..trmq&&c Envircnnrcnr, Pqamon I+&3 1968. Vol. 2, pp. 49-71.

P&&d in Great Britain

THE SIINULATION OF THE ATM~SPHE~C BOUNDARY

LAYER IN A WIND TIJNNEL
J. Awn-r and J. Cow
Central Ekctricity Restarch Laboratories,
Ckwe Road, Leathc&cad, Surrey, England

Abstract-A qualitativediscu&oxtof the physical nature of atmosphcrk boundary layer flows

leads to the conclusion that the primary aims in the simulation of these flows in a wiud tunnel
should bo to model the rckvant scales and i&nsitia of turbuknce. However, the simulation
of the variation of mean wind velocity with height is also desirable.
A system of bar&s and vortex generators is proposed as a mcaus of simulating turbuknt,
neutrally stabk* atmospheric boundary layers. Measurements of the intensities and spectra
of turbuknce in a small wind tunnel using such a system show that acceptable mprcscntations
of these features of the boundary layers cau be obtained. Therefore a scaled up version of
the system was installed in a Iaqcr Low Speed Wind Tunnel and the results of measure+
ments in this tunnel are presented here.

1. INTRODUCTION
MANV ENG~NEEXUNGproblems require an unders~d~g of the flow processes in
natural atmospheric winds. As these processes are not easily treated theoretically
Nomemdature:
= a universal constant m in a formula for diffusion;
i(n) = normaliz& power spectral density at frequency n (Eukrian);
FL(n) = normal&d power spectral density at frequency n &agmn&n);
11 =Lagra&nkngthscale;
i2 = Eukrian ler@h scale;
L = typical length scale of a structure;
& = Reynolds number;
T, t - time of travel of a particle;

u* = friction velocity = 2
J
u = longitudinal v&city Zmpouent;
u #lax = longitudinal velocity component in fiw stream;
Y = lateral velocity component;
W = smtical velocity component;
V = voltage output of constant temperature hot wire anemometer;
n - stmamwisc w-ordinate;
x = a nowlirncnsional &cquency parameter;
Y = lateral co-ordinate;
z = vertical coordinate;
20 - roughness length;
= rate of energy dissipation/unit mass per unit time;
1 = ‘I’aykr*s micro-scale of turbulence;
P = density;
r - shear stress;
V = kinematic viscasity;

K = van Kiirmh’s constant.

A mime on a velocity component denotes turbulent contribution and a bar denotes average with
respect to time.
D A 49
50 J.AumrrandJ.couNEAy

except by rather drastic si.mpli&ations, this understanding can only be gained by a

considerable amount of experimental work. At the Central Electricity Research
Laboratories (C.E.R.L.) the main interest is in the effects of atmospheric wind on the
static and dynamic loads on structures, such as chimneys and cooling towers, and on
the performance of cooling towers and the problem of smoke dispersal from chimneys.
The type of wind tunnel required for these investigations is very different from those
in normal aircraft practice. In particular, a wider speed range and a longer working
section are necessary. For ‘those reasons two large low speed wind tunnels have
been built at C.E.G.B., one at Marchwood Engineering Laboratories and one at
C.E.R.L., which is somewhat smaller and intended primarily for cooling tower
performance investigations.
In all of the problems mentioned above a satisfactory small scale model of the
atmospheric wind must be produced in the wind tunnel. One technique is to simuIate
a long stretch of terrain upstream of the model so that turbulence and the vertical
shear in the model wind develop naturally (JENSENand FRANCK, 1963). However,
in most cases this implies a prohibitively long wind tunnel, if the models used are
large enough to avoid excessive Reynolds number effects. DAVENIQRT(1966) and
CEWAK (1966) have both described tunnels in which the atmospheric boundary
Iayer is produced in this way. They have working sections about 100 ft long and the
boundary Iayer thickness is of the order of 3-4 ft depending on the terrain rough-
ness. Davenport remarks that even in these long wind tunnels a system of grids
might be an advantage so that the boundary layer would have more time for develop-
ment. In most existing tunnels the working section is too short to allow the boundary
layer to grow naturalfy.
The C.E.R.L. and ~ar~hwood tunnels represent a compro~se. It was assumed
that an approximation to the earth’s boundary layer could be produced by artificial
means within a distance of about 2 or 3 boundary layer heights. The terrain upstream
of the model to be tested would then be simulated for a distance of not less than ten
times the height of the model. This decision was based on the results obtained by
PANOFSKYand TOWNSEND(1964) which showed that a change in surface roughness
affected the flow within a wedge having a slope of about one in ten. Later work by
TOWNS&ND (1966) has shown that this is only true if the change of roughness is not
too great, implying that the simulated flow at the start of the roughness must be a
reasonable approximation to the required flow. At the time these wind tunnels were
designed it was not clear how this simulation was to be produced but it was hoped
that something satisfactory could be devised. For exampIe, the shear flow in the
natural wind can be produced by means of a grid of rods (OWEN and Z~XEWCZ,
1957). Turbulence of the required intensity can be produced by means of a coarse
grid screen, although this does not necessarily produce the required shear flow. It is
evident that a complete simulation of the atmospheric wind in all respects is very
difficult, if not even impossible.
Before deciding what type of grid should be used to produce the simulated anno-
sphere wind it is necessary to choose which aspects of the natural w&d must be
accurately represented and where some compromise is possible. At present the flow
processes are not well enough understood to give accurate quantitative answers to
this question. A general qualitative discussion can however be given and this will be
the subject of the first part of this paper.
 The Simulationof the AtmosphericBoudary Layer in a Wmd Tuned 51

The second part of the paper is concerned with a physical description of the flow
processes in the a~osphe~c wind and a discussion of how best to realize these at a
reduced scale in the wind tunnel. Systems of grids have been considered with the
objective of produ~g a reasonable approbation to the atmospheric boundary
layer flow within about two or three boundary layer heights from the start of the
working section.
On the basis of the discussion in the first part of this paper a system of grids and
vortex generators is proposed; physi~~y the flow processes behind this system are
believed to be an accelerated version of the flow process in a naturally developing
boundary layer.
Finally, in the third part of this paper the experimental rne~~rnen~ made in the
simulated boundary Iayer are described and discussed. Since the initial measurements
showed that an acceptable boundary layer might be represented, tkte system repro-
ducing it was sealed up and installed in the C.E.R.L. Low Speed Wind Tunnel. The
results of these tests are presented here and relate to a simulated boundary layer
which is sufficiently large to accommodate models at a scale of l/250 of full size.

