Atmospheric Environment 34 (2000) 3575}3583
Turbulence parameterisation for PBL dispersion models
in all stability conditions
G.A. Degrazia!,*, D. Anfossi", J.C. Carvalho#, C. Mangia$,
T. Tirabassi%, H.F. Campos Velho&
!Departamento de Fn& sica, Universidade Federal de Santa Maria, Santa Maria, Brazil
"CNR, Istituto di Cosmogeoxsica, Torino, Italy
#Departamento de CieL ncias Atmosfe& ricas, Universidade de SaJ o Paulo, Instituto AstronoL mico e Geofn& sico, SaJ o Paulo, Brazil
$CNR, ISIATA, Lecce, Italy
%CNR, ISAO, Bologna, Italy
&Instituto Nacional de Pesquisas Espaciais, LAC, SaJ o Jose& dos Campos, Brazil
Received 2 September 1999; received in revised form 5 December 1999; accepted 13 January 2000
Abstract
Accounting for the current knowledge of the planetary boundary layer (PBL) structure and characteristics, a new set of
turbulence parameterisations to be used in atmospheric dispersion models has been derived. That is, expressions for the
vertical pro"les of the Lagrangian length scale l and time scale ¹ and di!usion coe$cient K , i"u, v, w, are proposed.
i
i
i
The classical statistical di!usion theory, the observed spectral properties and observed characteristics of energy
containing eddies are used to estimate these parameters. The results of this new method are shown to agree with
previously determined parameterisations. In addition, these parameterisations give continuous values for the PBL at all
elevations (z )z)h, z ) and all stability conditions from unstable to stable, where h and z are the turbulent heights in
0
i
i
stable or neutral and convective PBL, respectively, and ¸ is the Monin}Obukhov length. It is the aim of this work to
present the general derivations of these expressions and to show how they compare to previous results. Finally,
a validation of the present parameterisation applied in a Lagrangian particle model, will be shown. The Copenhagen data
set is simulated. This data set is particularly suited for this validation, since most of the Copenhagen tracer experiments
were performed in stability conditions that are the result of the relative combination of wind shear and buoyancy forces.
As a consequence, a parameterisation scheme, able to deal contemporary with neutral and slightly convective condition,
is to be preferred. ( 2000 Elsevier Science Ltd. All rights reserved.
Keywords: Planetary boundary layer; Turbulence parameterisation; Statistical di!usion theory; Turbulent velocity spectra; Dispersion
models
1. Introduction
In the atmospheric dispersion models turbulence parameterisation is a key parameter. The reliability of each
model strongly depends on the way turbulent parameters
are calculated and related to the current understanding
of the planetary boundary layer (PBL). Most of the
turbulence parameterisation used in advanced dispersion
* Corresponding author.
models are based on PBL similarity theories (Hanna,
1982; Stull, 1988; Holtslag and Moeng, 1991; Kaimal and
Finnigan, 1994; Sun, 1993; Rodean, 1994): the turbulence
parameterisation is directly related to the basic physical
parameters describing the turbulence state of the PBL.
Throughout classical statistical di!usion theory
(Batchelor, 1949), it is possible to relate turbulent parameters to spectral distribution of turbulent kinetic energy
(TKE). Following this approach, Degrazia and Moraes
(1992), Degrazia et al. (1997, 1998) developed a model
for the turbulent spectra in both a purely shear and
1352-2310/00/$ - see front matter ( 2000 Elsevier Science Ltd. All rights reserved.
PII: S 1 3 5 2 - 2 3 1 0 ( 0 0 ) 0 0 1 1 6 - 3
3576
G.A. Degrazia et al. / Atmospheric Environment 34 (2000) 3575}3583
buoyancy dominated PBL, and proposed a formulation
for turbulent parameters to be used in air quality dispersion models in these extreme turbulent regimes.
However, in many real cases, dispersion of atmospheric
e%uents occurs in an intermediate PBL that is driven by
both shear and convective forcing.
The aim of this study is to use Taylor's di!usion theory
together with a model for turbulent spectra in a shearbuoyancy driven PBL, in order to generate a turbulence
parameterisation that gives continuous values for the
PBL at all elevations and all stability conditions from
unstable to stable. The very stable cases are excluded
since the Monin}Obukhov scaling does not apply in
these extreme conditions (Mahrt, 1999; Gryning, 1999).
Expressions for the vertical pro"les of the Lagrangian
length and decorrelation time scales and eddy
di!usivity are proposed to be used in atmospheric dispersion models. Employing simulations with a stochastic
Lagrangian model of turbulent di!usion, the new turbulent parameterisation is evaluated against ground-level
concentration data from atmospheric dispersion experiments that were carried out in the Copenhagen area
under neutral to unstable conditions (Gryning and Lyck,
1984, 1998) and is compared with Hanna (1982) parameterisation.
