturate$)
aZ
(11.13)
Equation for cloud water qc change:
-1 r aqc - condensation
-- - conversion and
aZ
coalescence to rainwater - entrainment (11.14)
(windy) (11.8)
Since ambient variables, such as qe, qs, and u, are Equation for rainwater qh change:
assumed to be constant with height in this method,
use of the method is limited to visible plume heights %h
-- - conversion and coalescence from
less than about 100 m aZ
cloud water - rainout - entrainment (11.15)
11-2.2 Numerical Approach for
Deep Visible Plumes The cloud-physics parameterizations of Kessler
(1969) can be used in the last two equations for
When the variables qe, q,, and u change with cloud water and rainwater.
height, models that use differential equations must be An interesting problem that can be studied with
used to follow the plume. Visible plume dimensions, the above model is the question of whether dry
cloud-water concentration, cloud formation, and rain- cooling towers are more likely to cause cloud
o u t can be predicted. Several models are available to formation than wet cooling towers. In a dry tower,
the interested reader (e.g, Wigley and Slawson, 1971; heat exchange is accomplished in a manner similar to
Hanna, 1976, 1981; We& 1974; Koenig, Murray, and that in an automobile radiator. Koenig, Murray, and
Tag, 1978). In general, these models are a marriage of Tag (1978) found that the plumes and clouds from
plume-rise and cloud-growth models and include the dry towers were much larger than those from wet
following equations, expressed in words. towers for the same total (latent t sensible) heat flux.
The reason for the difference is that the latent heat diameters [D (cm)] can be estimated by using the
from the wet towers is not all available, whereas the following equations (after Engelmann, 1968):
sensible heat from the dry towers is immediately
available. The model by Koenig (1979) has also been vt = 3.02 x lo5 D2 (D < 0.0093 cm)
used to simulate snowfall from wet cooling tower
plumes, a phenomenon that has been observed on = 6816 x D'. '" (0.0093 < D < 0.068 cm)
isolated occasions (e.g., Otts, 1976).
= 2155 x Do*746 (0.068 < D < 0.26 cm)
11-3 DRIFT DEPOSITION = 1077 x Do-224 (0.26 cm < D) (11.17)
Water deposition calculations are more difficult

Drift is circulating water drops that escape from
than particle deposition calculations because the
the cooling tower. Since the circulating water con-
water drops will shrink owing to evaporation if
tains impurities, drift should be minimized. Highly
relative humidity is less than 100%. A drop may
efficient drift eliminators, such as sinusoidal baffles,
evaporate completely before it hits the ground, and
which deflect and capture drift drops, can reduce
the very small particle remaining will not have
drift loss to a small fraction (<0.002%) of the
sufficient settling speed to fall through the turbu-
circulating water rate. As a result, environmental
lence. Thus, for relative humidities less than loo%,
effects of drift deposition on ground surfaces and
the change in mass (m) of the drop due to evapora-
vegetation are observed to be minimal except in the
tion must be accounted for (Mason, 1971):
immediate vicinity (<200 m) of mechanical-draft
cooling towers.
Drift deposition is calculated in much the same (11.18)
way as the dry deposition of particles because of
gravitational settling (see Chap. 10, Sec. 10-2.1). For where 6 = diffusion coefficient of water vapor (-0.24
drops with radii less than about 100 ,urn, the plume of cm2/sec)
drops is assumed to disperse in a Gaussian manner but qd = specific humidity a t drop surface
the axis of the drop plume is assumed to settle with p = air density, g/cm3
terminal speed (vt) with respect to the gaseous v = kinematic viscosity of air (-0.15 cm2/sec)
plume. The drift deposition flux is given by Re = Dvt/u (droplet Reynolds number)
Sh = [2 f 0.552 Re%/(6/u)Ih](Sherwood number)
The difference (qd - qe) can also be estimated by
using a formula &om Mason (1971):
where C(x,y,O) is the ground concentration of d r o p
lets. Generally, this calculation is made for drops in a
specified size range. A typical drift dropsize distribu-
tion is given in Table 11.1, which indicates that
84.5% of the drift mass is in drops with diameters less
than 200 p r n A plume calculation would be made for where R, = 8.32 x lo7 ergs mole-' O K - '
each of the nine dropsize classes with diameters less RH = relative humidity (on scale of 0 to 1)
than 200 pm. Terminal speed [vt (cm/sec)] for m, = mass of solute, g
Table 11.1 Drop-Size Distribution at Mouth of Chalk Point Tower"
/I
~~~ ~
Drop diameter Percent of Drop diameter Percent of Drop diameter Percent of Drop diameter Percent of
range,.um totalmass range,pm total mass range, wn total mass range, r m total maw
~~ ~~
10-30 13.8 150-180 350-400 0.9 800:900 0.8

30-50 28.8 180-2 10 400-450 0.7 900-1000 0.7
50-70 13.4 210-240 450-500 0.6 1000-1100 0.6
70-90 8.9 240-270 1.7 500-600 1.0 1100-1200 0.4
90-110 6.2 270-300 1.3 600-700 1.0 1200-1300 0.1
110-130 4.2 300-350 1.4 700-800 0.9 1300-1400 0.1
130-150 3.4
*From Environmental System Corporation, Cooling Tower Drift Dye Tracer Experiment, PPSP-CPCTP,1977.
M, = molecular weight of solute, g/mole -rw-

Mo = molecular weight of water, g/mole
'
i = van't Hoffs factor (-2 for sea salt)
Nu = 2 + 0.552 Re (v/kt)'
-c---
kt = thermal diffusivity of air = 0.24 cm2
sec-' at 0C
In these formulas the drop temperature is assumed to

equal the ambient wet-bulb temperature. Probably the
most satisfactory way to solve the evaporation p r o b
lem is by means of a numerical solution in time,where
a new drop size and settling speed are calculated from
Fq.11.5 Outline of a cooling tower vapor plume and
the above equations at each time step. a drift drop plume (for drops in a narrow size range).
For large drops with diameters greater than By pointA, all drift drops in this size range have
200 pm, deposition is calculated by a ballistic trajec- dropped out of the plume. In any intervd Ax, a
tory. Owing to these larger drops, maximum drift fraction [vg (Ax/u)/2R] of the drops breaks away.
deposition is usually observed at distances less than 1
or 2 km from the cooling tower. The smaller drops
each drop size and breakaway point, there is a unique
carry a larger fraction of the drift-water mass, but distance (x) from the cooling tower at which the drop
their deposition is spread over a much larger area. is calculated to strike the ground. Thus the drops in a
Trajectory calculations are generally made for about size range leaving the plume in the distance range x to
10 to 20 drop sizes, which represent the dropsize x + Axi will strike the ground between distances xa
class boundaries or midpoints from such data as those and Xb. They are spread over a crosswind distance
in Table 11.1. equal to Ox, where O is the plume angle (shown in
Drops are assumed to be initially distributed Fig. 11.6), which can be estimated from observed De
uniformly over the plume cross section. The plume
centerline (zC) is calculated as shown in Chap. 2,
Sec. 2-3, and the.bottom of the vapor plume (zb) is
given by the relation
Zb = ZC - R, (1 1.20)
TOWER
The vapor plume radius is found from
R, = Ro i.0.282, (11.21)
Fq.11.6 Sector m d k h drops in a given size range,
Consider drift drops of diameter D settling with speed
breaking away from &e plume at a certain distance,
vg in the plume. Since the interior of the plume is w i l l be deposited on the ground
probably saturated, evaporation need not be con-
sidered. A scheme' for calculating the fraction fi of
or from Pasquill-Gifford uy values. A typical value
drops that "break away" or settle out of the plume in
of O is 10" to 20". Then the drift deposition flux
distance interval Ax is shown in Fig. 11.5. The
( q j ) for the mass (Mj) of drops in that size range
fraction f i is estimated by using the equation
leaving the plume in the interval (Axi) is given by
(11-22) Mj fi
wij = e(.; - .,') (11.23)
which is the ratio of the distance the drops fall in the

interval Axi to the plume diameter at that distance. The total drift deposition pattern is obtained by
At each downwind distance interval Axi, the summing Eq. 11.23 over all dropsize classes and all
fraction fi of drops that break away will be different breakaway intervals.
for each of the ten or twenty drop-size classes. After Policastro, Dunn, and Breig (1978) have com-
breakaway, the drops may evaporate in the ambient pared the predictions of several drift deposition
air. Equations 11.18 and 11.19 are used to estimate models against some limited observations from Chalk
drop evaporation and changes in settling speed. For Point and have shown that there are six or seven
models that are roughly equal. Accuracies are 2. Environmental temperature is 25OC and relative
probably within a factor of 2 to 5. With all the humidity is 70%. Calculate the saturation deficit. A
possibilities for error, such as the breakaway and natural-draft cooling tower has a radius of 40 m and
evaporation calculations, it is easy to see why drift an initial temperature of 45C. The plume is
deposition models can never be highly accurate. saturated but contains no liquid water. What will be
the visible plume height in a calm environment?
Problems 3. A drift drop has a diameter of 50 pm. The
difference between the specific humidity at the drop
1. A single typical natural-draft cooling tower is surface and the environmental specific humidity is
located in a valley. Environmental temperature, wind 0.002 g/g. What is the rate of change of mass (m) of
speed, and relative humidity are 20C, 5 m/sec, and the drop as a result of evaporation?
SO%, respectively. What is the flux of water from this
tower (in g/sec)? Assuming that the tower effluent 4. Calculate the gravitational settling speed for
fills up the valley, which is 10 km wide and 1km water drops with diameters of 40,200, and 1400pm.
deep, what would be the flux of water from the Suppose these drops were released from a plume a t a
tower necessary to saturate the environment? What is height of 300 m in a wind speed of 5 m/sec. At what
the ratio of the actual tower flux to the flux distance from the release point would each drop
necessary to saturate the environment? strike the ground?
Air-Pollution Meteorology
in Complex Terrain
12-1 INTRODUCTION channeled up or down the valley, which reduces the
probability of plume impingement on the mountain-
In the past decade air-pollution meteorology in side.
complex terrain has emerged as a top issue. Part of Another favorable meteorological effect in com-
the reason for this is that many new power plants and plex terrain is the enhancement of turbulence due to
other industries are being built in the mountainous eddies that are set up by air passing over and around
western part of the United States. Furthermore, terrain obstacles. Panofsky. Egolf, and Lipschutz
many eastern industries are located near hills and (1978) found that, for a meteorological tower 500 m
ridges, and recent regulations require that pollutant downwind of a ridge, the standard deviation of
concentrations be calculated at the surface of these wind-direction fluctuations (00) was increased by a
terrain obstacles. Interest is also inspired by the factor of 2.5. Hanna (1980b) found that 0 0 was
requirement to calculate the impact of sources on increased by a factor of 1.6 during neutral conditions
distant national parks and forest preserves. when the wind direction was perpendicular to the
Complex terrain influences the trajectory and the valley (Fig. 12.1). The tower was located 2 to 10 km
diffusion of a plume. Does the plume strike a ridge, from the ridges to the northwest and southwest in
or is it deflected above the ridge? What methods
Fig. 12.1. More detailed observations of 00 were
should be used to estimate diffusion? Field data made at a network of eleven towers in the Geysers,
necessary to answer these questions are seriously Calif., geothermal area. The terrain consists of ran-
lacking. A few good field programs are now under way domly oriented 1000-m mountains and ridges. For
by the Environmental Protection Agency, the U. S. the eleven stations, Hanna ( 1 9 8 0 ~ )calculated the
Department of Energy, and the Electric Power median hourly 0 0 values from 5 days of observations
Research Institute. The techniques described in this at a height of 10 m. These 0 0 values are plotted
chapter provide guidance for estimating diffusion in against the hour of the day in Fig. 12.3, which shows
complex terrain. that nighttime 0 6 values are about 20" to 26" and
daytime values are about 30" to 35".Over flat terrain
12-2 METEOROLOGY nighttime 0 6 values are predicted to be only 5" or less
Even though it is possible that high-pollutant (see Tables 4.2 and 4.3). With extreme stabilities,
concentrations may occur in complex terrain (e.g., meandering may cause occasional high values of 00.
