Theor Appl Climatol (2011) 106:127–137

DOI 10.1007/s00704-011-0415-y

ORIGINAL PAPER

Experiments on integral length scale control in atmospheric

boundary layer wind tunnel
Kapil Varshney · Kamal Poddar

Received: 25 July 2010 / Accepted: 3 February 2011 / Published online: 13 March 2011
© Springer-Verlag 2011

Abstract Accurate predictions of turbulent charac- Nomenclature

teristics in the atmospheric boundary layer (ABL)
depends on understanding the effects of surface κ von Kármán constant
roughness on the spatial distribution of velocity, tur- w Slot width
bulence intensity, and turbulence length scales. Simu- d Displacement height
lation of the ABL characteristics have been performed f Frequency
in a short test section length wind tunnel to deter- fs Vortex shedding frequency
mine the appropriate length scale factor for modeling, St Strouhal number
which ensures correct aeroelastic behavior of structural u Fluctuating velocity in x-direction
models for non-aerodynamic applications. The ABL U Absolute velocity in x-direction
characteristics have been simulated by using various U Mean velocity in x-direction
configurations of passive devices such as vortex gen- Uz Mean velocity in x-direction at height z
erators, air barriers, and slot in the test section floor u∗ Shear velocity
which was extended into the contraction cone. Mean U ref Reference velocity
velocity and velocity fluctuations have been measured Uδ Freestream velocity
using a hot-wire anemometry system. Mean velocity, x Distance in the direction of the flow
turbulence intensity, turbulence scale, and power spec- y Spanwise distance from test section
tral density of velocity fluctuations have been obtained centerplane
from the experiments for various configuration of the z Vertical distance from wind tunnel floor
passive devices. It is shown that the integral length scale zref Reference height
factor can be controlled using various combinations of z0 Aerodynamic surface roughness length
the passive devices. Lux Longitudinal integral length scale of turbulence
Rux Longitudinal auto-correlation coefficient
S Length scale factor
Puu (f) Power spectrum of longitudinal velocity
fluctuations
α Power law exponent
δ Boundary layer thickness
K. Varshney (B) σu Standard deviation of u
Taitem Engineering, PC, 110 S Albany St,
Ithaca, NY, USA ν Air viscosity
e-mail: [email protected] AF Frontal area of all passive devices
AT Total surface area
K. Poddar λF Frontal area density
Department of Aerospace Engineering,
Indian Institute of Technology, Kanpur, m Model
Uttar Pradesh, 208016, India p Prototype
128 K. Varshney, K. Poddar

