Atmospheric Environment Vol. 32, No. 21, pp.

3775—3781, 1998
( 1998 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
PII: S1352–2310(98)00109–5 1352—2310/98 $19.00#0.00

ANALYSIS OF VARIOUS SCHEMES FOR THE ESTIMATION

OF ATMOSPHERIC STABILITY CLASSIFICATION
MANJU MOHAN* and T. A. SIDDIQUI-
Centre for Atmospheric Sciences, Indian Institute of Technology, New Delhi-110016, India

(First received 10 April 1997 and in final form 14 February 1998. Published August 1998)

Abstract—Atmospheric stability classification is required to quantify the dispersion capabilities of the

ambient atmosphere (i.e. the dispersion coefficients or the standard deviation of concentration distribution
in lateral and vertical directions, p and p , respectively) in the air quality models for concentration
predictions. Several different types of ystabilityzclassification schemes are given depending on the availability
of meteorological parameters and the related atmospheric processes in the lower part of the boundary layer.
In the present study, atmospheric stabilities are estimated from seven different stability classification
schemes based on Monin—Obukhov length, bulk and gradient Richardson numbers, temperature gradient,
wind speed ratio, etc. The estimated stabilities are compared with the Pasquill stability classification which
is very widely used and requires only routine data. The spread of stabilities in each scheme is large and
sometimes it covers the entire range of stabilities for a given Pasquill class. However, it has been observed
that the schemes based on Monin—Obukhov length and Richardson number gives reasonable comparison
than the rest of the schemes. The performance of each of these schemes has been discussed based on the
underlying physics. The study demonstrates that for a complete and exhaustive data set such as Kincaid,
there exists a wide variations among all atmospheric stabilities with Pasquill and one another in a manner
such that an entire range of stabilities may be obtained for a given Pasquill class. The implications of this on
estimated distance to maximum ground level concentrations are great and could affect the overall outcome
of conventional Gaussian models for industrial siting, planning and management. ( 1998 Elsevier Science
Ltd. All rights reserved

Key word index: Atmospheric stability, mechanical turbulence, convective turbulence, dispersion coeffi-
cients, Monin—Obukhov length, Richardson number, temperature gradient, wind speed ratio, ground level
concentrations.

1. INTRODUCTION and unimpeded mixing takes place underneath.

A modification in many of the recent models is that
Atmospheric stability is commonly evoked in many initial penetration of the lid from below is allowed.
dispersion studies as the single parameter used to In several models, this is also involved in the
define the turbulent state of the atmosphere or to parameterisation of the dispersion coefficients. How-
describe the dispersion capabilities of the atmosphere. ever, the focus of attention in the present study
For many applications the details of the turbulent is the various schemes for atmospheric stability
boundary layer are not known and recourse is made classification which define the turbulent state of
to simple categorisations of atmospheric conditions in the atmosphere and also reflect upon the dispersion
terms of different stabilities to broadly classify the capabilities of the atmosphere.
boundary layer’s dispersive properties. In the popular There are mainly six atmospheric stabilities desig-
Gaussian models, an index of atmospheric stability nated as A (highly unstable or convective), B (moder-
usually determines the functional form of the various ately unstable), C (slightly unstable), D (neutral),
algorithm which in turn calculates the concentration E (moderately stable), and F (extremely stable). Later,
of pollutants. The other important parameter which stability G is also included to represent low wind
has significant role to play in atmospheric dispersion nighttime stable conditions (Table 1). Historically,
is boundary layer depth. This is regarded as a lid these stability classes were proposed by Pasquill and
where no transport of pollutants occurs across the lid, Gifford and further improved by Turner leading to
the general use of stability indices 1—6 (corresponding
to stability classes A—F) known as Pasquill—Gifford—
Turner (PGT) classification of stabilities used in many
* Author to whom correspondence should be addressed.
- Present address: Energy & Environment Dept, Engin- plume models (Turner, 1970). A number of atmo-
eers India Ltd, R&D Centre, Sector-16, Gurgaon-122 001, spheric stability classification schemes have been
India. developed to assess the six PGT stability indices in
3775
3776 M. MOHAN and T. A. SIDDIQUI

Table 1. Modified Pasquill stability classes

Wind Daytime! incoming solar radiation (W m~2) Within 1 h Nighttime cloud amount (oktas)
speed before sunset
(m s~1) (m s~1) Moderate Slight ((300) Overcast or after sunrise" 0—3 4—7 8
strong (300—600)
('600)