where plumes intercept hillsides), several physical However, in the Geysers, 0 0 is consistently high for
processes are acting that tend to lower concentra- all stable conditions, which leads us to the conclusion
tions. The first of these is the tendency of wind to that terrain obstacles cause an enhancement of 00.
favor the "grain" of the terrain; it rarely goes across This conclusion is further confirmed by the data in
it. Many field studies show the validity of this Fig. 12.4, in which 0 6 is plotted against wind speed
statement. Figure 12.1 i s a topographic map of the ' for nighttime runs at a given station. At low wind
Widows Creek Steam Plant area in northeastern speeds, corresponding to the largest static stability
Alabama, where the major terrain obstacle is a linear [(g/T) M / d z ] at night, 00 is a maximum as a result of
250-m terrain step. Figure 12.2 is a wind-rose diagram these terrain effects.
for meteorological stations in the valley and on the Of course, complex terrain also causes changes in
mountain (Hanna, 1980b). The wind on the moun- surface-layer wind speed and direction which ad-
tain blows with nearly the same frequency from all versely affect pollutant concentrations. Pollutants
directions, but the wind in the valley is strongly emitted near the ground into the nighttime drainage
81
.OGICAL
305 rn < Elev. < 457 rn

457 rn < Elev. < 610 rn
Fig.12.1 Topographic map of area within 1Okm of Widows Creek Steam Plant. Meteorological
station locations ar2 given.
layer over sloping terrain may follow the drainage

flow downhill toward population centers. The thick-
ness (h) of the drainage layer on a simple slope is
suggested by Briggs (1979) to be given by the
formulas:
h = 0.05 x sin 0 (0> 20") (12.1)
h = 0.037 x 0% (0< 20") (12.2)
where x is the distance along the slope from the top

of the slope and 0is the slope angle in radians. For an
angle of 20" and a distance of 500 m, the thickness
(h) is about 9 m. Briggs (1975) used energy conserva-
tion principles to arrive at the following formula for
the characteristic wind speed in the drainage layer:
u = 2 . ~ ( s i n0) '(~x)' (12.3)

180
where H is the downward sensible heat flux (in units
of m2/sec3), which is given by
Fig.12.2 Annual (1977) wind rose for valley and -
plateau meteorological stations (61- level) at Widows H = -g WIT' (12.4)
Creek. Numbers are frequency per 22%"sector. T
AIR-POLLUTION METEOROLOGY IN COMPLEX TERRAIN 83
60
I 1 ~ 1 ~ 1 ~ 1 ~ 1 ~ 11 I ~ I ) 1 ) I ~ I ~ I
-
2 5 o L 7-
Fig. 12.3 Diurnal variation of ue at a height of 10 m for 11 meteorological towers in complex terrain
at the Geysers geothermal area. Hourly ue d u e s for 5 days were averaged together for each of 11
stations. The highest, median, and lowest of the 11 ug values at each hour are plotted.
m I . I I I I I I I a deeper drainage river.- An example of typical air

d
570 - - trajectories in a valley is shown in Fig. 12.5, which is
I
0
w.
also taken from the results of the Geysers experiment
(Nappo et al., 1980). This drainage flow has a depth
of 50 to 100 m and a speed of 2 to 3 m/sec and is
about 2 to 3 km downhill from the ridgetops. A few
kilometers farther down the valley, where it broadens
out and the slope decreases, the drainage pool is
observed to be about 2 0 0 m deep with a speed of
about 1m/sec.
Numerical or physical models can be used to
estimate wind flow over complex terrain. Most of the
work before 1975 has been reported by Egan (1975).
- I I I I I I I
Numerical models in use are all research-grade, and
O 0 I I
much developmental work needs to be done. The
0 1 2 3 4 5 6 7 8 9
WIND SPEED (u), m/sec biggest need is for field data, which would aid in the
development and testing of such models. In numerical
Fig. 12.4 Hourly average 00 vs. wind speed for models, it is difficult to obtain the necessary detail
nighttime (2100 to 0600) data from five observation. (small grid size) while still retaining a large enough
nights-at the Geysers geothermal area. This station is
located about halfway down a 1000-melevation ridge. domain size to cover the area of interest. For simple
two-dimensional hills, the potential flow theory
For typical values, H can be of the order of 0.001 appears to give good results.
m2/sec3 on a clear night. If /3 = 27 and x = 1km, Physical models of flow over complex terrain
then u = 1.8 m/sec, which is a value of wind speed have been reviewed by Hosker (1980), who points
often observed in drainage layers. Drainage from the out that the most obvious effects occur when the
ideal slopes considered in the above derivation usually flow is stratified. Blocking, lee waves, rotors, and
rises to converge into a narrower valley, which gives wake formation are some of the more interesting
phenomena that have been observed. The stability displaced slightly in a stable atmosphere. For FH < 1,
parameter most often used to classify this work is the motions tend to be limited to horizontal planes; i.e.,
internal Froude number: plumes will impact the surface and try to go around
FH
U (12.5)
hills rather than over them (Hunt, Snyder, and
Lawson, 1978). For FH > 1, the airflow is likely to
be over the top of the hill.
where U is the free-stream wind speed, H is the hill
height, and N is the Brunt-Vakiila frequency (in
rads/sec) :
12-3 DIFFUSION CALCULATIONS
(12.6)
As Pasquill (1974) suggests, observations of wind
where 8 is the potential temperature and N is the flow and turbulence parameters, if available, should
natural frequency of oscillation of an air parcel be used t o calculate diffusion. However, in most real
50
-50
E
i
2
t-
a
>
w
-I
w
-1 00
-1 50
-200
300 600 900 1200 IO
DOWNWIND DISTANCE, m
Fz.12.5 Composite view of nighttime neutral pilot balloon trajectories down Anderson Creek
canyon in the Geysers geothermal area. The solid line beneath all the trajectories is the underiying
terrain elevation. Presumably the top of the drainage flow is marked by the pojnt a t which the
trajectories pop u p o u t of the flow. (From C.J. Nappo, S. R. Hanna, and H. F. Snodgrass, Drainage
Wind Observations Using Neutral-Lift Balloons, in Second Joint Conference on Applications of Air
Pollutwn Meteorology, p. 496, American Meteorological Society, Boston, Mass., 1980.)
AIR-POLLUTION METEOROLOGY IN COMPLEX TERRAIN 85
,,-,,d
\ + h/2AND BRIGGS)
(EGAN
h,
Fig. 12.6 Illustration of plume height assumptions in Briggs (1973) (-) and Egan (1975) (- - -)
models for neutral and unstable conditions. The line ht + h is also shown.
PLUME AXIS
Fig. 12.7 Effect of upwind terrain on plume trajectories for gentle slopes and abrupt ridges.
situations we do not have the luxury of such detailed stable conditions (Pasquill-Gifford classes E-F; or
observations. In this case the empirical methods, F H < l), both modelers assume that the plume
which are based on models developed by Briggs maintains a constant elevation; thus the effective
(1973) and Egan (1975, 1979), described in the plume height (h) is reduced by terrain height (ht). If
following paragraphs can be used to estimate diffu- the terrain height is greater than the effective plume
sion from elevated point sources. height, the plume may impinge on it in E or F
First, will a plume impact on a terrain obstacle, or conditions.
will it ride up over the obstacle? If a terrain rise is If there is a terrain rise upwind of the source and
downwind of a source in neutral and unstable the average slope of the rise above the source exceeds
conditions (Pasquill-Gifford classes A-B-C-D; or 2%, downwash may be induced by the air flowing
FH > l), the plume tends t o ride up the slope while down over the terrain drop. For steep hills, it is
losing part of its effective stack height (h) relative to possible to get a cavity effect, i.e., a counter-
the ground. In this case Briggs (1973) suggests that h rotating eddy. The cavity would extend three to ten
should be reduced by the terrain height (h,) or h/2, hill heights downwind, which would possibly cause
whichever is the smallest reduction. Terrain height downwash and fumigation of a plume in this region.
(ht) is measured from the base of the source stack. Figure 12.7 illustrates these effects. For more precise
This concept is illustrated in Fig. 12.6. The hi2 estimates of concentration distributions in this case,
correction for high terrain is based on potential flow wind-tunnel or waterchannel modeling of the situa-
theory and wind-tunnel experiments, as mentioned in tion should be done since the trajectories depend
Sec. 12-2. Egans (1975) model is the same as that of heavily on details of the topography.
Briggs except that Egan suggests a reduction of ht/2 Chapter 4 shows that an image source can be used
rather than ht for terrain heights less than half the in the Gaussian plume model to account for reflec-
effective plume height. Thus the assumption by Egan tion from the ground surface. However, Egan et al.
would give slightly lower ground-level concentrations (1979) point out that this assumption leads t o an
at the surface of small hills than that by Briggs. In abrupt factor of 2 increase in axial plume concentra-
.
h Model with
reflection
Model wtihout/
reflect ion
ABRUPT TERRAIN PLUME

Fig. 12.9 Schematic view of an elevated plume mixed
RISE IMPACTION
X - c across a alley of width W during stable conditions.
Fig.12.8 Axial concentration variation with down-
wind distance as an abrupt terrain rise is approached.
- .
, Gaussian model with reflection. - -, Gaussian C=- Q (12.8)
model without reflection. - - -, recommended curve. uhW
tion from plumes impinging on steep terrain in stable where all variables are nighttime values.
(E-F; FH < 1) conditions. Therefore his model has
the requirement that axial plume concentration can
never increase with downwind distance (see Problems
Fig. 12.8). For this model to be applied, the solid
l . A highway follows a constant elevation con-
curve in the figure must first be calculated so that the tour along the side and very near the top of a
minimum concentrations (Cmin) at point xmin can
two-dimensional slope with an angle of 10. Carbon
be determined. From point xmin to the point at
monoxide emissions are 0.1 g sec- m-. Assume
which the plume strikes the terrain, the axial concen-
night conditions with a drainage wind of 1 m/sec. If
tration is assumed to equal Cmin.
the carbon monoxide mixes uniformly within the
The analysis of 06 observations in complex
drainage layer, what is the carbon monoxide concen-
terrain at the Geysers geothermal site described in
tration a distance of 200 m, 500 m, and 1000 m
Sec. 12-2 showed that 00 was close to what would be
down the slope from the highway?
expected over flat terrain during the day; however,
2. Consider an isothermal atmosphere with a wind
owing to terrain effects, it was enhanced during the
speed of 5 m/sec flowing over a hill of height 200 m.
night. In fact, nighttime 06 does not fall much below
Calculate the internal Froude number. What will
the neutral value. This result is consistent with the
happen to a plume heading toward the hill?
few limited diffusion experiments that have been
3. With the use of Briggss criteria, fill in the
conducted in complex terrain and supports the
following table. Assume neutral conditions.
assumption by Egan et al. (1979) that stable classes E
and F should be shifted to neutral. class D when Plume rise Predicted
selecting uy and uZ. Knowledge of the site (e.g., above level Hill height of
valley width) should be used to modify oy if terrain height plume above
necessary. During neutral and unstable conditions, (4,m (ht), m hill
flat-terrain oy and oz curves can be used. 200 10
Diffusion in valleys is limited when the valley
width (W) equals roughly 2uy. At night elevated 10 9.9
plumes could fill up the valley horizontally with very 100 200
little vertical diffusion (see Fig. 12.9). The highest 500 1000
concentration experienced by the valley walls would
0 100
be given by
(12.7) 4. Assume that an elevated point-source plume

will impact a steep mesa during very stable condi-
In the morning, break-up fumigation brings the tions. The source strength is 10 g/sec and the wind
pollutant to the valley floor when the stable layer is speed is 4 m/sec. Calculate the centerline concentra-
eroded from below by the heating of the ground. The tion at the point the plume impacts the mesa a
average concentration in this case is distance of 5 km from the source.
Long-Range Transport and Diffusion
13-1 INTRODUCTION time, in this case, roughly 1 hr. In the diffusion

equation,
Long-range transport and diffusion calculations
ac ac ac a ac
-
have been important for the nuclear industry since
the 1940s because of the dangers of bomb tests.
