You are on page 1of 22

Boundary-Layer Meteorol (2008) 126:103–124

DOI 10.1007/s10546-007-9222-5

ORIGINAL PAPER

Wind-tunnel Modelling of Dispersion from a Scalar Area


Source in Urban-Like Roughness

Frauke Pascheke · Janet F. Barlow · Alan Robins

Received: 6 March 2007 / Accepted: 25 July 2007 / Published online: 20 September 2007
© Springer Science+Business Media B.V. 2007

Abstract A wind-tunnel study was conducted to investigate ventilation of scalars from


urban-like geometries at neighbourhood scale by exploring two different geometries a uni-
form height roughness and a non-uniform height roughness, both with an equal plan and
frontal density of λ p = λ f = 25%. In both configurations a sub-unit of the idealized urban
surface was coated with a thin layer of naphthalene to represent area sources. The naphthalene
sublimation method was used to measure directly total area-averaged transport of scalars out
of the complex geometries. At the same time, naphthalene vapour concentrations controlled
by the turbulent fluxes were detected using a fast Flame Ionisation Detection (FID) tech-
nique. This paper describes the novel use of a naphthalene coated surface as an area source in
dispersion studies. Particular emphasis was also given to testing whether the concentration
measurements were independent of Reynolds number. For low wind speeds, transfer from
the naphthalene surface is determined by a combination of forced and natural convection.
Compared with a propane point source release, a 25% higher free stream velocity was needed
for the naphthalene area source to yield Reynolds-number-independent concentration fields.
Ventilation transfer coefficients wT /U derived from the naphthalene sublimation method
showed that, whilst there was enhanced vertical momentum exchange due to obstacle height
variability, advection was reduced and dispersion from the source area was not enhanced.
Thus, the height variability of a canopy is an important parameter when generalising urban
dispersion. Fine resolution concentration measurements in the canopy showed the effect of
height variability on dispersion at street scale. Rapid vertical transport in the wake of individ-
ual high-rise obstacles was found to generate elevated point-like sources. A Gaussian plume
model was used to analyse differences in the downstream plumes. Intensified lateral and
vertical plume spread and plume dilution with height was found for the non-uniform height
roughness.

F. Pascheke · J. F. Barlow
Department of Meteorology, University of Reading, Earley Gate, PO Box 243, Reading RG6 6BB, UK
F. Pascheke (B) · A. Robins
School of Engineering, University of Surrey, Guildford GU2 7XH, UK
e-mail: f.pascheke@surrey.ac.uk

123
104 F. Pascheke et al.

Keywords Area source · Area-averaged turbulent fluxes · Concentration measurements ·


Naphthalene sublimation · Urban heat and pollutant dispersion · Wind tunnel

1 Introduction

Dispersion in urban areas is due to two processes, acting on similar time scales, which com-
bine to determine pollutant concentrations: horizontal advection through the urban canopy
and turbulent exchange with cleaner air above (ventilation). A greater surface roughness gen-
erates high turbulence levels within the urban canopy layer, resulting in intensified dispersion
processes and canopy ventilation. On the other hand, form drag due to the buildings increases
the momentum exchange from the flow to the surface, resulting in a reduction of the advection
velocity through the canopy, which in turn increases the residence time of scalars within the
canopy. The relative magnitudes of these opposing effects determine whether urban canopy
ventilation is enhanced or reduced as roughness is increased (Britter and Hanna 2003).
The turbulent wakes shed behind individual buildings interact, resulting in highly efficient
mixing and diffusion processes for momentum and scalar properties. The scale of the largest
eddies involved in this process is determined by the building dimensions. Thus, the char-
acteristics of the turbulence generated within an urban canopy are highly dependent on the
morphology of the roughness elements, making it difficult to define universally applicable
parameterizations. Cheng and Castro (2002a) performed wind-tunnel flow measurements
over urban-like rough surfaces and found that an array with non-uniform height roughness
elements with the same plan area density, λ p , and frontal area density, λ f , as a uniform
height array is more efficient in generating surface stress. This indicates that parameteriza-
tions based on morphological descriptors that only take into account the packing density will
not adequately capture the contribution to turbulence intensity due to building height vari-
ability. In a similar way, the flow in and above the canopy is often described as isolated, wake
interference or skimming flow, with thresholds based on values of the frontal area index, λ f
(Oke 1987). However, based on systematic wind-tunnel experiments over a wide range of
urban-like roughness arrays, Hall et al. (1996) pointed out the importance of height variabil-
ity in inhibiting the skimming flow regime. One general conclusion to be drawn from these
studies is that uniformly spaced bluff bodies of uniform height produce a canonical flow that
is highly relevant to urban boundary-layer studies, but one from which certain dynamical
mechanisms acting in more realistic geometries may be absent.
More recently, Heist et al. (2005) studied the effect of a single high-rise building on
the residence times of pollutants in a regular array of cubical buildings. In the wind-tun-
nel experiments tracer gas was released from a point source upstream of the obstacle array.
After reaching steady-state concentrations, the source was turned off and concentration time
series were measured at a single point directly upwind of the tall building for different build-
ing heights. Subsequently, characteristic time scales of the pollutant venting were derived
from the recorded time series and found to decrease significantly as the building height was
increased. This finding suggests that ventilation is particularly enhanced by rapid vertical
dispersion in the wake of tall buildings. The skewness of the building height distribution
may thus become an important factor in a morphologically based parameterization of urban
canopy turbulence.
In urban areas, pollution may be emitted from point sources, but they are so numerous that
it is helpful to define an area source. The naphthalene sublimation technique provides a means
of directly simulating an area source and also measuring the total area-averaged transport of
scalars out of any given geometry. The ventilation efficiency of the surface can be quantified

123
Wind-tunnel Modelling of Dispersion from a Scalar Area Source 105

by calculating a transfer coefficient (equivalent to a Stanton number, in engineering terms).


A full description of the technique and its use for the study of scalar transfer from complex
surfaces is reviewed by Goldstein and Cho (1995). Recently, Barlow and Belcher (2002) and
Barlow et al. (2004) applied the method to study the ventilation of idealised two-dimensional
street canyons.
The aim of the present study was to investigate dispersion from a scalar area source in
urban-like roughness. A key objective was to investigate the effect of varying roof height on
scalar dispersion within and above the urban canopy. The rough surfaces chosen for study
were two of those studied by Cheng and Castro (2002a), having equal λ f but with respectively
uniform and non-uniform height buildings. The use of Flame Ionisation Detectors (FIDs) for
the measurement of naphthalene concentrations was required to pursue this objective. The
development of this technique, which is entirely novel, is described in Sect. 2.5. In Sect. 2.6,
a careful study of the effect of Reynolds number on the measurements is presented. Finally,
in Sect. 3, the transfer coefficients and concentration fields for the two surfaces are presented
and discussed.