1 Introduction parameters include turbulence intensity and integral

length scales.
Prediction of the distribution of wind velocity and tur- The natural wind environment varies with the loca-
bulence scales downwind of a rough terrain is important tion and time. Therefore, it is very difficult to define
as a boundary condition for the design of wind energy the characteristics of ABL that can be taken as a pro-
projects such as simulation of tall structures. Boundary totype. Counihan (1969, 1971, 1973, 1975) presented
layer characteristics varies as it interacts with different and discussed the data of various researchers point-
obstacles. According to the American Association for ing out various discrepancies, but does not prescribe
Wind Engineering, the estimated losses due to severe a design data. It was represented by Counihan that
wind events are more than $5 billion annually. The gov- if the atmospheric and flat plate boundary layer data
erning lateral load of tall buildings is a large part of the are compared on the basis of non-dimensional height,
cost of the building and high wind also affects the cost z/δ with δ = 300 m, very little agreement is obtained
of the maintenance of such buildings. The stability of for rural data but using δ = 600 m gives better agree-
tall structures depends on its shape, damping, stiffness, ment. Pasquill (1967) suggested that the mechanical
and mass distribution which must be meticulously con- reproduction of energy tended to approach zero at a
sidered for the optimal designing. height of 600 m, and therefore, 600 m is taken as the
Vortex shedding frequency ( fs ) can also affect tall upper limit of boundary layer length. Counihan sug-
structures. The Strouhal number (St = fs W/U c ) for gested that the boundary layer height is affected very
the vortex shedding gives frequency of the vortices little by changes in terrain roughness. ESDU (1972,
at which they shed from the tall structures, where U c 1974–1975, 1985) data are most commonly used as the
is vortex convection velocity (also known as vortex prototype data. In choosing what may be considered
celerity) and W is building width. Sax (1978) sug- as a practical and typical boundary layer height, some
gested that increasing the intensity of turbulence low- compromises must be made, e.g., the boundary layer
ers the frequency of vortex shedding and causes the height is clearly a function of both gradient wind speed
vortex shedding regimes to occur at lower Reynolds and surface roughness. On the basis of review data
number. Strouhal number also depends on blockage for high wind speeds (U > 5−7 m/s) which produce
ratios. Using the shedding frequencies fs of an ob- adiabatic conditions, a value of 600 m is recommended
ject from spectral density functions, the shedding fre- (Counihan 1975; ESDU 1972, 1974–1975) as represent-
quency parameter (non-dimensional frequency), F = ing the average height of both rural and urban bound-
fs D2 /ν is calculated, where ν is the kinematic vis- ary layers. Davenport and Isyumov (1967) suggested
cosity of the fluid and D is the characteristic length that boundary layer height is a function of terrain type
of the object. The shedding frequency parameter is a and proposed the standard heights for rural and urban
function of both the Reynolds number and Strouhal terrains.
number. The shedding frequency parameter increases There have been many meteorological investigations
linearly with the Reynolds number (Nishioka and Sato made to duplicate a “typical” ABL in a wind tunnel.
1974). Jensen (1958) demonstrated that a model of a tall build-
Velocity measurements in turbulent flows can be ing must be scaled down at the same geometric scale as
performed accurately in both transparent and opaque ABL. Neutral flow, in which buoyancy effects are ab-
fluids (Cramer et al. 2006; Kozmar 2010). Simulation sent, is readily produced in a wind tunnel. Because the
of the velocity profile of the atmospheric boundary wind tunnels do not have sufficient test section length,
layer (ABL) is relatively simple. However, it is not the therefore combinations of passive devices have been
velocity profile alone that needs to be simulated for used to artificially increase the boundary layer depth,
accurate prediction of wind-induced loads and aero- to create natural wind profiles and turbulence profiles.
elasticity in structures, but the level of turbulence and Counihan’s (1973, 1975) method to simulate the ABL
the integral length scales must be simulated accurately. in a wind tunnel is well-known. He demonstrated that
One of the most important parameters to be simulated the length of the test section must be four to five times
in relation to wind loads on structures is the Reynolds of boundary layer thickness to reproduce ABL in a
number (Re = Ud/ν) which is based on mean velocity wind tunnel. In his study, the effect of four simple
and the characteristic building dimension. However, vortex generator shapes, triangular, cranked triangular,
due to geometric limitations and power requirements, it plane elliptic, and elliptic wedge on the ABL were
is difficult to simulate Re in the wind tunnel. The turbu- examined in a boundary layer wind tunnel.
lence characteristics of the wind are usually represented In an urban terrain, building arrangements are quite
in terms of statistical and spectral parameters. These complicated. Therefore, different configurations of
Experiments on integral length scale control in atmospheric boundary layer wind tunnel 129

Table 1 Typical values of aerodynamic surface roughness length

passive devices must be used in the wind tunnel to
see the effect of ABL characteristics on such buildings, Terrain Roughness length (m)
which demonstrate the important features of the real- Rural 0.01–0.15
Suburban 1.0
world configurations. Indeed, wind tunnel experiments
Urban 2.0–4.0
using various configurations of passive devices give
a comprehensive understanding of the physical phe-
nomenon governing the turbulence transport in the
ABL, velocity distribution, and integral length scales Simulation of the ABL in the wind tunnel can be
which occur at different terrains in the real world. achieved by matching certain wind parameters ob-
The objective of the paper is to investigate how the tained from the field. Some of these parameters are
turbulent scales varies as the wind disperses through the power law exponent (α), the aerodynamic surface
a large staggered array of obstacles such as roughness roughness length (z0 ), the shear velocity (u∗ ), and the
elements, barriers, and vortex generators in order to turbulence intensity profile. However, the typical trend
gain a better understanding of the physical processes is to use log-law representation for the lower 50 m. It is
involved in phenomenon in a build-up urban areas. quite incorrect to deduce the power-law exponent from
A wood panel was used to extend the test section in the measurements at this height range. However, it is
the contraction zone. Two slots were provided in the used to represent the meteorological data over a much
inner region of the extended test section so that the air larger height, outside the first 50-m range. There are
blowing through the slots injected momentum directly sufficient data available to deduce the typical values of
toward the top of the boundary layer. In particular, aerodynamic roughness length in various terrains. The
18 different configurations of the passive devices have typical values of the roughness length are presented in
been used and measurements have been performed Table 1. Counihan (1975) showed that z0 depends on
in the national wind tunnel facility (NWTF). These the power law exponent.
experiments allowed to examine the problem of scaling Power law in the form
effects in relating small-scale laboratory experiments to  α
U z1 z1
full-scale field experiments. = (2)
U z2 z2
provides a good fit to most of the meteorological data,
particularly in the case of high winds. Since the power
2 Atmospheric boundary layer and aerodynamic
law can be mathematically manipulated more easily
roughness
than a log law, it tends to be used more often in
meteorological problems. Here 1/α can be taken as an
The ABL, or planetary boundary layer, is a thin layer
indication of the amount of turbulence present. The
adjacent to the earth’s surface. Across this layer, flow
value of power index for rural terrain is proposed as
adjusts itself from boundary conditions near the earth’s
0.143, but for urban terrain, mean velocity measure-
surface to the flow conditions outside boundary layer.
ments are much more difficult. The value of power
The ABL extends from few meters to few kilometers
index for urban terrain has a significant range starting
(1 to 2 km). The ABL can be divided in two layers.
from 0.23 to 0.45 (Kozmar 2010; Counihan 1975; ESDU
The lower surface layer of ABL is approximately 50 m
1985). Davenport and Isyumov (1967) suggested that
which rotates with earth and has a constant shear stress.
the value of power index that can be used are 0.23 and
The region above the first 50 m (up to approximately
0.40 for suburban and urban terrains, respectively.
1 to 2 km) is affected by earth’s rotation as well as
the terrain roughness (Kaimal and Finnigan 1994). The
lower part of the mean velocity profile can be best
represented by a log-law. By utilizing the flat-plate 3 Experimental procedure
boundary layer theory of Prandtl and Von Kármán, the
velocity profiles can be represented in the form Flow simulation data were obtained from the boundary
layer wind tunnel. Natural growth of boundary layer
 