2. DISCUSSION OF SOME OF THE PROBLEMS TO BE INVESTLGATED

The flow of the atmospheric wind over the Earth’s surface is similar in some respects
to the flow in a ~rb~ent bonds layer on a 3at plate. In most cases of practical
interest the Earth’s surface is aer~~~~~y rough so that the e&et of viscosity
on the Bow as a whole can be neglected. In a turbulent boundary layer with an aero-
dy~a~ca~y rough surface the flow is entirely determined by the geometry of the
obstacles on the surface. For the regions which are not close to an individual obstacle
the flow is statistically determined by the ~s~bu~o~ of the roughness length, zo,
which is a measure of the average height of the roughness elements. If the analogy
between the atmospheric wind and a turbulent boundary layer is accepted the flow
in the Earth’s boundary layer can be simulated in a model, provided geometrical
similarity is preserved and the roughmess length, zo, is reduced in propo~ion to the
geometric scale and provided that the Reynolds number is not reduced so much that
the surface is no longer aerod~a~~~y rough. This is the model law proposed by
JENSEN and FRANCK (1963). The disadvantage of this approach to the simulation of
the atmospheric wind is that a prohibitive wind tunnel length is required for the
developmeat of the boundary layer.
It is worth noting that the flow in the boundary iayer on a flat plate differs in two
main respects from the a~osphe~c wind. In the natural wind the direction of the
mean flow usually varies with height, and it has been suggested that this could effect
the dispersion of pollutant in the atmosphere ('RLDESLEY and WALLINGTON,1965).
No such change of direction is present in the flat plate boundary layer flow. It is
difBcult to see how this effect can be modelled in a wind turmel and the problem will
be ignored for the time being. The second main difference is that the buoyancy effects
due to potenti~ temperature gradients cause an increase or decrease in the turbulence
intensity depending on whetber the ~~~~e gradient is unstable or stable. This
is obviously of great importance in the problem of the dispersal of pollutant. DAVEN-
PORT (1963) has su~ested that in high winds a neutral a~osphere is obtained.
52 f.AmcrrrandJ.Cum

Rung the problem to the simulation of the neutral atmosphere is therefore

jus&ed in the study of static and dynamic loads on structures. Simulation of the
neutral a~osphere is an essential step towards the complete simulation of stable
and unstable atmospheres and it is therefore wo~hw~e discussing the problem of
dispersal of pollutant in the neutral atmosphere.
In the case of a neutral atmosphere in which the change iu direction of the wind with
altitude is ignored the problem is to produce a thick, rough waI1,turbulent boundary
layer in the shortest possible streamwise distance. Even with these restrictions a
complete simufation will be possible only if the boundary layer is allowed to develop
naturaliy as it Bows over a rough floor with the roughness length scaled as the geo-
metric scale. The shear flow in the atmosphere can be simulated by various forms of
shear grids aad, high turbulence intensities C~I.Ibe produced by means of coarse grid
screens. The question is whether the t~bule~~e ~teusity and the shear flow can be
simulated at the same time, As this is likely to prove difficult it is important to decide
what properties of the flow must be s~ulated accurately and where some ~ompro~se
can be made.
The structure of the atmospheric wind is extremely complex and in principle an
infinite number of parameters are required for a complete description of the ffow.
An adequate des~~ption of fiow would probably require a ~ow~edge of the fo~Iowing
parameters :
(If mean velocity ~st~butio~ as a function of altitude;
(2) intensity of each component of turbulence as a function of ahitude;
(3) integral scales of turbulence;
(4) micro-scales of turbulence (energy dissipation scale);
(5) correlations between components of turbulence particularIy the Reynolds
stress U’W’;
(6) the spectra of turbulence.

The scales of turbuience and the spectra of turbulence are to some extent a difFereat
description of the same thing. In some problems it is more conve~ent to work in
terms of the sperm and in others in terms of scales of turbulence. As the turbulence
spectra are more easily measured than the scales of turbulence the discussion will be
mainly in terms of the spectra.
The main problems to be studied are:
(1) dynamic loads;
(2) static loads ;
(3) pollutant dispersal;
{4) cooling tower performance.
It is not the intention of this paper to discuss buoyancy scaling and the modelling of’
the dynamic chara~e~stics of structure. The discussion is concerned only with the
effect of the structure of the atmosphere wind on each of the stated problems.

The requ~emen~ for modehing the a~osphe~ wind are very stringent in this
case. Not only must the intensity of each component of ~rbu~~ce, pa~icu~arly in
 the simulationof the AtmosphaicBoundaryLayerin a Wind Turmiel 53
the direction of the win& be representative of the corresponding intensities of tur-
bulence in the atmosphere but the spectra of turbulence must also be similar to the
spectra in the atmosphere. For particulariy simple structures with few modes of
vibration and low damping it is sufficient that the spectrum be represented at fre-
quencies near to the natural frequencies. However the natural frequencies may not
always be known and in many cases, cooling towers for example, there are a large
number of natural modes. It is therefore advantageous to have a simulated atmos-
pheric wind which has the correct spectral content at all frequencies.
The spectral content of the simulated atmospheric wind should therefore be accur-
ately represented with regard to:
(a) frequencies of the order of the resonant frequencies of the structure;
(b) scales of the order of the length scale of the structure, or of the order of the
wavelength of the normal modes of vibration;
(c) frequencies of the order of the Strouhal frequency of the structure.
The third requirement ensures that the energy conterit is simulated at frequencies
where the turbulence is likely to affkct vortex shedding from the structure. These
requirements imply that the energy spectrum of the atmospheric wind must be simu-
lated for full scale frequencies in the region of l/10 to 10 Hz and for eddy scales of
tie order of 10 to 1000 ft. At frequencies very much lower than this it is not necessary
that the atmosphere t~b~ence should be simulated. In this case however allowance
for very low frequency turbulence, i.e. gusts, would have to be made. If the turbulence
spectrum could be simulated right down to frequencies equivalent to about 1 cycle/hr
at full scale then the gusts of about 1 min duration would also be simulated in the
tunnel. All statistical results obtained in the tunnel could then be related to the
hourly mean wind. As the turbulence in the atmosphere is usually assumed to be
statistically stationary over this period (DAVENPORT,1963) and as wind data is more
easily available in terms of hourly means this would be a great convenience in
analysiug the results and comparing them with full scale data.
In almost all cases the micro scale, or energy dissipation scale of turbulence will be
far too small to affect the structure. In the model, however, the energy dissipation
scale will be larger co_mparedwith the size of the stature and if it is too large the
flow may be affected by viscosity, especially if the model has small but important
details of the order of the micro-scale of turbulence.
An estimate of the micro-scale of turbulence in the model can be obtained by
considering the energy balance. In the region close to the ground the production and
dissipation of absent energy are almost equal (T~WNSZND,1961). The energy
production depends on the shear stress and the velocity gradient:
zdu u*3
g2_-x=-
pdz KZ’
This represents the energy extracted from the mean flow to produce turbulent energy.
‘This will be passed by inert% interactions to the high wave numbers where it is
dissipated by viscosity. As the turbulence tends to isotropy at high wave numbers the
energy dissipation is given by TAYLOR(1935) as:
15vu’2
12
54 J.-mdJ.CouMfurr

Now if the roughness length, z,, is tied correctly and the turbulence structure is
fully developed :

.JF u*
U?lWX
'umux
willbe the same in the model and full scale flows (JENSEN
and FRANC& 1963).
Equating the dissipation and production we have:

