Vous êtes sur la page 1sur 8

Atmospheric Environment 33 (1999) 4001}4008

On the in#uence of the urban roughness sublayer


on turbulence and dispersion
Mathias W. Rotach
Swiss Federal Institute of Technology, Winterthurerstr. 190, CH-8057 Zu( rich, Switzerland

Abstract

The concept of the urban roughness sublayer is discussed and this lowest atmospheric layer over a rough surface is
shown to have a non-negligible vertical extension over typical urban surfaces. The existing knowledge on the turbulence
and #ow structure within an urban roughness sublayer is reviewed, focusing on the height dependence of turbulent #uxes
and a scaling approach for turbulence statistics, such as velocity variances, in the above-roof part of the roughness
sublayer. Finally, the implication of this turbulence and #ow structure upon dispersion characteristics is investigated. The
most prominent di!erence of explicitly taking into account the roughness sublayer in a dispersion simulation (as
compared to assuming a &constant #ux layer') is a clearly enhanced ground level concentration far downwind from the
source. For the example of a tracer release experiment over a (sub) urban surface (Copenhagen) it is shown that
introducing the roughness sublayer clearly improves the model performance.  1999 Elsevier Science Ltd. All rights
reserved.

Keywords: Turbulence; Reynolds stress; Local scaling; Urban dispersion modelling

1. Introduction turbulence structure over real urban or suburban surfa-


ces. Important di!erences between the two types of surfa-
On a mesoscale perspective an urban settlement can be ces are the characteristic density of roughness elements
considered a rough, warm &spot' within the surrounding and also their sti!ness (or #exibility to react upon drag
rural surfaces. Consequently, an internal boundary layer forces).
develops at the rural}urban interface (e.g., Oke, 1988) The RS can be regarded as the layer adjacent to the
that is both mechanical and thermal in origin. In this surface, wherein the #ow and turbulence is in#uenced by
contribution we restrict the discussion to areas within the individual roughness elements. Consequently, it has
urban environment that are far enough from this a fully three-dimensional structure. However, for many
transitional region so that the urban boundary layer practical applications it is convenient to consider spatial
(UBL) has replaced the former rural boundary layer. The averages of the variables of interest in order to retain the
lowest atmospheric layer of this UBL must be considered simplistic one-dimensional (vertical) description of #ow
as a roughness sublayer (RS). This layer, that constitutes and turbulence characteristics. Also, a detailed descrip-
the rough-wall counterpart of the viscous sublayer over tion of the e!ect of individual buildings can be avoided.
&smooth walls' (Raupach et al., 1991) has, unlike the On the other hand it is clear that from full scale obser-
latter, a vertical extension of several tens of meters over vational programs it is almost impossible to truly
typical urban settings and is therefore of importance determine spatially averaged properties. Here, the argu-
when modelling #ow or dispersion processes in urban mentation of Rotach (1993a) is followed by noting that
environments. an observation at a particular point in space represents
Most of the present knowledge on roughness sublayers a variety of upwind and downwind geometries due to
over rough surfaces stems from vegetated or arti"cial changing wind direction. Thus spatial averages can be
surfaces (see Raupach et al., 1991 for an excellent review). approximated by averaging over all wind directions of
In contrast, relatively little is known about the #ow and approaching #ow.

1352-2310/99/$ - see front matter  1999 Elsevier Science Ltd. All rights reserved.
PII: S 1 3 5 2 - 2 3 1 0 ( 9 9 ) 0 0 1 4 1 - 7
4002 M.W. Rotach / Atmospheric Environment 33 (1999) 4001}4008

