You are on page 1of 10

Environmental Modeling and Assessment (2006) 11: 267Y276 # Springer 2006

DOI: 10.1007/s10666-006-9041-y

General plume dispersion model (GPDM) for point source emission


Seema Awasthia, Mukesh Khareb and Prashant Gargavac
a
Department of Applied Mechanics, Indian Institute of Technology, New Delhi, India
E-mail: seema_phd2002@yahoo.co.in
b
Department of Civil Engineering, Indian Institute of Technology, New Delhi, India
E-mail: mukeshk@civil.iitd.ernet.in
c
Central Pollution Control Board, New Delhi, India
E-mail: prashant_gargava@hotmail.com

Gaussian-based dispersion models are widely used to estimate local pollution levels. The accuracy of such models depends on stability
classification schemes as well as plume rise equations. A general plume dispersion model (GPDM) for a point source emission, based on
Gaussian plume dispersion equation, was developed. The program complex was developed using Java and Visual basic tools. It has the
flexibility of using five kinds of stability classification schemes, i.e., Lapse Rate, PasquillYGifford (PG), Turner, Y and Richardson
number. It also has the option of using two types of plume rise formulations Y Briggs and Holland_s. The model, applicable for both rural
and urban roughness conditions, uses meteorological and emission data as its input parameters, and calculates concentrations of pollutant
at the center of each cell in a predefined grid area with respect to the given source location. Its performance was tested by comparing with
4-h average field data of continuous releases of SO2 from Dadri thermal power plant (Uttar Pradesh, India). Results showed that the
Turner scheme used with Holland_s equation gives the best outcome having a degree of agreement (d) of 0.522.
Keywords: Gaussian plume dispersion model, air pollution, stability classification schemes, plume rise, concentration, index of
agreement

1. Introduction and plume rise formulations, and providing the best com-
bination with an improved predictive ability. Considering
Increasing urban air pollution due to the continuous the above-mentioned limitations in Gaussian plume model,
growth of industries and vehicular traffic has given rise to a we tried to develop a model in which different stability
need for comprehensive monitoring accompanied by mod- classification schemes and plume rise formulations are used.
eling of air quality. It is not always feasible to monitor/ In the present study, a general plume dispersion model
measure the concentrations of species at various vulnerable (GPDM) based on Gaussian equation was developed. Unlike
points of a particular area due to high cost and the experi- other Gaussian-based dispersion models, which use only PG
mental difficulties involved. Prediction of pollution levels or Turner scheme in combination with Brigg_s equations,
resulting from a given emission is carried out with the help GPDM is capable of considering five kinds of stability clas-
of air pollution dispersion models, which compute atmo- sification schemes and two plume rise equations (Briggs and
spheric transport and dispersion of pollutants being emitted Holland) for urban as well as rural conditions for a point
into the atmosphere. At present, Gaussian plume model is source emission. This attribute gives GPDM more objectiv-
the most widely used model for point source emission pre- ity and flexibility in its approach and application in air
dictions. It describes the dispersion around a single source in pollution dispersion studies. It uses meteorological and
an open and homogeneous terrain under steady-state con- emission data as its input parameters. Concentrations are
ditions. Several authors have discussed the limitations of determined by using distance crosswind and distance
Gaussian air pollution dispersion models in light of their upwind from the receptor to each source. Pollutant con-
sensitivity to inputs [8], questionable accuracy [4], and limi- centrations are estimated at the center of each cell in a
tations regarding predictions [3]. Most Gaussian-based mod- predefined grid area with respect to the given source
els use the PasquillYGifford stability classification scheme, location. GPDM predictions were later compared with
which considers only ambient temperature gradient as its 4-h average field data at four receptor locations due to
variant. There are meteorological parameters, i.e., wind speed continuous releases of SO2 from Dadri thermal power plant
and direction, solar radiation, potential temperature gradient, situated in the state of Uttar Pradesh, India (site plan is
etc., which are taken into account by other stability classifi- shown in figure 1).
cation schemes (Lapse Rate, PasquillY Gifford, Turner, Y
and Richardson Number). In addition, plume rise formula-
tions other than Briggs equations were also investigated for 2. Air quality dispersion modeling
their suitability as per the existing field conditions.
Hence, there is need for a general form of model capa- A mathematical model is an assembly of concepts or
ble of incorporating various stability classification schemes phenomena in the form of one or more mathematical equa-
268 S. Awasthi et al. / General plume dispersion model (GPDM) for point source emission