L2, 22, RcFS

T
L-&i rl.2Fst!G*
where the s&ix N refers to the model, FS to full scale.
The requirement that the micro-scale of turbulence should be small compared with
the details of the model is therefore a minimum Reynolds numbers requirement. This
is already determined by the requirement for rough wall flow.
It is not known whether the Reynolds stress is of any direct importance in tiecting
the pressure field on a structure. Such an effect is possible as high Reynold stresses
would imply that high velocities tend to be associated with downward flows and low
velocities with upward flows. In this case one might expect to find that the average
force on a body at zero incidence to the mean flow would be in a downward direction.
Whether this effect exists, or if it does whether it is important is conjectural. It will be
argued Iater that the Reynolds stresses are of indirect importance in ensuring the
correct dodgy development of the boundary layer,
Very little work has been done on the effect of the wind shear on the dynamic
loads on structures. Work by Baines on static loads, (BASH’&1%3), has shown that
the incident vorticity in the wind wraps round the structure to form trailing vortices.
These produce downward flows at the face of the buying and si~~ant changes in
local pressures are obtained although the effect on overall load is not so large. It is
therefore reasonable to expect that effects of at least the same order are present in the
local dynamic loads.
The conclusion must therefore be that in the study of dynamic loads the primary
aims should be to produce turbulence with the correct intensity and the correct
energy content in the middle range of frequencies. It is not essential to simulate the
very low frequency end of the spectrum or the very high frequency end which car-
responds to the viscous dissipation region. Simulation of the velocity profile is
desirable if it can be done without compromising the turbulence structure.

The static load is simply the time-mean of the dynamic load and it is conceivable
that the turbulence could affect the time-mean load. This occurs for example at
certain Reynolds numbers where turbulence in the free stream affects the transition
from laminar to turbulent Bow in the boundary layer of a sphere. This couId happen
in tests on a modei cooling tower if one is forced to carry out the experiment at too
low a Reynolds number. There remains the possibility that the pressure gradient
imposed on the structure by the turbulence is su~cient to cause separation of a
turbulent boundary layer with a consequent effect on the mean pressures. There is
little evidence on cylindrical bodies in a turbulent stream at high Reynolds number
 The Simulationof the Atmospheric Boundary Layer in a Wind Tutu& 55

but measurements by VICKERY(1966) on a rectangular prism have shown signi&ant

effects of the turbulence on the pressures. This is presumably not a boundary layer
transition effect as the separations are &red by the sharp edges of the structure.
The effect of wind shear on static loads has been shown to be of importance and
in contrast to the situation with dynamic tests it may be of greater importance to
model the shear flow rather than the turbulence structure. The purpose of mean
pressure measurements is however to check whether the structure will fail. It is the
higher speed gusts of say 10 set duration that are responsible for the largest loads.
Gusts of significantly shorter duration will not give time for the aerodynamic loads
to reach their quasi-static values. The short duration gusts may of course be important
but here we are in the realm of dynamic testing. As a criterion for ftinre of structures
the static tests should therefore be made in a stream which simulates the extreme
gust velocity profile rather than the mean flow velocity profile. It has been suggested
(see report of the FERRYLWRGE COMM~I-REOF INQUIRY,1966) that the gust of 10 set
duration may have a much more uniform velocity distribution than the mean velocity,
thus is may be that tests in a uniform stream are more realistic as a criterion for static
failure than tests in a shear flow. This view was taken in the research work connected
with the inquiry into the collapse of the Ferrybridge “C” cooling towers. It is stated
in the above report that Meteorological Office data on 10 set gusts do show a varia-
tion with altitude but that this variation is much smaller than that of the mean flow.
In many cases results of great accuracy are not needed from static tests and an
indication of orders of magnitude are sufficient. It is not, or should not be, usual
practice to design structures with a small safety margin on the basis of static loads
alone. In view of this, tests in a uniform stream may be satisfactory and may even be
more desirable than tests in a graded wind. It is therefore considered doubtful whether
the measurement of static loads presents a real problem in simulating the Earth’s
boundary layer.

2.4. Dispersal of pollutant

An idea of the effect of the various turbulence parameters on pollutant dispersal
can be obtained by considering the formulation of the problem introduced by TAYLOR
(1920). Taylor’s main result is that the mean square dispersion (of particles):
[Yfi]” is given by:

[Y.-J2 = 2 (o’“) d / R, de dt,

where Y’ is the turbulence velocity ffuctuation in the direction y and Rc is the Lagran-
gian correlation coefficient between the velocities v’ (2) and v’ (t+n. In homo-
geneous turbulence the above quantities are not a function of position, in the real
atmosphere they are. Similar results to the above are obtained for each direction.
In the initial stages I+ will be close to unity and the plume dispersal depends only
on the intensity of turbulence. At large times the correlation R, will tend to zero. If
the integral converges a length scale:
56 J.AmmrandJ.cO-

can be d&.ned and the plume dispersal then depends on the Lagrangian Iengtb
scale li. Tbc corresponding quantity in the Eukrian description of the flow is:

where R, is the correlation coefficient between the turbulence velocity at points

separated by a distance y. There is no known exact relation between I, and & but
TAYLOR(1935) has su that they can be assumed proportional for grid tur-
bulence.
If this is so then the model atmosphere must have the same intensity of turbulence
in each direction as tbc real a~ospbere and the integral scale of ~b~~~, l,, must
be reduced as the geometric scale of the model.
As it is easier to measure the frequency spectrum the above results can be put into
terms of the Lagrangian spectrum using the Fourier transform PMCJXJILL, 1962).
This gives:

[Yn-J2= P’* T2 i F,,(n) “‘“;I:;;;““,

where r;y,)dn is the normalized energy content at frequency between n and n+dn.
For small times the whole speotrum contributes equally and the dispersion depends
only on the intensity but for larger times the lower frcquenoies become more signifkant
than the higher frtquenoics. In the limit only FL(O)contributes to Ye2and FL(O)is
the counterpart of II and iz in this formulation.
The Lagrangian spectrum is not easily measured but it is reasonable to assume
that if the Eulerian spectra of tbe model and the real atmospitere arc similar then so
wili the Lagrangian spectra.
Another effect which may be of importance has been discussed by Scruvx~ (1966).
A plume will lose buoyancy by mixing with the colder air surrounding it..This mixing
process will depend on the turbulent energy at scales rather smaller than the plume
diameter. The turbulence at soaks of the order of the plume diameter and above will
tend to break the plume into p&s which will lose buoyanoy more rapidly as they arc
subjected to thmedimensional mixing instead of two dimetional. The eddy scales
responsible for these processes will be of the same order as those of interest in the
study of dynamic loads on stru&urcs.
It is not easy to assess the &eots of the mean veiocity distribution on the concentra-
tion of the blurt, CZFBUK(1963) has applied to the concept of Lagrangian simi-
larity to the flow field to determine the efkct of the velocity proBe on tbc ground level
concentration. The steps leading up to the fkal result will not be repeated here but
Cermak finds that the maximum ground concentration is approximately propor-
tional to:
1
X,,-----=-
ZZf z )
0 ZO

w&m z"is the mean plume Mgbt and z0 the roughness length The function f is in
fact tb vcitity profile. In most cases the mean plume height will be in a region
where the wind speed is close to the gradient wind speed. The velocity profile will
 The Simulation of the Atmospheric Bomdary Layer in a Wind Tunnel 57

therefore not have a very great effect on the ground level concentration. The point
at which the maximum ground level concentration occurs is also affected by the
velocity profile :
lZ z
jt_=- - z+constant,
bits,
jf(20>d
where b is a universal constant. As this is an integral form the position of maximum
ground level concentration is not greatly affected by the form of the function f but
mainly by the displacement thickness of the Earth’s boundary layer. For the purpose
of plume diffusion studies there is therefore no need for an accurate simulation of
the velocity profile.

