THE STRUCTURE OF VELOCITY AND DENSITY INTERFACES
IN A WEAKLY TURBULENT STRATIFIED SHEAR FLOW
BY
GREGORY M. POWELL
A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF
THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1980
to my parents
ACKNOWLEDGEMENTS
Many people have contributed their time and efforts toward the
completion of this study. In particular the author would like to ex
press his gratitude to Dr. D. Max Sheppard for his continuous support
and guidance. The encouragement and contributions of Dr. Y. H. Wang
are also greatly appreciated. Special thanks go to Ivan Chou for
his assistance in carrying out the experiments and for the many hours
spent discussing key aspects of this study.
Thanks also go to Vernon Sparkman and the Coastal Engineering
Laboratory staff for their assistance with equipment for these experi
ments. The author wishes to thank Katherine Williams for her efforts in
typing the final manuscript and Rhonda Rogers, Rena Herb, Pat McGhee,
Kathy Fariello, Sue Green and Cathy Nell who contributed to the typing
of this manuscript in its early stages. Additional thanks go to
Lillean Pieter who drafted the figures for this study.
Finally, the author wishes to thank his wife Carol for her love,
support and patience.
This work was supported by the Fluid Mechanics Branch of the
Mathematical and Information Sciences Division of the Office of Naval
Research under Contract N0001468A01730016.
TABLE OF CONTENTS
PAGE
ACKNOWLEDGEMENTS . . . . . . . . ... .. .... . iii
LIST OF FIGURES . . . . . . . . ... . . . vi
LIST OF SYMBOLS . . . . . . . . ... .... . viii
ABSTRACT . . . . . . . . ... . . . . . xi
CHAPTER
I INTRODUCTION . . . . . . . . ... . . 1
Motivation . . . . . . . .... . 1
TwoLayer Flows . . . . . . . . 3
A Regime Concept For TwoLayer Flows . . . 5
Scope of This Investigation . . . . . . 7
II BACKGROUND DISCUSSION . . . . . . . . . 9
Introduction . . . . . . . . . 9
Laminar Flow . . . . . . . . . 10
Stability of Stratified Shear Flows . . . .. 11
Turbulent Entrainment Across Density
Interfaces . . . . . . . .... . 14
Dynamic Structure of Stratified
Mixing Layer . . . . . . . ... .17
Dimensional Analysis . . . . . . ... 19
Governing Equations . . . . . . ... 21
III EXPERIMENTAL APPARATUS AND PROCEDURE . . . . .. 25
Facilities . . . . . . . .... . 25
PAGE
Instrumentation . . . . . . . .... 28
Data Acquisition Procedure . . . . .... .32
Data Reduction Procedure . . . . . .... 34
IV RESULTS . . . . . . . . ... ........ .48
Introduction . . . . . . . .... 48
Density Structure . . . . . . .... .49
Mean Velocity . . . . . . . . ... 52
The Richardson Number . . . . . .... .54
Shear Stress Profiles . . . . . .... 55
Energy Spectra . . . . . . . . ... 57
V CONCLUSIONS . . . . . . . . . . . . 83
APPENDICES
A PROFILE SHAPING ELEMENT . . . . . . . .... .89
B INDIVIDUAL DENSITY PROFILES . . . . . . .... .92
C THE ERROR IN ASSUMING Dp/Dt = 0 . . . . . . . 173
D ENERGY SPECTRA . . . . . . .... . . . 177
E VERTICAL SHEAR STRESS PROFILE . . . . . .... 196
BIBLIOGRAPHY . . ... .. . . . . . . . 208
BIOGRAPHICAL SKETCH . . . .... . . . . . . . 212
LIST OF FIGURES
Page
Figur
1
2
.3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
e
THE EFFECT OF VERTICAL LENGTH SCALES ON
THE RICHARDSON NUMBER PROFILE . . . .
SCHEMATIC DRAWING OF THE OVERALL FACILITY .
CROSSSECTIONAL VIEW OF THE TEST SECTION .
UPSTREAM END OF TEST SECTION . . . .
DOWNSTREAM END OF TEST SECTION . . . .
BLACK BOX SCHEMATIC OF THE INSTRUMENTATION
SCHEMATIC OF THERMISTOR ELECTRONICS .. ...
CALIBRATION CURVE FOR THERMISTER . . .
CROSSSECTION OF CONDUCTIVITY PROBE .... .
SCHEMATIC OF CONDUCTIVITY PROBE ELECTRONICS
CALIBRATION CURVE FOR CONDUCTIVITY PROBE .
CONDUCTIVITY VOLTAGE VS. TEMPERATURE . .
HOT FILM CALIBRATION CURVES . . . . .
EXPERIMENTS AND GLOBAL PARAMETERS . . .
EXPERIMENTS AND MEASURED QUANTITIES . . .
NONDIMENSIONAL DENSITY PROFILE 12/6/78 . .
DIMENSIONLESS VAISALA FREQUENCY PROFILE 12/6/
NONDIMENSIONAL DENSITY PROFILE 10/13/78 . .
DIMENSIONLESS VAISALA FREQUENCY PROFILE 10/13
NONDIMENSIONAL DENSITY PROFILE 7/24/78 . .