2 Experimental Set-up

2.1 Boundary-Layer Wind Tunnel

The experiments were performed in the ‘A’ boundary-layer wind tunnel of the Environmental
Flow and Research Centre (EnFlo). The test section of the open-circuit wind tunnel is 0.9 m
wide, 0.6 m high and 4.5 m long, and the wind tunnel is equipped with an automated three-
dimensional (3D) traverse system. The wind tunnel set-up is identical to the flow experiments
of Cheng and Castro (2002a) using urban-like roughnesses labelled C10S and RM10S. In
each case, the entire tunnel floor was covered with repeated units of the specified roughness.
The test section blockage  and the pressure gradient along the test section were negligible:
 
Amodel ∂p 2δ
= ≈ 0.02, < 0.006,
Awindtunnel ∂ x ρUδ2
where δ denotes the boundary-layer thickness, and (X, Y, Z ) are the streamwise, lateral and
vertical coordinates respectively. The plane X = 0 is at the beginning of the roughness, Y = 0
denotes the test section centreline and the plane Z = 0 is at ground level. A ramp positioned
in front of the first roughness row provided a smooth transition between the contraction exit
and the rough surface. No additional devices commonly used to establish boundary-layer
conditions, such as vortex generators or turbulence grids, were used.

2.2 Roughness Surfaces

Two urban-like roughnesses with different geometries were used. Both surfaces were com-
posed of many repeated units, each 80 × 80 mm2 in area and made of moulded plastic
material (Acrylonitrile Butadiene Styrene). A single unit consisted of 16 uniform height
cubes (C10S), or non-uniform height cuboids (RM10S) respectively, which were arranged in
staggered arrays as shown schematically in Figs. 1 and 2. The cube side length was 10 mm.
For RM10S each cuboid had the same plan area of 10 × 10 mm2 but the height distribution
was approximately a Gaussian with five heights equally spaced from 2.8 to 17.2 mm and a
mean and standard deviation of 10.2 mm and 3.6 mm respectively. Hence for both geometries,
the plan area density and frontal area density were almost equal, i.e. λ p = λ f = 25%.

123
106 F. Pascheke et al.

X/H
-4 -2 0 2 4
40 4
10 10
Z
Y 10 10
20 2
10 10
X

Y [mm]
10 10

Y /H
0 0
10 10
40
10 10
2960 20 -20 -2
298 0 10 10
0
3000
10 10
-2 0
3020 -40 -4
2960 2980 3000 3020 3040
3040 -40
X [mm]

Fig. 1 Sketch of 3D and plan view of one unit for array with uniform heights (C10S). All dimensions in mm,
element heights are indicated on the plan view. Mean flow is along the x-axis

X/H
-4 -2 0 2 4
40 4

Z 10 10

Y 10 13.6
20 2
13.6 10
X
Y [mm]

13.6 6.4

Y/H
0 0
10 6.4
40
10 17.2
29 60 20 -20 -2
2980 6.4 13.6
0
3000
2.8 10
-20
3020 -40 -4
2960 2980 3000 3020 3040
3040 -40
X [mm]

Fig. 2 Sketch of 3D and plan view of one unit for array with non-uniform heights (RM10S). All dimensions
in mm, element heights are indicated on the plan view. Mean flow is along the x-axis

2.3 Naphthalene Sublimation Technique

To simulate a scalar area source, the naphthalene sublimation technique was used (Barlow
and Belcher 2002; Goldstein and Cho 1995). Naphthalene sublimes readily at room tempera-
ture, and if applied to a surface, a thin layer of saturated naphthalene vapour is rapidly formed
and maintained in the immediate vicinity of the surface. Given that the surface temperature
is uniform and constant for the time of an experimental run, the naphthalene vapour pressure
above the surface will also be uniform and constant, representing a constant concentration
boundary condition. The area- and time-averaged flux from the surface F is given by:
 
F = w T ρ S − ρ ∞ , (1)

where the overbar denotes a temporal


 average
 and angled brackets denote a spatial averages.
w T is the transfer velocity and ρ S − ρ ∞ is the difference between the mean naphthalene
vapour density at the active surface and the mean naphthalene vapour density in the free
stream. In sufficiently large laboratories, the naphthalene vapour density in the free stream is
negligibly small (ρ ∞ ≈ 0), as the following estimation demonstrates. The maximum back-
ground vapour density in the EnFlo laboratory was estimated to be ρ ∞ ≈ 5 × 10−8 kg m−3 ,

123
Wind-tunnel Modelling of Dispersion from a Scalar Area Source 107

based on room volume, air changes and typical naphthalene sublimation rate. Given that
the vapour density at the surface was typically ρ S ≈ 5 × 10−4 kg m−3 , ρ ∞ was deemed
negligible. The flux density F adjacent to the source is derived directly from the mass loss
m of naphthalene from the coated area A taking place during an experimental run of time
t:
m
F = . (2)
At
The mean naphthalene vapour density at the surface for a given surface temperature can
be calculated using the ideal gas law
eS
ρS = (3)
RN T S
where R N = 64.95 J mol−1 K−1 is the gas constant for naphthalene and T S is the surface tem-
perature in kelvin. The naphthalene saturation vapour pressure at the surface can be derived
from the following empirical expression (CRC 1993)
13.57− 3729
e S = 10 TS
, (4)
which is valid in the temperature range 250–330 K.
Combining Eqs. 1 and 2, the transfer velocity is given by
m
wT = . (5)
A t ρS
The relationship between transfer velocity and reference wind speed U is an independent
function for each surface geometry and source location:
wT = C T U (6)
where C T is the transfer coefficient. Barlow and Belcher (2002) demonstrated that the rela-
tionship is close to linear for sufficiently high wind speeds, thus allowing easy comparison
of the transfer coefficient for different urban surface geometries.
In terms of dispersion, the flux density F is equivalent to the source area emission rate
Q A . Implicit in Eq. 6 is the fact that the emission rate for naphthalene vapour is dependent
on wind speed, which contrasts with more traditional trace gases that are released at a fixed
emission rate. Hence, for sufficiently high wind speeds, the measured naphthalene vapour
density is invariant with wind speed (see Sect. 2.5.4 for discussion of this) and only requires
normalization by the surface vapour density to give
ρ
ρ N∗ = . (7)
ρS
For a fixed emission rate source, the vapour density would be normalized by a given
reference wind speed Ur e f and area emission rate Q A to give
ρ Ur e f
ρ F∗ = . (8)
QA
Recalling that F ≡ Q A and ρ ∞ ≈ 0, by substitution of Eqs. 1 and 6 into Eq. 7, it can
be shown that
ρ N∗ = C T ρ F∗ . (9)

123
108 F. Pascheke et al.

That is, the normalized naphthalene vapour density field need only be divided by the transfer
coefficient C T to be fully equivalent to the normalized vapour density field due to a fixed
emission rate source.

2.4 Area Sources

The objective was to measure fluxes and vapour densities due to an area source of one unit size
within the roughness array, with leading edge at X = 2960 mm and placed symmetrically
about the wind-tunnel centreline (Y = 0 mm: see Figs. 1 and 2 for dimensions). This is just
over one unit upstream of the location at which Cheng and Castro (2002a) reported dynami-
cal characteristics of the two surfaces ((X, Y ) = (3145, 0)). For each surface, an area source
unit with geometrical features as shown in Figs. 1 and 2 was manufactured from aluminium.
Aluminium has a high thermal conductivity, so that temperature gradients across the unit
surface were negligible. Each unit was briefly dipped into a beaker of molten naphthalene
to coat the ‘street’ surface only. This was to simulate an area source at ground level, and
highlights the advantage of this technique in being able to control precisely which surface is
‘active’. The reproducibility of the naphthalene layer thickness δnaph achieved by this method
was determined to be 0.7 ± 0.1 mm. Any spilling and misplaced naphthalene residuals were
accurately removed to ensure only the street was coated. Under the chosen experimental
conditions, a recoating of the surface was required after approximately three hours to ensure
full surface coverage at all times.