u∗ z in the wind tunnel requires long test sections (length
Uz = ln (1) about eight to ten times the height of the test section).
κ z0
However, for short test section wind tunnels, artificial
where κ is a non-dimensional parameter, known as the methods (placing the roughening devices on the up-
von Kármán constant. The value of κ is equal to 0.4 for stream) have to be adopted for simulation of the ABL
the atmospheric boundary layer. characteristics.
130 K. Varshney, K. Poddar

NWTF was a very low turbulence wind tunnel 2.25 m high. Wind speeds up to 80 m/s can be generated
specially designed for aeronautical applications. The by using a 12 blade blower, and the turbulence level
length of the test section was not sufficient for bound- in this wind tunnel is less than 0.8%. Temperature
ary layer depth to occur over the test location. There- stratification in the wind tunnel was negligible (<0.2◦ C)
fore, artifices have been used to make the boundary (Varshney and Panigrahi 2005).
layer grow. The wind tunnel was a closed-return type, Several configurations have been employed to simu-
and the test section is 8.75 m long, 3 m wide, and late the lower part of the atmospheric boundary layer

Fig. 1 Arrangements of simulation hardware in the wind tunnel test section. The simulation hardware include three castellated barrier
walls, five constant angled wedge-shaped vortex generators, total 72 roughness elements, and an extended test section in contraction
zone
Experiments on integral length scale control in atmospheric boundary layer wind tunnel 131