. . . 24
. . . 39
. . . 40
. . . 41
. . . 42
. . . 43
. . . 44
. . . 44
. . . 45
. . . 45
. . . 46
. . . 46
. . . 47
. . . 59
. . . 60
. . . 61
78 ...... 62
. . . 63
/78 . . .. 64
. . . 65
Figure
21 DIMENSIONLESS VAISALA FREQUENCY PROFILE 7/24/78
22 NONDIMENSIONAL DENSITY PROFILE 10/19/77 . . .
23 DIMENSIONLESS VAISALA FREQUENCY PROFILE 10/19/77
24 NONDIMENSIONAL DENSITY PROFILE 7/6/76 . . .
25 DIMENSIONLESS VAISALA FREQUENCY PROFILE 7/6/76
26 INTERFACIAL THICKNESS VS. KEULEGAN NUMBER ...
27 NORMALIZED MEAN VELOCITY PROFILE GROUP 1 . .
28 NORMALIZED MEAN VELOCITY PROFILE GROUP 2 . .
29 MEAN VELOCITY PROFILE COMPARISON . . . .
30 MEAN VELOCITY PROFILE, LOGNORMAL PLOT ..
31 NORMALIZED SHEAR STRESS PROFILE GROUP 1 . .
32 NORMALIZED SHEAR STRESS PROFILE GROUP 2 . .
33 COMPUTED SHEAR STRESS PROFILES GROUP 1 . .
34 INTERFACIAL FRICTION FACTOR VS. REYNOLDS NUMBER
35 VISCOUS SHEAR STRESS/TOTAL STRESS GROUP 1 ..
36 VISCOUS SHEAR STRESS/TOTAL STRESS GROUP 2 . .
37 VISCOUS SHEAR STRESS/TOTAL STRESS COMPARISON
Page
66
. . . 67
. . . 68
. . . 69
. . . 70
. . . 71
. . . 72
. . . 73
. . . 74
. . . 75
. . . 76
. . . 77
. . . 78
. . . 79
. . . 80
. . . 81
. . . 82
LIST OF SYMBOLS
Symbol Definition
D Distance from the bottom of the tank to the density
interface or the depth of a channel
De Equivalent pipe diameter = 4Rh
v Distance from bottom of the tank to the freesurface
6 Thickness of the density interface
6u Velocity boundary layer thickness
f and f* Undefined or experimentally defined functions
fi Interfacial friction factor
fo Sidewall friction factor
g Acceleration of gravity
g' Reduced gravity = gAp/pz
ys Molecular diffusion coefficient for salt in water
H Depth of the upper layer
K Keulegan number
k Molecular coefficient of diffusion (heat)
K Von Karman's constant
a Turbulent integral length scale
Ab Buoyancy length scale
kv Viscous length scale
X Kolmogorov turbulent microscale
Dynamic viscosity
viii
Symbol Definition
N BruntVaisala frequency
NMAX Maximum Vaisala frequency
N* Nondimensional Vaisala frequency N/NMAX
v Kinematic viscosity
Ve Csanady's (1978) effective viscosity
P Pressure
Po Time averaged pressure
P' Fluctuating pressure
Pe Peclet number based on turbulent length scale
Re Reynolds number
Ret Reynolds number based on turbulent velocity and
length scale
Ri Richardson number
Ri Gradient Richardson number 9g(P/
9 p(aU/l)2
Rh Hydraulic radius
p Fluid density
p Density of upper layer
p Density of lower layer
p' Fluctuating density
po Time averaged density
Ap Change in density across interface = P Pu
p* Nondimensional density excess = (pp )/Ap
S Salinity (mass of salt/unit volume)
T Temperature
t Time
Ti Interfacial shear stress
r Viscous shear stress
TT Total shear stress (viscous and Reynolds)
0 Keulegan stability parameter
U Time average velocity (xdirection)
U Free stream velocity (E > 15 cm)
U Erosion velocity
U* Friction velocity
U1 Turbulent velocity scale
u' Fluctuating velocity (xdirection)
V Time average velocity (ydirection)
v' Fluctuating velocity (ydirection)
W Time average transverse velocity (zdirection)
or half the width of the tank
w' Fluctuating velocity (zdirection)
x Coordinate axis direction of primary flow
y Coordinate axis vertical direction
z Coordinate axis transverse direction to flow
5 Vertical coordinate fixed to density interface
+
+ Nondimensional vertical coordinate fixed to
density interface (c = 5/6)
C* Nondimensional vertical coordinate fixed to
velocity interface (S*=(Sn)/Su)
n Vertical distance between velocity and density
interface
Viscous dissipation function
Def initi on
Symbol
Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment
of the Requirements for the Degree
of Doctor of Philosophy
THE STRUCTURE OF VELOCITY AND DENSITY INTERFACES
IN A WEAKLY TURBULENT STRATIFIED SHEAR FLOW
By
GREGORY M. POWELL
JUNE 1980
Chairman: Dr. Knox Millsaps
Cochairman: Dr. D. M. Sheppard
Major Department: Engineering Sciences
Most natural fluid bodies, the ocean, the atmoshpere, many lakes
and most estuaries, exhibit inherent stratification. Recently, with
the rapid development of coastal and offshore resources and with the
recognition of environmental quality as a fundamental consideration,
stratified flows have taken on a new importance for engineers and
hydrologists.
In this dissertation, the mixing region between two parallel and
homogeneous streams of fluid having different densities and velocities
is investigated experimentally. The flow is produced in a flume with a
test section 24.38 m long, 1.22 m high and 0.61 m wide by pumping the
upper 0.61 m deep layer in a closed circuit. The density of each layer
is controlled by the addition of sodium chloride in such a manner that
the lower layer is about 2.5 percent denser than the upper layer.
Tests are conducted at two downstream locations and at several
velocities between 2.7 cm/sec and 7.8 cm/sec. The mean velocity
and the fluctuating velocity components are measured with hot
film anemometers while the density structure is measured with a
high resolution conductivity probe. Profiles of mean velocity,
viscous stress, and Reynolds stress are obtained through the mixing
region. Energy spectra of the fluctuating velocity components
are also obtained at several points within the region and under
several flow conditions.
Over the range of conditions investigated, the interface moves
about and distorts a small amount in a temporally and spatially
random fashion. However, it is always distinct and readily
identifiable; there are no large convoluting motions of the inter
face. The velocity structure is similar in some ways to flow
over a smooth rigid surface.
The results show that for the conditions investigated the
interfacial boundary layer is marginally stable. When a well
developed turbulent boundary layer exists above the interface,
the Richardson number at the inflection point in the velocity
profile is of order one or less.
CHAPTER I
INTRODUCTION
Motivation
Studies of stratified1 flows are, with few exceptions, motivated
by two important facts. First, most natural fluid bodies are, to some
extent, stratified. The ocean, the atmosphere, many lakes, and most
estuaries exhibit inherent stratification; ostensibly, stratification
is the rule, not the exception. Secondly, when effluents are dis
charged into the water or air there is generally a difference in den
sity between the effluent and the receiving fluid; this produces a
stratified flow.
Most aspects of oceanic and atmospheric circulation are intimate
ly tied to the density structure of the fluid. For this reason, the
dynamics of stratified flows have, for many years, been important to
meteorologists and oceanographers. More recently, with the rapid
Stratification denotes a layered structure. There may be a finite
number of distinct layers in which case the system is referred to as
an "Nlayer" system. In the limit, the fluid may have an infinite
number of infinitely thin layers in which case the term "continuously
stratified" might be used. The property differentiating between lay
ers can in general be any property of the fluid. However, in this
paper, "stratified" will imply density stratification, since the den
sity is the ultimate factor affecting the dynamics of the flow. The
density variations can, in turn, result from temperature variations
and/or variations in dissolved or suspended constituents. Nothing
has been said regarding the orientation or extent of the layers with
respect to other layers or the gravitational field; in this paper,
the term is used loosely to imply a nonhomogeneous density structure.
development of coastal and offshore resources and with the recognition
of environmental quality as a fundamental consideration, the subject
has taken on a new importance for engineers and hydrologists.
Many of the world's large population centers are located on coast
lines near estuaries or on lakes. These centers depend on their
neighboring hydraulic systems for transportation, food, waste disposal,
recreation, and a multitude of other societal requisites. Many times
these requirements are mutually exclusive when the system is improperly
managed. The cornerstone of proper management is an accurate and reli
able model of the flow within the lake, estuary, or coastal zone.
Nevertheless, many mathematical models used today are unable to accu
rately describe the flows when stratification is present. In some
models, this is due to the use of vertically averaged equations of
motion. These equations cannot account for changes in flow direction
or speed within the fluid column. However, even those models that
attempt to account for stratification by employing a multilayer ap
proach suffer from an incomplete knowledge concerning the transport of
mass, momentum, and energy between layers in a stratified fluid. Conse
quently, the mathematical models available today require extensive
calibration and may be unreliable if significant variations are made
from the calibration condition. If the dynamics of natural fluid sys
tems are to be understood, modeled effectively and reliably, the effect
of stratification must be understood.
The consequences of stratification are generally threefold:
1. The mean flow pattern is altered by the redistribution of
inertia and pressure.
2. Turbulent energy is enhanced or decreased, depending on the
nature of the stratification. This, in turn, alters the
transfer of mass, momentum, and energy.
3. Stratification allows for the existence of gravity waves
internal to the fluid. These internal waves are capable of
transmitting energy and momentum across large distances with
little mixing or dissipation of energy.
Any one or all three of the above can have a profound effect on the
dynamics of the fluid system. For an extensive discussion of the many
phenomena associated with stratified fluids, the reader is referred to
Turner (1973), Phillips (1969), and Yih (1965). These books all have
lengthy bibliographies for further reference.
With the increasing impact man exerts upon nature and therefore
himself comes the need to better understand the dynamics of natural
fluid systems and the effects of density stratification.
TwoLayer Flows
Many studies of stratified flows, including this one, are con
cerned with various aspects of twolayer flows. This type of flow,
which is characterized by the presence of two thick layers of nearly
homogeneous fluid separated by a thin transition region in which the
density changes abruptly, continues to receive considerable attention
for a number of important reasons.
The first and most basic reason is the geometric simplicity of
the flow. As with all investigations of complicated phenomena, strat
ified flows must first be broken down into basic structural units;
when these units are understood, more complicated phenomena can be
considered.
Secondly, and perhaps more important in terms of the engineering
application, the twolayer flow closely represents many natural and
manmade occurrences. For example, saltwedge and fjordtype estuaries
often exhibit a density structure which is closely modeled as a two
layer flow. Additionally, when heated water or effluents from indus
trial processes are discharged into the environment the resulting flow
many times approximates a twolayer flow, the dynamics being control
led in large part by the processes taking place at the interface.
Finally, the formation and evolutionary development of the oceanic
thermocline and the atmospheric inversion are strongly dependent on
the mixing process at the thermocline or inversion. These phenomena
are often modeled as twolayer stratified flows. This is, of course,
a limited list of the possible occurrences; many other examples exist.