-+U-+v-=-K
at ax ay ax ax
However, during the past decade the problems of a a-+c -K,
+-K
a ac
- (13.1)
long-range transport of sulfur and nitrogen com- ay y ay az aZ
pounds from power plants and smog from urban areas
have become increasingly important. Sulfur from the turbulence contribution appears on the right-hand
distant sources reacts with water in the atmosphere to side. As the average wind field changes from hour to
form acid rain, which can cause serious environmental hour, the effect is felt in the advection terms u aC/ax
damage in such remote areas as the Adirondack and v aC/ay on the left-hand side, which also
Mountains of New York and the lakes of Scandinavia. contribute to variations in concentration C. If we
These effects are difficult and expensive to monitor, assume an averaging period of 1 day rather than 1 hr
although a few large programs are under way by the for the wind speed, then hour-to-hour variations
Organization for Economic Cooperation and Develop- should be parameterized as turbulence on the right
ment in Europe and the U. s. Department of Energy, side of Eq. 13.1. Thus turbulent diffusivities K,, K,,
U. S. Environmental Protection Agency, and Electric and K, or dispersion parameters ux, uy, and u, are
increasing functions of averaging time.
Power Research Institute in the United States. Several
So-called puff models claim to handle both
recent review papers exist on this subject, including
small-scale diffusion and large-scale meander (see, for
those by Bass (1980) and Eliassen (1980).
Numerous models are available, and some of these example, Heffter, 1980; Bass et al., 1979; Johnson,
Wolf, and Mancuso, 1978; Wendell etal., 1976).
models are described in this chapter. However, as yet
there is no definitive model for long-range transport Figure 13.1 shows how a plume is represented by a
and diffusion. There are simply too many missing
pieces of information, such as uy or K, coefficients,
at these scales. Since the Pasquill-Gifford uy and uz
curves are defined only to downwind distances of
about 10 km, long range is defined in this chapter
as any downwind distance greater than about 10 km.
At these distances wind speed and direction shears
and time and space variation in wind velocity become
important.
13-2 MODELING CONCEPTS

RELEASE
POINT
In practice, winds are always averaged over a
certain period, such as an hour. Turbulent diffusion is Fig. 13.1 Simulation of a plume by a series of puffs
then defined as the diffusion contributed by turbu- released at equal time intervals. It is assumed that
lent eddies with time scales less than the averaging a a t.
87
series of puffs. Puff release frequency is adjusted so K = lo4 m2 sec-. It is found that uy is not too
that there is roughly the amount of puff overlap crucial for the calculation of pollutant concentrations
shown on the figure. A wind field is needed, which averaged over a day or longer since total diffusion is
I would be based on radiosonde stations (spacing, dominated by meander for large averaging times. For
-300 km), surface stations (spacing, -50 km), or a these large averaging times, it is sufficient to know
mesoscale research grid (spacing, -20 km) of meteo- the wind-direction frequency distribution in 22 y2 -
rological towers. In models such as Shermans (1978), degree sectors. Sheih (1977) and Draxler (1979a)
the continuity equation is used to adjust the wind specify K, profiles to handle vertical diffusion. At
field so that it is mass consistent. The mixing depth short distances this is important for elevated sources;
must also be known since all models confine the however, at long distances ( > l o 0 km) the plume can
pollutants to the mixed layer. Transport wind veloc- be assumed to fill up the mixed layer uniformly
ity seems to be defined differently by each model; (Johnson, Wolf, and Mancuso, 1978). Draxlers
e.g., it equals 0.75 times the 850-mb wind in the (1979a) K, assumptions are listed in Table 13.1.
model by Johnson, Wolf, and Mancuso (1978) and
equals the concentration-weighted wind in the model Table 13.1 Draxlers Suggested Relation
by Heffter (1980). The observed winds must be Between K, and Stability Class
interpolated into other regions, generally using a
l/(distance) weighting, although some models use an [Below 100 rn, K,(z) = (z/lOO m)K,(100 m)]
additional weighting factor involving wind direction.
This situation is depicted schematically in Fig. 13.2. Stability class K,, m 2/see
Diffusion of the puff is handled quite crudely. .4 160
The first tenuous assumption is that puff diffusion is B 100
similar to plume diffusion, which is known to be C 70
wrong theoretically (see Chap. 6) but is still useful in D 15
applied models. Nevertheless, there is no information E 5.0
F 1.5
available on the spread of puffs at long range in the G 0.13
mixed layer, and it is necessary to fall back on plume
0s. Wendell et al. (1976) have extrapolated the
Pasquill-Gifford curves for uY and u, (Fig. 4.4) to On the other hand; removal is quite important
great distances-a procedure not recommended by since the half-lives for dry deposition, wet deposition,
Pasquill or Gifford. The Heffter (1980) model uses and chemical transformations of chemicals, such as
the simple expression uy = 0.5t (uy in meters and t in SO2, are typically on the order of a few days. Here
seconds) for all situations. Fay and Rosenzweig again, however, crude assumptions, such as vd = 1
(1980) and Johnson, Wolf, and Mancuso (1978) cm/sec for so2 a.id vd = 0.1 cm/sec for sulfates, are
a m m e that the Fickian law, U; = 2Kt, is valid, with made by most investigators. The conversion of SO2
to sulfate is generally assumed to take place at a rate
/
of about l%/hr. Wet removal rate is assumed to be
proportional to rainfall rate raised to some power;
e.g., Smith and Hunt (1978) use h (fraction removed
per second) = RH, where R is rainfall rate (in
mm/hr). For a moderate rain of 4 mm/hr, wet
I
removal takes place at a rate of 72%/hr. A slightly
?
1 i P f different formula is used by Johnson, Wolf, and
Mancuso (1978), with h = 0.6 x ~ O - R.~ When it is
raining, wet removal is quite effective; however, over
t a long time period, wet and dry removal are equally
effective.
Recommendations for the parameters discussed

Fie. 13.2 Example of wind vectors observed at eight above are as follows:
surface stations, which must be used to estimate the
wind vector at pointP. Most models weight each
observation by l / d 2 . The models by Draxler (1979b) uy = 0.3 (uy in m and t in sec) (13.2)
and Heffter (1980) weight, in addition, each observa-
tion by 1-0.5 lsin 91. K, (use Table 13.1) (13.3)
LONG-RANGE TRANSPORT AND DIFFUSION 89
Vd (SO2) = 1.0 cm/sec (13.4)
vd = 0.1 cmisec (13.5)
SO, + SOT2 conversion rate = 0.01 hr-' (13.6) 0 sc 0
X = lou4 RY2 (x in fraction sec-' and

R in mm/hr) (13.7)
Wind velocity: Use observed profiles,

weighted by concentration distribution
with height. Interpolate between
stations using l / r 2 weighting (13.8)
@ - Rawinsonde
With so many uncertain adjustable parameters, the
matching of predicted and observed concentration
patterns becomes quite easy. In these applications, release site
however, the model is not truly validated since 0 25 50 75100
validation requires testing of the model with an
independent data set.
Fig. 13.3 The Savannah River experiment sampling
arc and meteorological data locations. [From R. R.
13-3 APPLICATION TO AN INERT Draxler, Modeling the Results of Two Recent Meso-
TRACER scale Dispersion Experiments, Atrnos. Enuiron., 13:
1526 (1979).]
Some of the problems encountered with long-
range transport and diffusion models for sulfur are until it intercepts the 1200 GMT sounding. The
due to uncertainties in dry deposition, wet deposi- relations between the magnitude and direction of the
tion, and chemical transformations. The testing of surface wind (Vs), and average layer wind (VA) are
'
models by using inert tracers, such as Kr, a gas that determined and are arbitrarily assumed to apply at all
is released routinely from nuclear installations, or stations on the network. The advantage of using
man-made inert tracers, such as SF.5 or heavy surface winds is that they report hourly, in contrast
methanes, is useful. For these tracers, the observed to rawinsondes, which report at 12-hr intervals. For
concentration patterns are due solely to winds and example, this analysis may show that VJIVAI = 0.5
turbulence. Draxler (1979b) describes an application and that the direction of V, is 20" to the left of the
of his model to the Savannah River Laboratory (SRL) direction of VA. Next, a surface wind at the segment
area; the locations of rawinsonde stations, surface starting point is calculated by using a d-2 distance
stations, towers, release site, and a 90-km sampling weighting and a direction weighting (as)given by
arc are shown in Fig. 13.3. In addition to routine
*'
releases of Kr, two tracers, SF.5 and heavy methane as= 1 - 0.5 lsin (13.10)
(' 3CH4), were released from a 62-m stack over a 4 h r
where & is the angle between the wind direction and
period.
the line from the station to the segment starting
The model simulates the pollutant plume by a
point. Only angles with magnitudes of 90" or less are
series of puffs, whose trajectories are computed by
considered. Finally, a transport wind is determined
hourly advection segments. First, the rawinsonde by using the calculated surface wind and the relations
profile from the station nearest the segment starting
developed at the rawinsonde station.
point is used to estimate the average layer wind (VA)
Vertical diffusion is calculated from the one-
weighted by the concentration:
dimensional diffusion equation by using a vertical
n diffusivity equal to 7 m2/sec at a height of 100 m.
- ci vi This is the value Draxler was assuming for neutral
v A-- - x i (13.9)
conditions before he prepared the revised Table 13.1.
Horizontal diffusion is given by u,,(m) = 0.5t (sec).
where n is the number of levels up to the mixing Concentration contributions from each puff that
height. Mixing height is determined by following the passes a receptor during the sampling period are
maximum surface temperature up dry adiabatically summed, and an average concentration is calculated.
Observed and predicted SF6 concentrations on

the 90-km sampling arc are plotted in Fig. 13.4.
Problems
Draxler has adjusted K, to a value of 30 m2/sec so
that the maximum predicted concentration agrees 1. Assume mixing depths zi of 500 m, 1000 m,
with the observation. Neutral stability conditions and 2000m. For each of these mixing depths,
(Pasquill-Gifford) prevailed during this experiment, estimate the distance downwind at which u, = 0.3 zi
but a value of K, equal to 7 m2/sec produced a peak for stability classes A, B, C, D, E, and F (use Briggss
concentration that was much too high. The figure u formulas).
shows that the predicted plume is displaced about 20
2. What is the difference between uy used by
km from the observed peak, which indicates a
Heffter (1980) and that used by Fay and Roserizweig
wind-direction error of about 10 to 15.This error is
(1980) at downwind distances of 5 km, 1 0 k m ,
not surprising since the input wind observations are
50 km, 100 km, 500 km, and 1000 km? Assume wind
given only to the nearest 10. However, the net result
speed is 5 m/sec.
is a poor correlation between observed and predicted
concentrations at fixed points. We can conclude that 3. Crudely estimate the fraction of SOz remaining
the model did fairly well in estimating the crosswind in an air mass after it has crossed the Atlantic Ocean
concentration distribution but erred slightly in esti- from New York to London. Assume typical wind
mating plume direction. speeds and rainfall frequencies. Explain all your
assumptions.
4.A segment starting point in a trajectory calcula-
.
50
tion is at x = 0 , y = 0. Ten stations with the following
c)
E
m
c: coordinates are reporting the following wind veloc-
z 20
ities:
W
~
U
v) X X Speed, Direction,
U
0 Station (east), km (north), km m/sec degrees
z 10
0 1 -3.0 -1.0 8.0 190
+
a
a 2 -5.0 2.0 10.0 180
5.0 3 -2.0 5.0 11.0 230
V 4 0.5 2.0 9.0 200
z 5 2.0 9.0 3.0 280
8 Model predictions with
w various wind aswmDtions 6 6.0 6.0 2.0 300
$ 2.0
-Observed ........Standard .----Pibal 7 6.0 2.0 5.0 330
--- Surface ---
[r
w 8 3.0 -1.0 6.0 320
9 Tower
< 9 1.0 -2.0 5.0 270
1.a 10 -4.0 4.0 290
2.0
-50 -40 -30 -20 -10 0 10 20 30 40 50 60 70
NORTH SOUTH
CROSSWIND DISTANCE, km
Calculate the interpolated speed a t the segment
Fig. 13.4 Observed and predicted SF, concentration starting point by using
on the 9Wkm sampling arc at Savannah River. [From 1. l / d z weighting.