2.5. EApcts of the wind on cooling tower performance

The main problems here are:
(1) re-entrainment of warm air;
(2) effect of the external flow on the internal volume flow;
(3) change of plume buoyancy due to turbulent mixing.
These problems are closely related to those which have already been discussed.
The re-entrainment problem is essentially a near-field version of the plume dispersal
problem. In this case, however, the turbulence generated in the wake of the cooling
tower is probably more sign.ificant than the atmospheric turbulence and therefore it
is possible that the requirement for modelling the turbulence can be relaxed. The
effect of turbulence on the buoyancy cannot be neglected however.
The effect of the trailing vortices produced by the shear in the incident wind will
clearly affect the re-entrainment problem and may also be sign&ant in the effect of
the external flow on the internal volume flow.

2.6. Summary of atmospheric wind simulation requirements

The most difkult problem raised above occurs in the study of plume dispersal.
A great length of working section is needed downstream of the chimney to observe
the flow, this means that only a relatively short upstream fetch is available for the
development of the flows. In the case of plume dispersal problems a correct modelling
of the lateral and vertical turbulence intensity and the spectra of each component is
required, particularly at the low frequency end of the spectrum. Modelling the shear
in the incident wind is of less importance in this case.
For the study of dynamic loads on structures the spectrum and intensity of tur-
bulence must be modelled, particularly the horizontal component in the direction of
the wind. The middle range of frequencies is in this case the most important. Model-
ling the shear flow is desirable but secondary to the requirement for the correct
turbulence structure’. In the case of dynamic loads and other near field problems a
greater length of wind tunnel is available for flow development as the model can be
placed towards the rear of the working section.
The requirements for cooling tower performance studies are similar to those for
dynamic load problems, with perhaps a slight change of emphasis, but again the model
can be placed towards the rear ofthe working section.
58 3.~~d3.~~

The static load problem is in general less important than the above and if a satis-
factory s~ula~on of disc Ioads can be obtained the static loads can. e&y be
obtained as a special case.
It is ~on~l~d~ from the above that the main effort should be directed to the pro-
duction of the required structure of turbulence. If this is achieved then in those cases
where the shear flow is sign&ant, it can be allowed to develop naturally as a result of
the shear stresses set up by the turbulence.

3. FLOW PROCESSES IN THE AT~USP~E~r~ WIND

In this part the flow processes in the atmospheric wind, or, with the approximations
accepted earlier, a rough wall bounda~ layer on a flat plate are discussed. Probably
the best approach is to consider the energy balance within the boundary layer.
Turbulent energy is produced by the action of the Reynolds stresses on the velocity
gradient in the mean flow. This of course extracts energy from the mean flow and the
production of turbulence tends to be at low wave nears. Inertial interaction amongst
the turbulent eddies causes the energy to be passed from low wave numbers to high
wave numbers. Eventually the eddies at the highest wave numbers are dissipated by
viscosity to produce heat. If the Reynolds number is high enough the amount of
viscous dissipation depends on the energy transferred to the high wave numbers by
inertial action and not on the viscosity itself.
In the natural boundary layer most of the produc~o~ of turbulence takes place
in the region close to the ground and this is partly ~sipa~d and partly doused
out into the region away from the ground. The energy production and ~sipa~on are
almost equal so that the energy diGsed out is the small diBerence of two large
quartiles.
To date most approaches to the production of a shear flow have relied on local
energy ex~ac~on from the mean flow. The amount of turb~e~t energy production
is limited by the energy difference between the original uniform flow and the required
mean velocity at that level. Furthermore much of this energy will be dissipated, firstly
by inertial interaction and finally by viscosity, and we can have Mually no control
over this process. Compared with a natural boundary layer the intensity of ~rbuIence
in the outer part of the art&is1 boundary layer is likely to be too low unless there is
some mechanism for the transfer of the excess turbulent energy at near could level
to the outer region. This energy transfer will occur achy but the process will be
too slow for our purposes and some artikial mixing must be added. The mixing must
extend over the whole bounda~ layer and the soak of the turbulence produced by
the mixing process must be as large as possible.
It fo~ows that we must reiy for the majo~ty of the absent produc~o~ on the
regions close to the ground, this is the region where most of the energy is extracted
from the mean flow. Some production at higher levels may be possible but the ex-
change between mean flow energy and ~bulent energy prevents the production of
sufkient energy at this ievel without producing an excessive velocity defect in the
mean wind.
Once the artificial boundary layer has been produced the turbulence will decay
clearly at the low frequency end of the ~~ which will have its energy
passed to higher frequencies, unless there is some me&anism for the continued
produ~o~ of ~rbulen~e as the flow proceeds do~s~eam. In the natural boundary
 The Simulation of the A~&G@w&Boundary Layer in a Wiid ‘RI& 59