A third reason for studying twolayer flows, which has far
reaching implications for the study of continuously stratified fluids,
is the observance of steplike density structures in many continuously
stratified flows. Using high resolution instruments to document the
fine scale structure of continuously stratified fluids, several inves
tigators have found that in many natural fluid bodies the vertical
density and velocity profiles are not smoothly varying, as was origi
nally expected, but are steplike in appearance. Thick layers of
nearly homogeneous fluid are separated by thin transition regions in
which the velocity and density change abruptly.1 In the ocean this
The term "fine structure" is sometimes used in the literature when
referring to this type of stratification.
structure has been documented by Woods (1968), Woods and Wiley (1972),
and Osborne and Cox (1972). Woods referred to the nearly homogeneous
regions as "layers" and the high gradient regions as "sheets."1 This
fine structure has also been observed numerous times in the atmosphere;
see, for example, Browning (1971), Browning and Watkins (1970), and
Emmanuel (1973). The steplike structure has even been observed in
the fluff layer of a tidal channel where stratification is a conse
quence of fine suspended sediments.2 It is increasingly apparent that
this density structure is a widespread phenomenon, and the formation
depending little, if at all, on the stratifying agent. This step
like density structure might possibly be modeled as the superposition
of many twolayer flows.
A Regime Concept for TwoLayer Flows
In fluid mechanics, the concept of regimes is a useful tool for
describing or defining the characteristics of a flow. For example,
flows are often described as falling within the compressible or in
compressible, laminar or turbulent regimes. In the study of twolayer
stratified shear flows, a differentiation can be made on the basis of
interfacial dynamics. Observations reported by numerous investigators
Thorpe (1968), Koop (1976), Kantha (1975), and Browand and Wang (1972),
to mention just a few, reveal that interfacial instabilities can pro
duce two distinctly different modes of interfacial mixing.
In this paper the high gradient region may be alternatively referred
to as the interfacial region.
Seminar presented at the University of Florida by Dr. W. R. Parker
of the Institute of Oceanographic Sciences, November 18, 1977.
First, under conditions of high shear and/or low static stability,
large vortical structures, known as KelvinHelmholtz billows, form at
the interface. These large coherent structures produce an overturning
of the density interface, entraining fluid from the upper and lower
layers into the vortex spiral. Many studies, beginning with Helmholtz
(1868) and Kelvin (1871) and continuing to the present, have dealt
with various aspects of this instability. Recent experimental studies
in both liquids and gasesThorpe (1973, 1969), Roshko (1976), Koop
(1976), Browand and Winant (1973), Browand and Wang (1972), Scotti and
Corcos (1969)have greatly increased the understanding of the initia
tion, growth, and decay of this type of instability. Some of these
papers have excellent photographic plates showing the KelvinHelmholtz
billow. The correlation between experiment and theory strongly sub
stantiates the idea that these structures are the result of inviscid
instabilities. In this regime, the interface, in an unbounded flow,
diffuses rapidly with time and/or space until at some point it again
becomes stable.
The second regime exists when the shear is small and/or the static
stability is large. In this case the mixing seems to result from the
breaking of small wavelets which develop a sharp cusplike shape at the
crest and/or trough, ejecting fluid from one layer into the other. The
interface in this regime, unlike the first regime, remains distinct and
easily identifiable. The mixing process, in this case, more closely
approximates turbulent entrainment than diffusion. The origin and
nature of the motions which produce this mode of mixing are not as well
understood as those which produce the KelvinHelmholtz billow. They
might be the end result of viscous or inviscid instabilites which can
originate at the interface or at a distant location. Benjamin (1963)
has identified three types of instabilities which can arise in the
region of a flexible boundary. His results indicate that viscous
instabilities of the TollmeinSchlichting type can arise before the
KelvinHelmholtz instability. Browand and Wang (1973), however, have
proposed that the two different types of mixing result from two dif
ferent modes of an inviscid instability.
Thus, because of the dissimilarities between the two modes of
mixing and the fundamental questions which still remain, it is useful
at this time to study and model the two regimes separately. Eventual
ly, as more data are obtained, it may be possible to consider the two
regimesI through one comprehensive model.
Scope of This Investigation
The intent of this dissertation is to document experimentally, as
accurately as possible, the density and velocity structure of the inter
facial region in a weakly turbulent stratified shear flow. Previous
investigations, many of which are discussed in Chapter II, have either
provided qualitative descriptions of the mixing process, documented
interlayer transfer rates for mass momentum and energy or established
stability parameters for transition from laminar to turbulent flow.
Few have accurately measured the density and velocity structure in the
transition region between the layers. This study documents the time
averaged mean density and velocity profiles, the turbulent and viscous
shear stress profiles and the turbulent energy spectra through the
In this paper, when the terms "regime one" or "regime two" are used
alone, the reference will be to the conditions described above.
interfacial region of a weakly turbulent (regime 2) stratified shear
flow. The intent is to provide accurate baseline data needed to
establish sound theories describing the entrainment of fluid from one
layer to another or the shear stress at the interface between fluids
with different densities and velocities. These two generic problems
form the basis for more specific problems such as the dynamics of
saltwedge and fjordtype estuaries, the development and evolution of
the oceanic thermocline and the dynamics of atmospheric inversions.
In Chapter II, many papers published to date are categorized and
several important results from these studies are briefly discussed.
Chapter III describes the facilities, instrumentation, and procedures
used to obtain the data presented in Chapter IV. The last chapter
discusses the conclusions reached from this study and makes recommenda
tions regarding future work in this area.
CHAPTER II
BACKGROUND DISCUSSION
Introduction
Numerous articles on the dynamics of flow near a density interface
have been published in a wide variety of journals. Generally, these
studies fall into one of three categories: (1) studies of laminar
flow, (2) investigations into the dynamic stability of twolayer strat
ified shear flows, and (3) studies concerned with the turbulent trans
fer of mass, momentum, and energy across a density interface. This
last group can be further divided on the basis of where the turbulent
energy originates. Sometimes mixing is induced by a turbulent
energy source located far from the density interface. This situation
is usually characterized by the absence of mean flow vorticity in the
region of the interface. Other times the transfer results from the
production of turbulent energy at the interface. In this case, a
mean velocity gradient is required at the interface in order to main
tain production of turbulence. Still other times both energy sources
are present. This is the most common condition found in nature;
however, it is also the most complicated.
Studies of laminar flow and hydrodynamic stability are based on
firmly established and highly developed mathematical theories. The
theory of turbulent flow, however, has not yet been developed to the
point where complex flows can be predicted from basic equations.
Investigations of turbulent flow, therefore, must rely on experimental
observation augmented by less well developed mathematical models.
Although there is no direct link between linear stability theory
and fully developed turbulent flow, much insight into the origin of
turbulent motions can be obtained by examining the results of linear
stability theory. This is the primary reason for including, in this
chapter, the brief discussion on stability.
Laminar Flow
Keulegan (1944) and Lock (1951) analyzed the laminar boundary
layer between two streams of fluid with differing density and viscos
ity. The fluids were assumed to be incompressible, nondiffusive and
immiscible, such that the density and viscosity were discontinuous
across the interface. They used the standard boundary layer assump
tions and solved for the velocity in each layer by matching the solu
tions at the interface. As one would expect, their solutions show
that the boundary layer grows with the square root of the downstream
distance, the constant of proportionality depending on the viscosity
in the respective layer. The exact shape of the velocity profile is
dependent on the product of the density and viscosity ratios. Since
the requirement of laminar flow is seldom realized, studies of this
type have limited direct practical application toward problems in
engineering and geophysics. They are important, however, since they
supply the basic mean flow required for stability analysis, and they
form a limiting case upon which to build more applicable models.