R R. Draxler, Modeling the Results of Two Recent 2. l / d 2 and direction weighting.
Mesoscale Dispersion Experiments, Atmos. Enuiron.,
1 3 1528 (1979).]
Abramowirz, M., and I. A. Stegan (Eds.), 1964, -, 1974, Plume Rise from Multiple Sources, in
Handbook of Mathematical Functions, National Cooling Tower Environment-19?4, ERDA
Bureau of Standards Applied Math Ser. #55, Symposium Series, College Park, Md., Mar. 4-6,
GPO. 1974, Steven R. Hanna and Jerry Pel1 (Coordina-
Angeli, J. K., P. W. Allen, and E. A. Jessup, 1971, tors), pp. 161-179, CONF-740302,NTIS.
Mesoscale Relative Diffusion Estimates from -, 1975, Plume Rise Predictions, in Lectures on Air
Tetroon Flights, J. Appl. Meteorol., 10: 43-46. Pollution and Environmental Impact Analyses,
Barad, M. L. (Ed.), 1958, Project Prairie Grass: A W o r k s h o p F r o c e e d i n g s , B o s t o n , Mass.,
Field Program in Diffusion, Geophysical Research Sept. 29-Oct. 3, 1975, pp. 59-111, American
Papers, No. 59, Vols. I and II, Report AFCRC- Meteorological Society, Boston, Mass.
TR-58-235, Air Force Cambridge Research -, 1979, Analytic Modeling of Drainage Flows,
Center. ATDL Report 79/22, Atmospheric Turbulence
Barry, P. J., 1964, Estimation of Downwind Concen- and Diffusion Laboratory.
tration of Airborne Effluents Discharged in the -, 1981, Plume Rise and Buoyancy Effects, in
Neighborhood of Buildings, Report AECL-2043, Atmospheric Science and Power Production,
Atomic Energy of Canada, Ltd., Chalk River, Darryl Randerson (Ed.), DOE Report DOE/TIC-
0n tari 0. 27601, in press.
Bass, A., 1980; Modeling Long Range Transport and Businger, J. A., J. C. Wyngaard, Y. Izumi, and E. F.
Diffusion, in Proceedings of 2nd Joint Conference Bradley, 1971, Flux-Profile Relationships in the
on Applications of Air Pollution Meteorology, Atmosphere Surface Layer, J . Atmos. Sci., 28:
New Orleans, La., Mar. 24-27, 1980, American 181-189.
Meteorological Society, Boston, Mass. Byzova, N. L., Ye. K. Garger, and V. N. Ivanov, 1970,
-, C. W. Benkley, J. S. Scire, and C. S. Morns, Experimental Estimation of the Lagrangian Time
1979, Development of Mesoscale Air Quulity Scale of Turbulence, Izv. Atmos. Oceanic Phys.,
Simulation Models. Vol. I. Comparative Sensi- 6: 315-320.
Cagnetti, P., 1975, Downwind Concentrations of an
tivity Studies of Puff, Plume and Grid Models for Airborne Tracer Released in the Neighborhood of
Long-Distance Dispersion Modeling, Report EPA- a Building, Atmos. Environ., 9(8): 739-747.
600/7-79-XX, Environmental Protection Agency. Carpenter, S. B., F. W. Thomas, and F. E. Gartrell,
Batchelor, G. K., 1950, Application of the Similarity 1968, Full-Scale Study of Plume Rise at Large
Theory of Turbulence to Atmospheric Diffusion, Electric Generating Stations, Tennessee Valley
Q. J. R Meteorol. Soc., 76(328): 133-146. Authority, Muscle Shoals, Ala.
-, 1952, Diffusion in a Field of Homogeneous -, T. L. Montgomery, J. M. Leavitt, W. C.
Turbulence. 11. The Relative Motion of Particles, Colbaugh, and F.W. Thomas, 1971, Principal
Roc.cambridge Philos. Soc., 48: 345362. Plume Dispersion Models, TVA Power Plants, J.
Braham, R. B., B. K. Seely, and W. D. Crozier, 1952, Air Pollut. Control ASSOC.,21 : 491495.
A Technique for Tagging and Tracing Air Parcels, Carslaw, H. S., and J. C. Jaeger, 1959, Conduction of
Trans. Am. Geophys. Union, 33: 825833. ' Heat in Solids, 2nd ed., Oxford University Press,
Briggs, G. A., 1973, Diffuion Estimation for Small London.
Emissions, ATDL Contribution File NO. 79, Caughey, S. J., J. C. Wyngaard, and J. C. Kaimal,
Atmospheric Turbulence and Diffusion Labora- 1979, Turbulence in the Evolving Stable Layer, J.
tory. Atmos. S c i , 36: 1041-1052.
91
Chamberlain, A. C., 1975, The Movement of Particles 313-390, D. A. Haugen (Ed.), Science Press,
in Plant Communities, in Vegetation and the Princeton, N. J.
Atmosphere, Vol. 1, Chap. 5, pp. 155-203, J. L. Doran, J. C., T. W. Horst, and P. W. Nickola, 1978,
Monteith (Ed.), Academic Press, London. Variations in Measured Values of Lateral Diffu-
Chaudhry, F. H., and R. N. Meroney, 1973, Simi- sion Parameters, J. Appl. Meteorol., 17(6):
larity Theory of Diffusion and the Observed 825831.
Vertical Spread in the Diabatic Surface Layer, Draxler, R. R., 1976, Determination of Atmospheric
Boundary-Layer Meteorol., 3: 405-415. Diffusion Parameters, Atmos. Environ., 10:
Cramer, H. E., 1957, A Practical Method for Estimat- 99-105.
ing the Dispersal of Atmospheric Contaminants, -, 1979a, Estimating Vertical Diffusion from
in Proceedings of the First National Conference Routine Meteorological Tower Measurements,
on Applied Meteorology, Sec. C, pp. C-33-C-35, Atmos. Environ., 13: 1559-1564.
American Meteorological Society, Hartford, -, 1979b, Modeling the Results of Two Recent
Conn. Mesoscale Dispersion Experiments, Atmos.
Crane, G., H. A. Panofsky, and 0. Zeman, 1977, A Enuiron., 13: 1523-1533.
Model for Dispersion from Area Sources in Edinger, J. G., 1952, A Technique for Measuring the
Convective Turbulence, Atmos. Environ., 11: Detailed Structure of Atmospheric Flow, Geo-
893-900. physics Research Papers, No. 19, Air Force Cam-
Crawford, T. V., 1966, Predicting and Sampling bridge Research Laboratories, Geophysics Re-
Nuclear Clouds from the Viewpoint of Diffusion search Directorate, Cambridge, Mass.
Theory, USAEC Report UCRL-14983, University Egan, B. A., 1975, Turbulent Diffusion in Complex
of California, Lawrence Radiation Laboratory. Terrain, in Lectures on Air Pollution and Environ-
Crozier, W. D., and B. K. Seely, 1955, Concentration mental Impact Analyses, pp. 112-135,D. Haugen
Distributions in Aerosol Plumes Three t o Twenty- (Ed.), American Meteorological Society, Boston,
Two Miles from a Point Source, Trans. Am. Mass.
Geophys. Union, 36: 42-52. -, R. DErrico, and C. Vaudo, 1979, Estimating Air
Dana, M. T., and J. M. Hales, 1976, Statistical Quality Levels in Regions of High Terrain Under
Aspects of the Washout of Polydisperse Aerosols, Stable Atmospheric Conditions, Preprints, Fourth
Atmos. Environ., 10: 45-50. Symposium on Turbulence, Diffusion, and Air
Davies, R. W., 1959, Large-Scale Diffusion from an Pollution, Am. Meteorol. Society, Boston, Mass.
Oil Fire, in Advances in Geophysics, Vol. 6, pp. -, and J. Mahoney, 1972, Numerical Modeling of
413-414, F.N. Frenkiel and P. A. Sheppard Advection and Diffusion of Urban Area Source
(Eds.), Academic Press, New York. Pollutants, J. Appl. Meteorol., 11: 312-322.
Deardorff, J. W., 1970, A Three Dimensional Numeri- Ekman, V. W., 1902, On jordrotationens inverkan p i
cal Investigation of the Idealized Planetary vindstrommar i nafvet, Nytt. Mag. For Naturv.,
Boundary Layer, J. Geophys. Fluid Dyn., 1: 40: 37-63.
377410. Eliassen, A., 1980, A Review of Long-Range Trans-
-, 1974, A Three Dimensional Numerical Study of port Modeling, J. Appl. Meteorol., 19: 231-240.
the Height and Mean Structure of a Heated Engelmann, R. J-, 1968, The Calculation of Precipita-
Planetary Boundary Layer, Boundary-Layer tion Scavenging, in Meteorology and Atomic
Meteorol., 1 : 81-106. Energy-1 968, pp. 208-221, D. H. Slade (Ed.),
Demerjian, K. L., and K. L. Schere, 1979, Applica- USAEC Report TID-24190, U. S. Atomic Energy
tions of a Photochemical Box Model for 0 3 Air Commission, NTIS.
Quality in Houston, Texas, in Proceedings of Environmental Protection Agency, 1978, Guideline
Specialty Conference on OzonelOxidants: Inter- on Air Quality Models, OAQPS Guideline Series
actions with the Total Environment, Environ- No. 1.2-080, Report EPA-45012-78-027, Office
mental Protection Agency, Research Triangle of Air Quality Planning and Standards, Research
Park, N. C. Triangle Park, N. C.
Dickson, C. R., G. E. Start, and E. H. Markee, Jr., Environmental Systems Corporation, 1977, Cooling
1969, Aerodynamic Effects of the EBR-II Reac- Tower Drift Dye Tracer Experiment, Report
tor Complex on Effluent Concentration, Nucl. PPSP-CPCTP, Environmental Systems Corpora-
&f., 10: 228-242. tion, Knoxville, Tenn.
Donaldson, C. DuP., 1973, Atmospheric Turbulence Eschenroeder, A. Q., J. R. Martinez, and R.A.
and the Dispersal of Atmospheric Pollutants, in Nordsieck, 1972, Evaluation of a Diffision Model
AMS Workshop on Micrometeorology, pp. f o r Photochemical Smog Simulation, Final Re-
REFERENCES 93
port, Contract No. 68-02-0336 by General Re- -,and S. R. Hanna, 1973, Modeling Urban Air
search Corporation, Santa Barbara, Calif., for Pollution, Atrnos. Enuiron., 7: 131-136.
Environmental Protection Agency. Golder, D., 1912, Relations Among Stability Parame-
Falls, A. H., and J. H. Seinfeld, 1978, Continued ters in the Surface Layer, Boundary-Layer
Development of a Kinetic Mechanism for Photo- .kIeteorol., 3: 47-58.
chemical Smog, Enuiron. Sci. Technol., 12: Halitsky, J., 1963, Gas Diffusion near Buildings,
1398-1405. ASHRAE Trans., 69: 464-485.
Fay, J.. A., M. Escudier, and D. P. Hoult, 1969, A Hanna, S. R., 1968, A Method of Estimating Vertical
Correlation of Field Observations of Plume Rise, Eddy Transport in the Planetary Boundary Layer
Fluid Mechanics Laboratory Publication No. Using Characteristics of the Vertical Velocity
69-4, Massachusetts Institute of Technology. Spectrum, J. A trnos. Scz., 25: 1026.
and 5. J. Rosenzweig, 1980, An Andytical
1971, A Simple Method of Calculating Disper-
-, -:
Diffusion Model for Long Distance Transport of sion from Urban -4rea Sources, J. ilir PoZZut.
Air Pollutants, Atmos. Environ., 14: 355-366. Control ASSOC., 21 : 774-777.