Due to the lack of a comprehensive physically based Table 1


theory for the (spatially) average vertical structure of De"nitions of the roughness sublayer height, z , according to various
H
authors. &D' refers to the average inter-element spacing
the RS, observational data is often compared to the
predictions of Monin}Obukhov similarity theory, which Site/Characteristics z Author(s)
H
is, strictly speaking, only valid for the inertial sublayer
(note that over smoother surfaces, where the RS has Wind tunnel, various roughness' h#1.5D Raupach et al. (1980)
a negligible vertical extension, this layer is often called Tall vegetation 3D Garratt (1978)
Tall vegetation 4.5 h Garratt (1980)
surface layer). For the upper part of the RS (i.e. for z'h, Review 2}5 h Raupach et al. (1991)
where z is the physical height and h denotes the average
height of the canopy) this approach has become quite
standard for #ow over vegetated surfaces (e.g. Kaimal
and Finnigan, 1994) and was also followed in the few elements. It is clear that the average spacing of buildings
studies of full scale urban surfaces (see Section 3). How- in real cities is less well de"ned than that of roughness
ever, it will be pointed out in Section 3 that due to the elements in a wind tunnel experiment, so that in most
non-constancy of turbulent #uxes within the RS, a par- applications the simpler relationships that are based only
ticular form of local scaling is much more appropriate to on h, the average building height, are used. From the
describe the turbulence variables. Within the canopy (i.e., available recent full-scale observational programs the
in the lower part of the RS) it is often convenient to following can be deduced:
construct (properly scaled) average vertical pro"les for
the variables of interest. E Roth and Oke (1993) found evidence that their upper
In this contribution the concept of a roughness sub- level at z/h"2.65 was still within the RS over a subur-
layer is described in Section 2, where also its vertical ban surface in Vancouver.
extension for typical urban settings is discussed. In Sec- E The level at z/h"1.54 in the ZuK rich Urban climate
tion 3, the vertical turbulence and #ow structure in the Program (Rotach, 1993a) was clearly within the RS.
urban RS is investigated, making use of the results from From the same data set there are also indications that
an extensive "eld study within the Zurich Urban Climate z /h'2.1.
H
Program. Wherever possible, the results from this pro- E From the measurements over a suburban surface in
gram are compared to those from other (including arti"- Sapporo by Oikawa and Meng (1995) a value of z /h is
H
cial) rough surfaces. Finally, in Section 4, the e!ect of di$cult to derive. However, a maximum in Reynolds
including the turbulence structure of the roughness sub- stress was observed at a height of z /h+1.5 and this
H
layer upon dispersion of passive scalars is demonstrated. level might be regarded as an estimate for z for this
H
particular type of surface geometry.

2. The vertical extension of the Roughness Sublayer In Fig. 1 an attempt is made to display the various
layers over rough surfaces in a non-dimensional form, i.e.
From its de"nition, the roughness sublayer extends as a function of z/z (where z denotes the height of the

from the surface (z"0) up to a height z , at which the boundary layer) and z /h. The upper boundary of the
H
in#uence of individual roughness elements on the #ow inertial sublayer is chosen at z/z "0.1, therefore assum-

is &mixed up' by turbulence (Raupach et al., 1991), and ing that the argument of &double matching' is valid over
the #ow can be considered horizontally homogeneous if very rough surfaces too, i.e. the inertial sublayer is de-
the density, height and distribution of roughness ele- "ned as the region where simultaneously zz , with

ments do not vary over the upwind area of in#uence. The z the roughness length and zz is valid (see Tennekes

remaining part of the surface layer is usually termed and Lumley, 1972). For z an intermediate value of
H
inertial sublayer (IS). Note that the present de"nition of z "3h has been chosen (see Table 1 and the above
H
the RS is di!erent from that of Oke (1988) who has discussion for urban evidence). Note that any other as-
de"ned the lower boundary of the RS as the top of the sumption concerning z would simply move the corre-
H
urban canopy layer (UCL), i.e. roughly the average build- sponding line in Fig. 1 in the vertical direction or } if z is
H
ing height. Here, the more general concept of Oke et al. di!erent for plant canopies and urban type settlements
(1989) or Raupach et al. (1991) is adopted, in which the } give it another slope. The arrows denoted &city', &forest'
UCL is part of the RS. A number of criteria have been and &crop', respectively in Fig. 1 are based on typical
proposed in the literature to determine the height heights for the involved variables, for example z "