Stability of Stratified Shear Flows
Helmholtz (1868) and Kelvin (1871), interested in the generation
of water waves by wind, were the first to investigate analytically the
stability of stratified shear flows. Their studies examined the case
where the density and velocity are discontinuous at the interface and
uniform within each layer (the stratified vortex sheet). Later, Taylor
(1931), Goldstein (1931), and Holmboe (1962), to list some of the more
widely quoted examples, analyzed piecewise linear velocity and density
profiles. Holmboe (1962) and Miles (1961, 1963) extended the analysis
to include hyperbolic tangent velocity and density profiles. For a
lengthy discussion on the stability of parallel inviscid flows the
reader is referred to the review article by Drazin and Howard (1966).
Two important general observations grow out of these analytic
solutions. The first is Miles' 1/4 Theorem: if the gradient Richardson
number is everywhere greater than 1/4, the flow will be stable. This
is only a sufficient condition for stability; it states nothing about
when the flow will become unstable or if stability will persist below
1/4. The second observation is more qualitative in nature. The point
at which the flow becomes unstable and the nature of the instability
depend strongly on the details of the velocity and density structure
or, alternatively, the Richardson number profile. This second point is
reinforced by the results from experimental studies of twolayer
stratified flows. Figure 1 shows two hypothetical velocity and density
profiles and their respective Richardson number profiles. Although the
total change in velocity and density is the same for both cases, the
shapes of the Richardson number profiles are radically different. This
difference results solely from the diversities of the vertical length
scales.
Thorpe (1969, 1973), Scotti and Corcos (1969), and Delisi and
Corcos (1973) investigated experimentally the stability of flows where
the gradient Richardson number is a minimum at the interface (see
Figure 1). They found that Miles Theorem held and that Kelvin
Helmholtz billows formed when the interface became unstable. The in
vestigations of Browand and Wang (1972) and Wang (1975), however,
reveal a much different stability boundary in the Richardson number
wave number plane. It should be noted that in these experiments the
viscous boundary layer was much thicker than the diffusion layer.
Under these conditions, the profile of the gradient Richardson number
will exhibit a local, or possibly an absolute, maximum at the inter
face (see Figure 1). Koop (1976), using the same facility as Wang
(1975), demonstrated that two distinctly different modes of instability
form at the interface, depending on the initial conditions. These two
instabilities correlate well with the results of Hazel's numerical
study (1972) and the analytic study of Holmboe (1962). The character
of the interfacial disturbance created by these two instabilities
corresponds, at least qualitatively, to the two regimes discussed in
Chapter I.
To this point, only inviscid stability theory has been discussed;
the effects of viscosity need to be considered. Maslowe and Thompson
(1971) and Gage (1973) numerically analyzed the stability of stratified
shear flows, including the effects of both viscosity and heat conduc
tion. Their results support the idea that viscosity has a stabilizing
influence. It should be noted, however, that the velocity and density
profiles that were used again led to a Richardson number profile with
a minimum at the interface.
Benjamin (1963) investigated the stability of a flexible boundary
in an inviscid flow, and extended his results to the case of an
inviscid fluid flowing over a viscous fluid. His model assumed a
uniform density in both the upper and lower layers, with a Blasius
velocity profile in the upper layer and zero flow below the interface.
The Richardson number profile under these assumptions is undefined at
and below the interface. This model, nevertheless, closely approxi
mates many actual laboratory and geophysical flows. When the density
difference is large, as with air over water, Benjamin's results show
that the KelvinHelmholtz (KH) instability will be the dominant in
stability. However, when the density difference is small, Tollmien
Schlichting (TS) instabilities can arise at half the velocity required
for a KH instability. Since the TS instability requires energy dis
sipation to grow, viscosity must be considered a destabilizing factor
under certain conditions. To date, no one to the author's knowledge
has examined analytically or numerically the stability of an unbounded
viscous stratified shear flow where the Richardson number profile has a
maximum at the interface. It may be that under this condition, viscos
ity is a destabilizing factor, and TS instabilities can arise before
KH instabilities. This could explain the two regimes discussed in
Chapter I: regime one is undoubtedly the result of KH instabilities;
regime two might be the result of TS instabilities. As Benjamin (1963)
demonstrates, the KH instability extracts energy from the mean flow at
a much higher rate than the TS instability; therefore, if the KH in
stability exists, it will dominate.
Keulegan (1949) experimentally investigated the stability of a
twolayer viscous stratified flow. Using an approach which paralleled
Jeffreys' theory (1926) for the generation of water waves, Keulegan
derived a stability parameter,
1/3
S(vg Ap/p)
U
where v is the viscosity, g the gravitational acceleration, p the fluid
density, Ap the density difference across the interface, and U the
velocity of the upper layer. His experimental data showed that the
critical value for the generation of interfacial wavelets is:
e = 0.127, Re < 450
e = 0.178, Re > 450
where Re is the Reynolds number based on the hydraulic radius of the
flume. The instabilities observed by Keulegan had regime two
characteristics.
Turbulent Entrainment Across Density Interfaces
Many investigators, beginning with Rouse and Dodu (1955) and
Cromwell (1960), have studied, using oscillating grid tanks, the
problem of turbulent mass transport across density interfaces. Within
these tanks, turbulent motions are generated above and/or below a
stable density interface by oscillating a grid of rigid bars. This
experimental apparatus makes the measurement of entrainment rates rel
atively simple; however, since no mean velocity is present, measurement
of turbulent properties is difficult. Hopfinger and Toly (1976) were
able, nevertheless, to measure the turbulent fluctuations using a hot
film anemometer mounted on a rotating arm. Their results show that the
This parameter is sometimes referred to as the Keulegan number (K)
where K = e3
turbulent structure is closely tied to the amplitude and frequency of
oscillation and to the geometry of the grid. The turbulent energy is
concentrated in frequency bands close to the grid oscillation frequency
and harmonics of this frequency. Dispersion of the turbulent energy to
wider frequency bands is slow in this type of facility since there is
no mean velocity shear, and therefore only limited vortex stretching.
Turbulent motions produced in this manner can generate standing waves
at the interface which grow and break due to partial resonance with
the oscillator.
Despite these drawbacks some significant findings have resulted
from these experiments. Linden (1975) found that the rate of increase
in the potential energy of the fluid column is proportional to the rate
at which turbulent energy is supplied to the interfacial region. There
is no simple relationship, however, between the rate of change of po
tential energy and the energy input at the grid. He also found that as
much as half the energy supplied to the interface could be radiated
away by internal waves if the Vaisala frequency below the interface is
not zero.
Turner (1968) and Crapper and Linden (1974) studied the effect of
molecular diffusivity on the mixing rate. They found, for low turbu
lent Reynolds numbers,
Ret = U i < 50
that the mechanics of the mixing process and the rate of mixing were
dependent on the turbulent Peclet number:
Pe E ,U
k
where U,, is a turbulent velocity scale, a is a turbulent length
scale, k the coefficient of molecular diffusivity, and v the kinematic
viscosity. When Pe < 200, the interface has a diffussive core and
the thickness of the interface depends on the value of Peclet number.
When Pe > 200, there is no dependence on Peclet number; the
turbulent eddies completely penetrate the interface, and molecular
diffusion across the interface is insignificant. Turner (1968) also
found when Peclet number is large (the case when salt is used as a
stratifying agent) that:
Ue R3/2 Ri 'p
U1 PU
U 1
where Ue is the entrainment velocity and the Richardson number is
based on U,, and a.
Ellison and Turner (1959), Kato and Phillips (1969), and Moore
and Long (1971) have investigated the problems of turbulent entrainment
across a density interface in the presence of a mean velocity shear.
Their results support the relationship:
Ue I
= Ri
1
The discrepancy between the above two power laws has been the source
of much debate [see Long (1975) ]. It is not clear which expression
is correct or if, as Long (1975) argues, the difference is due to the
presence of a mean velocity shear. It is clear, however, that, in
general,
Ue1(
Ue f(Ri, Pe, Re) and
U1
that only when Reynolds number and Peclet numbers are large is
Ue
f(Ri)
1
Dynamic Structure of a Stratified Mixing Layer
To date, there is no model for the turbulent stratified halfjet
or mixing layer that is valid over a wide range of Richardson numbers.