2. Analytical derivation of the Lagrangian length scale,
decorrelation time scale and eddy di4usivity
Part of this derivation was already proposed by
Degrazia et al. (1998) and is resumed here for sake of
clarity of presentation. According to Weil (1989), the
dispersion rate of a particle ensemble in a turbulent #ow
can be expressed, for large travel time (tA¹ ), as
i
A B P
d x2
=
i "p2 o (t) dt"p2¹ ,
i
i i
i
dt 2
0
(1)
where i"u, v, w, x2 is the variance of particle positions,
i
p2 corresponds to the Lagrangian variance of the ith
i
component of the turbulent wind "eld, o (t) is the nori
malised Lagrangian autocorrelation function and ¹ is
i
the Lagrangian decorrelation time scale. Concerning the
horizontal ¹ components, the larger scales due to the
i
meandering (scales greater than the spectral gap of Van
der Hoven wind velocity spectrum (Stull, 1988)) are not
included.
Following Tennekes and Lumley (1972), the Lagrangian length scale l can be de"ned in terms of the Lagrani
gian variance and decorrelation time scale, namely
l "p ¹ .
(2)
i
i i
On the basis of Taylor's theory, Batchelor (1949) (see
also Degrazia and Moraes, 1992) proposed the following
relationships for the rate of dispersion for an ensemble of
particles for any travel time t:
A B
P
d x2
p2b = FE(n) sin(2pnt/b )
i " i i
i
i dn,
(3)
dt 2
2p
n
0
where b is the ratio of the Lagrangian to the Eulerian
i
time scales, FE(n) is the Eulerian spectrum of energy
i
normalised by the velocity variance and n is the frequency. Wandel and Kofoed-Hansen (1962), Angell
(1974), Pasquill (1974, p. 89) and Hanna (1981) have
suggested the following expression for b :
i
;
b "d ,
(4)
i
p
i
where ; is the mean wind speed and d is a constant
whose numerical value is given by a large number of both
experimental and theoretical works. In this paper, we use
the value d"0.55 estimated by Degrazia and Anfossi
(1998).
The "lter spectrum in Eq. (3) has its major passband
around zero frequency. In this case, the eddy di!usivity
for large times depends on the behaviour of the spectrum
near the origin, so that the concept of "ltering applied to
a one-dimensional turbulent velocity "eld allows to select
the energy-containing eddies (Degrazia and Moraes,
1992; Degrazia et al., 1996, 1997). This means that the
"lter chooses the characteristic frequency (nP0) describing these eddies. As a consequence, the rate of dispersion
becomes independent of the travel time from the source
and can be expressed as a function of local properties of
turbulence, as follows:
A B
d x2
p2b FE(0)
i " i i i .
dt 2
4
(5)
Therefore, with this information we can construct and
calculate various fundamental parameters (eddy diffusivities and Lagrangian length and time scales) associated to energy-containing eddies and that describe the
turbulent transport process in the PBL.
From Eqs. (1), (2) and (5), the Lagrangian length scale
can be expressed as
b p FE(0)
l" i i i
i
4
(6)
that yields a Lagrangian decorrelation time scale given
by
b FE(0)
l
(7)
¹" i" i i .
i p
4
i
It is worthwhile to point out the bene"ts of using the
parameterisation given by Eq. (5): Taylor's theory is valid
for homogeneous turbulence only, whereas Eqs. (5)}(7),
referring to the eddy di!usivity, Lagrangian length scale
and Lagrangian decorrelation time scale, respectively,
can be used in non-homogeneous turbulence as well, thus
resulting in more general application.
G.A. Degrazia et al. / Atmospheric Environment 34 (2000) 3575}3583
3. Turbulent velocity spectra in a shear/buoyancy
driven PBL
It is well known that turbulent dispersion in the PBL is
generated by two di!erent processes: mechanical and
thermal. The former is related to wind shear, and it is
most e!ective close to the ground. The latter results from
a buoyancy forcing mechanism and it is generally responsible for convective transport of momentum, heat or
other scalars. These two forcing mechanisms produce
a wide range of turbulent eddies and consequently a spectral distribution of turbulent kinetic energy (TKE) over
a broad range of scales.
The #ow patterns resulting from interactions among
shear-buoyancy turbulent eddies are quite complex, and
even if they fall between the extreme cases (purely mechanic or purely convective), they show structures which
are not present in either the two limiting cases (Moeng
and Sullivan, 1994).
An hypothesis of linear combination of the two mechanism can be assumed only when there is statistical independence between their Fourier components. This
happens when the energy containing wave-number
ranges are well apart for the two spectra. However, it has
been shown (H+jstrup, 1982; Berkowicz and Prahm,
1984; Moeng and Sullivan, 1994) that a simple TKE
model, based on neglecting the interaction e!ects between shear and buoyancy, can be regarded as a good
approximation for an intermediate PBL. Thus, assuming
the hypothesis of superposition of the e!ects produced by
two forcing mechanisms, we can write the dimensional
Eulerian spectra as
SE(n)"SE (n)#SE (n),
i
i"
i4
(8)
where the "rst term on the r.h.s. represents the buoyancy
production, the second one is the mechanical component,
the subscript b is for buoyancy, s for shear and i"u, v, w.
Following Degrazia et al. (1998) the dimensional convective turbulence spectra in the PBL can be written as
nSE (n)
1.06c f (t z/z )2@3
i" "
i
e i
,
w2
[( fH )#]5@3M1#1.5f/[( fH )#]N5@3
H
.i
.i
(9a)
where w is the convective velocity scale, f"nz/; is the
H
reduced frequency, ; is the mean wind speed in the
convective PBL, z is the height above the surface, ( fH )# is
.i
the reduced frequency of the convective spectral peak,
t "e z /w3 is the adimensional dissipation rate funce
" i H
tion, e is the buoyant rate of TKE dissipation and z is
"
i
the convective boundary layer (CBL) height.