using different combinations of passive devices which fourth-order polynomial curve fit through the points
included constant-wedge angle spires with height 2 m, was done to obtain the transfer function which was used
three castellated barrier walls of height 10 cm, and to convert measured voltages from the CTA with the
many roughness elements mounted on floor to produce hot wire probe into velocity sample with an uncertainty
initial momentum defect in the boundary layer (Fig. 1). of approximately 1.5%.
Each roughness element was a rectangular piece with
a width of 15 cm, length 0.5 cm, and height of 20 cm.
Roughness elements were placed in 12 staggered rows
4 Methodology
of six elements in each row with an average spacing of
30 cm in lengthwise x- as well as spanwise y-directions
The lateral homogeneity and the influences of the side
(Fig. 1). Roughness elements covered a total of 3.6 m
walls on the air flow have been investigated at 15
length of the test section. The frontal area density of
different lateral positions at the time of experiments.
the obstacle array varied from 0.06 to 0.10. For the
The velocity profiles around the center of the tun-
calculation of the frontal area density, the frontal area
nel were reasonably homogeneous with mean wind
(AF ) of each obstacle, area of each barrier, and area
speed uz . Decelerated wind speeds in the outer velocity
of each vortex generator were considered. Further, this
profiles next to the side walls were within acceptable
frontal area depends on the direction of the airflow. In
limits. In addition, turbulence structure remained sim-
our calculations, the frontal area density is defined as
ilar at all sufficiently high Reynolds number (>105 ).
the ratio of the frontal area and the total surface area
Surfaces considered in this study were sufficiently
(λF = AF /AT ). The normal to the long face of the array
rough and exhibited Reynolds number similarity. Ad-
was oriented toward the prevailing wind directions,
ditional measurements were made to assure Reynolds
resulting in a flat and homogeneous upwind.
number invariance, lateral uniformity, and longitudinal
An automated (computer controlled) three-axis tra-
uniformity.
versing mechanism has been used for traversing the hot
In order to obtain the integral length scale, the ran-
wire probe. It has a capability to move in three principal
dom behavior of a fluctuating velocity at a particular
directions (i.e., three-translation degrees of freedom)
point in space has been a point of discussion. For
with maximum possible displacement of 8.75, 2.8, and
example, for small objects, it would be reasonable to
1.70 m in the streamwise (x), spanwise (y), and vertical
assume that the wind force acting due to gustiness on
directions (z), respectively. Its positioning accuracy is
a point could be approximated as the force acting over
of the order of ±0.1 mm over entire range.
the whole object. However, for large constructions such
In this study, a hot-wire probe (P-11) connected
as low-rise buildings, bridges, or towers, this can no
with DANTEC streamline DT-11, constant tempera-
longer be the case. Therefore, there can be no direct
ture anemometer (CTA) was used to measure instan-
relationship between the random behavior of the wind
taneous velocity signals. The hot wire probe was made
from place to place along the structure. It is nonethe-
of wollaston platinum rhodium (90/10) wire, joined
less necessary to obtain the measure of distribution of
to the prongs by etching and soft soldering. Veloc-
gust, say, along a building; thus, statistical method must
ity fluctuations were sampled at 6,000 Hz. The soft-
be employed. Integral length scales of turbulence are
ware required for the sampling was developed using
measures of average size of turbulent eddy in the flow.
the application tool LabVIEW (a virtual instrumenta-
To obtain the integral length scale, auto-correlation
tion software) having properties such as selection of
coefficient, Rux , has been used in this study.
different channels, sampling rate, and buffer size. A
16-bit analog to digital converter was used to represent
the analog signal which had a resolution of 0.075 mV [U(t) − U(t)][U(t + τ ) − U(t + τ )]
Rux = 2
(3)
in the 0–5 V range. The higher the resolution, the U
higher the number of divisions the range was broken
into, and therefore, the smaller the detectable volt- Taylor’s formula for integral length scale is adopted
age changes. The range, resolution, and gain available in this work. With the assumption of frozen turbulence,
on a data acquisition board determined the smallest the integral length scale can be estimated from the
detectable change in voltage. Calibration between the integral time scale as
CTA output and the flow velocity was performed by
exposing the probe to a set of known velocities and T
then voltages were recorded. Track of the temperature Lux = U Rux (τ )dτ, (4)
was also kept during calibration and experiments. A 0
132 K. Varshney, K. Poddar

The area under the curve of Rux gives time scale A pitot-static pressure probe in conjunction with a
which on further multiplication with mean velocity manometer is also used to measure the free stream
produces integral length scale. Estimates of turbulence velocity. The pitot-static probe was placed at the same
scales depend significantly upon the length and degree traversing system side by side of the hot-wire probe.
of stationarity of the record being analyzed, which Distance of the probes from the floor was varied,
usually vary from experiment to experiment. and the data were collected at several locations. Mean
Using the procedure proposed by Cook (1978a, b) in velocity from velocity time series was calculated as
Eq. 5, a point was selected in the boundary layer where follows.
the longitudinal length scale as measured above was Mean velocity in x-direction:
maximum. This point was quite uniformly at a height
of 40 cm in the wind tunnel. T
Uz = U(t)dt, (6)
91.1(z)0.491 0
S= m
(5)
(Lux )1.403
m (z 0 )m
0.088
where U z is mean velocity and U(t) is instantaneous
velocity at time t.
At this height, the model values of (z)m , (z0 )m , Power-law index for the velocity profile is based on
and (Lx )m were assumed with m denoting model. A the form
trial scale factor S0 = 400 was assumed. With this  α
value, the corresponding full-scale values of (z) p , (z0 ) p , Uz z
= (7)
and (Lx ) p with p denoting full scale were calculated. U ref zref
Finally, the value of the longitudinal length scale of
turbulence (Lux )a with a denoting atmosphere was To present the data in standard non-dimensional
obtained. The value of (Lux )a /(Lux )m gave a corrected forms, shear velocity was calculated using the method
value of the scale factor S1 . Obtained values were used proposed by Perry and Joubert (1963). The log law for
again to get convergence. Integral scale factors for the velocity profile was assumed in the form
all configurations are presented in Table 2. Detailed  
procedure for calculating the integral length scale can U(z) 1 z−d
= ln (8)
be found in Cook (1978a). u∗ κ z0