2.5 Measurement Techniques

The EnFlo laboratory operates a comprehensive, in-house ‘virtual instrumentation’ package


for data acquisition, instrument control, data processing and analysis. The software tools
run in the LabVIEW (National Instruments) environment on Apple Macintosh computers.
In the present work, all electrical signals were digitised using a 16-bit data acquisition board
(USB-9211, National Instruments) and saved (and archived) as raw data.

2.5.1 Temperature

The naphthalene saturation vapour pressure is very sensitive to temperature changes (see Eq.
4). Therefore, constant laboratory temperatures are preferable and a precise measurement
of the temperature of the active surface is essential during each run. A computer controlled
procedure was developed specifically to keep the laboratory temperatures constant to within
±0.2◦ C. The surface temperature of the coated unit, as well as the ambient air temperature
inside the wind tunnel, were continuously recorded using negative temperature coefficient
10 k thermistors (RS 151-237, RS Components Ltd.). Both thermistors were calibrated
against a precision platinum resistance thermometer with an absolute accuracy better than
±0.01◦ C (calibration temperature range: 16–35◦ C). The surface thermistor was plugged into
a small recess in the bottom of the coated surface of the unit and itself coated with a heat
sink compound to ensure good thermal contact with the aluminium unit.

2.5.2 Mass

To measure the mass loss from the coated surface a mass balance (F4P, Stanton Instruments)
with an accuracy of ±0.1 mg was used. The coated surface unit was weighed prior to a run,
and then removed from the wind tunnel and re-weighed as quickly as possible after a run to

123
Wind-tunnel Modelling of Dispersion from a Scalar Area Source 109

minimise losses and hence errors due to natural convection. On average, the total duration
for the mounting, dismounting and weighing of the test unit took no longer than 90 s. Nev-
ertheless, the mass loss rate due to natural convection was determined from additional tests,
which included all steps in the procedure except the running of the wind tunnel. For typical
test runs with a free-stream velocity Ur e f between 5 and 20 m s−1 , the error in transfer veloc-
ities due to natural convection was determined to be < 0.8%. Figure 3 shows the systematic
relative error  F as a percentage of the measured flux. All subsequent transfer velocities were
corrected for the mass loss due to natural convection.

2.5.3 Velocity

The mean boundary-layer flow wind speed profiles were measured using a miniature hot
wire anemometer probe (55P11, Dantec) operated with a constant temperature bridge sys-
tem (Newcastle). The hot wire probe was calibrated against a Pitot-static tube connected to a
micro-manometer (FC012, Furness). During all experiments, the reference free-stream veloc-
ity
 was continuously  recorded with the Pitot-static tube, which was mounted at
X r e f , Yr e f , Z r e f = (2.7, 0, 0.45) m and its value was used to close the tunnel speed
control loop to maintain velocity within ±0.05 m s−1 .

2.5.4 Concentrations

The naphthalene vapour concentrations were measured using a fast flame ionisation detector
(HFR400 fast FID, Cambustion). Naphthalene vapour is a hydrocarbon gas (C10 H6 ) and when
hydrocarbons are introduced into a hydrogen flame in the detector, ions are produced and
can be detected with a high voltage ion collector. The current across the collector is directly
proportional to the ionisation rate and thus depends on the hydrocarbon concentration in the
sample gas. As the measurement of naphthalene vapour using a FID system is novel, the
following procedures were developed to ensure accurate concentration measurements.
Due to the lack of available naphthalene calibration gas, the FID was calibrated against
propane calibration gas at concentrations of 151 ppmv and 749 ppmv. An automated recalibra-
tion every 30 min was included in the experimental procedure; background concentrations
were also measured every five minutes. Any calibration drifts, and the rising background
concentration, were removed from the data using linear de-trending. For a given surface

1
0.9
0.8
0.7
0.6
ε F [%]

0.5
0.4
0.3
0.2
0.1
0
5 10 15 20 25
Uref [m s-1]

Fig. 3 Error in derivation of turbulent fluxes due to natural convection during test unit handling for different
free-stream velocities

123
110 F. Pascheke et al.

600
C propane
500
Linear fit (R2 = 0.9997)

C propane [ppm]
400

300

200

100

0
0 50 100 150 200 250
Cnaphthalene [ppm]

Fig. 4 Best local linear relationship between propane-equivalent and theoretically derived saturated naphtha-
lene vapour concentrations for the temperature range of 13–32◦ C

temperature, the concentration of naphthalene vapour at the surface, C S , can be expressed in


units of ppm by volume according to
eS
C S = 106 (10)
p
where p is the air pressure. A normalization of measured concentrations with the corre-
sponding saturated source concentration C N∗ = C/C S provides temperature independent
results.1
Since the concentration sampling system was calibrated against propane but the source
concentrations were calculated using Eqs. 3 and 10, a methodology was required to convert
the naphthalene source concentrations into propane-equivalent concentrations. A small plas-
tic cylinder was used for this, the inner walls of which were coated with molten naphthalene
(cylinder volume: V = 5 × 10−6 m3 , coated surface: A ≈ 23.5 × 10−6 m2 ). The cylinder
was placed in a temperature controlled water bath. Assuming that the vapour pressure inside
the cylinder was saturated after a reasonably short duration, concentrations and temperatures
were measured at the cylinder centre for a temperature range of 13–32◦ C. Figure 4 shows
the best local linear relationship between the measured and the theoretical concentrations for
the studied temperature range, given by
C pr opane = 2.27Cnaphthalene − 11.31 (11)
with R2= 0.9997. This result is only used to calculate the source concentrations, all of
which were in the range of Cnaphthalene ≈ 100 ppmv.

2.6 Reynolds Number Independence

Particular emphasis was given to testing whether flow and concentration measurements were
independent of Reynolds number. For different urban-like geometries, the Reynolds number
is usually based on U H , the mean velocity at obstacle height, and the mean obstacle height
H as a characteristic length scale of the obstacles giving
UH H
Re H = (12)
ν
where ν is the kinematic viscosity of air. Quoted critical values for an in-canopy flow to
be independent of viscous effects vary between 2,100 and 15,000 (Uehara et al. 2003), and
1 Note that C/C is equivalent to ρ/ρ in Eq. 7 when surface temperature T and air temperature T are
S S S
equal. Given that all experiments were performed under neutral conditions, this requirement was met.

123
Wind-tunnel Modelling of Dispersion from a Scalar Area Source 111

Table 1 Values of free-stream velocity Ur e f , corresponding spatially averaged velocity at building height
U H and the Reynolds number based on U H and building height H for the tested surfaces

C10S RM10S
Ur e f [m s−1 ] U H [m s−1 ] Re H U H [m s−1 ] Re H

24 5.3 3520 4.3 2880


16 3.5 2333 2.9 1933
8 1.8 1200 1.4 933

the range is presumably caused by different geometrical complexities and upstream flow
conditions of the individual studies.
In the EnFlo ‘A’ tunnel a maximum free stream velocity of 25 m s−1 could be realised.
Thus, additional tests to check for Reynolds number independence were performed for both
set-ups. Table 1 shows the reference velocity range that was covered in the following experi-
ments along with corresponding velocities at (mean) building height and Reynolds numbers
calculated using Eq. 12. The velocities at building height U H were derived from a spatially
averaged velocity profile for each surface. These were based on four individual profiles for
C10S and 24 for RM10S and fully agreed with the equivalent spatially averaged profiles
given by Cheng and Castro (2002a).
The ‘roof level’ wind speed, U H , is commonly used in urban air quality and related
meteorological work, even though its definition must always be somewhat imprecise. This
common use was one reason for adopting it here, rather than a more accurately measurable
quantity such as the wind speed at, say, 1.5H or 2H . Another reason in using U H lies in
providing an appropriate speed scale for the RM10S surface, where using a wind speed at
much greater height could well fail to characterise the deceleration over the single unit of
random roughness. Clearly, U H is a rather imprecise quantity both to measure and to use
but has some merit. Compared with the results in Table 1, the Reynolds numbers using, say,
U (Z = 1.5H ) calculated from the log-law fit to the spatially-averaged mean velocity pro-
files would be somewhat larger and the differences between the values for C10S and RM10S
somewhat greater.