The mechanical component of the dimensional spectrum is the same as the neutral limit of Degrazia and
Moraes (1992):
nSE (n)
1.5c f/2@3
i4 "
i e
.
u2
[( fH )n`s]5@3M1#1.5f5@3/[( fH )n`s]5@3N
H
.i
.i
(9b)
3577
where u is the friction velocity, the dissipation rate
H
/ "e kz/u3 is adimensionalized with surface layer scale
4
H
ing parameters, e is the mechanical rate of TKE dissipa4
tion, i is the von Karman constant and ( fH )n`s is the
.i
reduced frequency of the neutral or stable spectral peak.
For both Eqs. (9a) and (9b), c "a a (2pi)~2@3 and
i
i u
a "0.5$0.05 and a "1, 4, 4 for u, v and w compou
i
3 3
nents, respectively (Champagne et al., 1977; Sorbjan,
1989).
By analytically integrating the Eulerian spectra given
by Eqs. (9a) and (9b) over the whole frequency domain,
one can obtain the buoyant and mechanical wind velocity variances
P
p2 "
i"
=
1.06c t2@3w2 (z/z )2@3
i e H
i
SE (n) dn"
i"
[( fH )#]2@3
0
.i
(10a)
and
P
=
2.32c /2@3u2
i e H
(10b)
SE (n) dn"
i4
[( fH )#]2@3
0
.i
and, moreover, the total wind velocity variance (sum of
the two parts):
p2 "
i4
p2"a2w2 #b2u2 ,
(11)
i
i H
i H
where a and b are the buoyancy}force and shear}stress
i
i
terms, respectively, to be determined. In all these derivations it has been assumed that the Lagrangian and Eulerian variances of the turbulent wind "eld are equal. This
assumption is commonly made and is based upon the
fact that the turbulent kinetic energy is the same for both
approaches (Hanna, 1982, p. 177).
By means of the above hypothesis concerning the
statistical independence of convective and mechanical
spectra, we can handle each spectrum individually,
neglecting the interactions among shear}buoyancy
turbulent eddies. As a consequence, each turbulent
contribution to the normalised Eulerian spectra can be
scaled with the respective variance. Thus, we can write
SE
SE
(12)
FE(n)" i" # i4 .
i
p2
p2
i4
i"
Considering Eqs. (9a), (9b), (10a), (10b) and (12), the value
of the normalised Eulerian energy spectrum at the origin
can be given by
z
0.64z
FE(0)"
#
.
i
;( fH )# ;( fH )n`s
.i
.i
(13)
4. The turbulence parameterisation
It is the aim of this section to derive a turbulent
parameterisation giving continuous values for the PBL
at all elevations (z )z)h, z ) and all stability condi0
i
tions from unstable to stable. From Eqs. (4), (6) and (13)
3578
G.A. Degrazia et al. / Atmospheric Environment 34 (2000) 3575}3583
a Lagrangian length scale for all stability conditions and
that gives continuous values for all the PBL elevations
can be expressed as
GC
D
H
!¸ z
0.64
1@2 1
i
#
,
(14)
z !¸
( fH )# ( fH )n`s
i
.i
.i
where !¸/z is an average stability parameter for the
i
convective PBL. The term in square brackets has been
introduced in order to give a continuous transition from
the neutral to the convective conditions. From Eqs. (10a),
(10b) and (14) the eddy di!usivity K and Lagrangian
i
decorrelation time scale assume the following expressions, respectively:
l "0.14z
i
G
w ((!¸/z )z /¸)1@2(t z/z )1@3
i i
e i
H
[( fH )#]4@3
.i
/1@3u
e H
#
(15)
[( fH )n`s]4@3
.
K "p l "0.14Jc z
i
ii
i
H
and
G
z 0.14((!¸/z )z /!¸)1@2
l
i i
¹" i"
i p
[( fH )#]2@3w (t#z/z )1@3
Jc
i
.
i
H
e i
i
0.059
#
.
(16)
[( fH )n`s]2@3(/n`s)1@3u
.i
e
H
To construct turbulence parameterisations from Eqs.
(14)}(16) for PBL dispersion models it is necessary to
have expressions for w , u , t#, /n`s, ( fH )# and ( fH )n`s.
H H e e
.i
.i
For a convective PBL (t#)2@3+0.75 (Caughey and
e
Palmer, 1979; H+jstrup, 1982; Wilson, 1997), w "
H
(u ) (!z /i¸)1@3 and, recalling that ( f H )#"z/(j ) and
H0
i
.i
.i
that (j ) is the peak wavelength of the turbulent velocity
.i
spectra, ( f H)# expressions for i"u, v, w can be derived.