Table 2 Length scale factor using various configurations of passive devices

No. Configuration u∗ (m/s) α z0m (mm) z0 p (m) S
1 F.R., B.H. = 30 cm, V.G. = 5, w = 0 1.451 0.28 6.38 3.94 617
2 F.R., B.H. = 20 cm, V.G. = 5, w = 0 1.553 0.27 4.68 3.34 713
3 F.R., B.H. = 10 cm, V.G. = 5, w = 0 1.548 0.25 3.84 2.91 759
4 F.R., B.H. = 30 cm, V.G. = 5, w = 10 cm 1.462 0.29 6.51 3.94 605
5 F.R., B.H. = 20 cm, V.G. = 5, w = 10 cm 1.563 0.28 4.91 3.34 681
6 F.R., B.H. = 10 cm, V.G. = 5, w = 10 cm 1.568 0.26 4.12 3.06 743
7 F.R., B.H. = 30 cm, V.G. = 5, w = 20 cm 1.474 0.31 6.91 3.90 565
8 F.R., B.H. = 20 cm, V.G. = 5, w = 20 cm 1.601 0.29 5.18 3.38 652
9 F.R., B.H. = 10 cm, V.G. = 5, w = 20 cm 1.611 0.28 4.72 3.36 712
10 H.R., B.H. = 30 cm, V.G. = 5, w = 0 1.406 0.27 4.12 2.82 685
11 H.R., B.H. = 20 cm, V.G. = 5, w = 0 1.421 0.25 2.37 1.96 829
12 H.R., B.H. = 10 cm, V.G. = 5, w = 0 1.431 0.23 2.08 1.80 868
13 H.R., B.H. = 30 cm, V.G. = 5, w = 10 cm 1.461 0.28 4.28 2.88 673
14 H.R., B.H. = 20 cm, V.G. = 5, w = 10 cm 1.428 0.25 2.46 1.95 792
15 H.R., B.H. = 10 cm, V.G. = 5, w = 10 cm 1.440 0.24 2.44 1.98 812
16 H.R., B.H. = 30 cm, V.G. = 5, w = 20 cm 1.400 0.29 4.41 2.94 666
17 H.R., B.H. = 20 cm, V.G. = 5, w = 20 cm 1.431 0.27 2.95 2.20 744
18 H.R., B.H. = 10 cm, V.G. = 5, w = 20 cm 1.429 0.26 2.56 2.00 782
F.R. full roughness elements (72), H.R. half roughness elements (36), V.G. vortex generators, B.H. barrier height, 10 cm one plank of
10 cm with a gap of 10 cm from the ground, 20 cm two planks of 10 cm each with a gap of 10 cm between the two, 30 cm three planks
of 10 cm each with a gap of 10 cm between the two adjacent plank. w slot width in the extended test section in the contraction zone
Experiments on integral length scale control in atmospheric boundary layer wind tunnel 133

2.2
If d ≈ 0, this equation can be approximately written
as
    2.0
U(z) u∗ zU ref u∗ z0 U ref
= 5.75 log − 5.75 log
U ref U ref ν U ref ν 1.8 1:565
0.30 (Power law)
(9)
1.6
where U ref is the reference velocity taken at any conve-
nient height. In this study, U ref was taken at 1 m. With 1.4
this reference height, U z /U ref was calculated and plot-
ted against log(zU ref /ν). The value of u∗ was obtained

(z-d)/(zref-d)
1.2
by equating the slope of the best fit line for the upper
data with 5.75 u∗ /U ref . The value of z0 was obtained 1.0
from the intercept with the Y-axis by equating it to
(5.75 u∗ /U ref ) log(z0 U ref /μ). 0.8
The “intensity” of turbulence is a measure of am-
plitude of the velocity fluctuations, which occur in the
0.6
flow. Velocity fluctuations in the flow passing over a
point can be considered being caused by a superpo-
0.4
sition of eddies transported by the mean wind. The
simplest descriptor of atmospheric turbulence is turbu-
0.2
lence intensity.
Turbulence intensity in x-direction
0.0

0.0 0.2 0.4 0.6 0.8 1.0 1.2
u2 (z)
Iu (z) = (10) U / U ref
Uz
Fig. 2 Comparison of mean velocity profile (U δ = 18.23 m/s) in
where u is fluctuating velocity component in the 1:565 ABL simulation with the power law profile for α = 0.307.
x-direction. Uncertainties due to the finite number of measurements (samples
of the underlying velocity distribution) are shown by horizontal
bars