2.6.1 Flow Profiles

Mean flow profiles were measured at several positions within both roughness arrays using
a single hot wire. Turbulence intensities near both surfaces were much higher than 20%,
therefore, hot wire measurements inside the canopy could only be expected to give qual-
itative results. Figure 5 shows an example of normalized mean flow profiles measured at
(X, Y ) = (3005, −5) mm in the C10S roughness for a range of reference wind speeds.
Within the canopy, the measurement position was x = 1.5H downstream of a roughness
element, and therefore in the vicinity of the reattachment point at the end of the recirculation
zone. Thus, the flow characteristics at this location were likely to be more sensitive to Rey-
nolds number than elsewhere in the flow field. Figure 5b shows some evidence of changing
wind profile within the canopy only for the lowest reference wind speed (8 m s−1 ), but this
was not the case for all measurement positions within the canopy. Indeed, the discrepancies
shown in Fig. 5b might well be solely due to hot wire errors in a region of highly turbulent
flow with low mean velocity. Above the canopy at a height of Z ≈ 2H all profiles collapse

123
112 F. Pascheke et al.

a) 20 b)
Uref = 24 ms-1 1.4 Uref = 24 ms-1
-1 -1
Uref = 20 ms Uref = 20 ms
Uref = 16 ms-1 Uref = 16 ms
-1

-1
1.2
Uref = 12 ms Uref = 12 ms-1
15 Uref = 8 ms-1 Uref = 8 ms-1
1

Z/H
0.8
Z/H

10
0.6

0.4
5

0.2

0 0
0 0.5 1 0 0.1 0.2 0.3

U/U ref U/U ref

Fig. 5 Test for Reynolds number independence of wind profiles in C10S array at (X, Y ) = (3005, −5) mm.
(a) wind speed profile (b) lower part of the profile within an urban canopy

for the tested wind speed range and no wind speed dependencies were found for all studied
locations.

2.6.2 Concentration

In contrast to momentum transfer, the critical Reynolds number is not known for scalar
transfer. However, given a Schmidt number for molecular diffusion of naphthalene in air of
Scnaphthalene = 2.28 at 25◦ C, a significant difference to momentum transfer would not be
expected. Practically speaking, experiments should be performed at wind speeds such that the
Reynolds number dependence is small. Prior to using the naphthalene area source, a simpler
test was conducted for both surfaces using a continuous propane release from a point source
at the location (X, Y ) = (2965, −5) mm. Excessive source discharge rates may produce
vertical jets and thus influence the flow pattern in the given geometry. The average horizontal
velocity close to the ground was estimated to be approximately 0.05Ur e f and it was assumed
that a similar outflow velocity, W S , would be appropriate to ensure minimal effect on the
flow field. For each wind speed tested, the source strength was adjusted to keep a constant
ratio of W S /Ur e f = 0.04. (Note that at higher speeds, care had to be taken to correct for
sampling tilt due to increased drag on the FID unit, to ensure that the measurement location
of the needle was accurate to 0.2 mm).
Figure 6a shows concentration measurements at a number of locations within the canopy
for both surfaces. Concentrations are given in dimensionless form
C UH H 2
C∗, ps = (13)
Q
to allow for slight wind speed and source strength variations during the experiments; Q is the
emission volume flow rate. Overall, it can be seen that systematic changes of concentration
values occur for wind speeds below 16 m s−1 , but the effect is weaker for the RM10S surface.
This might be attributable to the more intense turbulent flow that was observed over the

123
Wind-tunnel Modelling of Dispersion from a Scalar Area Source 113

a) 0.12 b) 20 RM10S (3015, -5, 3)


RM10S (3015, -5, 3)
RM10S (3015, -15, 3)
RM10S (3015, -15, 3)
RM10S (3015, 5, 6)
RM10S (3015, 5, 6) 15
0.1 C10S (3015, -5, 3)
C10S (3015, -5, 3)
C10S (3015, -15, 3)
C10S (3015, -15, 3)
C10S (3015, 5, 6)
C10S (3015, 5, 6) 10

∆C [%]
C*, ps [ -]

0.08

5
0.06
0
0.04
-5
10 15 20 25 10 15 20 25
Uref [ms-1] Uref [ms-1]

Fig. 6 (a) Effect of wind speed on normalized concentrations from a point source at different locations below
canopy height for C10S and RM10S surfaces. (b) Percentage deviation from average concentration in the
wind speed range 16–24 m s−1 (grey area indicates ±1.5% reproducibility range)

RM10S surface by Cheng and Castro (2002a). Figure 6b shows the percentage deviations
from the average concentration in the wind speed range 16–24 m s−1 . The grey area indicates
the experimentally derived reproducibility of the concentration measurements (±1.5%). At
all wind speeds the averaging time was sufficient to ensure that the standard errors of the
observed concentrations were comparable with the experimental reproducibility.
The results also show a different dependence on Reynolds number for different measure-
ment locations. The left pointing triangle symbols in Fig. 6 denote concentration measure-
ments at (X, Y, Z ) = (3015, 5, 6) mm, which is 0.5H downstream of a roughness element.
Figure 6b shows a smaller dependence of concentration on wind speed for this location for
the RM10S surface when compared to C10S. The maximum vertical extent of a cuboid’s
recirculation zone is similar to the cuboid’s height. For the RM10S surface, the upstream
element is 6.4 mm high, and therefore the specified measurement position is close to the
top of the recirculation zone. Thus, changes of the cavity zone shape have less influence on
the flow characteristic at the given position. In the C10S roughness, however, the upstream
element has a height of 10 mm, so that the measurement location is within the recirculation
zone where the flow field may be more sensitive to wind speed.
The naphthalene sublimation technique can be compromised at low wind speeds because
transfer from the surface is then determined by a combination of forced and natural convec-
tion (Goldstein and Cho 1995). For the same reason, the low velocity limitation may add
to the Reynolds number dependence present for scalar concentrations measured above both
surfaces for wind speeds below 16 m s−1 .
Figure 7a shows the results from concentration measurements over a range of wind speeds
using the naphthalene area source described in Sect. 2.4 for both surfaces. Figure 7b shows the
percentage deviation from the average concentration in the wind speed range 20–24 m s−1 .
Generally, there is a more systematic dependence of concentration on wind speed for the
naphthalene area source compared to the point source (Fig. 6), and only for U > 20 m s−1
do variations appear to lie within the experimentally derived reproducibility of the measure-
ments, marked in grey (±3%). In comparing measurements made at Ur e f = 10 m s−1 , Fig.
7b shows percentage deviations from the high speed results of up to 40% for C10S and 30%
for RM10S in comparison to 15% and 10%, respectively for the point source. Again, the
magnitude of the deviation depends on measurement location with respect to recirculation
zones around the obstacles.
These tests indicated that concentration fields measured within the canopy using the
naphthalene area source showed undetectable changes when the reference wind speed was