.i
According to Kaimal et al. (1976), Caughey (1982) and
Degrazia and Anfossi (1998),
H
(j ) "(j ) "1.5z
.u
.v
i
and
C
(17)
A
B
A BD
z
z
(j ) "1.8z 1!exp !4
!0.0003exp 8
.w
i
z
z
i
i
so that
A B
A BD
B
A
z
f z
( fH )n`s"( f )n 1#a # #3.7 ,
.i
. *4
iG
K
(21)
where ( f )n is the frequency of the spectral peak in the
. i4
surface for neutral conditions, G is the geostrophic wind
speed and f "10~4 s~1 is the Coriolis parameter. Ac#
cording to Olesen et al. (1984) and Sorbjan (1989),
( f )n "0.045, ( f )n "0.16 and ( f )n " 0.33. Fur. u4
. v4
. w4
thermore, a "500 (Hanna, 1968, 1981); as a consew
quence of the Blackadar (1962) mixing length hypothesis
(i.e., the asymptotic length scale l +G/f is limited
=
#
by a constant value, equal for all the components) we
found a "3889 and a "1094. Then, by writing
u
v
f z/G"[ f z/(u ) ][(u ) /G] where (u ) /G is the neu#
#
H0
H0
H0
tral geostrophic drag coe$cient, Eq. (21) results
A
B
A
B
f z (u )
z
H 0 #3.7 .
( f H )n`s"( f )n 1#a #
(22)
.i
. i4
i (u )
G
K
H0
By considering (u ) /G"0.03 (Tennekes, 1982; Hanna,
H0
1982), Eq. (22) can be written as
f z
z
( f H )n`s"( f )n 1#0.03a # #3.7 .
(23)
.i
. i4
i (u )
K
H0
It is to point out that the meaning of continuous parameterisation for all stabilities in the present paper is the
following. At the same time stable and unstable conditions cannot coexist. Unstable and neutral e!ects may be
jointly take into account due to the contemporary presence of mechanical and convective turbulence. Stable and
neutral e!ects may also jointly be accounted for because
of the competition between wind-shear generated turbulence and stabilising e!ects of strati"cation. Therefore, in
unstable-neutral conditions the last term of Eq. (21) is set
equal to zero, whereas in the neutral-stable conditions
the term 1/( f H )# is set to zero.
.i
(18)
( f H)#"z/(B z )
(19)
.i
i i
with B " B " 1.5 and
u
v
!4z
8z
B "1.8 1!exp
!0.0003 exp
.
(20)
w
z
z
i
i
For a neutral or stable PBL, /n`s can be written
e
(Sorbjan, 1989) as /n`s"/n(1#3.7z/K), where
e
/n"1.25 and K"¸(1!z/h)(1.5a1 ~a2 ) (Degrazia and
e
Moraes, 1992) is the local Monin}Obukhov length. For
a shear-dominated stable boundary layer, a "1.5 and
1
a "1.0 (Nieuwstadt, 1984). Furthermore, for a neutral
2
C
or stable PBL, u2 "(u2 ) (1!z/h)a1 in which a "1.7
1
H
H0
for the neutral case (Wyngaard et al., 1974). Then, following Delage (1974), Stull (1988) and Sorbjan (1989), it is
obtained
5. Comparison with previous parameterisations
For horizontal homogeneity the PBL dynamics is
driven mainly by the vertical turbulent transport. Therefore, the present analysis will focus on the vertical exchange turbulent parameters.
In convective conditions (¸(0), the vertical Lagrangian length scale can be derived from Eqs. (14), (19) and
(20) as
A
B C
A BD
A B
z
1@2
4z
l# "0.25z 0.01 i
1!exp !
8
i
!¸
z
i
8z
!0.0003 exp
,
z
i
(24)
G.A. Degrazia et al. / Atmospheric Environment 34 (2000) 3575}3583
in which a typical value !¸/z "0.01 was used. In very
i
unstable conditions (50(!z /¸(300; Weil, 1988),
i
this value yields a Lagrangian length scale similar to the
mixing length proposed by Sun and Chang (1986) and
Sun (1993), whereas in weak to moderate convection the
magnitude of l# is smaller.
w
When the shear production is the only source of the
TKE (neutral boundary layer or shear-dominated stable
boundary layer, ¸'0), the vertical Lagrangian length
scale can be derived from Eqs. (14) and (21) by assuming
a "500 and ( f H )n "0.33, as
w
. w4
1 n`s
1
13.63
1
#
#
.
(25)
"
l
K
0.27z 5.43]10~4Gf~1
w
#
In neutral conditions, the vertical Lagrangian length
scale is proportional to z for small heights, while for
zPR l is proportional to Gf~1 (Blackadar, 1962).
w
#
For a shear-dominated stable boundary layer the numerical coe$cient in the r.h.s. of Eq. (25), 13.63, indicates
that the main e!ect of thermal strati"cation is to decrease
the size of the turbulence length scale and, consequently,
to increase the molecular dissipation. Therefore, the
small eddy size indicates that the structure of turbulence
does not respond directly to the ground conditions,
bringing to the #ow a z-less characteristic (Grant, 1992).
It is worth noting that the value of the numerical coe$cient of Eq. (25), 13.63, is in excellent agreement with the
value 14.0 proposed by Delage (1974). Furthermore, an
expression like (25) was also proposed by Stull (1988,
p. 208).
In unstable conditions (¸(0), the vertical eddy di!usivity can be derived from Eqs. (10a), (14) and (15):
AB
A
B C
A BD
A B
z
1@2
4z
K# "0.16w z 0.01 i
1!exp !
w
H i
!¸
z
i
8z 4@3
!0.0003 exp
.