)2.0 A A—B B C D F or G# F D
2.0—3.0 A—B B C C D F E D
3.0—5.0 B B—C C C D E D D
5.0—6.0 C C—D D D D D D D
'6.0 C D D D D D D D

! (Ref. Davies and Singh, 1985; Journal of Hazardous Materials 11, 1985). Excluding 1 h after sunrise and 1 h before sunset.
" Night was originally defined to include periods of 1 h before sunset and after sunrise. These 2 h are always categorised here
as D.
# Pasquill said that in light winds on clear nights the vertical spread may be less than for category F but excluded such cases
because the surface plume is unlikely to have any definable travel. However, they are important from the point of view of the
build up of pollution and category G (nighttime, 0 or 1 okta of cloud, wind speed 0 or 0.5 m s~1) has been added.

terms of different meteorological parameters. In the (c) Temperature gradient method.

last two decades or so considerable efforts have been (d) Gradient Richardson number (Ri) Method.
made to classify the stability and examples of this (e) Bulk Richardson number (Ri ) Method.
B
work appear in Golder (1972), USNRC (1974), Smith (f ) Monin—Obukhov (M—O) length method.
(1979), Sedefian and Bennet (1980) and Tagliazucca (g) Wind speed ratio method.
and Nanni (1983). All the classification schemes divide
The following subsections describe the various classi-
the stability into six or seven categories depending
fication methods which are used in this work for
upon the types of observations used. The choice of
defining the stability.
a particular scheme may have a significant impact on
final outcome from plume modelling studies on atmo- 2.1. º , solar insolation/cloud cover
spheric dispersion since these schemes generally do 10
not lead to the similar classification. Thus, a compara- This method requires a measurement of wind speed
tive study of the various classification schemes in- at 10 m and a measurement or estimation of solar
vogue has been made on the basis of extensive insolation during daytime and an assessment of cloud
measurements conducted in the year 1980—1981 dur- cover during nighttime. The stability class can then be
ing Electric Power Research Institute’s (EPRI) Plume known from Table 1 which is the modified Pasquill
Model Validation and Development (PMV&D) pro- stability analysis scheme (Davies and Singh, 1985).
ject at Kincaid power plant situated at Illinois, USA The direct solar insolation can be estimated by the
(EPRI EA-3074, 1983). following well-known astronomical relation which in-
cludes attenuation from various species present in the
atmosphere (Schayes, 1982).
R "S Cos Z (q!A )N exp(!A /Cos Z ) (1)
2. ATMOSPHERIC STABILITY CLASSIFICATION 4 0 N u K N
where S is the solar constant, Z is the zenith angle
As defined above, stability is a term applied quali- 0 N
and q is the transmissivity coefficient in clear air
tatively to the property of the atmosphere which gov-
including scattering. A is absorption by water va-
erns the accelerations of the vertical motion of an air K
pour, N is the absorption of radiation by clouds. The
parcel. The acceleration is positive in unstable atmo-
empirical formulations adopted for the estimation of
sphere (turbulence increases), zero when the atmo-
attenuation characteristics is considered to be similar
sphere is neutral and negative (deceleration) when the
as given by Manju Mohan and Sharma (1988). By
atmosphere is stable (turbulence suppressed). Several
considering the geometry of the celestial sphere,
parameters such as Monin—Obukhov length (¸),
Z can be defined as
gradient and bulk Richardson numbers, standard de- N
viation of wind direction fluctuations, etc. are used in Cos Z "Sin / Sin d#Cos / Cos d Cos u (2)
N
micrometeorology to determine the degree of stability
or instability. The different stability schemes can where / is latitude, d is the declination angle and u is
broadly be classified in terms of the following: the hour angle.