123
114 F. Pascheke et al.

a) 0.3
RM10S (3015, -5, 3)
b) 100 RM10S (3015, -5, 3)
RM10S (2995, 5, 3) RM10S (2995, 5, 3)
0.25
C10S (2995, -15, 3) 80 C10S (2995, -15, 3)
C10S (2995, 0, 3) C10S (2995, 0, 3)
0.2 C10S (2995, -5, 3) C10S (2995, -5, 3)
60

∆C [%]
CN [ -]

0.15
*

40
0.1
20
0.05

0
0
0 5 10 15 20 25 0 5 10 15 20 25
Uref [ms-1] Uref [ms-1]

Fig. 7 (a) Effect of wind speed on concentrations from naphthalene area source at different locations below
canopy height for C10S and RM10S surfaces. (b) Percentage deviation from average concentration in the
wind speed range 20–24 m s−1 (grey area indicates ±3% reproducibility range)

increased above 20 m s−1 , and hence Ur e f = 20 m s−1 was used throughout all measurements
of concentration fields presented in Sects. 3.2 and 3.3.

3 Results

3.1 Transfer Velocities

Transfer velocities wT were measured in order to quantify the effect of surface roughness
type on canopy ventilation from a ground level area source. The naphthalene sublimation
method as described in Sect. 2.3 was used for three different surface roughness configura-
tions. In addition to the C10S and RM10S surfaces, a local and mild roughness change was
simulated by replacing the active source unit within the C10S roughness by a single RM10S
unit, which will be denoted as CRM10S.
Figures 8 and 9 show transfer velocity wT as a function of wind speed, Ur e f and U H
respectively for each surface. Taking the complete wind speed range into account, a linear
relationship between transfer velocity and wind speed was observed for all tested roughness
geometries with R 2 ≥ 0.99 (Figs. 8a and 9a, and Table 2). Following the discussion in
Sect. 2.6 on the low wind speed limitation of the naphthalene sublimation method, it was

a) b)
0.04 0.04

0.03 0.03
wT [ms ]
wT [ms ]

-1
-1

0.02 0.02

C10S C10S
0.01 0.01 RM10S
RM10S
CRM10S CRM10S

0 0
0 5 10 15 20 25 0 5 10 15 20 25
-1 -1
Uref [ms ] Uref [ms ]

Fig. 8 Transfer velocities as a function of wind speed Ur e f derived from the naphthalene sublimation method.
Comparison of three different surface geometries: C10S, uniform cubes; RM10S, non-uniform cuboids;
CRM10S, single non-uniform cuboid unit within uniform cubes. (a) all wind speeds, (b) high wind speeds
only

123
Wind-tunnel Modelling of Dispersion from a Scalar Area Source 115

a) b)
0.04 0.04

0.03 0.03
wT [ms ]

wT [ms ]
-1

-1
0.02 0.02

C10S C10S
0.01 RM10S
0.01 RM10S
CRM10S CRM10S

0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
-1 -1
UH [ms ] UH [ms ]

Fig. 9 Transfer velocities as a function of wind speed U H derived from the naphthalene sublimation method.
Comparison of three different surface geometries: C10S, uniform cubes; RM10S, non-uniform cuboids;
CRM10S, single non-uniform cuboid unit within uniform cubes (U H from RM10S surface). (a) all wind
speeds, (b) high wind speeds only

Table 2 Observed linear relationships between transfer velocity wT and wind speed Ur e f or U H respectively,
for the complete wind speed range, and for high wind speeds only. Dynamical properties of each surface are
included for comparison

Dynamical Complete wind speed range High wind speeds


propertiesa wT = a Ur e f + c; wT = d Ur e f ;
wT = b U H + c; wT = e U H ;
All coefficients ×10−3 All coefficients ×10−3
u∗ UH
Ur e f Ur e f a b c R2 d e R2

C10S 0.058 0.22 1.43 6.52 6.9 0.99 1.78 8.10 0.92
RM10S 0.063 0.18 1.27 7.07 8.9 0.99 1.68 9.32 0.97
CRM10S − 0.18b 1.54 8.56 8.9 0.99 1.89 10.5 0.95

a Taken from Cheng and Castro (2002a), based on u determined in the inertial sublayer

b U from RM10S surface
H

concluded that the non-zero offset was a result of mixed convection processes at low wind
speeds. However, when data points for lower wind speeds were excluded, the measurements
were sufficiently well described by zero offset linear relationships, with R 2 ≥ 0.92 (Figs.
8b and 9b). The fact that some data points for the CRM10S surface could be included for
Ur e f < 20 m s−1 is perhaps explained by the increased turbulence due to the roughness
change dominating over molecular diffusion processes, and therefore leading to a weaker
dependence on Reynolds number. A repeatability of ±1.5% was determined for the fit param-
eters from repeated measurements for the RM10S roughness.
Focussing on the high wind speed results in Table 2, coefficients d and e are each equiva-
lent to the dimensionless transfer coefficient wT /U (Eq. 6), which represents the ventilation
efficiency of the surface. Using the free stream velocity Ur e f to derive the transfer coefficient
(coefficient d), the C10S roughness has a 6% higher ventilation efficiency than the RM10S
roughness. This might be explained by comparison with U H /Ur e f for the two surfaces (col-
umn 3): the wind speed at mean canopy height is 4% higher for the C10S surface. Coefficient
d for the CRM10S surface is 12.5% larger than for the RM10S surface, however the dynam-
ical properties were not known for this surface. Given that the configuration represents a
smooth-to-rough roughness change, a deceleration in streamwise flow would be expected

123
116 F. Pascheke et al.

at roof height, which might be compensated by lateral and vertical deflections of the mean
streamlines (by continuity), which in combination lead to better ventilation.
Using U H to derive the transfer coefficient (coefficient e) is equivalent to comparing
ventilation efficiencies for the same wind speed at rooftop. In this case, e is 15% higher
for the RM10S surface (note that the difference is about twice as large if wT is based on
the wind speed at a height of 1.5H ). This might be explained by the increased roughness
of the surface, as u ∗ /Ur e f is 10% higher for the RM10S surface. The dynamical properties
of the CRM10S surface were not known, but local increases in shear stress levels above those
of the RM10S surface might be expected at the transition between the C10S and RM10S
surfaces (Kaimal and Finnigan 1994; Cheng and Castro 2006b) and this could go some way
towards explaining the value of e, which is 30% higher than for the C10S surface. Even when
using U H calculated for the C10S surface instead of the RM10S surface, the value of e is
still 7% higher than for the C10S surface. Of course, U H is not a very precisely measured
quantity and uncertainty over its value should not be discounted. However, as already noted,
use of a more accurately measurable quantity, such as the wind speed at 2H , would underes-
timate the deceleration over the single random roughness unit of the CRM10S surface. The
higher values of both d and e for the CRM10S surface suggest that a localized increase in
the standard deviation of the building height is particularly effective in enhancing ventilation
from street level.