z (Table 1). These were derived from wind tunnel experi- 1000 m, h "20 m, h "10 m, h "1 m (or corre-
H    
ments (obviously, deriving rules to describe z from full- spondingly). Taking the crop as an example, it can be
H
scale measurements would require an enormous number seen from Fig. 1 that the vertical extension of the rough-
of observational programs) and can be seen from Table 1 ness sublayer is only a few meters, whereas the inertial
to rely on the distribution and geometry of roughness sublayer ranges from 3 up to 100 m (for the example
M.W. Rotach / Atmospheric Environment 33 (1999) 4001}4008 4003

Fig. 1. Sketch of the vertical extension of the various layers over


Fig. 2. Vertical variation of scaled (local) friction velocity from
rough surfaces and their variation with the non-dimensional
quantities z/z and z /h. A value of z /h"3 is assumed. For the the observations of Oikawa and Meng (1995), sub-urban (*) and
G G H Rotach (1993a), urban (*). For the de"nition of the reference
meaning of the arrows &city', &forest' and &crop' see text.
friction velocity, u , see Rotach (1993a). The data of Oikawa
HP
and Meng (1995) are re-scaled so that the maximum value is the
reference friction velocity. The solid line corresponds to a par-
given above concerning actual numbers). On the other ametrization for the data of Rotach (1993a).
hand, Fig. 1 reveals that the vertical extension of an
urban RS can amount to several tens of meters and might
} as opposed to rough vegetated surfaces } even com-
the measurement site de"nitely does not constitute
pletely "ll up the former rural surface layer ("RS#IS)
a smooth spot within a rougher surrounding. A detailed
so that no inertial sublayer is present. It is therefore
quadrant analysis showed (infrequent) sweeps to be re-
important to know the turbulence and #ow structure for
sponsible for most of the vertical momentum transport
a consistent interpretation of observed data or for practi-
close to roof level, similar to what can be observed over
cal applications such as dispersion modelling.
other types of rough surfaces (e.g. Raupach, 1981). Also,
the ratio between upward and downward transport of
momentum (&exuberance') is observed to strongly in-
3. Turbulence and 6ow structure crease (in its absolute value) when approaching the sur-
face, again similar to what is observed in vegetation
3.1. Reynolds stress canopies (e.g., Baldocchi and Meyers, 1988).
Evidence for a similar behaviour of Reynolds stress
One of the major results from the ZuK rich Urban cli- from other sites is discussed in the following. HoK gstroK m
mate program was the fact that Reynolds stress was et al. (1982) observed an increase in Reynolds stress
found to increase with height from very small values at between their 7 and 50 m levels, respectively, over an
z"d (d being the zero plane displacement) in the lower urban surface in Uppsala (S). However, their observa-
part of the RS towards a virtually constant value as tional tower was situated about 100 m downwind of
projected for the inertial sublayer aloft (Rotach, 1993a). the &edge' of the city. A simple estimate of the height
This increase in Reynolds stress (i.e. Ruw/Rz(0) was of the internal boundary layer, h , developing after this

found for all available simultaneous measurements at rough-to-smooth transition (h /z>"0.3(x/z>) , Kai-
  
two di!erent heights above roof level, and emerged also mal and Finnigan, 1994) yields h +12 m (taking for z>,
 
from a non-dimensionalized pro"le of Reynolds stress the larger of the two roughness lengths, 1 m according to
over the whole range of observations. Since this latter is HoK gstroK m et al., 1982). Thus, this increase in Reynolds
the result of averaging over all available runs represent- stress re#ects the adaptation to the new (non-urban)
ing approaching #ows from all possible wind directions, surface at the lower measurement level of 7 m, rather
it can be considered an estimate for a horizontally aver- than giving support for a roughness sublayer e!ect.
aged pro"le (Rotach, 1993a). For the site under consid- In a recent study over a suburban surface in Sapporo
eration, it can be ruled out that the increasing Reynolds (J), Oikawa and Meng (1995) also found Reynolds stress
stress is indicative for a local rough}smooth transition, as to vary with height (Fig. 2). Similar to the observations
4004 M.W. Rotach / Atmospheric Environment 33 (1999) 4001}4008