Figure 1. Site location of Dadri thermal power plant (source) and receptor locations.

tions, which approximate the behavior of a natural system or lutant at a certain point because of its simplicity. The
phenomena to predict the impacts or concentration of pa- model describes the transport and mixing of pollutants.
rameters under different types of current or future scenarios, The Gaussian plume dispersion equation has been derived
using readily available or measured input data [12]. Air from basic advectiveYdiffusive equation describing the
pollution modelling is defined as an organized and inter- transfer and diffusion of pollutants from instantaneous
active series of computational modules that can transform point source [12]. The concentration of pollutants (C) at
one set of databases (emissions, meteorology, geography, location (x, y, z) from a continuous elevated point source
etc.) into another set (concentration, depositions, health with an effective height of H is given by following
damage, etc.). The important and popular way of classifying equation
air pollution models is based on the model structure and
the approach used for the closure of the turbulent dif- @C @C þ divðCV Þ ¼ rðKrC Þ þ R
i
fusion equation [20], and widely used in urban air pol- @t
lution modeling [12]. The fundamental problem to all þ Qðt t0 Þðx  x0 Þð yy0 Þðzz0 Þ
modelling studies in air pollution is the identification of ð1Þ
function F, which would allow the prediction of the con-
centration of pollutants C (x, t) at any point in space x and where C(x, y, z, t) is the mass concentration of pollutant, V
time t, if the data on emissions and other meteorological is the wind velocity vector, K is the turbulent diffusivity
variables are given. tensor, R is the chemical and photochemical transformation
Models are useful to study the consequences of new of pollution, Q is the source term, t0 represents the time the
sources of air pollution or in changing the amount of pol- accident happens, and x0, y0, z0 are the emission source
lutants released into the air from existing emission sources. coordinates. At constant wind velocity and turbulent dif-
Granting of pre-construction permits under the Clean Air fusivity, equation (1) has an analytical solution in the form
Act often requires the use of models to estimate the air of Gaussian normal distribution
quality impact from proposed new and modified sources.
 2 ( !
Other applications involve assessing present air quality,
Q  12 yy  ðzr  He Þ2
defining cost-effective emissions reduction strategies and C ðx; y; zr ; He Þ¼ *e e
analyzing the effects of potential and actual accidental re- 2uy z 22y
leases of air toxics.
!)
2
 ðzr þ H e Þ
þe ð2Þ
22z
3. Gaussian plume dispersion model

The Gaussian plume model is the most commonly used where C is the pollutant concentration (mg/s), Q is the
dispersion model to estimate the concentration of a pol- emission rate from the source (mg/m3), zr is the receptor
S. Awasthi et al. / General plume dispersion model (GPDM) for point source emission 269

Table 1
Plume rise formulae.

Atmospheric condition Plume rise equation Variables used in the equation

Unstable or neutral Dh = 1.6F1/3x2/3uj1 for x G xf; (for x 9 xf, xf is downwind distance of final plume rise
(stability A, B, C, D) replace x with xf) = 119F0.40 for F Q 55 m4/s3
= 49F0.625 for F G 55 m4/s3
Stable (Stability E, F), Dh = 1.6F1/3x2/3uj1 for x G xf F is the buoyancy flux parameter (m4/s3)
when 1.84usj1/2x Q xf,
For x Q xf and F Q 55 m4/s3, Dh = 38.7F3/5uj1 u is the wind speed at the top of the stack (m/s)
For x Q xf and F G 55 m4/s3 Dh = 21.4F3/5uj1 x is the distance at which atmospheric turbulence
begins, to dominate entrainment (m)
For x Q 1.84usj1/2 s is the stability parameter (sj2)

height above ground (m), He is the effective stack height (m); x is the downwind distance from source to receptor
(m) (height of stack + plume rise), u is the mean horizon- (m), and y is lateral distance from source to receptor (m).
tal wind speed at He (m/s), y is horizontal dispersion A collection of information about Gaussian plume mod-
coefficient (m), z is the vertical dispersion coefficient els is given by Pasquill and Smith [15], Turner [11, 17].