3.2 Concentration Fields Above the Source Area

To obtain a more comprehensive picture of the ventilation mechanisms, the transfer velocities
derived for the C10S and RM10S surfaces were supplemented by high resolution concen-
tration measurements in two horizontal planes, at Z = 0.3H and Z = 1.2H , for Ur e f =
20 m s−1 . Figures 10 and 11 show contour plots of the normalized concentration fields,
the black dots indicating the underlying measurement positions. Figure 12 shows the area-
averaged concentrations for each height and surface. Note that all coordinates are normalized

C10S RM10S
z = 0.3H C*N [-]: 0.0E+00 5.0E-02 1.0E-01 1.5E-01 2.0E-01 2.5E-01 z = 0.3H C*N [-]: 0.0E+00 5.0E-02 1.0E-01 1.5E-01 2.0E-01 2.5E-01

a) 4 10 10 10 b) 4 2.8 10 2.8
10 10 10 10

10 10 10 10 13.6 10
2 2
10 10 13.6 10

10 10 10 13.6 6.4 13.6


Y/H
Y/H

0 0
10 10 10 6.4

10 10 10 10 17.2 10
-2 -2
10 10 6.4 13.6

10 10 10 2.8 10 2.8
-4 10 10 -4 10 10
-4 -2 0 2 4 -4 -2 0 2 4
X/H X/H

Fig. 10 Concentration field 3 mm above the active source area. (a) C10S surface, (b) RM10S surface. Numbers
indicate obstacle height in mm, black dots indicate measurement locations

123
Wind-tunnel Modelling of Dispersion from a Scalar Area Source 117

C10S RM10S
z = 1.2H C*N [-]: 0.0E+00 5.0E-02 1.0E-01 1.5E-01 2.0E-01 2.5E-01 z = 1.2H C*N [-]: 0.0E+00 5.0E-02 1.0E-01 1.5E-01 2.0E-01 2.5E-01

a) 4 b) 4

13.6
2 2
13.6

13.6 13.6
Y/H

Y/H
0 0

17.2
-2 -2
13.6

-4 -4
-4 -2 0 2 4 -4 -2 0 2 4
X/H X/H

Fig. 11 Concentration field 12 mm above the active source area. (a) C10S surface, (b) RM10S surface.
Numbers indicate obstacle height in mm, black dots indicate measurement locations

0.14
C10S Z=0.3H
0.12 C10S Z=1.2H
RM10S Z=0.3H
0.1 RM10S Z=1.2H
0.08
CN [ -]
*

0.06

0.04

0.02

0
-4 -3 -2 -1 0 1 2 3 4 5
X/H

Fig. 12 Area-averaged concentrations above the active source area for each row of roughness elements,
demonstrating the influence of rapid vertical dispersion

using the mean element height H with (X/H, Y /H ) = (0, 0) being the centre of the source
area.
Figure 10a shows the mean concentration distribution at Z = 0.3H for the C10S sur-
face. The concentration distribution shows a very regular pattern with increasing magnitude
in the direction of the flow (left to right). This is due to streamwise advection of naphtha-
lene vapour. A regular pattern of concentrations is observed around individual roughness
elements: lower concentrations ahead of the roughness element, and higher concentrations
behind. This is consistent with LES simulations of flow for the same experimental layout
(Z. Xie, pers. comm., 2007), which show higher velocities in the impact zone in front of a
roughness element, and lower velocities and strong updraughts in the wake of each roughness
element. Hence, the vertical transport is controlled by the repeated flow structures generated
by individual roughness elements.
Figure 10b shows that roughness element height variation has a clear influence on the
transport processes near the ground. The concentration pattern is no longer regular and the
streamwise increase in concentration due to advection is less continuous. In general, lower
concentrations are found in front of roughness elements and higher concentrations behind,

123
118 F. Pascheke et al.

but the interaction of the roughness element wakes influences the transport processes. This is
especially evident around the highest roughness element (17.2 mm). The low concentrations
in front of the roughness element and to the sides suggest localised patches of enhanced
wind speed, the pattern of which is similar to a horseshoe vortex, where a strong downwash
would be created at the roughness element front, which could also affect concentrations in
the wakes of the two upstream roughness elements.
Figure 11 shows the concentration patterns for Z = 1.2H , with the measurement plane just
above the roughness elements for the C10S surface. The striking difference for the RM10S
surface (see Fig. 11b) is that roughness elements intersect the measurement plane and large
concentrations are observed in their wakes. This leads to a large difference in concentrations
when averaged laterally for each of the four rows of roughness elements in the unit, which
is shown in Fig. 12 for each roughness element row.
Despite large differences in flow patterns at Z = 0.3H (leading to large differences
in row-averaged concentrations), the difference in the area-averaged concentration for the
whole unit is negligible (<5%). However, at Z = 1.2H , the area-averaged concentration is
18% larger for the RM10S surface than for the C10S surface. This is an example of vertical
transport in the wake of the taller roughness elements (so-called ‘rapid vertical dispersion’).

3.3 Concentration Fields Downstream of the Source Area

In addition to understanding dispersion mechanisms in the vicinity of the source area, con-
centration measurements were made in horizontal planes downstream of the source area to
determine far field dispersion characteristics. Figures 13, 14, 15 and 16, respectively show
the measurement planes at Z = 0.6H, 1.2H, 1.8H and 2.4H , respectively, the horizontal
extent of which were determined by the minimum detectable naphthalene vapour concentra-
tion. Some general features can be seen by looking at Figs. 13–16 in sequence. For Z = 0.6H
the plume concentration pattern is more elongated downstream for the C10S surface in com-
parison to the RM10S surface, which may be explained by increased wind speeds within
the roughness elements, leading to greater advection. The lateral spread of the plume is
greater for the RM10S surface, which is particularly clear at Z = 1.8H and Z = 2.4H . The
influence of rapid vertical transport in the wake of individual high roughness elements can

C10S RM10S
z = 0.6H C*N [ -]: 0.00E+00 8.00E-03 2.20E-02 3.80E-02 5.40E-02 z = 0.6H C*N [ -]: 0.00E+00 8.00E-03 2.20E-02 3.80E-02 5.40E-02

a) 10
b) 10

5 5
Y/H

Y/H

0 0

-5 -5

-10 -10
0 5 10 15 20 0 5 10 15 20
X/H X/H

Fig. 13 Concentration plume for Z = 6 mm = 0.6H . (a) C10S surface, (b) RM10S surface. Black dots
indicate measurement locations

123
Wind-tunnel Modelling of Dispersion from a Scalar Area Source 119

C10S RM10S
z = 1.2H C*N [ -]: 0.00E+00 8.00E-03 2.20E-02 3.80E-02 5.40E-02 z = 1.2H
*
CN [ -]: 0.00E+00 8.00E-03 2.20E-02 3.80E-02 5.40E-02
10 10
a) b)

5 5
Y/H

Y/H
0 0

-5 -5

-10 -10
0 5 10 15 20 0 5 10 15 20
X/H X/H

Fig. 14 Concentration plume for Z = 12 mm = 1.2H . (a) C10S surface, (b) RM10S surface. Black dots
indicate measurement locations

C10S RM10S
z = 1.8H C*N [ -]: 0.00E+00 8.00E-03 2.20E-02 3.80E-02 5.40E-02 z = 1.8H C*N [ -]: 0.00E+00 8.00E-03 2.20E-02 3.80E-02 5.40E-02

a) 10
b) 10

5 5
Y/H

Y/H

0 0

-5 -5

-10 -10
0 5 10 15 20 0 5 10 15 20
X/H X/H

Fig. 15 Concentration plume for Z = 18 mm = 1.8H . (a) C10S surface, (b) RM10S surface. Black dots
indicate measurement locations

clearly be seen in Fig. 14, which shows the concentrations at Z = 1.2H for both surfaces.
Concentrations in the wakes of roughness elements intersecting the measurement plane act
as point-like sources. The plume shape is thus strongly influenced by the locations of these
elevated sources. For the measurement planes at Z = 1.8H and Z = 2.4h, the plume shape
becomes more consistently Gaussian further downstream, indicating effective mixing.