from the ZuK rich site, the measurements from Sapporo (1993b) and later by Oikawa and Meng (1995). In these
reveal a maximum of Reynolds stress (expressed as the two studies it was shown that not only a function of local
local friction velocity, u in Fig. 2) above roof level (at scaling variables can be derived to describe the data;
H
z/h"1.54 in this case) with the norm of momentum rather, the relationships that are appropriate for the
transport (&uw) decreasing when approaching the &sur- surface layer can be used as long as local #uxes (and
face'. Again, in this experiment the measuring tower was stability measures) are employed. Taking into account
e!ectively located 30 m downwind of the suburban sur- that the above-cited earlier studies usually worked with
face. The above simple analysis yields h +4 m for only one level of observations, their common conclusion

z>"0.45 m (Oikawa and Meng, 1995) thus indicating in favour of surface layer scaling may therefore also be

that even the lowest measurement level (5.4 m) is above regarded as supporting evidence for the concept of local
the internal boundary layer height and the pro"le dis- scaling for the roughness sublayer. Local scaling was
played in Fig. 2 re#ects the characteristics of the (upwind) found to apply for the velocity and temperature variance
suburban roughness sublayer. as well as the non-dimensional gradients of wind speed
In another recent investigation in the city of Basel and temperature and also the relationship between the
(CH), Reynolds stress was also observed to be varying Richardson number and z/¸, where L is the Obukhov
with height with increasing Reynolds stress (norm) be- length. It is noteworthy that for the scaled vertical
tween the lowest two levels at z/h+1.6 and z/h+2.3, velocity variance, there seems to be a tendency for an
respectively, for most of the stability classes (Feigenwin- enhanced mechanical contribution as compared to the
ter et al., 1999). These observations were carried out in surface layer results, thus leading to systematically di!er-
the central part of the city and thus seem to con"rm the ent constants in the similarity relation (Rotach, 1993b;
"ndings from the ZuK rich study (even if the uppermost Roth, 1993).
level (z/h+3.5) might have su!ered from a distinctly Some other, more subtle aspects of RS turbulence,
di!erent source area than the two lower ones). such as the ratio between the correlation coe$cient of
Evidence for Reynolds stress varying with height in sensible heat and that for momentum and thus the rela-
a wind tunnel study, in which an urban surface was tive transport e$ciencies are not well represented by the
simulated, is given by Rafailidis (1997). In this study it is Monin}Obukhov similarity relations, but nevertheless
shown that Ruw/Rz in the vicinity of roof level is parti- the sparse available data indicates that the concept of
cularly dependent on the geometry of the roofs, with local scaling provides a useful approach to describe these
much larger vertical variability for slanted roofs than for turbulence characteristics (Roth and Oke, 1995).
#at ones. It must be noted that local scaling is only applicable
Although these studies all seem to con"rm the "nding in the upper part of the urban roughness sublayer,
that Reynolds stress increases with height in an urban i.e. roughly above roof level (nevertheless, Oikawa
RS, Fig. 2 that combines the results from Oikawa and and Meng (1995) reported good correspondence to
Meng (1995) and Rotach (1993a), shows that a possible Monin}Obukhov predictions at z/h"0.77). Closer
parametrization for the vertical characteristics of to the surface average vertical pro"les (scaled by, e.g., the
Reynolds stress will have to include (at least) some in- friction velocity from the inertial sublayer) as suggested
formation on the urban morphology. In addition, much by Rotach (1995) may prove to be useful.
work needs to be done on the vertical structure of other
important turbulence statistics such as sensible or latent 3.3. Mean yow
heat #ux for a consistent determination of (local, see
below) stability measures within the RS. Throughout the urban roughness sublayer the pro"le
of mean wind speed has a similar form as that in rough-
ness sublayers over vegetated surfaces. Wind speed varies
3.2. Variances and non-dimensional gradients only marginally within the lower part of the canopy, then
rapidly increases close to the roof level and further in-
In a number of earlier studies the turbulence character- creases above. However, unlike over vegetated surfaces,
istics in the lower urban boundary layer (mostly velocity the shape of the wind speed pro"le is very much depen-
variances) were found to only marginally deviate from dent on the direction of the approaching #ow relative to
the predictions of Monin}Obukhov similarity theory the dominant street canyon axis (Rotach, 1995). Above
(e.g., Bowne and Ball, 1970; Brook, 1972; Clarke et al., roof level, the local scaling concept can be used to show
1982; Steyn, 1982). HoK gstroK m et al. (1982) were the "rst to that the (observed and modelled) mean velocity gradient
recognize a possible height dependence of turbulent is smaller than what would be predicted by a constant
#uxes and therefore suggested a concept of local scaling #ux assumption, in which the friction velocity from the IS
for other turbulence statistics. Further evidence for the is employed as a scaling velocity. This is demonstrated in
requirement (and success) of the local scaling approach Fig. 3, where measured average wind speed pro"les from
within the roughness sublayer was presented by Rotach the ZuK rich site are compared to those obtained from
M.W. Rotach / Atmospheric Environment 33 (1999) 4001}4008 4005