5 Results and discussion

Power-law exponent improved with the increase
With each configuration, variations of mean veloc- in the barrier height. Slots in the extended test sec-
ity, streamwise turbulence intensity, and longitudinal tion imparted the same effect on the power-law ex-
length scales of streamwise turbulence with increasing ponent, helped in proper distribution of momentum
height were measured. However, for brevity, the data deficit across the thickness of the boundary layer. Slots
collected for configuration 7 (Table 2) will only be marginally improved the value of the exponent. In
discussed. configurations where full barrier height was used, the
value of power-law exponent was found slightly larger
Influence of passive devices on power-law exponent than that of the other two heights. For configurations
from 1 to 9 (Table 2) with full surface roughness, larger
Figure 2 shows the mean velocity profile when the values of exponent was obtained than that of with half
probe was mounted in the middle of the test section and surface roughness.
0.5 m downstream from the last row of the roughness As mentioned above, 3.4-m-long wood boards were
elements. Power-law exponent was calculated for each used in the contraction zone to extend the test section
configuration used in this study. A wide range of α, floor (Fig. 2). Two slots of 10 cm wide were provided in
starting from 0.23 to 0.31, was obtained (Table 2) which the boards in the inner region of the contraction zone.
is in good agreement with the data presented in ESDU Air passing through the slots in the tunnel floor injected
(1985). Lower values of α were obtained for relatively momentum directly toward the top of the boundary
smoother terrains, and the high values of α were ob- layer. A slight variation in the velocity was found in the
tained for very rough terrains. upper velocity profile.
134 K. Varshney, K. Poddar

Influence of passive devices on aerodynamic Influence of passive devices on turbulence intensity

roughness length
The primary purpose of the spires is to generate turbu-
Figure 3 shows the log-law representation of the mean lence; however, the tapered shape is also responsible
velocity profiles for configuration 7. Log-law repre- for creating a momentum deficit near the bottom of
sentation of the mean velocity profile with full sur- the test section. An inner turbulent shear layer was
face roughness shows a logarithmic law with a distinct then developed over an array of roughness elements of
“velocity-defect” or “wake” region above. It was ob- appropriate size to produce a flow having the desired
served that, with increase in barrier height, the wake turbulence structure. Barriers in conjunction with the
region becomes less distinct, and log law extends over spires were also used to develop a boundary layer of
a larger height. Aerodynamic surface length increased adequate depth and desired characteristics. Therefore,
with the increase in barrier height. The increase in with an increase in barrier height, turbulence inten-
aerodynamics surface length made the “defect-region” sity increased. Similarly, as mentioned above, slots in
more distinct, which was consistent with the naturally the extended test section in the contraction cone and
grown boundary layers. roughness elements on the floor also assisted to in-
The main purpose of barriers was to produce initial crease the boundary layer thickness and turbulence
momentum deficit near the ground level in order to level. Turbulence intensity increased with the increase
account for the smaller roughness length downstream. in slot width and number of roughness elements on
Two flow features, namely (1) a separation bubble the floor. All dimensions of the spires were obtained
downstream and (2) a pressure gradient downstream, by keeping the total blockage and rate of decrease of
are produced by barriers in the wind tunnel. The first
effect is detrimental but unavoidable. The separation
bubble effectively nullifies the effect of surface friction
700
in the bubble region between the barrier and downwind
of the bubble. This effect is reduced by castellating the Present Data [1:565]
Kozmar [4]
edge of the barrier. On the other hand, the adverse Counihan [8]
pressure gradient is instrumental because it prompts 600 Farrel-Iyenger [18]
the growth of boundary layer. In order to diminish the Reinhold [19]
Akins-Cermak [20]
adverse effects, roughness elements on the floor have ESDU [12]
been mounted to decay these characteristics and allow
500
boundary layer to recover before it is used. Aerody-
namic roughness length also increased with the increase
in slot width and roughness elements.
Full scale z (m)

400

14 300

12
u* = 1.474 m/s
10 z = 6.71 mm 200
Logarithmic law
U / u*

6 100

2 0
0 10 20 30 40
1 10 100 Turbulence intensities (%)
log (z)
Fig. 4 Longitudinal turbulence intensity profile. S = 565. Data
Fig. 3 Comparison of the mean velocity profile in 1:565 ABL are plotted against well-known other profiles as shown in the
simulation with the logarithmic law. Uncertainties due to the figure. Horizontal bars give uncertainty in longitudinal turbu-
finite number of measurements are shown by vertical bars lence intensity
Experiments on integral length scale control in atmospheric boundary layer wind tunnel 135

area with height as suggested by Tieleman et al. (1978).

The height of the spires was selected to maximize the Present data (1:565)
500
boundary layer thickness. Kozmar [4]
The intensity of turbulence is proportional to the Counihan [8]
Reinhold [19]
angular velocity of eddies. Eddies of equal size may
Akins-Cermak [20]
produce different intensities of turbulence, depending ESDU [12]
upon their rotational speeds. The energy content of a 400
large eddy is much greater than that of a smaller one
for a given intensity, and the energy is transferred from
larger eddies to smaller ones.