3.3.1 Lateral Concentration Profiles

Lateral mixing within the plume is further demonstrated by Figs. 17a and b, which show
concentrations as a function of Y /H within the plume, for each row of measurements
downstream of the source area at Z = 0.6H . The measured data are shown along with
fitted Gaussian curves (dashed lines) using Eq. 14:

123
120 F. Pascheke et al.

C10S RM10S
z = 2.4H C*N [ -]: 0.00E+00 8.00E-03 2.20E-02 3.80E-02 5.40E-02 z = 2.4H C*N [ -]: 0.00E+00 8.00E-03 2.20E-02 3.80E-02 5.40E-02

a) 10 b) 10

5 5

Y/H
Y/H

0 0

-5 -5

-10 -10
0 5 10 15 20 0 5 10 15 20
X/H X/H

Fig. 16 Concentration plume for Z = 24 mm = 2.4H . (a) C10S surface, (b) RM10S surface. Black dots
indicate measurement locations

a) 0.07 C10S
b) 0.07 RM10S
0.06 X / H = 5.5 0.06 X / H = 5.5
X / H = 7.5 X / H = 7.5
X / H = 9.5 X / H = 9.5
0.05 X / H = 11.5 0.05 X / H = 11.5
X / H = 13.5 X / H = 13.5
X / H = 15.5 X / H = 15.5
0.04 0.04
CN [ -]

CN [ -]

X / H = 17.5 X / H = 17.5
*

0.03 0.03

0.02 0.02

0.01 0.01

0 0
-10 -5 0 5 10 -10 -5 0 5 10
Y/H Y/H

Fig. 17 Lateral concentration distribution for each row downstream of source area centre at height Z = 0.6H .
(a) C10S, (b) RM10S surface. Dashed lines show best Gaussian fit using Eq. (14)

 
(Y /H − µ/H )2
C N∗ (Y /H ) = C N∗ , max ex p −  2 , (14)
2 σ y /H
where µ/H denotes the lateral plume centreline displacement and σ y /H is the lateral dis-
persion parameter.
It can be clearly seen that normalized concentrations for the same streamwise position are
lower for the RM10S surface, consistent with the observation of better lateral and vertical
dispersion made earlier. The comparison to the displayed best Gaussian fits shows that in
the C10S surface Gaussian concentration distributions are reached at shorter distances to
the source area. The R 2 parameter was used to access the suitability of the Gaussian plume
model. Based on a criterion of R 2 ≥ 0.99, a good representation of the lateral plume spread
by a Gaussian distribution is found for X/H ≥ 7.7. For the RM10S surface the lateral
distributions are initially inhomogeneous, with some evidence of a bimodal distribution at
X/H = 5.5. However, with the above criterion, a Gaussian plume shape is achieved from
X/H = 11.5 onwards.
The plume centreline displacement µ/H is shown in more detail in Fig. 18a and b. For
the C10S surface, the values scatter around a mean value of µ/H = 0.09 ± 0.05. In con-
trast, a height dependence is found in the RM10S geometry for each streamwise position
downstream of the centre source area, with negative µ/H values within the canopy and

123
Wind-tunnel Modelling of Dispersion from a Scalar Area Source 121

a) 0.6 C10S
b) 0.6 RM10S
Z / H = 0.6 Z / H = 0.6
0.4 Z / H = 1.2 0.4 Z / H = 1.2
Z / H = 1.8 Z / H = 1.8
Z / H = 2.4 Z / H = 2.4
0.2 0.2
µ / H [ -]

µ / H [ -]
0 0

-0.2 -0.2

-0.4 -0.4

-0.6 -0.6
5 10 15 5 10 15
X/H X/H

Fig. 18 Plume centreline displacement for each row downstream of the source area centre for each measure-
ment height. (a) C10S, (b) RM10S surface

slightly positive µ/H values at Z = 2.4H . Considering the locations of the individual high
roughness elements present in the RM10S geometry, this lateral shift of the plume centreline
displacement can be attributed to the additional sources in the wakes of the tallest roughness
elements.
The downstream growth of the normalized lateral plume spread parameter σ y /H is shown
in Figs. 19a and b. Two main differences can be found for the two surfaces The lateral dis-
persion parameters are significantly larger for the RM10S surface, while another notable
difference is the absence of a height dependence for the RM10S surface: the dispersion
parameters of the individual vertical measurement planes collapse to the same value. In
contrast, for the C10S surface, the lateral plume spread decreases with height. This marked
difference may be due to enhanced vertical mixing within the roughness sublayer of the
RM10S surface. The dispersion processes within the canopy differ substantially from those
in the boundary layer above. Initial mixing and diffusion within the canopy depend largely on
the immediate presence of obstacles (i.e. wake interference processes with scales determined
by the obstacle dimensions). In the boundary layer above, dispersion processes depend on
the mean and turbulent boundary-layer characteristics and σ y can be related to the turbu-
lence intensity σv /U . Although the Gaussian plume model gives a good description of the
plume concentrations within the considered canopies (as demonstrated in Figs. 17a and b),
the derived plume spread parameters do not necessarily match those for the boundary-layer
above. As a consequence, adjustment with height takes place for the C10S surface. It can be
assumed that the numerous turbulent wakes generated within the RM10S canopy penetrate
deeper into the boundary-layer above. Thus, the associated lateral mixing processes happen
to remain constant within the investigated height range. The initial height distribution at

a) 4.5 C10S
b) 4.5 RM10S
Z / H = 0.6 Z / H = 0.6
Z / H = 1.2 Z / H = 1.2
4 Z / H = 1.8 4 Z / H = 1.8
Z / H = 2.4 Z / H = 2.4
σY / H [ -]

σY / H [ -]

3.5 3.5

3 3

2.5 2.5
5 10 15 5 10 15
X/H X/H

Fig. 19 Lateral plume spread for each row downstream of source area centre. (a) C10S, (b) RM10S surface

123
122 F. Pascheke et al.

X/H = 5.5, and to a lesser extent at X/H = 7.5, can be attributed to the additional sources
in the wakes of taller buildings acting at different heights.