START

Input = U, Ta, α , std. dev., Dia (stack), Tsi, Vsi, Q, Zi,, P, L, xs, ys, m, n,
k, l, MH, time, R/U, Stab. Scheme(1/2/3/4/5), Br/Ho equation

Stability Class
(A,B,C,D,E or F)

Exponent in Power law eqn.(n) xi , yi If xi < 0, upwind location

Wind Speed at stack height (hs)

Buoyancy Flux Parameter (F)

Downwind distance at Final plume rise (xf)

Brigg’s Equations Holland’s Equation

Plume Rise
(∆h)

Hei = hs
Hei = hs + ∆h

Wind Speed at effective height Hei

Figure 2. Flow chart of GPDM.


270 S. Awasthi et al. / General plume dispersion model (GPDM) for point source emission

Plume
If Hei < MH,
Penetration
P=0
(P)

ZI’ = Z - hs If Hei > MH,


P = Z i’/∆h

Q m = Qs( 1 – P ) Qm = Qs

σ=e(I+J(Lnxi)+K(Lnxi)^2)

Terrain Type Rural


σyi , σzi

σ=Lxi (1+Mxi)n
Urban

σyi , σzi

C mn
Concentration in mth column and nth row of the grid, using Gaussian Equation

Figure 2. Continued.

A number of applied Gaussian models are fully presented and u is the ambient wind speed (m/s). Values of the
in Guidelines on Air Quality Models, edited by EPA correction factor (cf) are based on stability of the
(1986) and in EPRI (1982). atmosphere, as given in Schnelle and Dey [16].
A buoyant plume, rising into a well-mixed atmosphere
capped by a layer of stable air, may partially or completely
4. General plume dispersion model penetrate the elevated stable layer. In the GPDM, the
fraction (P) of the plume that penetrates the elevated stable
The GPDM algorithm estimates short-term pollutant con-
layer is estimated using the following conditions [5]:
centrations from a point source at a user-specified receptor
grid in a flat terrain. The model can predict pollutant con- (i) No penetration
centrations using any one of the five stability classification
schemes (Lapse Rate scheme, PasquillYGifford scheme, P = 0, if Zi/He Q 1.5;
Turner scheme, SigmaYTheta scheme and Richardson num-
ber) and any one of the two plume rise formulations (Briggs (ii) Total penetration
and Holland). It also takes into account urban and rural
roughness conditions for estimating vertical and horizontal P = 1, if Zi/He e 0.5;
dispersion coefficients. Table 1 summarizes plume rise for-
mulations for bent-over, hot buoyant plumes from elevated (iii) Partial penetration
point sources, as given by Briggs (1969) [1, 6, 7].
Another plume rise equation is the Holland formula [16] P = 1.5 j Zi/hi, if 0.5 G Zi/hi G 1.5.
cf *ð1:5ds *vs þ 1:08F Þ where hi is the predicted plume rise and Zi = Z j hs, Z
h ¼ ð3Þ
u is the stable layer aloft and hs is the stack height (m); Q
where ds is the diameter of the stack (m), vs is the stack exit indicates the modified source strength given by {Qs(1 j
velocity (m/s), F is the buoyancy flux parameter (m4/s3) P)}, where Qs is the emission rate on top of the stack.
S. Awasthi et al. / General plume dispersion model (GPDM) for point source emission 271

receptor

R U

yi α
M θ
P
source
xi

2 l
k

1 2 3 ………….. m
L

Figure 3. Typical grid parameters used in GPDM.