layer this is provided by the action of the Reynolds stresses on the mean velocity
profile. The production is
sdu
(per unit mass per unit time).
s=Pz
From this relationship it is clear why most of the production is concentrated close
to the ground. It is also evident that the continued production of turbulence requires
the Reynolds stresses and the velocity profile to be simulated.
This does not mean that in the initial stages an accurate velocity protile and shear
stress dis~bution are needed, it is sufficient that both r and &/G% should be of the
required order of magnitude in the region close to the ground. Furthermore if the
Reynolds stresses are simulated in the artificial boundary layer the flow will eventually
tend to the required natural boundary layer flow, although this process tends to be
rather slow in most cases.
Several methods have been published for the simulation of the shear flow in a
wind tunnel. These include a grid of rods (OWEN and ZIENK.IEWICZ, 1957), a curved
gauge screen (BAIIW, 1963) and a grid of flat plates (S~OM, 1962). None of these
methods were intended to simulate the turbulence properties of the atmosphere. If
the requirement for high intensity turbulence is to be met supplementary turbulence
generators will be needed. These cannot be placed upstream of the grid as the turbu-
lence will be damped on its passage through the grid. Nor can they be placed down-
stream as this would destroy the velocity protie. The only possibility of modifying
any of the above systems is to place the turbulence generators on to the grid elements
themselves, a grid of flat plates is the most convenient in this respect. This is the
approach followed by LKMJ (1966) at Bristol University. In order to achieve a
reasonabIe velocity profile approximately ten plates will be required and initially
the scale of turbulence will be of the order of the plate spacing. There would be some
subsequent interaction to produce larger scales of turbulence but as there is no
mechanism for the production of large Reynolds stresses one would not expect this
process to proceed rapidly. In view of this the turbulence production will be relatively
small (compared with a natural boundary layer) and there would be a net decay of
turbulence. This is probably not serious in itself but the lack of turbulence production
implies a deficiency of energy at the low frequency end of the spectrum which is so
important in plume studies. It was therefore decided not to follow this approach for
the simulation of the atmosphere in the C.E.R.L. low speed wind tunnel. Instead it
was decided to concentrate on the production of high intensity turbulence with large
Reynolds stresses and to allow the velocity proHe to develop naturally as a result. This
does require a comparatively large upstream fetch but in those cases where the velocity
profile is important the model can be placed towards the rear of the working section.
Some comprehensive turbulence measurements behind a grid of about 30 rods
have recently been presented by RYE (1966) and to some extent the above arguments
are borne out. A linear shear flow was produced; initially the correlation coefficient
~/JiitZ J@” was quite small but eventually a value of about O-4 was reached. At
this stage however the turbulence utensil was quite small. The scale of turbulence
WASof the order of 1.5 to 2 times the rod spacing. This suggests that there is a funda-
mental difficulty in the use of shear grids as a means of simulating atmospheric
turbulence. An increase of the rod spacing would lead to larger scales of turbulence
60 J.AmmrandJ.(lwrmim

but there would inevitably be some loss of control over the velocity profile. It has been
argued in the previous section that this is the more desirable alternative.
The system adopted was a simple barrier or “wall” of appropriate height placed
across the floor of the tunnel followed by a set of vortex generators with their in-
cidences in alternate directions. The arguments which lead to the adoption of this
scheme will be presented here, the extent to which these arguments are valid will be
considered in the discussion of the experimental results. The proposed system pro-
duces large Reynolds stresses’ and also gives the required mixing, As the vortex
generators are at alternate incidences the total stream-wise vorticity is zero (strictly
there may be a small net vorticity if the incidences are unequal) but there is con-
siderable stream-wise vorticity in local regions. As this pattern of flow is consistent
with Townsend’s description of the flow in a natural boundary layer (TOWNSEND,
1957) and the flow processes are consistent with the requirements set out above this
system appeared promising. It is not possible however to set the vortex generators at
arbitrarily large incidence to produce very powerful mixing as their drag would then
cause an excessive velocity defect in the outer half of the boundary layer.
The height of the boundary layer developed by this system will be fractionally
greater than the height of the vortex generators. An idea of the height of the wall
can be calculated, as the momentum thickness of the required boundary layer is
related to the total system drag by an equation which is very similar to von K&m&n’s
momentum integral equation. For a seventh root velocity profile it is found that the
wall height should be about one-tenth of the boundary layer thickness. The vortex
generators were chosen rather arbitrarily as 45” deltas; a reasonably low aspect ratio
is required so that the vortices separate from the leading edge. Exotic planforms like
ogee or gothics could be used but there is no need for a stable vortex pattern; in
fact a too stable vortex pattern would be disadvantageous. Some control over the
Reynolds stresses can probably be obtained by adjusting the vortex generator incidence
and presumably the vortex generator spacing has some effect on the scales of turbu-
lence. As a starting point the vortex generator spacing was chosen equal to the height.
This is rather arbitrary but seemed plausible.
One disadvantage of the present approach is that intially the turbulence structure
will not be homogeneous in the lateral direction; the vortices may tend to take up a
too stable position. As the vortex generators are placed behind the wall they will
be subjected to a non-uniform stream and their lift will fluctuate thus causing the
strength and position of the trailing vorticity to fluctuate. This will alleviate the non-
homogeniety in the lateral direction.
The essential feature of the present approach is that it attempts to copy as far as
possible the physical processes which occur in a natural boundary layer. The diBerence
between the natural boundary layer and the artificial one is that in the former the
turbulence production is gradual whereas in the present case most of the turbulent
energy is produced rapidly in the region close to the wall. The stream-wise vorticity
which develops gradually in the natural boundary layer is suddenly introduced by
the vortex generators thus producing the large Reynolds stresses. From this point on-
ward the two flows are quantitatively similar and will develop in a similar manner.
4. EXPERIMENTAL PROGRAMME
4.1. Tunnel details
Simulation of the Earth’s boundary layer and the accompanying measurements
 The Simulation of the Atmospheric Boundary Layer in a Wiid Tunnel 61

were carried out in the C.E.R.L. Low Speed Wind Tunnel whose working section
measured 15 ft x 5 ft., is 36 ft long, and consists of removable and interchangeable
sections on all sides. The tunnel wind-speed was monitored by a pitot-static tube
mounted in the roof in the upstream end of the working section. The velocity at this
point was 35 ft/sec which corresponded to a fan rotational speed of 500 rev/min.
It should be noted that this was the pitot-static tube reading with the boundary layer
producing equipment in the tunnel.

4.2. Model details

The method used for reproducing the atmospheric boundary layer is a scaled up
version of that used by ARhirrr (1966). For those tests the scale was 1/6OOOthand for
the present tests 1/25Oth assuming a full scale boundary layer thickness of 1000 ft.
A 6 in. high barrier was positioned at the upstream end of the working section and
immediately downstream of this were four triangular vortex generators 4 ft high and
4 ft apart and spaced about the longitudinal centre-line of the tunnel. At a later
stage in the tests a 2 in. dia. tube was positioned 25 in. above the barrier (FIG. 1).
Surface roughness was simulated by use of a ‘Y-ego” construction set. Standard
base boards, measuring 1% in. x 1% in. were first screwed to the tunnel floor and a
uniform roughness pattern was built up on these using bricks measuring lt in. x
3 in. x# in. The roughness was simulated downstream to the leading edge of the
turn-table. A greater degree of surface roughness can easily be produced by building
up on the existing brick pattern.