Full scale z (m)

Figure 4 shows the longitudinal turbulence intensity 300
of longitudinal velocity fluctuations measured in the
middle of the wind tunnel, 0.5 m downstream from the
last row of the roughness elements at different locations
in the vertical direction. Assuming a model scale of
200
565, the data are presented in terms of full-scale height.
Measured turbulence intensity profile is plotted against
the full-scale data up to 300 m proposed by ESDU
(1985) for z0 values of 1.5 m. Solid line shows the
full-scale data correlation bands which differ by 10% 100
from the mean values of experimental data. Measured
data are in good agreement with the data presented by
Kozmar (2010), Counihan (1975), Farell and Iyengar
(1999), Reinhold et al. (1978), and Akins and Cermak 0
(1976). 0 50 100 150 200 250 300
x
L u (m)
Influence of passive devices on model scale factor
Fig. 5 Longitudinal integral length scale profile, S = 565. Hori-
zontal bars give uncertainty in longitudinal length scale
Counihan (1975) proposed a graph to determine the
value of integral length scale with height (>10 m)
for different roughness lengths, corresponding to four
different terrains. It can be noted in this graph that the et al. (1978), and Akins and Cermak (1976). Mea-
value of integral length scale increases with height up to sured values of longitudinal length scales with full-scale
240 m in the ABL; however, above 240 m, it decreases heights are in good agreement with the data presented
and is independent of roughness length. Below 10 m in the width of the ESDU correlation bands for z0 =
height, it can be assumed that Lux decreases rapidly 1.5 m. There is always some scatter in atmospheric
with decrease of height. Counihan also proposed a integral length scale data which are reflected in the
graph to determine the value of integral length scale at width of the ESDU correlation bands. Though there
a reference height 30 m for different values of rough- are couple of measurement data which are outside the
ness length. full-scale data correlation bands proposed by ESDU,
The aeroelastic behavior of the structures in at- but most of the points fall within the proposed band.
mospheric boundary layer can be predicted by scaled At the model scale of 1:565, many of the measured
model experiments in the wind tunnels even when Re longitudinal integral length scales are closer to smaller
do not match, provided the scale factor for the longitu- Lux limit of the ESDU band. As shown in Fig. 5,
dinal length scale of turbulence, Lux , matches closely integral length scales are within the ESDU range up
with the actual physical model scale. The longitudinal to 250 m, and above 250 m, the results are slightly
length scales for turbulence were calculated at different below the suggested ESDU data. The data presented by
heights for each configuration. Kozmar (2010), Reinhold et al. (1978), and Akins and
Figure 5 shows the longitudinal integral length scales Cermak (1976) show a similar trend with the present
in terms of full-scale heights. A model scale of 1:565 data. In the data presented by Kozmar (2010) and
was obtained using the procedure described above. Akins and Cermak (1976), few measured points in the
Obtained results were also compared against Kozmar lower boundary layer are outside the band whereas for
(2010), Counihan (1975), ESDU data (1985), Reinhold upper boundary layer (>80 m), the points reside in
136 K. Varshney, K. Poddar