a so-called &urban' (or &rough wall') simulation the turbu-


lence structure of the roughness sublayer is explicitly
taken into account and the lowest part of the model
domain is considered in two parts, namely the roughness
sublayer and the inertial sublayer above. In contrast, in
the &non-urban' (or &smooth wall') simulation the lowest
part of the model domain is described entirely in terms of
surface layer similarity (as it is common for essentially all
dispersion models). Both the simulations were performed
using a Lagrangian stochastic dispersion model (Rotach
et al., 1996). For a low level source, the following charac-
teristics are observed:

E the downwind distance of the peak ground level con-


centration is slightly farther away from the source in
the urban case.
E the peak concentrations are very close to each other
for both simulations.
E the roll-o! of the surface concentration after the peak
is much less pronounced in the &urban' simulation
leading to a factor of two between the two simulations
Fig. 3. Average pro"les of scaled wind speed from the ZuK rich at some 20 source heights downwind from the source.
Urban climate Program. For the two curves, the non-dimen- E both the average height and width of the plume in-
sional gradient of wind speed was numerically integrated down- crease less (with increasing time or distance from the
ward from the uppermost measurement. &calculated' refers to the
source) in the &urban' case.
local scaling approach, i.e. using the height dependence of
Reynolds stress, whereas &constant #ux' uses the inertial sublayer
Reynolds stress to derive a constant friction velocity (from It is immediately clear that for elevated sources (e.g.
Rotach, 1993a). 1d2 corresponds to the average (over all wind single industrial plants) these e!ects will be less pro-
direction sectors) zero plane displacement. nounced (Rotach, 1997). However, in view of the fact that
in typical cities signi"cant contributions to the total
emissions arise from a large number of small sources,
which are predominantly close to the surface, it is argued
a numerical procedure. These latter pro"les were cal-
here that the e!ects of RS turbulence are important for
culated by numerically integrating the non-dimensional
modelling dispersion in an urban environment.
gradient of wind speed (U "Ru /Rz ) kz/u , where k is the
H The results from the above hypothetical simulations
von Kàrmàn constant) downward starting from the up-
may now be compared to those of a real situation. The
permost level of measurement. It can be seen from Fig. 3,
Copenhagen tracer experiment was selected as one of
that taking into account the local variation of Reynolds
the &standard data sets' for a dispersion model inter-
stress (and thus the local friction velocity) is su$cient to
comparison exercise in a series of workshops on &har-
explain the smaller observed velocity gradients as com-
monisation within atmospheric dispersion modelling for
pared to the assumption of constant #uxes. Indirectly,
regulatory purposes'. Tracer release was at 115 m above
Fig. 3 is of course again an indication for the success of
a (sub) urban part of Copenhagen (DK) under meteoro-
local scaling within the upper part of the urban RS, as
logical conditions of forced convection (see Gryning and
U was used according to the semi-empirical predictions
Lyck, 1984 for details). All participating operational dis-
of Monin}Obukhov similarity theory, but in connection
persion models underestimated on average the ground
with the local momentum #ux (i.e. u ).
H level concentrations (Olesen, 1995). Similarly, an average
underestimation was found for simulations with a one-
dimensional (Tassone et al., 1994) or a two-dimensional
4. Dispersion of pollutants within the roughness sublayer (Rotach et al., 1996) Lagrangian stochastic dispersion
model, respectively. In the above terminology, all these
The mean #ow and turbulence structure within the simulations correspond to the &non-urban' setting. Intro-
urban RS as summarized above has some important ducing the turbulence characteristics of the roughness
consequences on the characteristics of dispersion for sublayer in the two-dimensional Lagrangian dispersion
passive scalars. Rotach (1997) has compared two cases model (i.e. performing the corresponding &urban' simula-
for which identical input information is assumed to be tions) Rotach and de Haan (1997) showed that the average
given (&measurements' from the inertial sublayer): in mean bias can be removed (Fig. 4) and all the statistical
4006 M.W. Rotach / Atmospheric Environment 33 (1999) 4001}4008