2.2. p , standard deviation of the wind direction

(a) Wind speed at 10 m (º ), solar insolation/
10 h
cloud cover. fluctuation method
(b) Standard deviation of the horizontal wind di- This method of stability classification is one of the
rection fluctuation (p ) method. two recommended by the US Nuclear Regulatory
h
 Various schemes for atmospheric stability classification 3777

Commission (USNRC). Table 2 presents the state- and

ment of USNRC Correlation between p and Pasquill
h / "0.74#4.7Z/¸
stability following Sedefian and Bennet (1980). ) for Z/¸'0 (6)
/ "1#4.7Z/¸
.
2.3. ¹emperature gradient (*¹) method
/ and / in the above equations can be estimated by
The observed temperature differences are used in ) .
using the graph of Golder (1972) which relates Pas-
the format °C 100 m~1 i.e. the environmental lapse
quill stabilities to M—O length and the roughness
rate. The temperature gradient limits associated with
length, Z . Once / and / are calculated for different
each of the Pasquill stability class are represented in 0 ) .
values of ¸; equation (4) could be used to prescribe the
Table 2 (Sedefian and Bennett, 1980). This scheme
limits of Ri associated with each of the Pasquill stabil-
requires the temperature measurements at two levels.
ity categories.
2.4. Richardson number (Ri) method Roughness length for the present data set has been
taken as 0.1 m based on measurements. The values
The wind speed and temperature measurements at
calculated in this manner are presented in Table 3.
the two heights are used to calculate the gradient
Richardson number from the following relation:
2.5. Bulk Richardson number (Ri ) method
g(*h/*Z) B
Ri" (3) The bulk Richardson number is defined as
¹(*u/*Z)2
here h is potential temperature and u is the wind speed, g(*h/*Z)ZM 2
Ri " (7)
¹ is screen temperature and g the acceleration due to B ¹uN 2
gravity. The limits which are used to classify Richar-
dson number with different stabilities are calculated where ZM is usually taken to be the geometric mean
from the relation height [(Z Z )1@2"22 m]. Here Z is 10 m, Z is
1 2 1 2
50 m and u is the mean wind speed at the upper level.
(Z/¸/ )
Ri" ) (4) Ri can now be determined using Ri with the set of
B
/2 limits calculated in equation (4) by the following rela-
.
where / and / are nondimensional temperature tionship:
) .
and wind profiles, respectively, and given by Businger
RiZM 2(*uN )2 Ri/2
(1973) as Ri " " .
B uN 2(*Z)2 [ln(Z/Z )!t]2
/ "0.74(1!9Z/¸)1@2 0
) for Z/¸(0 (5)
/ "(1!15Z/¸)1@4 / "0.74#4.7Z/¸ (8)
. )

Table 2. p , d¹/dZ and º* limits for stability classification

h R
Passquill stability p (°) d¹/dZ (°C/100 m) º
h R
A 22.5(p d¹/dZ(!19 º (1.186
h R
B 17.5(p )22.5 !1.9)d¹/dZ(!1.7 1.186)º (1.207
h R
C 12.5(p )17.5 !1.7)d¹/dZ(!1.5 1.207)º (1.258
h R
D 7.5(p )12.5 !1.5)d¹/dZ(!0.5 1.258)º (1.59
h R
E 3.75(p )7.5 !0.5)d¹/dZ(!1.5 1.59)º (2.29
h R
F 2.0(p )3.75 1.5)d¹/dZ(4.0 2.29)º (3.0
h R
G p )2.0 d¹/dZ*4.0 º *3.0
h R
Note. º* : Limits have been estimated for observation site at Kincaid (with roughness length 0.1 m).
R

Table 3. Ri and Ri limits for stability classification estimated for the observation site at Kincaid with roughness
B
length (0.1 m)

Stability Richardson number Bulk Richardson

class number formulation
Businger Businger—Hick
formulation formulation

A Ri(!2.038 Ri(!5.34 Ri(!0.023

B !2.038)Ri(!0.75 !5.34)Ri(!2.26 !0.023)Ri (!0.011
B
C !0.75)Ri(!0.18 !2.26)Ri(!0.569 !0.011)Ri (!0.036
B
D !0.18)Ri(0.083 !0.569)Ri(0.083 !0.036)Ri (0.0072
B
E 0.083)Ri(0.16 0.083)Ri(0.196 0.0072)Ri (0.42
B
F 0.16)Ri(0.18 0.196)Ri(0.49 0.042)Ri (0.084
B
G Ri*0.18 Ri*0.49 Ri *0.84
B
3778 M. MOHAN and T. A. SIDDIQUI