There are, however, two limiting cases that have been studied and
modeled extensively. First, as the Richardson number approaches zero,
the homogenous turbulent halfjet results. At the other extreme, the
Richardson number approaches infinity and the static stability of the
interface becomes so large that the interface acts much like a solid
plate. The flow of air over water is an example of this extreme case.
The turbulent structures exhibited by these two limiting flows are
vastly different. For example, the eddy viscosity is almost constant
near the center of the homogeneous halfjet, while at a rigid plate
the eddy viscosity approaches the fluid viscosity. When a turbulent
stratified twolayer flow is in regime one, its dynamic structure is
more closely approximated by the homogeneous halfjet than flow over
a flat plate; however, when in regime two the stiuation is reversed.
This dissertation is concerned with the dynamic structure of a regime
two stratified mixing layer. Therefore, the many aspects of a regime
one mixing layer will not be discussed. Readers interested in this
type of flow are referred to Roshko (1976) and Brown and Roshko (1974).
Lofquist (1960) made extensive measurements of velocity and den
sity through the interfacial region of a regime two stratified flow.
Because hot film anemometers were in their infancy at the time his
work was done, Lofquist was able to directly measure only the pressure
gradient and the mean velocity and density profiles. The interfacial
shear stress was calculated using the equations of motion in conjunc
tion with the above measurements. All of Lofquist's experiments
reveal a velocity structure that closely resembles the flow over a
flat plate.
Csanady (1978) demonstrated, using Lofquist's (1960) data, that
a "Law of the Wall" could be formulated for density interfaces. The
"Law of the Interface," however, differs from the traditional "Law of
the Wall" in that it depends on two parameters, Reynolds number and
Keulegan number. Csanady chose to formulate the "Law of the Inter
face" as,
U(z) 1 n(zU ) + B(K),
U K V e(K)
where U, is the interfacial friction velocity, K the Von Karman con
stant, z the vertical coordinate, and ve an effective viscosity
defined by,
2
U,
Ve dU
dz Interface
B(K) and v (K) are constants which depend on the turbulent Keulegan
number defined by,
Apgv
K = 
pU*3
where v is the kinematic viscosity, Ap the density difference across
the interface, and p the fluid density.
Dimensional Analysis
In this section, a functional form of the velocity profile in the
region of a density interface is developed from dimensional reasoning.
The approach parallels the classical analysis of turbulent flow over a
rigid surface, except that the roughness length used in the analysis
of a rigid surface is replaced by a buoyancy length scale.
Under the Boussinesq approximation, and assuming the velocity
profile in the region of the density interface is only a function of
the fluid properties, the shear stress at the interface, gravity and
the vertical coordinate, the following holds:
U = f(p P, g', U., E)
where:
p is the fluid density of the lower layer;
p is the fluid viscosity (assumed to be equal in upper and
lower layers);
g' is the reduced gravity = gAp/pZ;
g is the acceleration of gravity;
Ap is the density jump across the interface;
U. is the interfacial friction velocity = Ti/p~
T. is the interfacial shear stress;
C is the vertical distance from the interface; and
U is the mean velocity.
The above independent variables (excluding C) can be combined to
form two length scales: a viscous length scale,
v PU*
and a buoyancy length scale,
U 2
Lb = *
A physical interpretation of the buoyancy length scale is as
follows: Lb is the vertical distance an inviscid particle of density
p would travel, in a medium of density p Ap, if it was given an
initial vertical velocity U,. Therefore, if U, is a measure of the
turbulent velocity scales, zb is a measure of the vertical distortions
of the interface caused by the turbulent motions within the flow.
Paralleling the approach used in the analysis of flow over a
rigid surface, two alternative functional expressions for the mean
velocity profile near a density interface are:
U ( d) p U g "
U f U*'
and
U(O) f*[ L *
where
u is a Richardson number,
pU.*
Sis a Reynolds number,
and
= = K is a Keulegan number.
When K >> 1, the viscous length scale dominates the buoyancy length
scale. Therefore, the distortions of the density interface (the
interfacial wavelets) will be deeply imbedded within a viscous sublayer
and should have little influence on the flow. Under this condition,
the first functional expression would appear most appropriate. When
K << 1, the distortions of the interface would be much larger than the
viscous sublayer and would therefore dominate the flow. Under this
condition, the second expression would appear most appropriate.
The results from Lofquist (1960) and from this study (Figure 30)
indicate the existence of a region above the interface where the veloc
ity profile has a logarithmic shape. It should therefore be possible
to represent the velocity profile in this region by:
U(E) = A(K) an ( pU ) + B(K) K > 1
U, +
In the case of flow over a rigid surface,
A(K) = 1/K
where K is Von Karman's constant, there is no reason to assume, in the
case of flow over a density interface, that A is constant. The flux
of mass through the interface could alter the turbulent structure and
therefore the value of A.
The above expression is similar in form to the "Law of the Inter
face" given by Csanady (1978), except that Csanady assumed A = I/K
and replaced the fluid viscosity with an "equivalent viscosity." The
author feels the above expression is a more physically meaningful form.
The functions A(K) and B(K), as well as the range of over which this
expression holds, remain to be determined, Finally, it seems reason
able to assume that as K becomes large, A(K)=I >/K.
Governing Equations
The governing equations for turbulent stratified flow are derived
from the differential form of the continuity and NavierStokes equations
using Reynolds decomposition technique. Specific concern here is for
twolayer flow in a retangular channel where the mean velocity is nearly
unidirectional. The xaxis is positive in the primary flow direction
while the yaxis is positive upward and the zaxis is perpendicular to
the x and y axes using a righthanded coordinate system.
If the flow is assumed to be nearly two dimensional and free of
secondary currents then at the center of the tank (z=O), where the mea
surements presented in Chapter IV were made, the continuity equation is
U aV2.1
+ = 2.1
ax ay
where U and V are the time averaged mean velocities in the x and y
directions, respectively.
If the pressure throughout the fluid is assumed to be hydrostatic,
the ycomponent momentum equation is simply the hydrostatic equation
aPo
p =g po(x,y), 2.2
where Po and po are the time mean pressure and density, respectively, and
g is the acceleration of gravity.
Finally, if third order correlations and correlations between veloc
ity and density fluctuations are assumed small, then the mean advective
accelerations, the horizontal pressure gradient, the side wall shear
stress and the interfacial shear stress are the only significant terms
in the xcomponent momentum equation. Furthermore, if po is approxi
mately equal to pu, the Boussinesq approximation holds and the xcompon
ent momentum equation is
aU2 + UV 1 Po + Txy + Dxy 2.3
ax ay pu ax ay az
where
DU
T = u'w' + U
xz PU 3Z '
is the shear stress maintained by the side walls and
= u'v' + aU
xy p + y
is the shear stress maintained by the interface.
Equation 2.3 can be further simplified if one assumes that the flow
does not vary with distance downstream and the side wall stresses are
insignificant. Under these conditions equation 2.3 becomes
1 apo xy 2.4
Pu ax ay
Appendix E contains the full form of the governing equations and
presents in detail the derivation of equations 2.1 through 2.4. In addi
tion, a seven parameter expression for the shear stress profile at the
center of the tank (z=O) is developed from equations 2.1 through 2.3
assuming a selfsimilar velocity and density profile. This expression
for the shear stress will be discussed further in Chapter IV.
 Pu 
4
 Pu
 +
 pu+Ap
to C
Thick Density Interface
7 8u
Thin Density Interface
FIGURE 1:
THE EFFECT OF VERTICAL LENGTH SCALES ON
THE RICHARDSON NUMBER PROFILE
Ri
 Pu+ Ap .
CHAPTER III
EXPERIMENTAL APPARATUS AND PROCEDURE
Facilities
The internal wave tank at the University of Florida's Coastal and
Oceanographic Engineering Laboratory was used in this study. The tank
was designed as a multipurpose stratified flow facility; consequently,
it had available many operational features which were not needed for
this experiment. Only those aspects of the facility pertinent to the
execution of this study will be discussed here. For a detailed de
scription of all the capabilities of this tank, see Sheppard, Shemdin,
and Wang (1973).