-, 1973, Description of ATDL Computer Model
Frenkiel, F. N., and I. Katz, 1956, Studies of for Dispersion from Multiple Sources, in Indus-
Small-Scale Turbulent Diffusion in the Atmo- trial Air Pollution Control, pp. 23-32, Ann Arbor
sphere, J. Meteorol., 13: 388-394. Science Publishers, Ann Arbor, Mich.
Gifford, F. A., Jr., 1955, A Simultaneous -, 1974, Meteorological Effects of the Mechanical
Lagrangian-Eulerian Turbulence Experiment, Draft Cooling Towers of the Oak Ridge Gaseous
Monthly Weather Rev., 83: 293-301. Diffusion Plant, in Cooling Tower Enuiron-
-, 1959a, Statistical Properties of a Fluctuating ment-Z974, ERDA Symposium Series, College
Plume Dispersion Model, in Advances in Geo- Park, Md., Mar. 4-6, 1974, Steven R. Hanna and
physics, Vol. 6 , pp. 117-138, F. N. Frenkiel and Jerry Pel1 (Coordinators), pp. 291-306, CONF-
P. A. Sheppard (Eds.), Academic Press, New 740302, NTIS.
York. -, 1975, Relative Diffusion of Tetroon Pairs Dur-
-, 195933, Computation of Pollution from Several ing Convective Conditions, paper presented at
Sources, Int. J. Air Water Pollut., 2 : 109-110. First Conference on Regional and Mesoscale
-, 1 9 6 1 , Use- of Routine Meteorological Modeling, Analysis and -Prediction, Las Vegas,
Observations for Estimating Atmospheric Disper- Nev., American Meteorological Society, Boston,
sion, Nucl. Saf., 2(4): 47-57. Mass.
-, 1968, An Outline of Theories of Diffusion in the -, 1976, Predicted and Observed Cooling Tower
Lower Layers of the Atmosphere, in Meteorology Plume Rise and Visible Plume Length at the
and Atomic Energy--1968, pp. 66-116, D. H. John E. Amos Power Plant, Atrnos. Enuiron., 10:
Slade (Ed.), USAEC Report TID-24190, U. S. 1043-1052.
Atomic Energy Commission, NTIS. -, 1978, Diurnal Variations of the Stability Factor
-, 1970, Atmospheric Diffusion in an Urban Area, in the Simple ATDL Urban Dispersion Model, J .
in- Proceedings of 2nd Congress o f the. Interna- Air Pollut. Control ASSOC.,28: 147-150.
tional Radiation Protection Association, -, 1980a, Users Guide for ATCOOL Cooling
Brighton, England, May 3-8, 1970, W. Englund Tower Plume Model, Report ERLTM-ARL-80,
(Ed.), Air Pollution Control Association. Environmental Science Services Administration,
-, 1975, Atmospheric Dispersion Models for Air Resources Laboratory.
Environmental Pollution Applications, in Lecy -, 1980b, Measured u,, and 06 in Complex Terrain
tures on Air Pollution and Environment Impact near the TVA Widows Creek, Alabama, Steam
Analyses, Boston, Mass., Sepf. 29-Oct. 3, 1975, Plant, Atmos. Environ., 14: 401.
pp. 35-58, American Meteorological Society, -, 1980c, Diurnal Variation of Horizontal Wind
Boston, Mass. Direction Fluctuations 06 in Complex Terrain at
-, 1976, Turbulent Diffusion Typing Schemes-A Geysers, Calif., Boundary-Layer Meteorol., 58:
Review, NUCLSa$, 17: 68-86. 1305-1309.
-, 1977, Tropospheric Relative Diffusion Observa- -, 1981, Atmospheric Effects of Energy Produc-
tion, J. Appl. Meteorol., 16: 311-313. tion, in Atmospheric Science and Power Produc-
-, 1981, Horizontal Diffusion in the Atmosphere: tion, Darryl Randerson (Ed.), DOE Report
A Lagrangian-Dynamical Theory, DOE Report DOE/TIC-27601, in press.
LA-8667-MS, Los Alamos Scientific Laboratory, -,and F. A. Gifford, 1977, Application of the
NTIS. ATDL Sample Urban Dispersion Model to Frank-
furt, West Germany, in Proceedings of Seventh of the International Conference on Air Pollutants
NA TOICCMS Meeting, Louvaine le Neuf, and Their Effects on the Terrestrial Ecosystem,
Belgium, ATDL Contribution No. 77/17, Atmo- Banff, Alberta, Canada, May 10-17, 1980, S. V.
spheric Turbulence and Diffusion Laboratory. Krupa and A. H. Legge (Eds.), John Wiley &
-, G. A. Briggs, J. Deardorff, B. A. Egan, F. A. Sons, New York.
Gifford, and F. Pasquill, 1977, Summary of Rec- -, 1981, Flow and Diffusion Near Obstacles, in
ommendations made by the AMs Workshop on Atmospheric Science and Power Production,
Stability Classification Schemes and Sigma Darryl Randerson, (Ed.), DOE Report DOE/TIC-
Curves, Bull. Am. Meteorol. SOC.,58: 1305-1309. 27601, in press.
Haugen, D. A. (Ed.), 1959, Project Prairie Grass: A Huber, A. H., and W. H. Snyder, 1976, Building Wake
Field Program in Diffusion, Geophysical Research Effects on Short Stack Effluents, in preprints of
Papers, No. 59, Vol. 111, Report AFCRC-TR- Third Symposium on Atmospheric Turbulence,
58-2.35, Air Force Cambridge Research Center.
Diffusion, and Air Quality, Raleigh, N. C.,
Hay, J. S., and F. Pasquill, 1959, Diffusion from a Oct. 19-22, 1976, pp- 235-242, American
Continuous Source in Relation to the Spectrum Meteorological Society, Boston, Mass.
and Scale of Turbulence, in Advances in Geo-
Hunt, J. C. R., W. H. Snyder, and R. E. Lawson, Jr.,
physics, Vol. 6, pp. 345-365, F. N. Frenkiel and
1978, Flow Structure and Turbulent Diffusion
P. A. Sheppard (Eds.), Academic Press, New
Around a Three-Dimensional Hill, Fluid Modeling
York.
Study on Effects of Stratification. Part I. Flow
Hayes, S. R., 1979, Performance Measures and Structure, DOE Report EPA-600/4-78-041,
Standard for Air Qunlity Simulation Models, Environmental Protection Agency.
Report EF78-93R2, Environmental Protection
Agency. Irwin, J. S., 1979a, Estimating Plume Dispersion-A
Recommended Generalized Scheme, in preprints
Heffter, J. L., 1965, The Variation of Horizontal
of Fourth Symposium on Turbulence, Diffusion,
Diffusion Parameters with Time for Travel
Periods of One Hour or Longer, J. Appl. and Air Pollution, Reno, Nev., Jan. 15-18, 1979,
pp. 62-69, American Meteorological Society,
MeteoroL, 4(1): 153-156.
Boston, Mass.
-, 1980, Air Resources Laboratories Atmospheric
Transport and Dispersion Model (ARL-A TAD), -, 1979b, A Theoretical Variation of the Wind
Report ERLTM-ARL-81, National Oceanic and Profile Power Law Exponent as a Function of
Atmospheric Administration, Air Resources S u r f a c e Roughness, and Stability, Atmos.
Laboratory. Enuiron., 13: 191-194.
Hogstrom, U., 1964, An Experimental Study on Islitzer, N. F., and R. K. Dumbauld, 1963, Atmo-
Atmospheric Diffusion, Tellus, 16(2): 205-251. spheric Diffusion-Deposition Studies over Flat
Terrain, Int. J. Air Water Pollut., 7(11-12):
Holzworth, G., 1972, Mixing Depths, Wind Speed,
and Potential for Urban Air Pollution Throughout
999-1022.
the Contiguous United States, Environmental Johnson, W. B., E. Shelar, R. E. Ruff, H. B. Singh,
Protection Agency Publication NO. AP-101. and L. Salas, 1975, Gas Tracer Study of Roof-
Horst, T. W., 1977, A Surface Depletion Model for Vent Effluent DiffLlsion at Millstone Nuclear
Deposition from a Gaussian Plume, Atmos. Power Station, Atomic Industrial Forum Report
Environ., 11: 41-46. AIF/NESP-O07b, Stanford Research Institute,
Palo Alto, Calif.
-, 1979, Lagrangian Similarity Modeling of Verti-
cal Diffusion from a Ground Level Source, J. -, R. C. SMarew, and D. B. Turner, 1976, Urban
AppL MeteoroL, 18: 733-740. Air Quality Simulation Modeling, in Air P o l b -
tion, 3rd ed., Vol. 1, Chap. 10, pp. 503-562, A. C.
Hosker, R. P., Jr., 1979, Empirical Estimation of
Stern (Ed.), Academic Press, New York.
Wake Cavity Size Behind Block-Type Structures,
in preprints of Fourth Symposium on Turbu- -,D. E. Wolf, and R. L. Mancuso, 1978, Long
lence, Diffusion, and Air Pollution, Reno, Nev., Term Regional Patterns and Transfrontier Ex-
Jan. 15-18, 1979, pp. 603-609, American Meteo- changes of Airborne Sulfur Pollution in Europe,
rological Society, Boston, Mass. Atmos. Enuiron., 12: 511-527.
-, 1980, Practical Application of Air Pollution Kaimal, J. S., et al., 1977, Turbulence Structure in
Deposition Models-Current Status, Data Re- the Convective Boundary Layer, J. Atmos. Sci.,
quirements, and Research Needs, in Proceedings 33: 2152-2169.
REFERENCES 95
Kao, S. K., and L. L. Wendell, 1968, Some Charac- Machta, L., H. L. Hamilton, Jr., L. F. Hubert, R. J.
teristics of Relative Particle Dispersion in the List, and K. M. Nagler, 1957, Airborne Measure-
A-tmospheres Boundary Layer, Atmos. Environ, ments of Atomic Debris, J. Meteorol., 14(2):
2: 397-407. 165-175.
Kazanski, A. B., and A. S. Monin, 1957, The Form of Mason, B. J., 1971, The Physics of Clouds, Clarendon
Smoke Jets, Izv. Atmos. Oceanic Phys., 8: Press, Oxford, England.
1020-1033. Meroney, R. N., 1979, Turbulent Diffusion near
Kessler, E., 1969, On the Distribution and Continuity Buildings, in Engineering Meteorology, E. J. Plate
of Water Substance in Atmospheric Circulations, (Ed.) (to be published).
Meteorol. Monogr., 10. -,and B. T. Yang, 1971, Wind-Tunnel Study on
Koenig, L. R., 1979, Anomalous Cloudiness and Gaseous Mixing Due to Various Stack Heights and
Precipitation Caused by Industrial Heat Rejec- Injection Rates Above an Isolated Structure,
tion, Report R-2465-DOE, The Rand Corpora- Report CER71-72RNM-BTY16, Colorado State
tion. University, Fluid Dynamics and Diffusion Labo-
-, F. W. Murray, and P. M. Tag, 1978, Differences ratory.
in Atmospheric Convection Caused by Waste Meyer, J. H., T. W. Eagles, L. C. Kohlenstein, J. A.
Energy Rejected in the Forms of Sensible and Kagan, and W. D. Stanbro, 1974, Mechanical
Latent Heats, Atmos. Environ., 12: 1013-1020. Draft Cooling Tower Plume Behavior: Measure-
Lamb, R. G., 1979, The Effects of Release Height on ments, Models, Predictions, in Cooling Tower
Material Dispersion in the Convective Planetary Environment- 1974, ERDA Symposium Series,
Boundary Layer, in Fourth Symposium on College Park, Md., Mar. 4-6, 1974, Steven R.
Turbulence, Diffusion, and Air Pollution, Reno, H a n n a a n d J e r r y Pel1 (Coordinators),
Nev., Jan. 15-19, 1979, pp. 27-33, American pp. 307-352, CONF-740302,NTIS.
Meteorological Society, Boston, Mass. Monin, A. S., and A. M. Obukhov, 1953, Dimen-
Lange, R., 1978, ADPIC-A Three Dimensional sionless Characteristics of Turbulence in the
Particle-In-Cell Model for the Dispersal of Atmo- Layer of Atmosphere near the Ground, Dokl.
spheric Pollutants and Its Comparison t o Regional Akad. Nauk SSSR, 93: 257-267.