The dispersion parameters, y and z, were estimated GPDM uses power law profile to estimate the wind speed
using following equation [13], [9].
 n
uz hz
 ¼ ðLxi Þ * ð1 þ Mxi ÞN ð4Þ ¼ ð5Þ
ug hg
where xi is downwind distance (km) and L, M, N are con- where uz is the wind velocity at point z (m/s), ug is the wind
stants based on stability of the atmosphere and terrain type. velocity at ground station (m/s), hz is the height of point z

Table 2
Stack emission data at Dadri thermal power plant.

No. Stack Stack Stack Exit Emission


of units height (m) diameter temperature velocity load of SO2
(m) (K) (m/s)

4 210 4.5 400 23.34 1.02  109 mg/m3


272 S. Awasthi et al. / General plume dispersion model (GPDM) for point source emission

Table 3
Details of four receptor locations.

Receptor Sampling stations Distance from Direction Wrt, Receptor


source (km) North (-) height (m)

R1 Urja Bhawan 1.90 323 10


R2 Nidhauli 3.15 44 03
R3 Khangora 1.90 100 08
R4 Khatana 4.30 188 04

(m), hg is the ground station height and n is the power law 6. GPDM algorithm
constant.
The computer code for GPDM is developed in Java
language (figure 2). Input parameters used in the code are
5. Input parameters described below.

Input parameters used in GPDM are described below. (a) Meteorological parameters
Ug – Surface wind velocity (m/s);
(a) Meteorological data comprising wind speeds and di- U – Wind velocity at stack height;
rections, ambient temperature, mixing height, standard Uf – Wind velocity at final rise;
deviation, cloud cover, ambient temperature gradient. a – Wind direction with respect to North;
These data are taken hourly. MH – Mixing height (input parameter) (m);
(b) Source data comprising height and diameter of the Tai – Ambient temperature (input parameter)
stack, emission rate and exit velocity. (K);
Ambgrad – Ambient temperature gradient (input pa-
(c) Grid size and source location Y it includes size of the
rameter) (K/m);
grid and unit cells, location of the source in the grid in
 – Standard deviation of wind direction (in-
Cartesian co-ordinate system.
put parameter)
GPDM estimates the SO2 concentration of at the center
of each cell by using the hourly meteorological and source (b) Stack parameters
data. The user specifies the size of the grid and its unit cell. hsi – Stack height (m);
GPDM estimates 1-, 4- and 8-h average concentrations hdia – Stack diameter (m);
[18]. The output is in the form of screen print file that Q – Source strength (mg/m3);
shows the grid with concentration at the center of each cell. Tsi – Stack gas temperature (K);
It also shows the criteria used for calculating stability class Vsi – Stack exit velocity (m/s)
and plume rise including their values. The predicted con-
centrations are presented in the Ftext file_ in tabular form. (c) Grid parameters (figure 3)
Moreover, the tables may be exported and subjected to fur- P – Length of the grid along x-axis (m);
ther data processing such as graphical presentation, etc. by L – Length of the grid along y-direction (m);
third-party software. k – Cell size of each unit in the grid along x-axis (m);

Table 4
Index of agreement (d) for various combinations at four receptor locations.

S. no. Stability scheme Plume rise equation Combination no. Index of agreement d for receptor location Average d
R1 R2 R3 R4
1 Lapse Rate Brigg_s 11 0.472 0.480 0.548 0.483 0.496
2 PasquillYGifford Brigg_s 12 0.475 0.477 0.454 0.483 0.472
3 Turner Brigg_s 13 0.393 0.493 0.625 0.484 0.499
4 Richardson no. Brigg_s 14 0.472 0.483 0.449 0.484 0.472
5 Sigma Theta Brigg_s 15 0.472 0.479 0.449 0.483 0.471
6 Lapse Rate Holland 21 0.472 0.507 0.594 0.483 0.514
7 PasquillYGifford Holland 22 0.475 0.480 0.457 0.483 0.474
8 Turner Holland 23 0.426 0.528 0.649 0.484 0.522
9 Richardson no. Holland 24 0.439 0.511 0.488 0.481 0.480
10 Sigma Theta Holland 25 0.475 0.479 0.449 0.483 0.472
S. Awasthi et al. / General plume dispersion model (GPDM) for point source emission 273