Figure 2, a schematic drawing of the overall facility, shows the
locations of the main components. The test section is 24.38 meters
long, 1.83 meters high, and 0.61 meter wide in the lower twothirds of
the tank, while the upper onethird is 0.91 meter in width. Figure 3
shows a crosssectional view of the test section. In this study, the
bottom onethird is filled with salt water, and the middle onethird
with fresh water. The upper onethird of the tank is a wind tunnel
designed for the generation of wind waves and windinduced currents;
it was not used during this experiment. The sides of the test section
are constructed from onehalf inch (1.27 cm) thick glass panels, allow
ing visual observation of phenomena from both sides of the tank.
The flow is produced by pumping the upper layer in a closed cir
cuit. The pumping system consists of two centrifugal pumps connected
in parallel. The 60 hp pump can produce, in the test section, veloci
ties approximating 0.5 m/s. However, at low flow rates (velocities
less than 8 cm/s) this pump heats the water excessively and is diffi
cult to control. The second pump is only 5 hp but it can produce flows
up to 8 cm/s; therefore, for low flow rates the small pump was used to
minimize thermal stratification in the tank.
The saline solution is mixed before each experiment in one of
two 4,000 gallon (15,140 liter) storage tanks; the fresh water is
stored in the other. The saline solution is usually dyed with red
food color, so that the interface between the fresh and salt water is
easily identified. From the storage tanks the water is fed by gravity
into the internal wave tank. The salt water enters the tank through a
horizontal slot in the bottom; its inflow rate can be regulated from
0 to 5 gal/min (0.32 liter/sec) to an accuracy of + 0.05 gal/min
(+ .0032 liter/sec) using a Fischer and Porter tapered tube flow
meter. Fresh water from the city water supply can be continuously
added to the upper layer through a tap in the 12 inch (30.48 cm) pipe
which is located upstream of the pumps. The free surface level is
maintained by the discharge of fresh water across a weir at the down
stream end of the test section. This continuous exchange of fresh
water forestalls any significant change in the upper layer density due
to the entrainment of salt water across the density interface. The
maximum fresh water inflow rate is limited to 12 gallons/min
(0.76 liter/sec).
The entrance to the test section has been significantly modified
for this experiment from that described by Sheppard et al. (1973).
The new entrance section was designed with two criteria in mind:
(1) to allow for a smooth transition in the flow where the fresh upper
layer meets the lower salt layer, and (2) to produce a velocity pro
file which is straight, parallel, and as close as possible to the
fully developed profile for flow in a rectangular channel. The intent
of the second criterion is to minimize longitudinal variations in the
velocity profile. Figure 4 shows the general layout of the entrance
section. The profile shaping element is constructed from 880 pieces
of 3/4inch (1.91cm) I.D. PVC pipe. Each tube is cut to a prescribed
length based on its position within the tank crosssection. For de
tails of the construction see Appendix A.
The transition plate, which is just downstream of the profile
shaping element, is horizontal at their point of contact; the down
ward slope, as well as the curvature, increases with distance down
stream. This design was employed to insure that the upper and lower
layers meet smoothly and with no significant mixing, even though the
vertical position of the interface may vary several centimeters due to
setup and entrainment. The end of the transition plate is always at
least 25 cm below the interface.
The exit from the test section has also been modified for this
study. Figure 5 shows the layout at the downstream end of the tank.
This design minimizes saltwater entrainment and blockage close to the
interface while separating the upper and lower layers. By rotating
the spliter wedge about its pivotal point, the vertex of the wedge can
be corredpondingly adjusted to match longterm changes in the vertical
position of the interface. The vertically slotted weir, located 10 to
20 cm downstream of the wedge vertex, is designed to enhance turbulent
mixing. This eliminates the buildup of a wedge of intermediate den
sity water at the downstream end of the tank, thus reducing blockage
effects near the interface.
Instrumentation
The instrument system for this investigation consists of six in
dependent transducers, nine signal conditioners, a traverse mechanism,
and an analogtodigital (A/D) tape recorder. The output signal from
each transducer is passed through one or more signal conditions and is
sampled and recorded onto magnetic tape by the Kinemetrics Model DDS
1103 A/D tape recorder. Figure 6 shows the general arrangement of the
instrument package.
The Kinemetrics A/D recorder can simultaneously sample up to
16 channels of data at any one of 75 predetermined sample rates between
.05 and 1,000 samples/sec. The input voltage, which must be between
5.0v and +5.0v, is resolved into 4,096 discrete values. Consequently,
the input resolution for this unit is limited to +0.0012 volts.
The traverse mechanism is driven by a variable speed electric
motor which allows the speed to be continuously controlled between
0.02 cm/sec and 2.0 cm/sec. It can move, vertically, a total distance
of 71.0 cm. The probes for transducers two through six, which measure
temperature, conductivity, and velocity, are mounted at the end of the
traverse support rod.
The first transducer, which indicates the vertical position of
the traverse support rod, consists simply of a 10turn, lOkohm
variable resistor in series with two fixed resistors; the potentiameter
is attached to the gear drive of the traverse mechanism. The unit has
a linearity of 0.05 percent over the range of the output voltage from
5.0v to +5.0v. The signal from the position indicator goes to a low
pass filter with 20Hz cutoff frequency. The output from the filter
is then input to the Kinemetrics tape recorder. Since this voltage
can be determined only to within +0.0012 volt by the A/D recorder, the
resolution of the position indicator is limited to +0.085 mm.
The second transducer is a thermistor; its electronic schematic
diagram is shown in Figure 7. The probe, Model SF242A68 manufactured
by Victory Engineering Company, has a sensor only 1.5 mm in diameter
and a time constant of 0.4 second in still water. The thermistor was
calibrated against a mercury thermometer with a resolution of
+0.05 degrees Fahrenheit. The thermistor resolution is assumed, there
fore, to be the same. A typical calibration curve for this transducer
is shown in Figure 8, and although the electronic circuit was designed
to produce a linear output, a small degree of nonlinearity can be seen
in the calibration curve. To compensate for this, the calibration
points were fitted with a second order polynomial (three coefficients),
using the method of least square error. The difference between any
one of the calibration points and the curve fit value was always less
than the resolution stated above. The output from the thermistor
passes to a lowpass filter with a cutoff frequency of 20 Hz; the out
put from the filter is recorded by the A/D recorder.
The electrical conductivity of the fluid is measured by a third
transducer. Knowing the conductivity and temperature, the density of
the fluid can be determined. The conductivity probe, shown in
Figure 9, is constructed of acrylic plastic and has one stainless steel
and one platinum electrode. When the probe is operating, the surround
ing fluid is continuously drawn into the interior chamber through a
0.5 mm slit and a 0.3 mm diameter vertical hole at the bottom of the
probe. The fluid exits through the plastic tube connecting the probe
and the vacuum pump. The electrical current density between the elec
trodes is small except in the vertical hole, where it becomes very
large. Thus, the resistance between the electrodes at any instant is
governed by the conductivity of the fluid in the vertical hole, not by
the resistance at the fluidelectrode interface. This reduces the
effects caused by electrode contamination and plating. Figure 10 shows
the electronic circuit diagram for the transducer. The horizontal slit
in the probe is designed to enhance the selected withdrawal* of the
surrounding fluid. For more on the characteristics of this unit see
Sheppard and Doddington (1977).
The major advantage of this design over the socalled "single
electrode" conductivity probe used by previous investigators is that
the D.C. drift problem has been virtually eliminated while still main
taining the high vertical resolution needed for the study of density
stratified fluids. There are, however, two important disadvantages.
First, the time response for this probe is much slower than the single
point conductivity probe because the fluid must be physically drawn
through the hole in the probe. Based on a comparison of temperature
and conductivity time series data, a dropoff in the response curve is
*When a sink is placed in a stably stratified fluid, it tends to draw
fluid only from its own level provided the velocity is not too large.
believed to occur at about 2 Hz. Second, the probe is physically larg
er than a singlepoint conductivity probe; therefore, its presence will
alter the flowfield for a greater distance. In turn, this necessi
tates a greater spacing between the hot film probes and the conductiv
ity probe. As a consequence of both the above, correlations between
density and velocity fluctuations measured with this system are of
limited significance.