Tracer Studies, J. AppL Meteorol., 17: 320-329. -, and A. M. Yaglom,. 1971, Statistical Fluid
Lettau, H., 1970, Physical and Meteorological Basis Mechanics, Mechanics of Turbulence, Vol. I,
for Mathematical Models of Urban Diffusion, in J. Lumley (Ed.), The MIT Press, Cambridge, Mass.
Roceedings of Symposium on Multiple Source Moore, D. J., 1974, Observed and Calculated Magni-
Urban Diffusion Models, Air Pollution Control tudes and Distances of Maximum Ground Level
Official Publication No. AP 86, Environmental Concentration of Gaseous Effluent Material
Protection Agency. Downwind of a Tall Stack, Adv. Geophys., 18B:
Long, P. E., and D. W. Pepper, 1976, A Comparison 201-221.
of Six Numerical Schemes for Calculating the
Advection of Atmospheric Pollution, in Proceed- Nappo, C. J., 1974, A Method for Evaluating the
ings of the Third Symposium on Atmospheric Accuracy of Air Pollution Prediction Models, in
Roceedings of the Symposium on Atmospheric
Turbulence, Diffusion, and Air Quality, Raleigh,
Diffusion and Air Pollution, Santa Barbara, Calif.,
N. C., Oct. 19-22, 1976, pp. 181-186, American
Meteorological Society, Boston, Mass. Sept. 9-13, 1974, pp. 325-329, American Meteo-
rological Society, Boston, Mass.
MacCacken, M. C., and K. E. Grant, 1975, Livermore -, 1979, Relative and Single Particle Diffusion
Regional Air Quality Model (LJRAQ-I),in Ameri- Estimates Determined from Smoke Plume Photo-
can Meteorological Society Conference on Re- graphs, in Fourth Symposium on Turbulence,
gional and Mesoscale Modeling, Analysis, and Diffusion, and Air Pollution, Reno, Nev.,
Rediction, Las Vegas, May 6-9, 1975, American Jan. 15-19, 1979, pp. 4547, American Meteo-
Meteorological Society, Boston, Mass. rological Society, Boston, Mass.
McElroy, J. L., and F. Pooler, 1968, St. Louis -,S. R Hanna, and H. F. Snodgrass, 1980,
Dispersion Study, Report AP-53, U. S. Public Drainage Wind Observations Using Neutral-Lift
Health Service, National Air Pollution Control Balloons, in Proceedings of 2nd Joint Conference
Administration. on Applications of Air Pollution Meteorology,
McMahon, T. A., and P. J. Denison, 1979, Empirical New O r l e a n s , La., Mar. 24-27, 1980,
Atmospheric Deposition Parameters-A Survey, pp. 495498, American Meteorological Society,
Atmos. Environ., 13: 571-585. Boston, Mass.
Otts, R. E., 1976, Locally Heavy Snow Downwind American Institute of Aeronautics and Astro-
from Cooling Tower, NOAA Tech. Memo. NWS nautics, New York.
ER-72, National Oceanic and Atmospheric Prahm, L. P., and 0. Christensen, 1977, Long Range
Administration. Transport of Pollutants Simulated by a 2D
Overcamp, T. J., 1976, A General Gaussian Ps eudospectral Dispersion Model, J. Appl.
Diffusion-Deposition Model for Elevated Point Meteorol., 16: 896-910.
Sources, J. Appl. Meteorol., 15: 1167-1171. Randerson, D., 1972, Temporal Changes in Hor-
Pack, D. H., and J. K. Angell, 1963, A Preliminary izontal Diffusion Parameters of a Single Nuclear
Study of Air Trajectories in the Los Angeles Basin Debris Cloud, J. Appl. Meteorol., 11 : 670-673.
Derived from Tetroon Flights, Mon. Weather Reid, J. D., 1979, Markov Chain Simulations of
Rev., 91: 583-604. Vertical Dispersion in the Neutral Surface Layer
Pandolfo, J. P., and C. A. Jacobs, 1973, Tests of an for Surface and Elevated Releases, Boundary-
Urban Meteorological Pollutant Model Using CO Layer Meteorol., 16: 3-22.
Validation Data in the Los Angeles Metropolitan Reynolds, S. D., and P. M. Roth, 1973, Mathematical
Area Vol. I. CEM Report 490a, Contribution Modeling of Photochemical Air Pollution. Pt. 1.
68-02-0223, prepared by The Center for the Formulation of the Model, Atmos. Environ., 7:
Environment and Man, Inc., for the Environ- 1022-1061.
mental Protection Agency. Richardson, L. F., 1926, Atmospheric Diffusion
Panofsky, H. A., H. Tennekes, D. H. Lenschow, and Shown on a Distance-Neighbour Graph, Proc. R.
J. C. Wyngaard, 1977, The Characteristics of SOC.(London), Ser. A , 110: 709.
Turbulent Velocity Components in the Surface Roberts, 0. F. T., 1923, The Theoretical Scattering
Layer Under Convective Conditions, Boundary- of Smoke in a Turbulent Atmosphere, Proc. R.
Layer Meteorol., 11: 355-361. SOC.(London), Ser. A , 104: 640-654.
-,C. A. Egolf, and R. Lipschutz, 1978, On Sagendorf, J., and C. R. Dickson, 1974, Diffusion
Characteristics of Wind Direction Fluctuations in Under Low Wind Speed and Inversion Conditions,
the Surface Layer, Boundary-Layer Meteorol., Technical Memorandum 52, National Oceanic and
15: 439-446. Atmospheric Administration, Environmental Re-
Pasquill, F., 1961, The Estimation of the Dispersion search Laboratories, &r Resources Laboratory.
of Windborne Material, Meteorol. Mag., 90: Sehmel, G. A., 1980, Particle and Gas Dry Deposi-
33-49. tion: A Review, Atmos. Environ., 14: 983-1012.
-,1974, Atmospheric Diffusion, 2nd ed., John Seneca, J., 1955, Measures de diffusivite turbulent6
Wiley & Sons, New York. sur des flocons de fume&, J. Sci Meteorol., 7:
-, 1975, The Dispersion of Materials in the Atmo- 221 -225.
spheric Boundary Layer. The Basis for Generaliza-
Sheih, C. M., 1974, Application for a Statistical
tion, in Lectures on Air Pollution and Environ- Trajectory Model t o the Simulation of Sulfur
mental Impact Analyses, pp. 1-34, American Pollution over Northeastern United States,
Meteorological Society, Boston, Mass. Atrnos. Enuiron., 11: 173-178.
-, 1976, Atmospheric Dispersion Parameters in
Sherman, C. A., 1978, A Mass Consistent Model for
Gaussian Plume Modeling: Part II. Possible Re-
Wind Fields over Complex Terrain, J. Appl.
quirements for Change in the Turner Workbook
Meteorol., 17: 312-319.
Values, Report EPA-600/4-760306, U. S. Envi-
ronmental Protection Agency. Shir, C. C., 1973, A Preliminary Numerical Study of
Paulson, C. A., 1970, The Mathematical Representa- Atmospheric Turbulent Flows in the Idealized
tion on Wind Speed and Temperature Profiles in Planetary Boundary Layer, J. Atmos. Sci., 30:
the Unstable Atmospheric Surface Layer, J. Appl. 1327-1339.
Meteorol., 9: 857-861. -, and L. J. Shieh, 1974, A Generalized Urban Air
Peterson, K. R., 1968, Continuous Point-Source Pollution Model and Its Application to the Study
Plume Behavior out to 160 Miles, J. Appl. of SO2 Distributions in the St. Louis Metro-
Meteorol., 7: 217-226. politan Area, J. Appl. MeteoroL, 13: 185-204.
Policastro, A. J., W. E. Dunn, and M. Breig, 1978, Simon, C., and W. Proudfit, 1967, Some Observations
Evaluation of Theory and Performance of Salt of Plume Rise and Plume Concentration Distrib-
Drift Deposition Models for Natural Draft Cool- uted over N.Y.C., Paper 67-83 of the 60th Annual
ing Towers, in Proceedings of the 2nd AIAAI Meeting of the Air Pollution Control Association,
ASME Thermophysics and Heat Transfer Confer- Cleveland, Ohio, June 11-16, 1967, Air Pollution
ence, Palo Alto, Calif., May24-28, 1978, Control Association.
REFERENCES 97
Slinn, W. G. N., 1974, The Redistribution of a Gas U. S. Department of Commerce, 1968, Environ-
Plume Caused by Reversible Washout, Atmos. mental Data Service, Climate Atlas of the United
Environ., 8: 233-239. States.
Smith, E. J., and K. J. Hefferman, 1956, The Decay Van der Hoven, I., 1968, Deposition of Particles and
of the Ice-Nucleating Properties of Silver Iodide Gases, in Meteorology and Atomic Energy-
Released from a Mountain Top, Q. J. R 1968, pp. 202-207, D. Slade (Ed.), USAEC Re-
Meteorol. Soc., 8 2 : 301-309. port TID-24190, U. S. Atomic Energy Commis-
Smith, F. B., 1967, The Eulerian-Lagrangian Time
sion! NTIS.
Scale Relationship in One-Dimensional Turbu-
lence, in Proceedings of the U. S. Atomic Energy -, 1976, A Survey of Field Measurements of
Commission Meteorological Information Meeting, Atmospheric Diffusion Under Low Wind Speed,
Chalk River, Ontario, Canada, Sept. 11-15, Inversion Conditions, Nucl. Saf., 17(2): 223-230.
1967, CONF-670931. Venkatram, A., 1978, An Examination of Box
-, 1972, A Scheme for Estimating the Vertical Models for Air Quality Simulation, Atmos.
Dispersion of a Plume from a Source near Ground Environ., 12: 2243-2250.
Level, in Proceedings of the Third Meeting o f the Weil, J., 1974, The Rise of Moist Buoyant Plumes,J.
Expert Panel on Air Pollution Modeling, Report Appl. Meteorol., 13: 435-443.
NATO-CCMS-14, North Atlantic Treaty Organiza-
tion, Brussels. Wendell, L. L., D. C. Powell, and R. L. Drake, 1976,
-, and J. S. Hay, 1961, The Expansion of Clusters A Regional Scale Model for Computing Deposi-
of Particles in the Atmosphere, Q. J. R . Meteorol. tion and Ground Level Air Concentration of SO2
SOC.,87(371): 82-101. and Sulfates from Elevated and Ground Sources,
-,and R. D. Hunt, 1978, Meteorological Aspects in preprint Volume, Third Symposium on Atmo-
of the Transport of Pollution over Long Dis- spheric Turbulence, Diffusion, and Air Quality,
tances, Atmos. Environ., 12: 461-477. Raleigh, N. C., Oct. 19-22, 1976, pp. 318-324,
Smith, M. E., 1951, The Forecasting of Micrometeo- -4merican Meteorological Society, Boston, Mass.
rological Variables, Meteorol. Monogr., 4: 50-55. Wigley, T. M. L., and P. R. Slawson, 1971, On the
-, 1968, Recommended Guide for the Prediction Condensation of Buoyant, Moist, Bent-Over
of the Dispersion of Airborne Effluents, 1st ed., Plumes, J. Appl. Meteorol:, 10: 259-263.
American Society of Mechanical Engineers, New
Wilson, D. J., 1976, Contamination 0.f Building Air
York. Intakes from Nearby Vents, University of Alberta
Sutton, 0. G., 1932, A Theory of Eddy Diffusion in
the Atmosphere, R o c . R. SOC.(London), Ser. A , Department of Mechanical Engineering Report
135: 143. No. 1, University of Alberta, Edmonton, Alberta,
-, 1953, Micrometeorology, McGraw-Hill Book Canada.
Company, New York. -, 1979, Flow Patterns over Flat-Roofed Buildings
Taylor, G . I., 1921, Diffusion by Continuous Move- and Application to Exhaust Stack Design,
ments, Proc. London Math. SOC.,20: 196. ASHRAE Trans., 85(2): 284295.