where 1616-8, 1984). The field data were collected around

t"!4.7Z/¸ for Z/¸'0 Kincaid Power Plant in the central Illinois, approxim-
ately 40 km southeast of Springfields, USA. The
1#/~1 l#/~2
t"2 ln . #ln . Kincaid power plant is surrounded by flat farmland
2 2 with some lakes and terrain at elevation of 180 m
n a.m.s.l. The roughness length is approximately 0.1 m
!2 tan~1(/~1)# for Z/¸(0. (9)
. 2 (EPRI EA-3077, 1984). Apart from the several EPRI
reports the data have been described and extensively
During unstable conditions the Businger’s formula-
used by several authors (Hanna and Chang, 1992;
tion is used as in equation (8); however, Hicks (as
Hanna and Paine, 1989 and Turner et al., 1991).
given by Carson and Richards, 1978) formulation is
The instrumentation involved had fixed and mobile
used for the stable case which is
aerometric stations, 100 m meteorological tower,
4.25 1 acoustic Doppler sounder, double-theodolite T-sonde
/~8" # (10)
. Z/¸ (Z/¸)2 system, routine weather observations, etc. The
measurements included, among others, the para-
which leads to meters such as temperatures at 10, 50 and 100 m; wind
4.25 1 speed, wind direction and r at 10, 30, 50 and 100 m;
t(Z/¸)"!7 ln(Z/¸) # #0.7. (11) h
Z/¸ 2(Z/¸)2 surface pressure, relative humidity, net radiation,
cloud cover, mixing height, etc.
This form of / is applicable for Z/¸*0.5. Table 3 In the present study, in all 188 hrs of observations
.
gives the set of limits for all the Pasquill stabilities were considered. The height levels for estimating
with respect of Richardson number (Businger and gradients in the computations are taken as 10 and
Businger—Hicks formulation) and bulk Richardson 50 m which are similar to the ones used in an earlier
number. study by Sedefian and Bennet (1980).

2.6. Monin—Obukhov length (¸) method

The Monin—Obukhov length is usually defined by 4. ANALYSIS OF RESULTS FROM DIFFERENT STABILITY

¸"!oC ¹º3 /kgH (12) CLASSIFICATION SCHEMES

1 *
where o is the density of air at temperature ¹, C is The atmospheric stability has been assessed from
1
specific heat capacity at constant pressure, º is the seven different schemes namely, Pasquill, M—O
*
friction velocity, k is Von-Karman constant and H is length, p (10 m), Gradient Richardson number, Bulk
h
the sensible heat flux. The estimate of stability class in Richardson number, temperature gradient and
terms of ¸ and Z has been taken from Golder (1972). º scheme. For gradient Richardson number, two
0 R
different classification schemes are used based on
2.7. ¼ind speed ratio (º ) method Businger and Businger—Hicks formulation for / and
R .
The ratio of the 50 and 10 m wind speeds is used / . As the Businger and Businger—Hicks formulations
)
here which can be represented by a power-law form are similar during unstable conditions, the difference
in results are seen only during stable conditions. As
º .
º 50 "(Z /Z )PU "(50/10)PU (13) mentioned earlier, the field experiments were conduc-
Rº . 2 1
10 ted from March 1980 to June 1981 by EPRI. In all
here P and thus º for each stability class is deter- 188 h of data have been utilised from four different
6 R months, viz., April, May, June and July. The hourly
mined by the following relation (Panofsky and
Prasad, 1965) classification of atmospheric stability from the above-
/ mentioned schemes has been compared with the
P" . (14) Pasquill scheme as this is among the first known
6 ln(ZM /Z )t
0 classifications and still widely used. The percentage
which fits power-law formulae to wind profiles well. variation of rest of the stability schemes with resect to
The Businger formula for / was used in these calcu- Pasquill A—G is shown in Tables 4(a—g). The distribu-
.
lations. Table 2 gives a set of limit of P for each tion of total number of observations in each Pasquill
U
stability class. stability class, viz. A, B, C, D, E, F and G is 20, 36, 46,
46, 11, 25 and 4, respectively.
In general, all schemes show a wide variation of
3. DATA FOR EVALUATION stabilities in comparison to a prescribed stability from
the Pasquill scheme. The variation generally goes
The data set used for the stability classification is from one to two categories lower or higher than the
taken from the field experiments conducted by EPRI, Pasquill class and often spreads over the entire range
USA from March 1980 to June 1981 under the Field of stabilities from A—G.
Measurement Plan (FMP) for Plume Model Valida- In general, the M—O length scheme shows the best
tion & Development (EA-3064 Research Project comparison among all the schemes except for Pasquill
 Various schemes for atmospheric stability classification 3779