4.3. Instrumentation and measurements

The mean and fluctuating values of the velocity at any point in the boundary layer
were measured by a Dansk Industri Syndikat A/S (D.I.S.A.) X-wire probe used with
two D.I.S.A. constant-temperature anemometers. The two wires of the cross probe
had previously been calibrated and these calibration curves were represented on the
variable diode function generators (VDFG) of a P.A.C.E. TR 20R Analogue Com-
puter. The cross probe was carried on the tunnel traverse gear which was built into
and carried by one of the interchangeable roof box sections. The probe could be
traversed on any one of the three mutually perpendicular axes of the tunnel working
section.
The outputs from the two D.I.S.A. constant temperature anemometers were fed
into the P.A.C.E. Analogue Computer. There, the linearized outputs of the VDFG’s
were adjusted to give equal sensitivity and then d8erenced to obtain P, u‘ and v,
v’ or ii, U’ and ti, v depending on whether the probe was aligned in the horizontal
or vertical plane respectively. A gain could be applied to the components u’, f’
(or w’) in order to use the full dynamic range of the computing amplifiers and they
could also be recorded on a Epsylon F.M. magnetic tape recorder for later analysis
to obtain the frequency spectra, These fluctuating components were also monitored
continuously on a dual-beam Tektronix storage oscilloscope. In order to obtain the
r.m.s. values of the fluctuating velocities, i.e. Jzi’2, Jv”, JS’2 the individual components
were squared and integrated on the same P.A.C.E. circuit. Similarly the Reynolds
stress was obtained by time integration of the product u’w’, noting that the individual
components had gain factors applied to them at an earlier point in the circuit. The
P.A.C.E. output of integrated and mean values was read off from a digital voltmeter.
62 J.-andJ.Coumim

Initial measurements hdkated that an integration time of 200 set was satisfactory,
based on the assumption that repeatability to within 5 per cent for mean square
readings should be satisfactory for such m~uremen~. However during the initial
stages of the test programmes it was found that Iarge errors were occurring in measure-
ments of the mean velocity due to changes in the ambient temperature. As a result of
observations it was concluded that the maximum allowable drift during a complete
traverse of the boundary layer was of the order of 1°C. Consequently the tunnel was
normally run for about an hour before taking measurement in order to ensure that
the ambient temperature was sensibly constant throughout the building which sur-
rounds the tunnel. When such conditions were obtained the output voltages from
the constant temperature anemometers were backed off so that there was a zero
voltage input into the VDFG’s at zero velocity. On commencing the tests the outputs
of the VDFG’s were equalized when the probe was in the free stream flow.
The maximum extent of a vertical traverse was 50 in., and this was divided into
intervals of 0.5 in. at the lower end of the boundary layer increasing to 3 in. at the
outer part. Due to the various degrees of turbulence through the boundary Iayer the
gain setting, on traversing from top to bottom of the boundary layer, was continually
reduced to be consistent with the maximum output of 10 V from the P.A.C.E.
Measurements of all components were taken on the tunnel centre-line and at
distances of 2 and 4 ft from it, 6 ft behind the vortex generators. Some measurements
were repeated with the vortex generator incidence set at 5” and 15” as well as the
nominal 10” setting. The tests were repeated with a 2 in. dia. bar tied at a height of
25 in. above the barrier. The final set of me~uremen~ were taken 12.5 ft dosses
of the vortex generators, i.e. coincident with the leading edge of the turntable.
The turbulence measurements presented here have not been corrected for free
stream turbulence which is of the order of O-50 to l-00 per cent for all components.

4.4. Experimental reds

The results of the tests are presented in FIGS. 2 to 9. The results of FIG. 2 serve as
an indication of the order of repeatability with the probe horizontal and vertical.
The accuracy to which the mean square values could eventually be measured was
increased to 2-3 per cent due to adoption of a systematic experimental procedure
and maintaining the ambient temperature change to within 1°C.

4.41. Vdocity profiles. The measured velocity pro&s are presented in FK, 3 ;
the 4th Power Law profile is also plotted for comparison. There is a marked diflbrence
between the pro%les measured at the various spanwise stations. The proBe measured
directly behind one of the generators clearly shows the effects of the vortex shed by it.
In general there is excessive loss in velocity in the inner sections of the boundary
layer and not enough defect in the outer section. It was thought that the addition of
a 2 in, dia. tube 25 in. above the barrier would increase the velocity defect in the outer
portion of the boundary layer, but this defect was owed only at the section 4 ft
from the tunnel cent@-line. At this section the pro&s approximates to that of the
+th Power Law over a Iarge part of the boundary layer height. The tube had little
effect on the profile at the centre-line.
At the section coincident with the turn-table leading edge the e&t of the vortex
is siili apparent in the traverse in fine witi the vortex generator.
(Facing p. 62)
 The Simulation of the Atinospbuic Boundary Uycr in a Wii Tunnel 63

C- on cmtre line C-4ft =4ft offccntre line

12’4ft aft of V.GT.E.- V.G. at IO*
2 in. dia. tube 25 h above barrier
.
.

c . c
0

0 40 0
Velocity profile Longitudinal turbulence
0 Q
o = Robe wirer horizantol
* i p&e wires vertical Q Q
t
o a
Q P

d
0
D
0
0
a
0
0
a

FIO. 2 Variation of velocity and turbulence intensity with height.

6ft oft of V.G.TE. -

6ft aft of V.G.fE. + 2in.D. bar -- -
12’4ft oft of V.G.T.E.+PinD.bor -------
+th power law profile - Wocity profile

FIG.3. comparison of velocity profiles at various spanwise and axial stations.

4.4.2. Longitudinal turbul~ce. The longitudinal turbulence measurements are

presented in FIG. 4. Six feet from the generator trailing edge the percentage turbulence
in the lower part of the boundary layer decreases on moving away from the tunnel
ccntre-line. In general, at all sections at which measurements were made, there is an
64 J. ARMIIT and J. tIbmmm
excess of turbulence in the lower sections of the boundary layer and insu&ient in
the outer parts. The addition of the 2 in. dia. tube produces an improvement in the
distribution only at the section 4 ft from the ttmnel oentr&ne and between 20 and
40 in. from the datum level at this section. The additional turbulent energy due to the
presence of the tube is not large and the effect seems to be mostly a redistribution of
the turbulent energy aheady present, thus showing that the region close to the ground
must be relied upon to prqduce the major part of the turbulence. Below the tube
there is generally a small reduction in turbulence intensity at all of the sections

6ft aft of VG.T.E.-

611 aft of ‘4.G.T.E.e 2in.O.k --
Wz R aft of VG.T.E.+2in.D.bar ----

FIG.4. Comparison of longhdiaal turbukncc prdlcs at various spanwiscand axial stations.

traversed. It is assumed that the redistribution of the turbulent energy is produced by

deflection downwards of part of the lifting vortex from the generator as a result of
interference by the 2 in. dia. tube. In this region the vortex flow is such as to then
entrain local turbulent energy upwards into the section 4 f%from the tunnel eentre-
line, thus producing local increases in turbulent energy.
At the station farther downstream the distributions in the outer se&ions of the
boundary layer have improved. On the eentre-line section the improvement is not as
great as at the other two sections, as the turbulenoe in the outer part of the boundary
layer is still too low.
4.4.3. Lateral rurbuience. The lateral turbulence m- ts are shown in
FIG. 5. The magnitudes of the percentage turbulence are as expected in general lower
than those of the longitudinal turbulence, e.g. 12 and 15 per cent respectively at the
datum level. The observations made co nceming the lor&udiMl turbulence also
apply in the case of the lateral turbulence, i.e. there is an excess of turbulence at lower
Ievels which is only a&cted at one station by addition of the 2 in. dia. tube. The trends
o~~~~s~on~~~~~~~yo~*
4.4.4. VerGxxl ttstfflknc+. Theresultsoftheae- are presented in FiG. 6.
The distributions in this case are slightly difhmt from the two other comets in
 TllcsimulatioooftbeAtmqphaicBomldmyLayainawdTuMel 65

as much as there is an increase in the turbulence level from the datum level up to a
height of 8 in. Above this height the turbulence decreases as in the previous cases.
The turbulence levels also tend to be higher on the centrc-line section and at all scc-
tions there is an excess of turbulence in the lower section of the boundary layer. The

6ft aft of V.G.‘I:E. -

6ft aft of V.G.T.E. + 2in.D.b~ ---
12hftaftofV.G.T.E.+2in.D.bar --a--
9-b \LatCrul tlhblce

of lateral turbulence pro&s at various

Fro. 5. Comparison and axial
spanwise stations.