(26)
z
i
The eddy di!usivity resulting from Eq. (26) yields a wellbehaved exchange coe$cient like that one given by
Holtslag and Moeng (1991); this last is a non-local eddy
di!usivity derived from countergradient terms by using
large eddy simulation data (Degrazia et al., 1997).
When the shear production term is the only input to
the reservoir of turbulent energy (neutral boundary layer
or shear stable boundary layer, ¸'0), the vertical eddy
di!usivity can be derived from Eqs. (15), (10b) and (23), as
Kn`s"
8
0.4(1#3.7z/K)1@3u z
H
.
z 4@3
1#15f z/(u ) #3.7
#
H0
K
C
D
(27)
In the neutral limit (¸PR), from Eq. (27) we obtain:
0.4(1!z/h)0.85(u ) z
H0 .
Kn "
8
[1#15f z/(u ) ]4@3
#
H0
(28)
3579
In the highly stable limit (z/¸PR), eddy vertical
motion is strongly limited by the positive strati"cation
and Eq. (27) results
0.4(1!z/h)3@4(u ) z
H0
K4 "
w 1#3.7z/¸(1!z/h)5@4
(29)
and so we expect eddy sizes to be limited entirely by the
stability and not by the distance from the surface. This is
referred to as local-height-independent (z-less) scaling
and will occur for ¸)z)h. In this case, Eq. (29) yields
K4 "0.11(u ) ¸ and eddies will scale with ¸ rather than
w
H0
with z as in the neutral case. At larger values of z/¸ and
z/h local similarity concepts are needed and Eq. (29) can
be written as
A B
z 2
K4 "0.11(u ) ¸ 1!
,
w
H0
h
(30)
where now z is large enough for surface #uxes to be
locally irrelevant.
In convective conditions (¸(0), according to Eqs.
(16), (18) and (19), the following expression for the vertical
decorrelation time scale can be obtained:
A
B C
A BD
A B
z
z
4z
1@2
¹# "0.39 i 0.01 i
1!exp !
w
w
!¸
z
H
i
8z 2@3
.
(31)
!0.0003 exp
z
i
For very unstable conditions (z /¸+100), in the central
i
region of the CBL, Eq. (31) can be approximated by
z
¹# +0.33 i .
(32)
w
w
H
It is worth noting that Hanna (1981) found, from Eulerian and Lagrangian observations, that for elevated
heights in the CBL, the vertical Lagrangian decorrelation
time scale was given by
z
(33)
¹# +0.17 i .
w
p
w
Following Hicks (1985) and Weil (1988), for elevated
heights in the CBL (z'0.1z ), the turbulence structure
i
can be idealised as vertically homogeneous with
p +0.59w . As a consequence, the last equation can be
8
H
written as
z
¹# :0.29 i
(34)
8
w
H
showing that the present derivation (Eq. (32)) is in good
agreement with Hanna's result.
When the buoyancy term is zero or negative, the shear
production term is the only input to the reservoir of
turbulent energy. In this case, the vertical decorrelation
3580
G.A. Degrazia et al. / Atmospheric Environment 34 (2000) 3575}3583
time scale can be derived from Eqs. (16) and (23) as
0.19z
¹n`s"
.
w
u [1#15f z/(u ) #3.7z/K]2@3(1#3.7z/K)1@3
H
#
H0
(35)
In the neutral limit (¸PR), we obtain from the previous equation
0.19z
¹n "
w (u ) [1#15f z/(u ) ]2@3(1!z/h)0.85
H0
#
H0
that can be also written as
(36)
0.27z/p
8
¹n "
.
(37)
w [1#15f z/(u ) ]2@3
#
H0
We notice that Eq. (37) has the same form as that one
given by Hanna (1982).
In the highly stable limit (z/¸PR), eddy vertical
motion is strongly limited by the positive strati"cation
and Eq. (35) results:
0.19z
.
(38)
¹4 "
w (u ) [1#3.7z/K](1!z/h)3@4
H0
Eddies will scale with ¸ rather than with z as in the
neutral case and will occur for ¸)z@h. In this case, Eq.
(38) yields ¹4 "0.051¸/(u ) .
w
H0
At larger values of z/¸ and z/h, local similarity holds
and the decorrelation time scale can be written as
¹4 "0.051
w
¸(1!z/h)1@2
.
(u )
H0
(39)
6. Test of the proposal parameterisation with the
Copenhagen tracer experiment
In Section 4, we have derived the expressions (Eqs.
(14)}(16) which allow the parameterisation of l , K and
i i
¹ for the PBL at all elevations (z )z)h, z ) and all
i
0
i
stability conditions from unstable to stable. In Section 5,
we have shown how this new parameterisation correctly
recovers important results previously obtained by other
researchers. It is the aim of this section to test our
parameterisation in a practical application and to show
how it works.
The actual atmospheric conditions only very rarely are
such as to be correctly represented by purely convective
or stable or neutral situations. In the majority of cases
the stability conditions and, consequently, the dispersive
conditions, are a combination of wind shear and buoyancy forces. One of the main peculiarities of the present
parameterisation is to be able to deal with such situations. As a consequence, to test our new parameterisation, we decided to simulate the Copenhagen tracer
dispersion experiment (Gryning and Lyck, 1984, 1998).