3.3.2 Vertical Concentration Profiles

Figures 20a and b show the measured vertical normalized concentrations for each row of
measurements downstream of the source area centre at the position closest to the centreline
(Y /H = 0.5 or Y /H = −0.5, respectively). The measured data are shown along with the
best fit of the Gaussian profile given in Eq. 15:
 
∗ ∗ (Z /H − η/H )2
C N (Z /H ) = C N , max ex p − , (15)
2 (σz /H )2

where η/H is a zero plane adjustment for optimum curve fitting. In both surfaces, Gaussian
distributions with R 2 ≥ 0.99 were found at all downstream positions. In the RM10S surface
the profiles have smaller concentration gradients with height, indicating more efficient verti-
cal dispersion. This finding is supported by the downstream growth of the normalised vertical
plume spread parameter σz /H shown in Fig. 21a, which is somewhat larger for the RM10S
surface. It is also notable that for both surfaces at X/H = 5.5 the smallest concentrations
can be found at Z /H = 2.4. The plume concentration fields in Figure 16a and b clearly show
that maximum concentrations at Z /H = 2.4 occur further downstream.
Figure 21b shows the effective plume height in more detail, where the adjustment occurs
with fetch and height depending on the given in-canopy morphology. Figure 21b indicates
that the plume centre rises more rapidly between the first two rows for the RM10S surface.
However, given the ‘noisiness’ of this parameter, these differences might be subject to fitting
errors.

a) 3 b) 3
C10S RM10S
X / H = 5.5 X / H = 5.5
2.5 X / H = 7.5 2.5 X / H = 7.5
X / H = 9.5 X / H = 9.5
X / H = 11.5 X / H = 11.5
X / H = 13.5 X / H = 13.5
2 2 X / H = 15.5
X / H = 15.5
X / H = 17.5
X / H = 17.5
Z/H

Z/H

1.5 1.5

1 1

0.5 0.5

0 0
0 0.02 0.04 0.06 0.08 0 0.02 0.04 0.06 0.08
C*N [ -] C*N [ -]

Fig. 20 Vertical concentration distribution for each row downstream of source area centre at Y /H = 0.5 or
Y /H = −0.5 respectively. Dashed lines show fitted Gaussian profiles (Eq. 15). (a) C10S, (b) RM10S surface

123
Wind-tunnel Modelling of Dispersion from a Scalar Area Source 123

a) 2.5 C10S
b) 1
C10S
RM10S RM10S

2
σZ / H [ -]

η / H [ -]
0.5
1.5

0
0.5
5 10 15 5 10 15
X/H X/H

Fig. 21 (a) Vertical plume spread for each row downstream of source area centre for the C10S and the
RM10S surface, (b) vertical displacement of concentration maximum for each row downstream of the source
area centre for the C10S and the RM10S

4 Conclusions

Naphthalene coated surfaces were used to represent area sources at street level and thereby
to study ventilation from two different urban-like geometries at neighbourhood scale: one
with uniform height roughness elements, the other with non-uniform heights. Area-averaged
scalar fluxes were measured and the results represented in terms of a dimensionless transfer
coefficient wT /Ur e f . Naphthalene concentrations were measured with high spatial resolu-
tion in two horizontal planes above the active source area to capture the scalar transport at
street scale. Coarse mesh concentration measurements at four different heights in the wake of
the active surface were also conducted to study differences in the plume spread arising from
the different geometrical surface characteristics. A novel Fast Flame Ionisation Detection
technique was developed to measure naphthalene vapour concentrations.
For both of the geometries studied, the plan area density as well as the frontal area density
was λ p = λ f = 25%. Two different height distributions were studied, one uniform and
the other randomly (Gaussian) distributed but with the same mean height. The difference
in height distribution had a significant effect on the ventilation efficiency of the geometries
considered. The transfer coefficient was calculated using two different reference velocities.
Using U H , which represents the flow at the top of the canopy, resulted in an increased ven-
tilation coefficient for the non-uniform height array. Using the free-stream velocity, Ur e f ,
which is unaffected by the surface roughness, yielded a larger transfer coefficient for the
uniform height roughness. The latter definition of the transfer coefficient is probably a more
useful representation of ventilation from the canopy as the incident flow at higher elevation
drives the exchange and is largely unaffected by changes in the surface morphology. The
results demonstrate that, whilst there is enhanced vertical momentum exchange due to height
variability, the dispersion from a limited area source within the canopy is not enhanced.
This is because the advection is reduced. The height variability of a canopy must therefore
be considered as an important parameter when generalising urban dispersion characteristics
based on morphometric methods—its influence on both advection and dispersion should be
acknowledged.
Fine resolution concentration measurements close to the source elucidated the effect
of height variability on dispersion at street scale. Both array geometries generated similar
area–averaged concentration values close to the active surface at Z = 0.3H but the area-
averaged concentration at Z = 1.2H was 18% higher for the non-uniform surface. The
important role of individual high-rise roughness elements on local ventilation processes was
seen in the concentration measurements at both heights. At street level, the tallest roughness

123
124 F. Pascheke et al.

elements affected the flow around nearby roughness elements, showing that the lateral extent
of their influence is large. Surface released scalars were also transported upwards in the wake
of individual high roughness elements, thus generating point-like sources at elevated heights.
Coarse resolution concentration measurements downstream of the source showed sig-
nificant differences in the location and shape of the plume. The non-uniform height array
generated additional turbulence within the canopy, resulting in more efficient mixing pro-
cesses. As a consequence, the lateral and vertical plume spread was increased and plume
dilution was intensified in and above the RM10S surface.
The results presented show the potential of using a naphthalene area source in wind-tunnel
studies of urban dispersion problems. Bulk fluxes of scalars can be derived from the naphtha-
lene sublimation technique to quantify area-averaged ventilation. The novel measurement of
naphthalene vapour concentrations using a Fast Flame Ionisation Detector technique proved
to be sensitive enough to capture variability of concentration fields within the urban arrays.
Thus information on the exchange processes between neighbourhood and street scales could
be gained. The combined use of the sublimation method of flux determination and FID plume
mapping has been found to be a powerful research tool. Its utility will be further enhanced in
future studies by concurrent measurements of the flow field, for example using pulsed wire,
laser Doppler anemometry (LDA) or particle image velocimetry (PIV).

Acknowledgements The authors are very grateful for the outstanding technical support from Paul Hayden,
Tom Lawton and Allan Wells at the EnFlo laboratory. The project was funded by the Engineering and Physical
Sciences Research Council, grant number GR/S71798/01

References

Barlow JF, Belcher SE (2002) A wind tunnel model for quantifying fluxes in the urban boundary layer. Bound-
ary-Layer Meteorol 104:131–150
Barlow JF, Harman IN, Belcher SE (2004) Scalar fluxes from urban street canyons. Part I: Laboratory simu-
lation. Boundary-Layer Meteorol 113:369–385
Britter RE, Hanna SR (2003) Flow and dispersion in urban areas. Ann Rev Fluid Mech 35:469–496
Cheng H, Castro IP (2002a) Near wall flow over urban-like roughness. Boundary-Layer Meteorol 104:229–
259
Cheng H, Castro IP (2002b) Near-wall flow development after a step change in surface roughness. Boundary-
Layer Meteorol 105:411–432
CRC Handbook of Chemistry and Physics (1993) Lide DR (ed) 74th edn. CRC Press Inc.
Goldstein RJ, Cho HH (1995) A review of mass transfer measurements using naphthalene sublimation. Exp
Thermal Fluid Sci 10:461–434
Hall DJ, Macdonald R, Walker S, Spanton AM (1996) Measurements of dispersion within simulated urban
arrays - A small scale wind tunnel study. BRE Client Report 178/96, Build. Res. Establ., Garston, Watford,
UK
Heist DK, Brixey LA, Perry SG, Bowker GE (2005) Residence time measurements in an array of build-
ings. In: Proceedings of PHYSMOD 2005, Int. workshop on physical modelling of flow and dispersion
phenomena
Kaimal JC, Finnigan J (1994) Atmospheric boundary layer flows: Their structure and measurement. Oxford
University Press, New York, 289 pp
Oke T (1987) Boundary-layer climates. Routhledge, London, 450 pp
Uehara K, Wakamatsu S, Ooka R (2003) Studies on critical Reynolds number indices for wind-tunnel exper-
iments on flow within urban areas. Boundary-Layer Meteorol 107:353–370

123

You might also like