A calibration curve of conductivity voltage versus density for a
fluid temperature of 75.0 degrees Fahrenheit is shown in Figure 11,
while Figure 12 shows the effect of temperature on conductivity voltage
for two different levels of salt concentration. Using these two plots,
the density can be determined to within +5.0 x 105 gm/cc.
The output voltage from this transducer is again input to a low
pass filter with a 20Hz cutoff frequency, and the output from the
filter is again recorded on magnetic tape.
The last three transducers consist of three hot film sensors and
their respective anemometers. The first two sensors are part of an
xtype hot film probe, ThermoSystems Model 124360w NaC1. This probe
is used to measure both the mean velocity and the fluctuating compon
ents of velocity in the plane of the sensors. The third hot film,
Model 1232w NaC1, is a wedgetype probe used only for mean velocity
measurements. All three sensors use ThermoSystems Model 1050 general
purpose anemometers.
The output from each anemometer is first passed to a Thermo
Systems Model 1057 signal conditioner. This conditioner suppresses
the excessive D.C. Voltage to make the signal compatible with the in
put requirements of the A/D recorder. From the conditioner, the
signal is sent to a lowpass filter with a cutoff frequency of 17 Hz
and from there the signal goes to the A/D recorder where it is digi
tized and recorded onto magnetic tape.
The purpose of the lowpass filters is to prevent contamination
of the signal by high frequency noise, specifically 60cycle noise,
which is picked up from the large pumps, electric motors, and other
support equipment in the laboratory. The lowpass filters used with
the position, temperature, and conductivity transducers are fourth
order Butterworth filters with the 3 db point at 20.2 Hz. At 10 Hz,
the power loss through these filters is less than two percent. Sixth
order Butterworth filters are used with the hot film anemometers. The
3 db point for these filters is at 17.3 Hz, which assures that the
power loss at 10 Hz is less than one percent. Spectral analysis of
the velocity data indicates that for these experiments, negligible
energy exists above 7 Hz; therefore, these lowpass filters have little
or no effect on the true signal from their respective transducers.
Data Acquisition Procedure
Two different techniques were employed to obtain data during these
experiments. The first is referred to as a constant speed traverse
(CST), while the second is called a stepped traverse (ST).
The primary objective of the constant speed traverse was to
obtain, in a quasiinstantaneous manner, the structure of the density
transition region. This was accomplished by moving the probe package
through the interface at a nearly constant speed between 0.5 mm/sec
and 1.5 mm/sec. As the probes moved down, the position, conductivity,
and thermistor voltages were each sampled and recorded at the rate of
25 Hz. Since the thickness of the interfacial region was about 1.5 cm,
it took no more than 30 seconds to traverse the entire transition re
gion. The second objective of the CST was to determine the position
of the interface, defined as the point where the density equals the
average density between the upper and lower layers. The position of
the interface was then plotted as a function of time and all other
measurements were referenced to the interface.
The stepped traverse was used to obtain information on the struc
ture of the turbulent boundary layer in the region of the density
interface. With this method the probes were placed at a predetermined
position and the time series data from the thermistor, the conductivity
probe, and the three hot film anemometers were recorded. The time
series were sampled at the rate of 15 Hz over a period of 540 seconds
a total of 8,100 samples. While these data were being recorded on
magnetic tape, the position voltage was recorded in the lab book for a
later determination of the exact probe position. The probes were then
moved to another position and the time series data were again recorded.
This procedure was repeated until sufficient points were obtained
(15 to 20) to determine the profile through the transition region.
The drop in the position of the interface over the 540 seconds
required to record the time series data at each point was much less
than 0.2 mm. This is insignificant when determining the position of
the probe relative to the interface. However, the total time required
to complete an ST profile was as long as three hours; care was taken
to account for the drop in the interface over this time interval.
The steps followed in taking data were:
1. The salt water was allowed to flow in under the fresh water
at the rate of 3.5 gal/min; at this rate 15 hours were re
quired to fill the tank. The filling process began the night
before the tests.
2. The instruments were calibrated and mounted on the traverse
just prior to the start of the experiment.
3. The pump was started and the flow increased to the velocity
desired. This required as much as one hour since it had to
be done slowly to prevent surges in the tank and seiching of
the interface.
4. The pump was run for an additional hour or more before any
data were taken. This allowed time for any transients to die
out.
5. A CST was taken first, followed by an ST profile, which was
followed in turn by another CST. This sequence of three
profiles was usually repeated several times; each ST was
preceded and followed by a CST.
Data Reduction Procedure
The data sets were reduced on the IBM System/370 computer at the
University of Florida using the Fortran IV language. The first step in
the reduction process was to convert the instrument voltages into phys
ically significant values. Because of nonlinearities in most of the
calibration curves (the position indicator is the only instrument which
is linear over the entire range of operation), this must be done before
any other numerical computations. To make this conversion, the least
squares method was used to fit an Nth order polynomial to the cali
bration points for each instrument. Since the position indicator is
linear, N = 1 was used for this instrument. The thermistor and the
conductivity probe, however, required N = 2 and N = 4, respectively
(see Figures 8 and 11). The polynomial equations were then used within
the computer program to convert instrument voltages to physical values.
In the conversion of conductivity voltage to density, corrections were
made for changes in temperature. Figure 12 shows the effect of temper
ature on the conductivity voltage for two solutions with different
salinities. The shape of the hot film calibration curves (see Fig
ure 13) and the fact that below a critical velocity the velocity can
no longer be measured accurately make it desirable to fit these cali
bration points with a sequence of straight lines rather than a poly
nomial. If the hot film voltages dropped below the critical values
(points 1 in Figure 13) more than five percent of the time, the data
were discarded. Variations in the fluid temperature and density were
accounted for when the velocity was computed.
In the reduction of the CST data, it was desirable to have the
data output at nearly uniform intervals of about 0.5 mm (vertical
separation distance). To accomplish this, m values of sampled data
were averaged to produce one output point. The value of m depended on
the speed of the traverse at the time, but it usually fell between
10 and 30. This numerical filter, which reduced the effects of high
frequency electrical noise (specifically 60cycle), was applied to all
three channels of CST data (position, temperature, and conductivity).
Determination of the Vaisala frequency at a point requires the
density gradient, dp/dz, at the point. The nature of differentiation
is such that the calculated value of the density gradient can fluctuate
greatly due to small errors in the value of the density and/or the
position. To lessen this problem, the density gradient at a point was
calculated by first fitting a leastsquares quadratic to five points in
the density profile (the point of interest and two points on.either
side). The density gradient was then calculated by evaluating the
derivative of the quadratic at the point of interest. The thickness
of the interface 6 was determined from,
6 g Ap
p NMAX2
where NMAX is the largest value of the Vaisala frequency taken from
the core of the interfacial region.
The final result from the CST data was a tabulation of tempera
ture, specific gravity, and Vaisala frequency versus position. The
computer program then plotted these profiles using the Gould
(series 5100) plotter. The program also printed out the position of
the interface, the value of the density which defines the interface,
and the exact time that the sensors crossed the interface.
The raw data from the ST runs consisted of four or five time
series: thermistor voltage, conductivity voltage, and the output
voltage from two or three hot film anemometers versus time. The first
step in reducing these data was to convert each value of thermistor
voltage to temperature in degrees Fahrenheit. The temperature at each
point in time was then used to convert the conductivity voltage to the
fluid density (the calibration polynomials discussed above were used
to make both these conversions). Transformation of the hot film volt
ages to instantaneous velocities requires knowledge of the instantaneous
fluid temperature and density around the hot film sensors (see
Figure 13). It should be noted that the thermistor, the conductivity
probe, and the hot film probes were separated horizontally by a small
but finite distance ( = 2 cm). This horizontal separation meant that
high frequency fluctuations in the temperature and conductivity at the
respective probe positions were not related in general to the high
frequency fluctuations in temperature and density at the hot film
sensors. If these fluctuations are not filtered out they may induce
fluctuations in the velocity measurements which do not actually exist.