Copenhagen data set is particularly suited for this validation since most of these tracer experiments were performed in stability conditions that are the result of the
relative combination of wind shear and buoyancy forces.
Besides ground-level concentrations (g.l.c.s) at 3 arcs in
the range 2000}6000 m from the source, located at 115 m
above the ground, Copenhagen data set includes wind
speed at 10 and 115 m, p and p at 115 m, ¸, (u ) and
v
w
H0
z . All these data are on an hourly basis.
i
The tracer dispersion was simulated by the Lagrangian
particle model LAMBDA (Ferrero et al., 1995; Ferrero
and Anfossi, 1998a, b). For a detailed presentation and
discussion of LAMBDA we refer to the quoted papers.
Here we shall merely repeat the key points.
The actual version of LAMBDA is based on the generalised Langevin equation (Thomson, 1987), whose coe$cients are obtained by solving the Fokker}Planck
equation, and satis"es the well-mixed condition. It can
use as input higher-order moments of the atmospheric
probability density function (PDF) of wind velocity. In
the present application, LAMBDA used a Gaussian
PDF on the horizontal plane and a Gram}Charlier PDF,
truncated to the fourth order, in the vertical.
As far as the meteorological and turbulence input to
LAMBDA is concerned, we proceeded as follows. Wind
speed at 10 and 115 m were used to calculate the coe$cient for the exponential wind pro"le. p , p and ¹
v w
i
vertical pro"les were prescribed both with the present
parameterisation and with the Hanna's (1982) one
(neutral conditions). This last parameterisation was
chosen for the comparison because it is well known and
widely used in the specialised literature.
The third- and fourth-order momentum of the vertical
velocity (w3) and (w4) were prescribed according to Weil
(1990) and Ferrero and Anfossi (1998a), respectively.
p and p pro"les, either obtained with Hanna' scheme
v
w
or with the present scheme, were multiplied by a proper
coe$cient (di!erent for the two schemes) in order to
exactly match their observed value at 115 m.
LAMBDA simulation results are presented in Tables
1}3. These results are based on a model evaluation performed on all the pairs of observed and predicted values,
grouping all together the various tracer experiments performed. The statistical parameters considered are: mean
value (MEAN), standard deviation (S.D.), correlation
coe$cient (R), percentage of model predictions falling
within a factor of two (FA2), fractional bias (FB) and
normalised mean square error of the distribution
(NMSE).
Table 1 shows the results relative to the concentration
maxima on the arcs normalised by the emission rate
(ARCMAX/Q). We can see that the results of the present
parameterisation are signi"cantly better than those obtained with Hanna scheme, except for the correlation
coe$cient. Table 2 refers to the cross-wind integrated
G.A. Degrazia et al. / Atmospheric Environment 34 (2000) 3575}3583
3581
Table 1
Statistical indexes for ARCMAX/Q
Run
Mean
S.D.
R
FA2
FB
NMSE
Observed
Hanna (1982)
Present parameterisation
632.7
946.9
622.9
460.3
548.1
420.1
*
0.92
0.90
*
74
96
*
0.40
!0.02
*
0.24
0.10
Run
Mean
S.D.
R
FA2
FB
NMSE
Observed
Hanna (1982)
Present parameterisation
448.7
424.9
507.1
239.3
175.7
289.9
*
0.78
0.60
*
91
100
*
!0.05
!0.15
*
0.12
0.08
Run
Mean
S.D.
R
FA2
FB
NMSE
Observed
Hanna (1982)
Present parameterisation
382.0
174.3
400.1
159.8
84.8
161.7
*
0.90
0.70
*
65
100
*
!0.62
0.04
*
0.71
0.10
Table 2
As in Table 1 but for C /Q
y
Table 3
As in Table 1 but for p
y
concentrations at the various arcs normalised by the
emission rate (C /Q). In this case the quality of the results
y
obtained with the two parameterisations is comparable.
Finally, Table 3 illustrates the results relative to the
cross-wind concentration standard deviations (p ). ExY
cluding the correlation coe$cient, it can be seen that the
proposed parameterisation outperforms the Hanna
scheme.
In conclusion it can be said that the general agreement
between observed and predicted tracer g.l.c.s is good and
that the present turbulence parameterisation generally
performs better than the classical one.
7. Conclusions
In this paper we presented a turbulence parameterisation for dispersion models in all stability conditions,
excluding the very stable conditions. The present turbulence parameterisation is based on Taylor's statistical
di!usion theory, in which the shear buoyancy PBL
spectra are modelled by means of a linear combination of
the convective and mechanical turbulent energy. Such an
approach is strictly valid when the mechanical and convective wavelength peaks are well apart, so that the
concept of localness can be applied. This means that in
the turbulent spectrum, the energy #ux predominantly
involves scales of comparable size. The parameterisation
gives continuous values for the PBL at all elevations and
all stability conditions from unstable to stable conditions.