Fig. 4. Crosswind integrated ground level concentrations, CW as Fig. 5. Vertical plume spread for one of the experiments (Octo-
observed in the Copenhagen tracer experiment and predicted by ber 19, 1978) during the Copenhagen tracer releases as a func-
a two-dimensional Lagrangian stochastic dispersion model. (*) tion of distance from the source. (solid line) &urban' simulation
no roughness sublayer present in the simulation (&non-urban' (see caption of Fig. 4), (dashed line) &non-urban' simulation
simulation), (*) simulation includes the turbulence structure of (dotted line) Pasquill class C after Hanna (1982) or Boubel et al.
the roughness sublayer (&urban' simulation). (1994) recommended for urban conditions.

measures for model performance are better for the &ur- operational dispersion models for the Copenhagen data
ban' simulations. Although the improvement is only mi- set as mentioned above. However, a systematic variation
nor due to the relatively large source height, the fact that of surface characteristics (average building height and
the mean bias is reduced (now close to zero, see Rotach density, height of the RS,...) will be necessary in order to
and de Haan, 1997) while at the same time the scatter (as derive explicit &urban' parameterizations for, e.g., p or
X
measured through the normalized mean square error) is p which are appropriate for speci"c city structures
W
diminishing, shows that the improved performance of the and/or stability regimes.
&urban' simulation is not the result of mere &tuning'.
Rather, the improved representation of the physical
properties of the #ow close to the surface is responsible 5. Summary and conclusions
for this result and this is likely to increase the credibility
of dispersion simulations over urban surfaces. In this contribution the (still sparse) knowledge on the
Having shown the superiority of the &urban' simula- turbulence structure within the urban roughness sub-
tions for one single tracer experiment and anticipating layer from full scale observations is reviewed. The aver-
that similar results will be obtained when performing age turbulence structure of the urban roughness sublayer
similar comparisons using tracer data from other urban can best be characterized through the non-uniformity of
experiments, the results from the Lagrangian particle turbulent #uxes with height. Accordingly, traditional sur-
model may be used to derive parameterizations for &ur- face layer scaling cannot be appropriate in this region. It
ban' plume parameters such as the vertical and the hori- is argued that a speci"c form of local scaling, namely
zontal plume widths, p and p , respectively, which are a formulation that employs the known dimensionless
X W
required in simpler, operational dispersion models. As an functions from the surface layer, but in connection with
example Fig. 5 compares p for one of the Copenhagen the local #uxes, can successfully be used within the upper
X
tracer release experiments (experiment of Oct. 19, 1978, part of an urban roughness sublayer (i.e. in its above-roof
see Gryning and Lyck, 1984) from the &urban' and the region). Furthermore, the e!ect of including the turbu-
&non-urban' simulations. Clearly, the increase of p with lence structure of the roughness sublayer upon dispersion
X
distance from the source of both the &urban' and the of passive scalars is demonstrated using a two-dimen-
&non-urban' simulations is less pronounced than that sional Lagrangian stochastic dispersion model. It is
predicted by the appropriate Pasquill-type C curve for shown that maximum concentrations and the distance
urban conditions (Hanna, 1982; Boubel et al., 1994). This from the source of their occurrence are not severely
explains the partly severe underestimation of some of the altered by the presence of a roughness sublayer.
M.W. Rotach / Atmospheric Environment 33 (1999) 4001}4008 4007