10 10

5
5

0
0

-5
-5 0 5 10
-5
-5 0 5 10
Figure 4. QYQ plot of receptor, R3 Y Khangora (PasquillYGifford and
Brigg_s equation).
Figure 6. QYQ plot of receptor, R3 Y Khangora (Turner and Brigg_s
Equation).
l – Cell size of each unit in the grid along y-axis (m);
xsi – x-Coordinate of the source;
ysi – y-Coordinate of the source.
If xsi G (m j 0.5) k,  = [(270 j a j tanj1(Y/X)]; Else,
The y-axis is parallel to North and the origin is at the
 = [(90 j a j tanj1(Y/X)]
lower left corner of the grid.
where Y = (n j 0.5) l j ysi and X = (m j 0.5) k j xsi
a – Wind direction with respect to North;
m – Number of cells in x-direction, i.e., 1, 2, 3, . . . , m; (d) Other input parameters
n – Number of cells in y-direction, i.e., 1, 2, 3, . . . , n Terrain – Rural or urban (1/2);
Receptor_s co-ordinate (calculated by the GPDM) = Zri – Receptor height (m);
{(m j 0.5) k, (n j 0.5)l}; (assuming the receptor to be Scheme – 1 Y Lapse Rate, 2 Y PasquillYGifford, 3 Y
at the center of each cell in the grid) Turner, 4 Y SigmaYTheta, 5 Y Richardson
xi – Downwind distance of receptor from the source = Number;
R cos  (if xi G 0, receptor is in upwind direction, else Author – Briggs or Holland plume rise equation (1/2);
downwind) Time – Time of observation in hours for Turner
yi – Crosswind distance from the source = R sin  scheme

10 10

5 5

0 0

-5 -5
-5 0 5 10 -5 0 5 10

Figure 5. QYQ plot of receptor, R3 Y Khangora (Lapse Rate and Brigg_s Figure 7. QYQ plot of receptor, R3 Y Khangora (Sigma Theta and Brigg_s
equation). equation).
274 S. Awasthi et al. / General plume dispersion model (GPDM) for point source emission

10 10

5 5

0
0

-5
-5 -5 0 5 10
-5 0 5 10

Figure 9. QYQ plot of receptor, R3 Y Khangora (PasquillYGifford and


Figure 8. QYQ plot of receptor, R3 Y Khangora (Richardson no. and Holland equation).
Brigg_s equation).

9. Results and discussion


7. Performance evaluation of GPDM
GPDM performance is evaluated by comparing 4-h
Index of agreement d was used to evaluate the pre- average concentrations at four receptors around Dadri
dictive performance of GPDM [19]. Index d allows for thermal power plant with observed data of SO2 concen-
sensitivity towards differences in observed and predicted trations monitored at the same time and same places. All
values as well as proportionality changes (d = 0 indicates 10 combinations of five stability classification schemes
no agreement; d = 1 indicates perfect agreement) and two plume rise formulae were used in the testing to find
the most suitable combination. Index of agreement d ob-
P 2
Cpi  Coi tained for various combinations at all receptor locations are
d ¼ 1  P   2 ð6Þ presented in table 4. It is observed that for all combina-
abs Cpi  Com þ absðCoi  Com Þ

where Coi is the observed concentration, Cpi is predicted


concentration, and Cpm and Com are the averages of the 10
computed and observed concentrations, respectively.