Table 4a. M—O Length Scheme (%) Table 4f. Bulk Richardson No. Scheme (%)

A B C D E F G A B C D E F G

A 85 15 — — — — — A 100 — — — — — —
B 41.7 38.9 19.4 — — — — B 91.7 2.7 — — 2.8 — 2.8
C 6.5 84.8 8.7 — — — — C 73.9 15.2 4.3 4.3 — — 2.2
D — 6.5 19.6 54.3 8.7 4.3 6.5 D 21.7 39.1 6.5 4.3 10.9 6.5 10.9
E — — — 9.1 45.5 18.2 27.3 E — — — — 9.1 36.4 56.4
F — — — — 16 4 80 F — — — — — — 100
G — — — — — — 100 G — — — — — — 100

Table 4b. º Scheme (%) Table 4g. p (10 m) Scheme (%)

R h
A B C D E F G A B C D E F G

A 80 5 15 — — — — A 65 — 10 20 — 5 —
B 41.7 13.9 13.9 27.8 — 2.8 — B 41.7 16.7 5.6 8.3 13.9 5.6 8.3
C 37 2.2 23.9 32.6 4.3 — — C 2.2 2.2 13 34.8 15.2 19.6 13
D 28.3 4.3 15.2 32.6 10.9 6.5 — D 16.9 4.3 6.5 17.4 23.9 10.9 21.6
E 18.2 — — 18.2 27.3 9.1 18.2 E 18.2 — — 18.2 19.1 18.2 36.4
F 4 — — 20 48 24 4 F 4 20 8 20 28 8 12
G — — — — 25 75 — G 50 — — 25 — 25 —

Table 4c. d¹/dZ Scheme (%) due to there being a well-mixed boundary layer dur-
ing highly convective conditions. However, as the
A B C D E F G
stability increases (this in the entire text means that
A 15 30 20 35 — — — the stability increases from unstable to neutral to
B 33.3 11.1 11.1 33.3 8.3 — 2.8 stable i.e. A—G), the decoupling of the surface layer
C 10.9 10.9 21.7 45.7 8.7 — 2.2 from the layer above takes place. Also, º is more
D 4.3 2.2 6.5 47.8 30.4 4.3 4.3 R
E — — — 9.1 9.1 27.3 54.5 directly associated with mechanically generated tur-
F — — — — 20 28 52 bulence; the converse is true for the d¹/dZ scheme,
G — — — — — — 100 which would be associated more with thermally gen-
erated turbulence. In comparison to the above two
schemes, the scheme based on Ri and Ri shows better
Table 4d. Ri Scheme: Businger formulation (%) B
performance which may be attributed to the fact that
A B C D E F G these represent ratio of mechanical-to-convective tur-
bulence (Table 4d—f ). Though less, the variations do
A 95 5 — — — — — exist here both for Ri and Ri which are more pro-
B 50 30.6 13.9 — — — 5.6
B
nounced during stable conditions in comparison to
C 8.7 32.6 39.1 17.4 — 2.2 — convective cases which could again be explained due
D — 15.2 26.1 41.3 8.7 2.2 6.5
E — — — 18.2 27.3 18.2 36.4 to the decoupling of layer and reduced turbulence
F — — — — 8 4 88 during stable stratification in comparison to convec-
G — — — — — — 100 tive conditions. However, as Ri and Ri are functions
B
of height, it might be improper to use a single value of
these to represent the turbulent characteristics of the
Table 4e. Ri Scheme: Businger—Hicks Formulation (%) whole layer in question. In addition, different forms of
/ and / could be used (Yaglom, 1976). As men-
A B C D E F G . )
tioned earlier, two different formulations of / and
.
A 95 5 — — — — — / are used during stable situations. Here the results
B 50 30.6 13.9 — — 2.8 2.8 )
indicate better performance of Businger—Hicks for-
C 8.7 32.6 39.1 17.4 — 2.2 —
D 2.2 15.2 23.9 41.3 10.9 2.2 4.3
mulation in comparison to Businger. A comparison
E — — — 18.2 36.4 9.1 36.4 between Ri and Ri methods reveal a better perfor-
B
F — — — — 12 28 60 mance of Ri method for all stabilities in comparison to
G — — — — — — 100 Ri method. Considering the physics and the formula-
B
tion of the two, no significant difference is found
except that the mean wind at the upper level is used in
stability F where the tendency has been to categorise Ri whereas wind speed gradient is used in Ri. It is
B
one stability higher i.e. stability G (Table 4a). The thought that the later may represent the surface
º method as well as the temperature gradient boundary layer in a better manner.
R
methods show poor correlation with Pasquill schemes Pasquill stability has also been compared with the
except for stability A (Table 4b and c). This may be p (10 m) scheme (Table 4g). In fact the p method has
h h
3780 M. MOHAN and T. A. SIDDIQUI