Furthermore, the derived turbulent parameters are
well-behaved and presented in form of similarity pro"les,
using the velocity (u , w ) and characteristic length
H H
scales (z , h, ¸, K). The parameterisation adequately rei
produces important results previously obtained by other
approaches. The formulation employs the empirical relationship between stability and wavelength peak for the
turbulent kinetic energy spectra. Therefore, the turbulence parameterisation is expressed in terms of the energy-containing eddies, which are responsible for the
main turbulent transport process in the PBL. The nature
of the subject is not suited for a direct check between
experiment and model. However, by using a Lagrangian
particle model and a data set of di!usion experiments
performed in a PBL generated both by thermal and
3582
G.A. Degrazia et al. / Atmospheric Environment 34 (2000) 3575}3583
mechanical forcing, the turbulent velocity variances and
the Lagrangian decorrelation time scales derived in the
present work were compared to the ones obtained by
Hanna (1982) scheme. This last scheme was chosen since
it is frequently used by the scienti"c community. Results
obtained are quite encouraging. Therefore, the new turbulent parameters may be suitable for applications in
regulatory air pollution modelling.
Acknowledgements
This work was partially supported by CNPq,
FAPERGS and CAPES.
References
Angell, J.K., 1974. Lagrangian}Eulerian time-scale relationship
estimated from constant volume balloon #ights past a tall
tower, turbulent di!usion in environmental pollution. In:
Landsberg, H.E., Van Mieghem, J. (Eds.), Advances in Geophysics, Vol. 18A, pp. 419}431.
Batchelor, G.K., 1949. Di!usion in a "eld of homogeneous
turbulence I: Eulerian analysis. Australian Journal of Scienti"c Research 2, 437}450.
Berkowicz, R.R., Prahm, L.P., 1984. Spectral representation of
the vertical structure of turbulence in the convective boundary layer. Quarterly Journal of the Royal Meteorological
Society 110, 35}52.
Blackadar, A.K., 1962. The vertical distribution of wind and
turbulent exchange in a neutral atmosphere. Journal of Geophysical Research 67, 3095}3102.
Caughey, S.J., 1982. Observed characteristics of the atmospheric
boundary layer. In: Nieuwstadt, F.T.M., van Dop, H. (Eds.),
Atmospheric Turbulence and Air Pollution Modelling.
Reidel, Dordrecht, pp. 107}158.
Caughey, S.J., Palmer, S.G., 1979. Some aspects of turbulence
structure throught the depth of the convective boundary
layer. Quarterly Journal of the Royal Meteorological Society
105, 811}827.
Champagne, F.H., Friehe, J.C., La Rue, J.C., Wyngaard, J.C.,
1977. Flux measurements, #ux estimation techniques,
and "ne-scale turbulence measurements in the unstable surface layer over land. Journal of Atmospheric Science 34,
515}530.
Degrazia, G., Anfossi, D., 1998. Estimation of the Kolmogorov
constant C from classical statistical di!usion theory. Atmo0
spheric Environment 32, 3611}3614.
Degrazia, G., Anfossi, D., Campos Velho, H.F., Ferrero, E., 1998.
A Lagrangian decorrelation time scale for non-homogeneous
turbulence. Boundary-Layer Meteorology 86, 525}534.
Degrazia, G.A., Campos Velho, H.F., Carvalho, J.C., 1997. Nonlocal exchange coe$cients for the convective boundary layer
derived from spectral properties. Contributions to Atmospheric Physics 70, 57}64.
Degrazia, G.A., Moraes, O.L.L., 1992. A model for eddy di!usivity in a stable boundary layer. Boundary-Layer Meteorology 58, 205}214.
Degrazia, G.A., Moraes, O.L.L., Oliveira, A.P., 1996. An analytical method to evaluate mixing length scales for the planetary
boundary layer. Journal of Applied Meteorology 35,
974}977.
Delage, Y., 1974. A numerical study of the nocturnal atmospheric boundary layer. Quarterly Journal of the Royal
Meteorological Society 100, 351}364.
Ferrero, E., Anfossi, D., 1998a. Comparison of PDFs, closures
schemes and turbulence parameterisations in Lagrangian
Stochastic Models. International Journal of Environment
and Pollution 9, 384}410.
Ferrero, E., Anfossi, D., 1998b. Sensitivity analysis of
Lagrangian Stochastic models for CBL with di!erent PDF's
and turbulence parameterisations. In: Gryning, S.E.,
Chaumerliac, N. (Eds.), Air Pollution Modelling and its
Applications XII, Vol. 22. Plenum Press, New York, pp.
673}680.
Ferrero, E., Anfossi, D., Brusasca, G., Tinarelli, G., 1995. Lagrangian particle model LAMBDA: evaluation against tracer
data. International Journal of Environment and Pollution 5,
360}374.
Grant, A.L.M., 1992. The structure of turbulence in the nearneutral atmospheric boundary layer. Journal of Atmospheric
Science 49, 226}239.
Gryning, S.E., 1999. Some aspects of atmospheric dispersion in
the strati"ed atmospheric boundary layer over homogeneous terrain. Boundary-Layer Meteorology 90, 479}494.
Gryning, S.E., Lyck, E., 1984. Atmospheric dispersion from
elevated source in an urban area: comparison between tracer
experiments and model calculations. Journal of Climate Applied Meteorology 23, 651}654.