8. Emissions and meteorological data used in the study


5
The location of Dadri thermal power plant (i.e., source)
and four receptor locations are shown in the site plan
(figure 1). Details of receptor locations are shown in
table 2. Source characteristics including stack_s height,
diameter, exit velocity, temperature and emission rates are
given in table 3. Meteorological data comprising wind 0
speed, wind direction, ambient temperature and its gradi-
ent, mixing height, etc. have been recorded continuously
for 12 days at hourly duration in the month of March
(2001) by Central pollution Control Board (CPCB),
Delhi. These data were substituted in the GPDM to -5
-5 0 5 10
estimate 1- and 4-h average SO2 concentrations at four
receptor locations shown in the site plan. Additionally, 4-
h average SO2 concentrations were also recorded at all Figure 10. QYQ plot of receptor, R3 Y Khangora (Lapse Rate and Holland
receptor locations. equation).
S. Awasthi et al. / General plume dispersion model (GPDM) for point source emission 275

10 10

5 5

0
0

-5
-5 -5 0 5 10
-5 0 5 10

Figure 12. QYQ plot of receptor, R3 Y Khangora (Sigma Theta and


Figure 11. QYQ plot of receptor, R3 Y Khangora (Turner and Holland Holland equation).
equation).

tions, index of agreement d is in the range of 0.471Y0.522. ment in the predicted and measured normalised concentra-
This shows that agreement in the predictions and the data tion distributions. It is observed that estimation results
is Fgood_ on average, although internal variation in the using the Turner scheme in combination with the Holland
predictions versus data is substantial. However, such a formula (figure 11) are closer to the one-to-one line
large scatter is not uncommon for comparisons of field compared to the other combinations. However, for low
data with model predictions. One obvious reason for concentrations, the model under-predicts the measured
this is the stochastic nature of atmospheric turbulence. A data. Under-prediction for low concentrations is caused
model prediction represents an ensemble average value, by the excessive number of zero concentrations in the
whereas any measured short-time averaged concentration predicted data, compared with the measured values.
value reflects a specific realization of the ensemble.
Turner scheme in combination with Holland plume rise
formulation has been found to be best suitable prediction
combination for the particular site chosen for the study. 10
Using this combination, GPDM has yielded overall 52.2%
error-free predictions, i.e., d = 0.522. This may be attri-
butable to the fact that Turner scheme is more objective in
nature compared to the other schemes, because it considers
wind speed as well as solar radiation in defining the
5
stability of the atmosphere. Index of agreement d for the
receptor location R3 (Khangora) represents the maximum
for all combinations in comparison with other receptors.
Based on field data (wind speed and direction), it was
observed that this receptor lies in the predominant wind
direction (during the experiment), thereby giving the best 0
error-free predictions.
Performance of the model was further examined with
the help of QYQ (quantileYquantile) plots for the receptor
R3 (Khangora) for all 10 combinations (figures 4Y13).
These plots describe whether the predicted and measured -5
-5 0 5 10
normalised concentration distributions are similar [14]. If
the distributions of model predictions and observations
were identical, the points would lie on the one-to-one line Figure 13. QYQ plot of receptor, R3 Y Khangora (Richardson no. and
[10]. All the quantileYquantile plots show a good agree- Holland equation).
276 S. Awasthi et al. / General plume dispersion model (GPDM) for point source emission