been considered as the basis of comparison with vari- of the other schemes mentioned above. The plausible
ous other methods in an earlier study by Sedefian and reasons could be that this classification may vary from
Bannett (1978). This is because the direct determina- one site to other and the involvement of only a single-
tion of dispersion coefficients p (lateral) and p (verti- level measurement, which may not necessarily repres-
y z
cal) could be based on p and p which are standard ent the characteristics of the entire surface boundary
h e
deviation of lateral and inclination angles of wind layer. It may be argued here that p could be corre-
h
direction, respectively. However, as shown in Table 4, lated well empirically with the lateral dispersion para-
the p method shows a poor comparison with the meters, but, this may not be true to the same extent
h
Pasquill as well as with the rest other schemes (viz. with the atmospheric stability which may partly be
M—O length, Ri and Ri ). Also it is to be pointed out explained from the above-cited reasons. There may,
B
that the p classification may vary from place to place however be some unexplained reasons to this needing
h
and be influenced by local topographical and climatic further study in this direction.
features. This, together with the fact that single-level As expected, large variations often covering the
and unidirectional representation of turbulence in the entire range of stabilities have serious implications
form of p may not incorporate the dynamics of the while predicting the downwind distance to maximum
h
whole mixing layer, could explain the poor compari- ground level concentration (glc) as well as the max-
son of this scheme with the Pasquill scheme in com- imum glc itself. An approximate estimate of this could
parison to others. be obtained from the graphical format developed by
For stability A, all schemes compare extremely well Turner (1970) based on Gaussian plume formulation.
(80% and above) except for p and temperature gradi- Distance to maximum downwind concentration for
h
ent scheme. For stability B the best comparison is an effective stack height of 100 m based on above
with ¸ scheme followed by Ri. For stability C, the best formulation is approximated as 0.48, 0.78, 1.4, 5.4, 6.2
comparison is with scheme ¸ followed by Ri,º and and 14 km for stabilities A—F, respectively. By further
R
d¹/dZ schemes. p and Ri schemes show poor com- increasing the effective stack height, the above vari-
h B
parison. For stability D, again the best comparison is ation would also be enhanced further. As this distance
with ¸ scheme followed by d¹/dZ, Ri and is an important parameter for industrial siting and
º schemes. Again, the p and Ri schemes show poor designing of optimal stack height, etc., the above vari-
R h B
comparison. In case of stability E, the ¸ scheme and ations cannot be overlooked. Thus, based on the
the Ri (Businger—Hicks) schemes perform well while present study it could be demonstrated that at least
B
for stability F, Ri scheme based on Businger—Hicks for one data set which is detailed and extensive, there
formulation, d¹/dZ scheme followed by º scheme exists a great deal of discrepancies in atmospheric
R
compares reasonably well. It is to be pointed out that stability classifications from various widely used
¸ scheme shows stability G (80%), for Pasquill class schemes which would significantly affect the concen-
F, which may be attributed to the fact that the classi- tration pattern based on air quality models. It is quite
fication adopted from Golder (1972) as such does not likely that the other data sets may reveal the similar
distinguish between F&G stabilities. Stability G com- trend.
pares well with ¸, Ri and Ri schemes. The rest of the
B
schemes always show higher stabilities. As far as dis-
persion is concerned, the great deal of meandering 5. CONCLUSIONS
during low-wind conditions represented by stability
Atmospheric stability has been estimated from
G should lead to relatively large spread and in turn
seven different schemes invogue based on the exten-
the larger than expected values of dispersion coeffi-
sive field experiments at Kincaid, U.S.A. and the per-
cients. In general, schemes based on ¸ and Richar-
formance of each of these schemes against Pasquill
dson number (gradient and bulk) perform well while
classification has been explained on the basis of under-
comparing with Pasquill classification. Except for
lying physics. The wide variation of stabilities, very
stability D, the performance of d¹/dZ scheme is un-
often covering the entire range, has serious implica-
satisfactory when compared with Pasquill classifica-
tions in terms of concentration predictions based on
tion. The º scheme shows good comparison for
R Gaussian or related air quality models and it is sug-
stability A (80%) while for other stabilities the com-
gested that appropriate choice be made in selecting
parison is not so satisfactory. The p method shows
h a particular methodology for this purpose.
somewhat reasonable comparison only during stabil-
ity A (65%).
From the above discussion it could be inferred that REFERENCES
the schemes which are represented by either the mech-
anical turbulence, viz. º or the thermal turbulence, Berkowicz, R. and Prahm, L. P. (1982) Evaluation of the
R
viz. d¹/dZ would perform poorly in comparison to profile method for estimation of surface fluxes of
the other schemes which are represented by both momentum and heat. Atmospheric Environment 16,
2809—2819.
thermal as well as mechanical turbulence (viz. ¸, Ri, Businger, J. A. (1973) Turbulent transfer in the atmospheric
Ri ). Contrary to the general expectations the surface layer. Workshop on Micrometeorology, American
B
p scheme does not give a good comparison as some Meteorological Society 67—120.
h
 Various schemes for atmospheric stability classification 3781