6ft aft of VG.TE.

6ft aft of b!G.T.E.+PinD.bar -- L
12h ff off of VGTE.*2in &bar ----- vediil fubuiali~~ Ua

same~~areproducadbythe2in.tubeandtheresultsatthe~stationshow
that the distributions at the outer sections of the tunnel are bcgiming to approximate
to the required conditions.
E A
66 J. Amrrr and J. -

The rclativc magnitudes of all three of the turbulent components are also of the
right order.
4.4.5. Reynoi& stresses. The Reynolds stresses are shown plotted in FIG. 7. The
steep gradients at the lower sections of the boundary layer indicate that the flow is
still being accelerated, i.e. mixing is still incomplete at this stage. The addition of the
2 in. dia. tube 25 in. above the barrier results in an increase of Reynolds stress above
25 in. and a reduction below this height at the section 4 ft from the tunnel centre-
line. In general the same observations apply here as in the case of the turbulent
components.

6ft oft of ‘d G.T E.

6ft oft of VG.T.E. t Zin. D. bar ---
12’$ft oft of V.G.T.E.+2in.D.bar -----.
Reynolds stress x7* &,

u’y. x72
4J2_ ‘U,
Fro. 7. Reynolds stmas profiles at various spmwisc and axial stations.

At the section 12.5 ft down&ream at 2 and 4 ft from the tunnel centre-line the stress
distributions tend to approach those required, but that of the centre-line shows the
flow to be still accelerating.
4.4.6. Vortex generator incidence. The effects of changing the vortex generator
incidence is shown in FIGS. 8 and 9. On the centre-line there is very little change in
either the velocity progle or longitudinal turbulence. At the section 2 ft from the
tunnel centre-line the effect of reducing the incidence from 10” to 5’ is clearly seen in
the outer section of the boundary layer. Due to reduction in strength of the lifting
vortex at 5” incidence the irregularities in both velocity and turbulence profiles have
been considerably reduced without affecting the lower sections of the pro&s.
It is possible that the effects of changing incidence may be more apparent further
downstream, but this point has not been investigated.
4.4.7. Energy spectra. The turbulence measurements in the low speed wind tunnel
have not yet been fully analysed to give spectra of turbulence. The prebminary runs
arein~ntwith~~ttstsmadeinasimulatedboundarylayer2in.thickand
the results of these tests are shown in FIG. 10. Due to the scale factor of 24 and the
spesdfadorof~6betweenthettstsinthelowspeedwindtunndandthesmallwiDd
tuMelthefrequencies shown in FIG. 10 should be divided by 40 to give the correspond-
ing result in the low speed wind tunnel. The spectra were measured at a point 3 in.
 The Simulationof the AtmosphericBoundaq Laya in a Wind +hnael 67

bebind the vortex gencratcxs at B he&&t of 8 mm above the ground rouglmess elc-
mcnts, these dimensions correspond to 6 ft and 8 in. respectively at the scaks used in
the bw speed wind tunnel. The two sets of results are for .and non-linear

6ft aft af CIGXE.

hot wire anemometers. The l~~~~d~~ ~rb~en~ spectrum is cornpared with the
spectra pqmed by ~A~RT’(l~3~ and HARRIS(1!?63),the ordinate has been

6ft aft of VGIE.

YG.setting: 5’ --
100 -

adjusted to fit the data. A scale factor of 80 was taken ia fitting the full scak data
to the model results so that t&52 in. ~~~ layer ~0~~~~ to about 1300 ft at
full scale. This is 8 typicai value for the nt height in a hi& wind condition.
68 and J. CouNIfuN
J. ARMITT
The model boundary layer is seen to be a good representation of the fith scale results.
At the higher frequencies the model data falls below the full scale, this could be due
to viscous dissipation. The higher Reynolds number tests in the low speed wind tunnel
should resolve this point but this work has not yet been carried out due to lack of
time.
The curve shown for the vertical spectrum is that proposed by PANOF~KYand
MCCORMKK(1960). There is no free parameter in fitting this curve as it is a function
of the height z. Ln the energy containing range around 300 Hz the spectrum in the

FREQUENCY(Hz.L06SCALE)

FIO. 10. TurbuIencespectra at a height of 8 mm above the ground.

model tests is well above the full scale curve. The total energy in the vertical turbu-
lence is known to be too high at the level where this was measured (see FIG. 6) and if
the spectrum had been measured at a lower level the spectral density might have been
too low. There is a prospect that better agreement will be obtained at a downstream
measuring station as the total turbulent energy has reduced to a more reasonable
level. The lateral spectrum fails between the other two in a plausible manner.
It should be noted that there was signScant turbulence in the free stream of the
small tunnel in which these spectra were measured. With pessimistic assumptions the
correction to the vertical turbulence spectrum will be not more than 15 per cent in
the region close to the ground.

4.5. Summaryof expwimentai results

The proposed system produces a flow which has a turbuha structureapproximat-
ing io the atmospheric wind over fairly open country, HARRIS(1966). The intensity
of the three composts of turbulence at a distance of 12.5 ft from the vortex mtor
tmilingedgeisrepmaentativeoffullscal e conditions. Sign&ant lateral non-u&
f&m&y is present but a reasonably uniform region about 5 fI in width and 2 ft in
 The Simulation of the Atmospheric Fhmdary Layer in a Wind Tunnel 69

depth is available and this should be satisfactory for tests on models at about 1/25Oth
scale. It should be noted that the models will be at 18 to 20 ft from the vortex generator
trailing edge and extrapolations of the trends in FIGS. 5 to 8 suggest that some further
improvement can be expected. Measurements at stations further downstream are
intended in future tests.
The spectra of turbulence are in fair agreement with those proposed for the atmo-
spheric wind in neutral conditions.
The mean velocity proties show greater non-uniformity than the turbulence structure
and the improvement as the flow continues downstream is slow. This illustrates the
fundamental difficulties referred to in Section 3.