more on the turbulence characteristics than on the

mean velocity profile. The principal turbulence char-
acteristics that matter are variation of longitudinal tur-
bulence intensity with height and integral length scale.
With each configuration, variations of mean veloc-
ity, stream wise turbulence intensity, and longitudinal
length scales of stream wise turbulence with increasing
height have been measured.
Based on results presented in this paper, it is very
clear that the effective length-scale factor decreases as
we increase: (a) the barrier height, (b) the slot-width in
the extension board into the contraction cone, and (c)
the roughness element distribution on the test section
floor. The maximum length-scale factor is obtained in
configuration no. 7 (Table 2), where 20 cm slot and
30 cm of barrier have been used with full roughness
elements distributed on the entire test section length
Fig. 6 Longitudinal power spectrum of the velocity fluctuations upstream of the measuring point. Furthermore, it is
shown that the integral length scale factor can be con-
trolled using different combinations of passive devices
which will be conducive to simulate ABL of various
the band. Similarly, data presented by Reinhold et al.
rough-to-smooth terrains.
(1978) also show similar behavior. It can be noted in
Table 2 that the model scale factor improves with the
slot width and height of the castellated barrier walls. Acknowledgements The authors are thankful for the financial
support provided by the Aerospace Engineering Department,
It can be noted in the Table 2 that the model scale Indian Institute of Technology, Kanpur. We express our thanks
factor S decreased with the increase in barrier height. to the technical staff of the department for manufacturing simu-
It was found that increase in slot width decreased the lation hardware used in this study.
model scale factor significantly. Two 10-cm-wide slots
in the extended test section decreased the model scale
factor by 20.6%. The maximum length-scale factor is
References
obtained in configuration no. 7 (Table 2), where 20 cm
slot and 30 cm of barrier have been used with full
Akins RE, Cermak JE (1976) Wind pressures on buildings, fluid
roughness elements distributed on the entire test sec- dynamics and diffusion laboratory. Report CER76-77REA-
tion length upstream of the measuring point. JEC15, Colorado State University, Fort Collins
The spectrum of longitudinal velocity components is Cook NJ (1978a) Determination of model-scale factor in wind-
tunnel simulations of the adiabatic boundary layer. J Wind
presented in Fig. 6. The power spectrum agree well with
Eng Ind Aerodyn 2(4):311–321
Kolmogorov −5/3 inertial subrange. Cook NJ (1978b) Wind-tunnel simulation of adiabatic at-
mospheric boundary layer by roughness, barrier and mixing
device method. J Wind Eng Ind Aerodyn 3:157–176
Counihan J (1969) An improved method of simulating an at-
6 Conclusions mospheric boundary layer in a wind tunnel. Atmos Environ
3:197–214
Counihan J (1971) Wind tunnel determination of the roughness
As mentioned above, the aeroelastic behavior of the
length as a function of the fetch and the roughness density
structures in atmospheric boundary layer can be pre- of three-dimensional roughness elements. Atmos Environ
dicted by scaled model experiments in the wind tunnels 5:637–642
even when Reynolds number does not match, provided Counihan J (1973) Simulation of an adiabatic urban boundary
layer in a wind tunnel. Atmos Environ 7:673–689
the scale factor for the longitudinal length scale of tur-
Counihan J (1975) Adiabatic atmospheric boundary layers: a re-
bulence, Lux , matches closely with the actual physical view and analysis of data from the period 1880–1972. Atmos
model scale. The longitudinal length scales for turbu- Environ 9:871–905
lence have been calculated at different heights in this Cramer A, Varshney K, Gundrum T, Gerbeth G (2006) Exper-
imental study on the sensitivity and accuracy of electric po-
study.
tential local flow measurements. Flow Meas Instrum 17:1–11
It has been established that the aeroelastic response Davenport AG, Isyumov N (1967) The application of at-
of structures and dispersion characteristics depend mospheric boundary layer wind tunnel to prediction of wind
Experiments on integral length scale control in atmospheric boundary layer wind tunnel 137

loading, wind effects on buildings and structures. University Nishioka M, Sato H (1974) Measurements of velocity distrib-
of Toronto Press, Ottawa, pp 210–230 utions in the wake of a circular cylinder at low Reynolds
ESDU Data Item No. 72026 (1972) Characteristics of wind numbers. J Fluid Mech 65(1):97–112
speed in lower layers of atmospheric near the ground: strong Pasquill F (1967) Wind structure in the atmospheric boundary
winds—natural atmosphere. Engineering Science Data Unit, layer. Discuss on Archit Aero Phil Trans R Soc A 269:321–
London 554
ESDU Data Item No. 74030, 74031 (1974–1975) Characteristics Perry AE, Joubert PN (1963) Roughness boundary layers in ad-
of atmospheric turbulence near the ground. Engineering verse pressure gradients. J Fluid Mech 17:193–221
Science Data Unit, London Reinhold TA, Tieleman HW, Maher FJ (1978) Simulation of the
ESDU Data Item No. 85020 (1985) Characteristics of at- urban neutral boundary layer for the model study of wind
mospheric turbulence near the ground. Engineering Science loads on tall buildings. Report VPI-E-77-12. Department of
Data Unit, London Engineering Science and Mechanics, Virginia Polytechnic
Farell C, Iyengar AKS (1999) Experiments on the wind tunnel Institute and State University, Blacksburg
simulation of atmospheric boundary layers. J Wind Eng Ind Sax MJ (1978) Fluctuating pressure field on a Yawed rigid cylin-
Aerodyn 79:11–35 der. Ph.D. thesis, Graduate School of The University of
Jensen M (1958) The model law phenomena in natural wind. Minnesota, Minnesota
Ingenioren 2(4):121–128 Tieleman HW, Reinhold TA, Marshall RD (1978) Wind tun-
Kaimal JC, Finnigan JJ (1994) Atmospheric boundary layer nel simulation of atmospheric surface layer for the study of
flows: their structure and measurement. Oxford University wind loads on low-rise buildings. J Wind Eng Ind Aerodyn
Press, New York 3(1):21–38
Kozmar H (2010) Scale effects in wind tunnel modeling of an Varshney K, Panigrahi PK (2005) Artificial neural network con-
urban atmospheric boundary layer. Theor Appl Climatol trol of a heat exchanger in a closed flow air circuit. Appl Soft
100:153–162 Comput 5:441–465