However, for emissions at small heights above roof level, Garratt, J.R., 1978. Flux pro"le relations above tall vegetation.
the evolution of the concentration downwind of the max- Quarterly Journal of Royal Meteorological Society 104,
imum is signi"cantly a!ected. Bearing in mind that pollu- 199}211.
tant concentrations in urban environments often result Garratt, J.R., 1980. Surface in#uence upon vertical pro"les in the
from the sum of a large number of (relatively weak) atmospheric near surface layer. Quarterly Journal of Royal
Meteorological Society 106, 803}819.
sources, rather than being the result of emissions from
Gryning, S.-E., Lyck, E., 1984. Atmospheric dispersion
one large source, this indicates that it is important to
from elevated sources in an urban areas: Comparison be-
explicitly include the roughness sublayer in urban disper- tween tracer experiments and model calculations.
sion models in order to improve the quality of their Journal of Climatology and Applied Meteorology 23,
predictions. 651}660.
The above considerations have shown that over urban Hanna, S.R., 1982. Applications in modelling, In: Nieuwstadt,
surfaces the RS has a non-negligible vertical extension F.T.M., van Dop, H. (Eds.), Atmospheric Turbulence and Air
and } in connection with its modi"ed turbulence struc- Pollution. Reidel, Dorrecht, 358 pp.
ture as compared to the surface layer } also can signi"- HoK gstroK m, U., BergstroK m, H., Alexandersson, H., 1982. Turbu-
cantly a!ect the dispersion characteristics (among others) lence characteristics in a near-neutrally strati"ed
in such an environment. However, since the observa- urban atmosphere. Boundary-Layer Meteorology 23,
449}472.
tional evidence on the #ow and turbulence structure in
Kaimal, J.C., Finnigan, J.J., 1994. Atmospheric Boundary Layer
this layer is still sparse, more experimental and also
Flows. Oxford University Press, Oxford, 289 pp.
theoretical work will have to be done in this area. In Oikawa, S., Meng, Y., 1995. Turbulence characteristics and
particular, the question of the in#uence of urban mor- organized motion in a suburban roughness sublayer. Bound-
phology upon average pro"les of momentum #ux and ary-Layer Meteorology 74, 289}312.
that of probably also non-constant vertical pro"les of Oke, T.R., 1988. The urban energy balance. Progress in Physics
heat (and moisture) #uxes will have to be addressed in and Geography 12, 471}508.
more detail. Oke, T.R., Cleugh, H.A., Grimmond, S., Schnid, H.P., Roth, M.,
1989. Evaluation of spatially-averaged #uxes of heat, mass
and momentum in the urban boundary layer. Weather and
Climate 9, 14}21.
Olesen, H.R., 1995. The model validation exercise at mol: over-
Acknowledgements view of results. International Journal Environmental Pollu-
tion 5, 761}784.
The present work was partly "nanced by the Swiss Rafailidis, S., 1997. In#uence of building areal density and roof
Federal Department of Education and Sciences (BBW) shape on the wind characteristics above a town. Boundary-
and the Swiss Federal Department of Environment, For- Layer Meteorology 85, 255}271.
est and Landscape (BUWAL) through a project in the Raupach, M.R., 1981. Conditional statistics of Reynolds stress in
framework of COST 615 (citair). rough-wall and smooth-wall turbulent boundary layers.
Journal of Fluid Mechanics 108, 363}382.
Raupach, M.R., Thom, A.S., Edwards, I., 1980. A wind-tunnel
study of turbulent #ow close to regularly arranged
rough surfaces. Boundary-Layer Meteorology 18,
References 373}397.
Raupach, M.R., Antonia, R.A., Rajagopalan, S., 1991. Rough-
Baldocchi, D.D., Meyers, T.P., 1988. Turbulence structure in wall turbulent boundary layers. Applied Mechanics
a deciduous forest. Boundary-Layer Meteorology 43, Reviews 44, 1}25.
345}364. Rotach, M.W., 1993a. Turbulence close to a rough urban surface
Boubel, R.W., Fox, D.L., Turner, D.B., Stern, A.C., 1994. Funda- part I: Reynolds stress. Boundary-Layer Meteorology 65,
mentals of Air Pollution, Third ed., Academic Press, New 1}28.
York, 574 pp. Rotach, M.W., 1993b. Turbulence close to a rough urban surface
Bowne, N.E., Ball, J.T., 1970. Observational comparison of rural part II: variances and gradients. Boundary-Layer Meteoro-
and urban boundary layer turbulence. Journal of Applied logy 66, 75}92.
Meteorology 18, 1072}1077. Rotach, M.W., 1995. Pro"les of turbulence statistics in and
Brook, R.R., 1972. The measurement of turbulence in a city above an urban street canyon. Atmospheric Environment
environment. Journal of Applied Meteorology 11, 443}450. 29, 1473}1486.
Clarke, C.F., Ching, J.K.S., Godowich, J.M., 1982. A study of Rotach, M.W., 1997. The turbulence structure in an urban
turbulence in an urban environment EPA Technical Report, roughness sublayer. In: Perkins, R.J., Belcher, S.E. (Eds.),
EPA 600-S3-82-062. Flow and Dispersion through Groups of Obstacles. Claren-
Feigenwinter, Ch., Vogt, R., Parlow, E., 1999. Vertical structure don Press, Oxford, pp. 143}155.
of selected turbulence characteristics above an urban Rotach, M.W., de Haan, P., 1997. On the urban aspect of the
canopy. Theoritical and Applied Climatology 62 (1}2), Copenhagen data set. International Journal of Environ-
51}63. mental Pollution 8, 279}286.
4008 M.W. Rotach / Atmospheric Environment 33 (1999) 4001}4008