5. PROPOSED MODIFICATIONS TO THE SYSTEM

The main faults in the system appear in the mean velocity profile and are mainly
due to the fact that the vortex produced by the generators has rolled into a concen-
trated vortex which is at too great a height for effective mixing in the early stages of
development. It was originally expected that the high turbulence levels would pre-
vent this. A modification of the vortex generator system is being considered. It appears
that the double delta or an ogive planform which was originally rejected as unneces-
sarily complex would be more satisfactory. Furthermore a lower aspect ratio would
be desirable.
Another possible approach would be to use a two stage vortex generator system.
A set of 2 ft vortex generators would be placed behind the barrier (the 2 in. tube will
be removed) these could be set at a fairly large incidence as one can afford a certain
amount of vortex drag at this level. With the vortices closer to ground level the mixing
should be more effective. The 4 ft vortex generators would then be placed behind
these, the incidence would then be smaller as the problem of the vortex rolling up is
less severe at a 5” incidence.

FURTHER WORK

Further work on atmospheric simulation is intended and will be carried out in a

small wind tunnel currently being built. The work in the low speed wind tunnel is
too time-consuming to allow an extensive study of every parameter of interest. As
the frequency of the turbulence is so low (most of the energy is at frequencies below
10 Hz) long integration periods are necessary and a typical run takes several hours
for one vertical traverse through the boundary layer. The small tunnel will allow
quicker runs with fewer failures due to the problem of temperature drift In addition
to a study of the effects of changes in configuration, measurements of space correla-
tions are intended. There is a noticeable intermittency in the turbulence in the outer
half of the boundary layer and it is intended to compare this with results for a natural
laboratory boundary layer.
A further investigation that is proposed is to set the vortex generators to a common
incidence to see if the lateral shear of the atmospheric wind can be simulated. The
presence of the tunnel walls will give problems here and it is likely that only the centre
potion of the flow will be usable.
70 J.~~dJ.~~

7. CONCLUSIONS

It is concluded from a discussion of a number of practical problems in the effects

of the atmospheric wind that the simulation of the turbulence structure is of greater
importance than the simulation of the shear flow. The confbcting requirements of
producing large scale turbulence whilst retaining close control of the velocity profile
imply that a compromise is necessary. This problem is aggravated by the fact that the
mean velocity profile in a boundary layer (even a genuine boundary layer) is slow to
recover from any non-uniformity. A second problem is that the energy balance in a
boundary layer is such that any grid system which produces the correct velocity
profile will probably, in the initial stages at feast, have excessiveturbulent energy near
the ground and a deficiency in the outer half.
The proposed system of a barrier followed by a set of vortex generators at alternate
incidences produces the required intensity and spectra of turbulence. The turbulence
is initially of too high intensity near the ground but the redistribution of turbulent
energy is fairly rapid and the turbulence structure improves as the flow proceeds
downstream. The mean velocity profile shows non-uniformities which do not recover
quickly enough. It was recognized at the outset that some compromise in the velocity
profile is necessary if an acceptable turbulence structure is to be obtained but it is
believed that with suitable modifications to the vortex generator system some improve-
ment can be achieved.

REFERENCES
A,wrrr (1966) The. simulation of the atmospheric boundary layer in a wind tunnet. C.E.R.L. Note
RDILIN/83/66.
BAINES(1%3) E&&s of velocity distribution on wind loa& and flow patterns on buildings. Paper 6,
Symposium16, Int. Conf. on Wind Effects on Buiidiws and Structures hebi at NPL. H.M.S.O.,
London.
CERMAK (1963) Lagrangian similarity hypothesis applied to diffusionin turbulent shear flow. J. nirid
Mech. 15,49-64.
CERMAK(1966) Paper presentedat meeting on Ground Wind Load Problems in Relation to Launch
Vehicles. Langley Research Centre, Hampton, Virgi& 7/8th Jan.
DAVENFQRT (1%3) The relationship of wind strwhxre to wind load& Paper 2, Smsiw 16,
rnt. Cl?& on wild E$kcts on WdfrtlrrElJand s- hdd ot NPL H.ws.0, LQll&n.
DA~EKRXT (1966) The relation&p of wind structure to wind loadin& Paper prewted at meeting
on Ground Wind Load Problems in Relation to Launch V&i&s. I..%@ey Reswch Qntre,
Hampton, Vi&&a. 718th Jan.
FERRY~RIDCE C!owux?xs~(1966) Report of the &d&s of the Gmmittee of impdry C.&G.B.
IfARRa (lW3) E.R.A. Report SPljT14 (unpublished).
HARRIS(1966) E&A, Report SPl/T28 (unpublished).
JENSENand FRANCX(1963) Modcl-scole Tests in Turbde~t W&d Part 1, phmommo Depemic?lton
the windspavad Da&hTecbnical Press, copenhagixl.
LLOM)(1966) The vtion of Shear flow in a wind tunnel. Q& R. met. Sue. 93,79-96.
OwENand Zmrxnnwcz (19!!7’) The production of uniform &ear flock in a wind tunnel. J. Ruid
Mech. 2,521-531.
PANO- and MccoRlcw=r (1960) The spectrum of vertical velocity near the surfas. Q. Jl. R. met.
sac. sq 495-303.
PAHOFSKY and Tw (1964) chpnge of temh and the wind wofile. Q. Jl. R. met.
SUG.9@,147-155.
PAS~UILLF. (1962) Atmospheric D#&on. Van Nostrand, New York.
 The Simulation of the Atmospheaic
Bo~dary Layerin a Wind Ttmd 71

Rare (1966) Resultsof an attempt to gexwate a homogewous abut shear 30w. J. Ruiii nmk
2s, 97-m.
~~ (1960) ~~ Luycr i%ei.wy.~~w-~ New York.
Sawm (1966) On tbe breakdown of cbimm?y plumes into discrete p&Es. Air & FKzt.~o@&r.int. J.
x0,41 l-417.
&ROM(1962) Scaltsmodel wind tmmel studies ODatmospheric -ion. &xee&gs 7th Ny&u&cs
Conference.
TAYLOR(1920) Diffusion by corAm.~ow mcwments. 1920, Proc. to&. Mdh. Sm. !3er,2,20,196-211.
TAYLOR(1935) Statistical theory of ~b~~~. Parts I-IV. Bvc. R. Sot, A ‘151,421478.
TOWNSEND (1937) The turbulenl botmdary layer. Proc. LU.T.A.M. Sag OFIlSmm&aryLayer
~S~~~* springer, Berlin.
TOWNEND(1961) ~~~b~~ layers and waif turb&nce. J. Ffuid Meek. XX,97.
Towmpn, (1966) The fIow in a turbulent bamxiaxy layer after a change in surface roughness.
.A F&&iMe&. 26,255.
TYLDESLEY and WALLINGXON (1965) The effect of w&d shear aad verticai diBi&on on horizontai
dispersion. Q. Jt. R Met. Sot. 9, M-174.
VXCKERY (1966) Flustuating lift and drag oa a k&g cylinder of square cross-section in a smooth and
in a turbuknt stream. 3, &Wd Neck 23,481.