Carson, D. J. and Richards, J. R. (1978) Modelling surface Sedefian, L. and Bennet, E. (1980) A comparison of turbu-
turbulent fluxes in stable conditions. Boundary ¸ayer Met- lence classification schemes, Atmospheric Environment 14,
eorology 14, 67—81. 741—750.
Davies, M. E. and Singh, S. (1985) Thorney Island: its Smith, F. B. (1979) The relation between Pasquill stability
geography and meteorology. Journal of Hazardous Mater- and Kazanski-Monin stability (in neutral and unstable
ials 11, 91—124. conditions), Atmospheric Environment 13, 879—881.
EPRI EA-3064 Research Project 1616-8 (1984) Plume Model Smith, F. B. (1990) Atmospheric structure. Proceedings of Air
Validation and Development Field Measurements-Plains Pollution Modelling for Environmental Impact Assessment,
site. International Centre for Theoretical Physics, Tieste, June,
EPRI EA-3077 Research Project 1616-9 (1984) Diagnostic 1990.
Validation of Plume Models at a Plain site. Tagliazucca, M. and Nanni, T. (1983) An atmospheric diffu-
Golder, D. (1972) Relation among stability parameter in the sion classification scheme based on the Kazanski—Monin
surface layer., Boundary ¸ayer Meteorology 13, 47—58. stability parameter, Atmospheric Environment 17(11),
Hanna, S. R. and Chang, J. C. (1992) Boundary layer para- 2205—2211.
meterisations for applied dispersion modelling over urban Turner, D. B. (1970) Workbook of atmospheric dispersion
areas. Boundary ¸ayer Meteorology 58, 229—259. estimates. Office of Air Program Pub. No. AP-26, Envir-
Hanna, S. R. and Paine, R. J. (1989) Hybrid Plume Disper- onmental Protection Agency, USA.
sion Model (HPDM) development and evaluation. Jour- Turner, D. B., Bender, L. W., Paumier, J. O. and Boone, P. F.
nal of Applied Meteorology 28, 206—224. (1991) Evaluation of the TUPOS air quality dispersion
Mohan, M. and Sharma O. P. (1988) A model for ambient air model using data from the EPRI Kincaid field study.
quality analysis using routine meteorological observa- Atmospheric Environment 25A, 2187—2207.
tions, Atmospheric Environment 22, 651—661. US Nuclear Regulatory Commission (1974), On site me-
Panofsky, H. A. and Prasad, N. B. (1965) Similarity theory teorological program, Regulatory Guide 1.23, 6pp.
and diffusion. Intl. Journal of Air ¼ater Pollution 9, Yaglom, A. M. (1970) Comments on wind and temperature
419—430. flux profile relationships. Boundary ¸ayer Meteorology 11,
Pasquill, F. (1974) Atmospheric Diffusion. 2nd ed. New York, 89—102.
Halsted Press. Zapert, J. G. (1991) Evaluation of dense gas simulation
Schayes, G. (1982) Direct determination of diffusivity profiles models, EPA report no. 450/4-90-018, Office of Air Quality
from synoptic reports. Atmospheric Environment 14, Planning and Standards, ºS EPA, Research Triangle Park,
741—750. NC, USA.