-, 1948, Dynamics of a Mass of Hot Gas Rising in -, and D. D. J. Netterville, 1978, Interaction of a
Air, USAEC Report MDDC-919 (LADC-276), Los Roof-Level Plume with a Downwind Building,
Alamos Scientific Laboratory, NTIS. Atmos. Envuon., 12(5): 1051-1059.
Turner, D. B., 1967, Workbook of Atmospheric
Dispersion Estimates, Public Health Service, Wyngaard, J. C., 1975, Modeling the Planetary
Boundary Layer. Extension to the Stable Case,
Publication 999-AP-26, Robert A. Taft Sanitary
Engineering Center, Cincinnati, Ohio. Boundary-Layer MeteoroL, 9: 441-460.
-, 1979, Atmospheric Dispersion Modeling: A -, 0. R. Cote, and K. S. Rao, 1974, Modeling the
Critical Review, J. Air Pollut. Control ASSOC.,29: Atmospheric Boundary Layer, Ado. Geophys.,
502-519. 18A: 193-211.
Abramowitz, M. (Ed.), 52 Eagles, T. W. (cited as Meyer et al.), 77 Irwin, J. S., 7, 31, 32
Allen, P. W.,43 Edinger, J. G., 43 Islitzer, N. F., 48
Angell, J. K., 43 Egan, B. A., 53,83,85 Ivanov, V. N., 43
Egan, B. A. (cited as Hanna et al.), 27 Izumi, Y. (cited as Businger et al.), 7, 54
Egolf, C. A., 81
Barad, M. L. (Ed.), 48 Ekman, V. W., 53 Jacobs, C. A., 62
Barry, P. J., 24 Eliassen, A., 87 Jaeger, J. C., 50
Bass, A., 87 Engelmann, R. J., 78
Batchelor, G. K., 41,42 Jessup, E. A., 43
EnvironmentalProtection Agency, 25 Johnson, W. B., 24,61,87
Benkley, C. W. (cited as Bass et d.), 87 EnvironmentalSystems Corporation, 78
Bradley, E. F. (cited as Businger et al.), Eschenroeder, A. Q., 62
7, 54 Kagan, J. A. (cited as Meyer e t d),77
Braham, R. B., 43 Escudier, M., 14 Kaimal, J. S., 7,8
Breig, M., 79 Kao, S. K., 43
Briggs, G. A., 11,14,16,17,19,29,30, Falls, A. H., 73 Katz, I., 43
82,85 Fay, J. A., 14, 88 Kazanslu, A. B., 43
Briggs, G. A. (cited as Hanna et al.), 27 Frenkiel, F. N., 43 Kessler, E., 77
Businger, J. A., 7,54 Koenig, L. R., 77,78
Byzova, N. L., 43 Kohlenstein, L, C. (cited as Meyer et al.),
Garger, Y e K., 43 77
Gartrell, F. E. (cited as Carpenter et al.),
Cagnetti, P., 22 17,27
Carpenter, S. B., 17,27 Gifford, F. A., 24,25,27,29,33,34, Lamb, R. G., 49
Carslaw, H. S., 50 39,42,43,46,58, 59,60 Lange, R., 54
Caughey, S. J., 8 Gifford, F. A. (cited as Hanna et d.), 27 Lawson, R. E., Jr., 84
Chamberlain, A. C., 67 Golder, D., 27 Leavitt, J.M. (cited as Carpenter et al.), 27
Chaudhry, F. H., 47 Grant, K. E., 62 Lenschow, 1). H.(cited as Panofsky
Christensen, O., 54 et al.), 7, 54
Colbaugh, W. C. (cited as Carpenter Lettau, H., 57
Hales, J. M., 71 Lipschutz, R, 81
et d),27 Halitsky, J., 22,23 List, R. J. (cited as Machta et al.), 43
Cote, 0. R. (cited as Wyngaard et al.), 7 Hamilton, H. L,J . (cited as Machta Long, P. E., 54
Cramer, H. E., 27,28 et al.), 43
Crane, G., 54 Hanna, S. R, 17,27,43, 55, 58, 59,60,
Crawford, T. V., 43 74,77,81,83,84 MacCracken, M. C., 62
Crozier, W. D., 43 Haugen, D. A. (Ed.), 29 McElroy, J. L., 30
Hay, J. S., 39,41,43 McMahon, T. A., 69,70
Dana, M. T., 71 Hayes, S. R., 63 Machta, L., 43
Davis, R W., 43 Hefferman, K. J., 43 Mahoney, J., 53
Deardorff, J. W., 7,49 Heffter, J. L., 43, 87 Mancuso, R. L., 87
Deardorff, J.W.(cited,as Hanna,etd.), 27 Hogtrom, U., 43 Markee, E. H., 24
Demerjian, K. L., 58 Holzworth, G., 6 Martinez, J. R., 62
DErrico, D.,85 Horst, T. W., 31,38,39,46,47,48,69 Mason, B. J., 78
Denison, P. J., 69,70 Hosker, R. P., Jr., 19,21,29,68,71, Meroney, R. N., 24,47
Dickson, C. R., 24, 30 72,83 Meyer, J . H., 77
Donaldson, C., 5,50 Hoult, D. P., 14 Monin, A. S., 6, 43,46
Doran, J. C., 31,38,39 Huber, A. H., 24 Montgomery, T. L. (cited as
Drake, R. L,47 Hubert, L. F. (cited as Machta et al.), Carpenter et al.), 27
Draxler, R. R,31,38,88,89 43 Moore, D. J., 17
Dumbauld, R. K., 48 Hunt, J. C. R, 84 Morris, C. S. (cited as Bass et al.), 87
Dunn, W. E., 79 Hunt, R. D., 88 Murray, F. W., 77
98
AUTHOR INDEX 99
Nagler, K. M. (cited as Machta et al.), 43 Rosenzweig,J. J., 88 Tennekes, H. (cited as Panofsky et al.),
Nappo, C. J., 44,63,83,84 Roth, P. M., 62 7, 54
Netterville, D. D. J., 22 Ruff, R. E. (cited as Johnson et al.), 24 Thomas, F. W., 17,27
Nickola, P. W., 31,38,39 Thomas, F. W. (cited as Carpenter et al.),
Nordsieck, R A., 62 27
Sagendorff, J., 30
Salas, L. (cited as Johnson et al.), 24 Turner, D. B., 27,61,62,64,65
Obukhov, A., 6 Schere, K. L.58
Otts, R. E., 78 Scire, J. S. (cited as Bass et al.), 87
U. S. Department of Commerce, 2
Overcamp, T. J., 69 Seely, B. K., 43
Sehmel, G. A., 69,70
Seinfeld, J. H., 73 Van der Hoven, I., 30,67
Pack, D. H., 43 Seneca, J., 43 Vaudo, C, 85
Pandolfo, J. P., 62 Sheih, C. M., 62,88 Venkatram, A., 58
Panofsky, H. A., 7,54,81 Shelar, E. (cited as Johnson et al.), 24
Pasquill, F., 9,10,25,27,29,31, 33, Sherman, C. A., 88
38,39,46,50, 52,84 Shir, C. C., 55,62 Weil, J., 77
Pasquill, F. (cited as Hanna et al.), 27 Simon, C., 15 Wendell, L. L., 43, 87
Paulson, C. A., 7 Singh, H. B. (cited as Johnson et al.), 24 Wigley, T. M. L., 77
Pepper, D. W., 54 Sklarew, R. C., 61 Wilson, D. J., 20, 21, 22
Peterson, K. R., 43 Slawson, P. R., 77 Wolf, D. E., 87
Policastro, A. J., 79 Slinn, W. G. N., 71 Wolf, M. A., 71
Pooler, F., 30 Smith, E. J., 43 Wyngaard, J. C., 7,8,54
Powell, D. C., 87 Smith, F. B., 29,41,52,88 Wyngaard, J. C. (cited as Businger
Prahm, L. P., 54 Smith,M. E., 27,29,59,60 et al.), 7, 54
Proudfit, W., 15 Snodgrass, H. F., 83,84 Wyngaard, J. C. (cited as Panofsky
Snyder, W. H., 24,84 et al.), 7,54
Stanbro, W. B. (cited as Meyer et d),77
Randerson, D., 43 Start, G. E., 24
Rao, K. S. (cited as Wyngaard et al.), 7 Stegan, I. A. (Ed.), 52 Yaglom, A. M., 46
Reid, J. D., 10,40 Sutton, G., 25,38,50 Yang, B. T., 24
Reynolds, S. D., 62
Richardson, L. F., 42 Tag,P. M., 77
Roberts, 0. F. T., 43, 51 Taylor, G. I., 12,30, 36 Zeman, O., 55
Acid rain, 8 7 Cloud microphysics, 77 Eddy conductivity, 7, 54
Adiabatic temperature gradient, 2, 3, 12 Complex terrain air-pollution Eddy diffusivity, 5, 7, 42, 54, 55
Air trajectories in a valley, 81-84 meteorology, 30, 81-86 Eddy dissipation rate, 8, 16, 42, 44
Airflow around obstacles, 19-24 Computers, 40, 53-54 Eddy energy spectra, 8, 9, 41, 42
APKAC model, 6 4 Concentration in cavity, 24 Eddy viscosity, 5, 6
Area sources Conjugate power law, 5 2 Effective radius of momentum,
continuous, 57-66 Continuity equation, 5 temperature, and moisture plume, 77
instantaneous, 50, 51 Continuous sources, 25-40, 46-49 Effective stack height, 11, 19, 22, 85
Atmospheric Turbulence and Diffusion Convective scaling velocity (w.), 7, 32, 4 8 Ekman spiral, 5
Laboratory (ATDL) model for urban Conversion from SO2 to sulfate, 73, 88, 89 Emissions, importance of, 5 7
diffusion, 59-61 Cooling towers, 17, 74-80 Entrainment in rising plume, 12, 13, 77
Averaging time Coriolis force, 2, 5, 6., 55 Environmental Protection Agency (EPA)
effect on regional scale turbulence, 87 Correlation coefficient, 8, 36-40, 63 models, 6 2
effect on u, and uZr25, 27 Critical wind speed for maximum C, Equation of motion, 2, 77
17, 22 Equation of state, 1
Ballistic trajectory of drift drop, 79 Crosswind-integrated concentration, Error analysis, 63
Bent-over plume 47, 48 Eulerian turbulence, 9
buoyancy conservation equation, 1 3 CRSTER model, 65 Evaluation of models, 63
entrainment velocity, 1 2 Exponential chemical decay rate, 72, 7 3
Deposition rate
momentum donservation equation, 1 3
of gases, 68
rise in stable environment, 1 4 Fickian diffusion equation, 25, 51, 88
trajectory near source, 13, 1 4 of SO2 and sulfate, 88
Deposition speed (vd), 68-70 Filter function for spectrum, 6 , 38
volume flux, 11 Finite difference approximations, 53, 5 4
Diffusion
p, 9, 10, 39, 40 First law of thermodynamics, 2, 77
around obstacles, 19-24
Bias, 63 Flushing time, 5 7
calm conditions, 2 5
Box model Frankfurt, West Germany, 5 7
daytime PBL, similarity theory, 48, 4 9
diffusion from area sources, 57-59 Friction velocity, 6 , 16, 17, 46, 5 4
removal processes, 7 3 Gaussian model, 25-35
puff, 41-45, 51 Froude number, 8 4
Breakaway mechanism for drift drops, 7 9 Fumigation, 17, 32, 33, 86
similarity models, 46-49
Breakup model for plume rise, 1 7
statistical models, 36.40
Brookhaven National Laboratory, 27, 2 8 Gamma function, 5 2
surface layer, similarity theory, 46-48
Brownian diffusion, 69 Gaussian distribution of turbulent
urban, 57-66
Brunt-VaidE frequency, 4, 12, 14, 84 speeds, 4 0
Diffusion equation, 5, 50-56
Buildings, diffusion around, 19-24 Gaussian formula
analytical solutions, 50-53
Buoyancy flux of plume, 11
long range, 8 7 plume, 24-35
numerical solutions, 53-56 puff, 4 2
Carbon monoxide urban areas, 62 General circulation, 1, 2
area source model, 6 1 Diffusivity coefficient, 5-7 Geostrophic wind speed, 2, 6
deposition speed, 70 large scale (Ky), 42, 88 Geysers, Calif., 81, 8 4
street canyon model, 61-62 vertical (KJ, 51, 54, 55, 88 Gradient transport models, 50-56
Cavity, building, 19-24 Dimensional analysis, 42, 46-49 Gravitational settling, 67-68
CDM model, 6 4 Distribution functions, 63
CDMQC model, 65 Downwash, 19, 2 3 Hadley cells, 1
Chemical kinetic equations, 72-73 induced by terrain, 85 Half-life of SO*, 6 7
Chemical removal, 72-73, 88, 89 Drainage flow, 81-84 Higher order closure, 55, 56
Chemical transformation rate, Drift deposition, 78-80 Highway models, 61, 6 4
SOpulfate, 73, 88, 89 Dry cooling towers, 77-78 HIWAY model, 6 4
Closure assumption for plume rise, 1 2 Dry deposition, 67-71, 88 Hydrostatic equation, 1
100
SUBJECT INDEX 101