10. Conclusions [3] M.M. Benarie, The limits of air pollution modelling, Atmos. En-
viron. 21 (1987) 1Y5.
A general plume dispersion model for a point source [4] M.R. Beychok, How accurate are dispersion predictions?, in:
Hydrocarbon Processing (Gulf Publishing, Houston, TX, 1979).
emission was developed to estimate the pollutant concen- [5] M.R. Beychok, Fundamentals of Stack Gas Dispersion (published
tration in the downwind direction. The model is based on by the author, Irvine, CA, 1995).
the Gaussian equation. Particular attention was given to [6] G.A. Briggs, Diffusion estimation for small emissions in environ-
classify the atmospheric stability and plume rise calcula- mental research laboratories, Air Resources Atmospheric Turbu-
tion. The option of using five kinds of stability classifica- lence and Diffusion Laboratory, Annual Report of the USAEC,
Report ATDL-106, National Oceanic and Atmospheric Adminis-
tion schemes and two kinds of plume rise formulations are tration, Oak Ridge, TN (1973).
included in the model. It imparts more objectivity and [7] G.A. Briggs, Plume rise and buoyancy effects, in: Atmospheric
flexibility in terms of its approach and application in air Science and Power Production, ed. D. Randerson, U.S. Department
pollution dispersion studies. of Energy, DOE/TIC 27601 (1984) 327Y366.
The model has certain limitations, as follows. It does not [8] M.D. Carrascal, M. Puigcerver and P. Puig, Sensitivity of Gaussian
plume model to dispersion specifications, Theor. Appl. Climatol.
take into account the inversion conditions chemical transfor- 48 (1993) 147Y157.
mation, wet deposition or inhomogeneous terrain. The [9] A.G. Davenport, The application if statistical concepts to the wind
accuracy of GPDM was examined by comparing its predic- loading of structures, Proc. Inst. Civil Eng. 19 (1961) 449Y473.
tions against one analytic solution. For the selected example [10] P. Goyal and T.V.B.P.S. Ramakrishna, Dispersion of pollutants in
convective low wind: a case study of Delhi, Atmos. Environ. 36
case, the error caused by the numerical method was nil.
(2002) 2071Y2079.
Model predictions were also compared with the field data [11] S.R. Hanna, G.A. Briggs and R.P. Hosker Jr., Handbook on
at four receptor locations surrounding Dadri thermal power Atmospheric Diffusion, Atmospheric Turbulence and Diffusion
plant in Uttar Pradesh, India. Index of agreement d for all Laboratory, NOAA, USA Technical Information Centre, US De-
combinations lies in the range of 0.471Y0.522. The most partment of Energy (1982).
suitable combination is that of Turner scheme and Holland [12] M. Khare and P. Sharma, Modelling Urban Vehicle Emissions
(WIT Press, Southampton, Boston, 2002).
equation (d = 0.522). There is substantial internal variation [13] R.W. McMullen, The Change of Concentration Standard Deviation
in predicted and observed concentrations, but the quanti- with Distance (JAPCA 1975).
leYquantile plots for the receptor located at Khangora (R3) [14] J. Nikmo, J.P. Tuovinen, J. Kukkonen and I. Valkama, A hybrid
show that the agreement of the normalized concentration plume model for local-scale atmospheric dispersion, Atmos.
distributions is good for all combinations. However, the Environ. 33 (1999) 4389Y4399.
[15] F. Pasquill and F.B. Smith, Atmospheric Diffusion, 3rd edn (Wiley,
comparison of model predictions with one dataset only New York, 1983) p. 437.
cannot be considered as a complete model validation. [16] K.B. Schnelle and P.R. Dey, Atmospheric Dispersion Modelling
Compliance Guide (McGraw-Hill, Europe, 1999).
[17] D.B. Turner, Workbook of Atmospheric Dispersion Estimates: An
Introduction to Dispersion Modelling (Lewis Publishers, Boca
References Raton, FL, 1994).
[18] U.S. EPA, Appendix H: Recommendations for Estimating Con-
[1] G.A. Briggs, Plume rise predictions, in: Lectures on Air Pollution centrations of Longer Averaging Periods from the Maximum One-
and Environmental Impact Analyses. Workshop Proceedings, Hour Concentration for Screening Purposes. U.S. Environmental
Boston, MA (American Meteorological Society, Boston, MA, Protection Agency, Research Triangle Park, NC, USA (2001).
1975) pp. 59Y111. [19] C.J. Willmott, On the validation of models, Phys. Geogr. 2 (1981)
[2] G.A. Briggs, Plume Rise, USAEC Critical Review Series, TID- 84Y194.
25075, Clearinghouse for Federal Scientific and Technical Infor- [20] P. Zannetti, Air Pollution Modelling (Van Nostrand-Reinhold, New
mation Materials of IAEA Meeting (1969). York, 1990) p. 444.

You might also like