Rotach, M.W., Gryning, S.-E., Tassone, C., 1996. A two-dimen- a patchy urban surface. Journal of Atmospheric Science 52,
sional stochastic lagrangian dispersion model for daytime 1863}1874.
conditions. Quarterly Journal of the Royal Meteorological Steyn, D.G., 1982. Turbulence in an unstable surface layer over
Society 122, 367}389. suburban terrain. Boundary-Layer Meteorology 22,
Roth, M., 1993. Turbulent transfer relationships over an urban 183}191.
surface. II: integral statistics. Quarterly Journal of the Royal Tassone, C., Gryning, S.-E., Rotach, M.W., 1994, A random-
Meteorological Society 119, 1105}1120. walk model for atmospheric dispersion in the daytime
Roth, M., Oke, T.R., 1993. Turbulent transfer relationships over boundary layer. In: Gryning, S.E., Millan, M.M. (Eds.), Air
an urban surface. I: spectral characteristics. Quarterly Journal Pollution Modelling and ist Application X, NATO, vol 18.
of the Royal Meteorological Society 119, 1071}1104. pp. 243}251. Challenges of Modern Society.
Roth, M., Oke, T.R., 1995. Relative e$ciencies of Tennekes, H., Lumley, J.L., 1972. A First Cours in Turbulence.
turbulent transfer of heat, mass and momentum over MIT Press, Cambridge, MA, 300 pp.

Vous aimerez peut-être aussi