Inertial subrange influence on Physical models of flow over complex 0,
puff diffusion, 42 terrain, 83, 8 4 in Gaussian equation, 25-35
Instantaneous plume snapshot. 44 Planetary boundary layer, 4-6 from uc method, 28, 31
Instantaneous source. 41 -45, 5 1 Plume, bent-over (see Bent-over stability class method, 27-30
Intensity of turbulence, 41 plume) Similarity theory
Inversion. 3, 14-16 Plume impaction on terrain. 8 4 for diffusion, 46-49
Plume meander. 30: 3 4 for spectra, 8
Jet stream. 2 Plume penetration of elevated inversion, for wind and temperature
14, 15 profiles, 6. 7
Plume rise Snowfall from cooling tower plume, 78
K coefficients, 5.7, 50-56, 62, 88 from cooling towers, 74, 76 SO? deposition speeds, 7 1, 88
Krypton 85, 89. 90 determined by ambient turbulence, Source depletion model, 68
15-17 Source effects on airflow and diffusion,
Lagrangian-Eulerian relations, 9, 10, 40 general, 1 1.18, 46 19-24
Lagrangian time scale, 37-40, 43, 49, 51 limited by ambient stability, 14 Sources close to building, 22
Lagrangian turbulence, 9 multiple sources. 17 Sources upwind of building, 22
Langevins equation, 4 3 trajectory near source, 13. 14 Spectra, 8, 38, 39
Lapse rates. 3. 12 Potential temperature, 3 Split-H concept. 2 4
Latent heat, 3, 7 4 Prairie Grass experiment. 29. 40. 4 8 Stability
Line source. 51 Primary pollutants. 72 classification schemes, 27, 85, 86
Long-range transport and diffusion, 87-90 Psychrometric chart, 74. 76 general definition, 3
Los Angeles model for CO, 6 2 PTDIS model, 6 4 limits to plume rise. 1 4
PTMAX model, 6 4 . term s. 12-14
PTMTP model. 6 4 Stable lapse rate, 3
Mass change of drift drops. 78 Puff diffusion. 41-45, 51 Stable profiles of u,, uv,and
Maximum ground-level concentration. Puff models of long-range diffusion, aw,8
11, 17, 22-24, 32, 33 87, 88 Stack aerodynamic effect, 19
Mean plume height (i), 46-48 Statistical models, 36-40
Mechanical-draft cooling towers, 7 4 Stokess law, 6 7
Mesoscale eddies, 43 Rainfall effect on deposition, 71, 72
Street canyon model, 6 1
Mixing layer height (z,), 6-8, 27, 31, R.4hI model, 65
Surface buoyancy flux, 6, 16, 17
48, 55 Raveliswood power plant, 15
Mixing ratio, 3 Recirculation zones, 19-21
Model evaluation, 63 . Regional Air Pollution Study
Taylor entrainment assumption, 12, 1 3
Models, EPA, 64, 65 (St. Louis), 6 1
Taylors statistical theory, 36-38, 41, 51
Moist adiabatic temperature gradient, 3 Release time, 41
Temperature gradient
Momentum flux, turbulent, to ground, 6 Removal mechanisms, 67-73
Momentum flux of plume, 11, 1 2 Resistance analogy for deposition, 69 - as indicator of stability class, 27
in surface layer, 7
Monin-Obukhov length (L), 6, 27, Reynolds averaging, 5
Tennessee Valley Authority, 27, 2 9
47-49, 5 4 Richardson number, 4, 7, 27
Terminal speed of water drops, 78
Monte Carlo model, 36-40 Roughness length, 6, 2 9
Tilted plume model, 68
Multiple sources, plume rise, 17 Time scale of turbulence, 9, 30, 37-40
Sampling time, 33, 38, 39, 41 Top-hat model, 11, 12
Narrow plume hypothesis, 59 Saturation deficit, 76, 77 Trajectory models
National Climatic Center, 2 Savannah River Laboratory experiment, for long-range transport, 87-90
Natural-draft cooling tower, 7 4 89, 90 for urban diffusion, 6 2
Neutral lapse rate. 3 Scavenging coefficient, 7 1;73 Travel time, 36-41
Neutral plume rise, 1 6 Screening model, 58 Turbulence
Neutral profiles of uu, uv, Second-order closure, 5, 55, 56 fluxes, 4, 5 , 50, 55
and uW,8 Secondary pollutants, 7 2 influence on plume rise, 15, 1 6
Nuclear Regulatory Commission, 27 Sector model, 3 4 intensity, 8
Numerical cooling tower plume ue, 30. 31 velocity fluctuations u,, u,,
models, 77 00, 8. 30, 31, 39, 81-83 and uwr8
Numerical instabilities, 53 uuprofies in PBL, 8
0 , profiles in PBL, 8
uvuse in statistical models, 36-40 UNAMAP models, 62, 64-65
PAL model, 65 u%profies in PBL, 8 Unstable lapse rate, 3
Partial reflection model, 69 by Unstable profiles of uu, uv, and
Particle-in-cell method, 5 4 from diffusion equation, 52 uW.8
Particle trajectory model, 40 in Gaussian equation, 25-33 Urban diffusion, 30, 57-66
Peak analysis for model evaluation, 63 for long-range diffusion, 88
Peak-to-mean concentration ratio, 33, 3 4 peak to mean, 33, 3 4
Performance measures, 63 puff, 41-45 Valley limited diffusion, 86
Photochemical box model, 58 frurn 00 method, 27-3 1 VALLEY model, 65
Photochemical smog, chemical stability class method, 27-30 Vents, diffusion from, 22-24
description, 7 3 statistical models, 36-40 Vertical diffusivity, 54, 55
r .. . ..
.
1 -
Vertical distribution of concentration, Virtual source, 14, 17 Wavelength of maximum energy, 8

47,4a Visible plume dimensions for cooling Weighting scheme for winds, 88, 90
Vertical plume tower, 76-77 Wet deposition, 71, 72
buoyancy conservation equation, 12 Volume flux of plume, 11, 12 Wet removal of sulfur, 88
entrainment velocity, 12 von Ka'rmin's constant, 6 Widows Creek Steam Plant, 81, 82
momentum conservation equation, 13 Wind direction shear, 5, 6
rise in stable environment, 14 Wind profile in surface layer, 6, 7
trajectory near source, 13 Wake cavity, 20-24 Wind-speed power law, 32
volume flux, 11 Washout ratio, 71, 72 Wind-tunnel experiments, 19-22
DISC LA1M E R
This b o o k was prepared as an account of work sponsored b y an agency o f the

United States Government. Neither the United States Government nor any agency
thereof, nor any o f their employees. makes any warranty, express or implied, or
assumes any legal liability or responsibility for t h e accuracy, completeness, or
usefulness of any information, apparatus, product, o r process disclosed, or
represents that i t s use would n o t infringe privately owned rights. Reference herein
t o any specific commercial product, process, or service b y trade name, trademark,
manufacturer, or otherwise, does n o t necessarily constitute or imply i t s
endorsement, recommendation, o r favoring b y the United States Government or
any agency thereof. The views and opinions of authors expressed herein d o not
necessarily state or reflect those o f the United States Government or any agency
thereof.
* U S . GOVERNMENT PRINTING OFFICE: 1981 -740-145/2802

DOE TIC-11223 (Handbook On Atmospheric Diffusion)

Uploaded by

Copyright:

Available Formats

DOE TIC-11223 (Handbook On Atmospheric Diffusion)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

DOE TIC-11223 (Handbook On Atmospheric Diffusion)

Uploaded by

Copyright:

Available Formats

DOE/TIC-11223

Atmospheric Turbulence and Diffusion Laboratory

Jean S. Smith, Publication Editor

International Cop>-right. 0 1:. S. Department o f Energy. 1982, under the

Library o f Congress Cataloging in Publication Data

Includes bibliographical references and index.

Available as DE82002045 (DOE/TIC-11223) for $10.75 from

National Technical Information Service

DOE Distribution Category UC-11

Printed i n the United States of America

6-3 Similarity Approach . . . . . . . 42 9-5.2 Trajectory Models . . . . . 62

1-1 INTRODUCTION North of 30" N latitude, the direct Hadley sell

The earths surface exerts a drag on the atmo-

Next, consider the continuity equation for A:

- A simple expression for the variation of the wind

where K is a diffusivity coefficient (in units of

1-4.3 Similarity Theory Gives Wind and L= U*

the top of the mixedlayer (Zi). GEOSTROPHIC WIND SPEED

tions). The absolute value of L can be thought of as (1.40)

where subscript "0" refers to a surface value and fn (1.41)

In stable conditions the following solution is easily

The eddy diffusivity for momentum (Km) and for

(1.39) = 0.37 (0.96 < ?< 1) (1.45)

= 1.3 (1)-: (stable) (1.50)

So far we have been discussing Eulerian turbu- (1.64)

2-1 INTRODUCTION 2-2.1 Definitions

Figure 2.1 is a schematic drawing of a vertical

The review article by Brigs (1975) provides

ENVIRONMENTAL STABILITY. s $(%+ 0 01 OC/rn)

VERTICAL PLUME BENT-OVER PLUME

Environmental stability (s) plays a prime role in v, = olw (vertical plumes)

The entrainment velocity (ve) is the effective

Momentum conservation 2-3 PLUME TRAJECTORY NEAR

2-2.3 Set of Equations for Bent-Over 2-3.1 Vertical Plumes

"2/3 LAW" FOR BUOYANT PLUMES

MOMENTUM FLUX ENHANCEMENT DUE TO BUOYANCY (Fox/Mu)

At night a deep (100 to 200m) stable layer

---- PLUME TOP

A fraction (1 - P) of the plume reflects off the

zi < 2.5 Fo (2.22) On cloudy, windy days or typical sunny summer

Fig. 2.4 Diagram of buoyant plume penetrating an elevated inversion.

During convective conditions the neutral value for

Many sites have several stacks that are close

MGC = 0.234 5 Q (2.28)

For the unstable model, the critical wind speed

Diffusion calculations would be greatly simplified

to be perpendicular to the building face; if it is not,

UPWIND VELOCITY PROFILE

zI, HIGH TURBULENCE ZONE BOUNDARY

zll, ROOF WAKE BOUNDARY

the roof is unlikely, and if the plume clears the wake

cavity does not extend to this height, a fact obviously

For h > H + 1.5 f , Forb < H,

he = h (3.9) he = h - 1.5 f (3.11)

Finally, the plume is assumed to be trapped in the C= Q cWH)u (3.14)

(c) Concentrations in and Downwind of Cav-

4-1 WHY USE THE GAUSSIAN MODEL?

Other dispersion models, such as the K-model, the

T a b l e 4.1 Meteorological Conditions Defining

K, = O.~U,Z~-~~/" (8.22) known. This method of solution is known as "first