Gryning, S.E., Lyck, E., 1998. The Copenhagen tracer experiments: reporting of measurements. RIS" National Laboratory, ISBN 87-550-2395-9. 54 p (only available in
an electronic version at: http://www.risoe.dk/rispubl/
VEA/ris-r-1054.htm).
Hanna, S.R., 1968. A method of estimating vertical eddy transport in the planetary boundary layer using characteristics of
the vertical velocity spectrum. Journal of Atmospheric
Science 25, 1026.
Hanna, S.R., 1981. Lagrangian and Eulerian time-scale in the
daytime boundary layer. Journal of Applied Meteorology 20,
242}249.
Hanna, S.R., 1982. Applications in air pollution modelling.
In: Nieuwstadt, F.T.M., Van Dop, H. (Eds.), Atmospheric Turbulence and Air Pollution Modelling. Reidel,
Dordrecht, pp. 275}310.
Hicks, B.B., 1985. Behavior of turbulence statistics in the convective boundary layer. Journal of Climate and Applied Meteorology 24, 607}614.
Holtslag, A.A., Moeng, C.H., 1991. Eddy di!usivity and countergradient transport in the convective boundary layer. Journal
of Atmospheric Science 48, 1690}1698.
H+jstrup, J., 1982. Velocity spectra in the unstable boundary
layer. Journal of Atmospheric Science 39, 2239}2248.
Kaimal, J.C., Wyngaard, J.C., Haugen, D.A., Cote', O.R., Izumi,
Y., Caughey, S.J., Readings, C.J., 1976. Turbulence structure
in the convective boundary layer. Journal of Atmospheric
Science 33, 2152}2169.
Kaimal, J.C., Finnigan, J.J., 1994. Atmospheric Boundary Layer
Flows. Oxford University Press, Oxford, 289 pp.
G.A. Degrazia et al. / Atmospheric Environment 34 (2000) 3575}3583
Mahrt, L., 1999. Strati"ed atmospheric boundary layer. Boundary-Layer Meteorology 90, 375}396.
Moeng, C.H., Sullivan, P.P., 1994. A comparison of shear and
buoyancy driven planetary boundary layer #ows. Journal of
Atmospheric Science 51, 999}1022.
Nieuwstadt, F.T.M., 1984. The turbulent structure of the stable
nocturnal boundary layer. Journal of Atmospheric Science
41, 2202}2216.
Olesen, H.R., Larsen, S.E., H+jstrup, J., 1984. Modelling velocity
spectra in the lower part of the planetary boundary layer.
Boundary-Layer Meteorology 29, 285}312.
Pasquill, F., 1974. Atmospheric Di!usion. Wiley, New York,
429 pp.
Rodean, H.C., 1994. Notes on the Langevin model for turbulent
di!usion of `markeda particles. Techn. Report UCRLID-115869, Lawrence Livermore National Laboratory,
122 pp.
Sorbjan, Z., 1989. Structure of the Atmospheric Boundary
Layer. Prentice-Hall, Englewood Cli!s, NJ, pp. 317.
Stull, R.B., 1988. An Introduction to Boundary Layer Meteorology. Kluwer Academic Publishers, Boston, 666 pp.
Sun, W.Y., 1993. Numerical simulation of a planetary boundary
layer. Part I: cloud-free case. BeitraK ge zur Physik den AtmosphaK re 66, 3}16.
Sun, W.Y., Chang, C.Z., 1986. Di!usion model for a convective
layer. Part I: numerical simulation of convective
boundary layer. Journal of Applied Meteorology 25,
1445}1453.
3583
Tennekes, H., 1982. Similarity relations, scaling laws and spectral dynamics. In: Nieuwstadt, F.T.M., Van Dop, H. (Eds.),
Atmospheric Turbulence and Air Pollution Modelling.
Reidel, Dordrecht, pp. 37}68.
Tennekes, H., Lumley, J.L., 1972. A First Course in Turbulence.
MIT Press, Cambridge, MA, 300 pp.
Thomson, D.J., 1987. Criteria for the selection of stochastic
models of particle trajectories in turbulent #ows. Journal of
Fluid Mechanics 180, 529}556.
Wandel, C.F., Kofoed-Hansen, O., 1962. On the Eulerian-Lagrangian transform in the statistical theory of turbulence.
Journal of Geophysical Research 76, 3089}3093.
Weil, J.C., 1988. Plume Rise. In: Venkatram, A., Wyngaard, J.C.
(Eds.), Lectures on Air Pollution Modeling, American Meteorological Society, Boston, 390 pp.
Weil, J.C., 1989. Stochastic modeling of dispersion in the convective boundary layer. In: van Dop, H. (Ed.), Air Pollution
Modelling and its Applications VII. Plenum Press, New
York, 620 pp.
Weil, J.C., 1990. A diagnosis of the asymmetry in top}down and
bottom}up di!usion using a Lagrangian stochastic model.
Journal of Atmosphere Science 47, 501}515.
Wilson, K.D., 1997. A three-dimensional correlation/spectral
model for turbulent velocities in a convective boundary
layer. Boundary-Layer Meteorology 85, 35}52.
Wyngaard, J.C., Cote, O.R., Rao, K.S., 1974. Modelling of the
atmospheric boundary layer. Advances in Geophysics, Vol.
18A. Academic Press, New York, pp. 193}212.