Also, since the temperature and density corrections were applied uni
formly to both channels of the hot film data, these fluctuations may
have resulted in unrealistic values of the velocity correlations. To
reduce the effects from this source of error, the temperature and
density were averaged over a one second period centered about the time
point of interest. The averaged values were then used to compute the
velocity at the time of interest. Mathematically, this process is as
follows:
1 i+n 1 i+n
T. = T(t=tj) p E p(t=tj)
2n+l j=in 2n+l j=in
V(t=t) = V[HFV(t=ti), Ti, i]
Where HFV is the hot film voltage, V the velocity, T the tempera
ture, and p the fluid density. The functional relationship between V,
HFV, T, and p is shown in Figure 13 and (2n+l)At = 1 second.
In order to estimate the magnitude of this potential source of
error a simple experiment was conducted. Identical time series records,
for a point in the interfacial region, were reduced by two different
methods. With the first method the instantaneous velocity and all the
correlations were computed using a constant value of temperature and
density (the average values of temperature and density for the
540 second time period). The second method used the instantaneous
values of temperature and density to compute the instantaneous
velocities and the velocity correlations. A comparison of the results
showed that the relative difference between the values was always less
than 0.1 percent. It was concluded, therefore, that fluctuations in
the temperature and density do not result in a significant source of
error.
39
E II
o a
E =
oo
LL L
\ E
E
U)=1
LL
7M J
L 0
L
C1\ rn 0
r oo
0 E
,8 1 rO7 '\' Li 
S 0 / <0
o U
0o I( \\\\ 2 o
d> 0 20
UU,
LOU o U)
roa !r > \
a ^ Lu
a) >0
a)a
0 0J t CL U
0 (n)
4 5
^^M &'1L ? o
interface / 183cm
Steel ...._/...
Frame
Stations: 3,6,10,11,15,18
0 10 20 30 40 50 Centimeters
Scale
FIGURE 3: CROSSSECTIONAL VIEW OF THE TEST SECTION
8
I
31 
Do
C)
cc
CLL
_8
E
am
I
s
N Q
toSu
.0
F
LI
IL
LI)
LL)
C)
CI
LL)
LL
~
FIGURE 7: SCHEMATIC OF THERMISTOR ELECTRONICS
90r
701
4.0
L T= 74.8489+ 3.6498 V0.9488V2
I I I
0.0
2.0
2.0
THERMISTOR VOLTAGE
CALIBRATION CURVE FOR THERMISTOR
Thermistor Probe(20K)
Model SF242A68
250K
Output
40 K
10 K
15 Volts d.c.
4.0
..... I I
OP Amps,
MC 1741G
60h
FIGURE 8:
E o
J U) .) Cl
Q oJooU)
02
C.))
0o .0
E o
oE E t *
oc
Sa
0) CD
z
liU
VO
I C)
w CD
W" Q
CD
LL
Temperature = 75.0 F
0.0
5 4 3 2 I 2 3 4 5
5 4 3 2 I 0 I 2 3 4 5
FIGURE 11:
CONDUCTIVITY VOLTAGE
CALIBRATION CURVE FOR CONDUCTIVITY PROBE
35r
Curve # 2
Slope: 0.037
AS.G.= 1.92%(o 70.0 F
\.
3.01
2.5[
3.5
1 4.0
 4.5
Curve # I
Slope: 0.060
AS.G.=0.25% ( 70.0F
2.0
SI I I I I I
FIGURE 12:
70 80 90
TEMPERATURE, F
CONDUCTIVITY VOLTAGE VS. TEMPERATURE
2.5
2.0
1.5
1.0
0.5
 5.0
5.5
47
4.0
0 3.0
0
LJ _Hot Film
H Channel # I
< i Sensor
Temperature 100.601
> 2.0
U
"^Point I
O.5 1.0 1.5 2.0
.45 c m4 ] .45
(V/p ) ) 
S sec gm
4.0
U
0
o 3.0
S_ HotFilm
Channel# 2
j r Sensor
> Temperature 100.6 0
LL 2.0
""Point I
I.0
0.5 1.0 1.5 2.0
.45 r cm4 .45
(V/p) +
(V) sec gm
FIGURE 13: HOT FILM CALIBRATION CURVES
F
F
CHAPTER IV
RESULTS
Introduction
The data presented in this chapter are the results of those
experiments considered accurate, consistent, and reliable. The pri
mary factor in making this determination was the mean velocity. If
the mean velocity at a point varied significantly over a long period
of time or if the velocity profile away from the interface varied
radically with position, the data were discarded as invalid.
Figure 14 gives a chronological listing of the experiments and
the corresponding global parameters which define the state of flow.
Figure 15 lists some of the experimentally derived quantities for each
test.
Five experiments were conducted at a location 800 cm downstream
from the entrance while four others were conducted at 1600 cm. The
overall density difference was maintained between 2.2 and 2.7 percent
while the mean velocity varied between 2.7 cm/sec and 7.8 cm/sec. The
Reynolds number for the upper layer, based on the equivalent pipe
diameter, was always greater than 2.0 x 104. Thus, the upper layer
was always above the critical Reynolds number for turbulent flow,
while the flow deep within the bottom layer was always laminar.
Density Structure
The physical appearance of the density interface is similar to
the surface of water at the point when wind first starts to generate
waves, except that the time scales are longer. Small amplitude wave
lets having large wave numbers and short crests move along the inter
face with an apparently random distribution in both space and time.
The wave lengths, crest lengths and amplitudes of these waves increase
very slowly with distance downstream. Many of the wavelets form sharp
crests, break and eject salt water into the upper layer. At times,
the ejected salt water falls back to the interface before diffusing
completely. The structure of the density interface and the dynamics
of the mixing process clearly indicate that this flow is always within
regime two (see Chapter I). Although the interface is continuously
moving, it always remains distinct and easily identifiable with no
large coherent overturning structures present.
Appendix B contains the quasiinstantaneous profiles of density,
temperature and Vaisala frequency obtained by the CST technique.
Because the density interface is in a state of agitation there is con
siderable variance between density profiles, even between consecutive
profiles taken under the same conditions. Some of the profiles
exhibit what appear to be double interfaces or unstable regions.
These structures may be aberrations resulting from the passage of a
wave during the 10 to 30 seconds required for the probes to traverse
the interfacial region or they may actually exist. When the structures
exist, they appear in all profiles taken during the particular experi
ment. When the structures result from the passage of a wave, they
appear and disappear at random. The density plots from October 26, 1978,
and May 3, 1977, illustrate the first case while those from
October 13, 1978, illustrate the latter. The small perturbations in
the density profiles found deep within each layer are most likely the
result of noise on the conductivity signal. These fluctuations, most
pronounced in the salt layer where the conductivity probe is least
sensitive, translate into large fluctuations in the Vaisala frequency
profiles.
Due to the large variance between individual profiles, a more
meaningful measure of the density structure is obtained by first non
dimensionalizing the density and Vaisala frequency profiles, and then
forming an ensemble average by overlaying several of these plots.
Figures 16, 18, 20, 22 and 24 show the overlayed density profiles
while Figures 17, 19, 21, 23 and 25 show the corresponding Vaisala
frequency plots. The nondimensional density, p*, is defined by
p(/6) pu
p*(5/6) ( /
(pk PU
where pu is the density of the upper layer and p the density of the
lower layer. The vertical coordinate E is nondimensionalized by the
interfacial thickness parameter, 6 defined by
d = 29 Ap
S(P + Pu) NMAXZ
where g is the acceleration of gravity and NMAX is the largest value
of Vaisala frequency from the core of the interfacial region. The
dimensionless Vaisala frequency, N*, is obtained by dividing the
Vaisala frequency by NMAX. The solid line on each of the nondimen
sionalized density plots is given by
P*/) = 1 [l tanh (2)
while the solid line on the nondimensionalized Vaisala frequency plots
is given by
N*(S/6) = sech (2E/6)
These functional representations of the density and Vaisala frequency
are generally a good fit to the experimental data. The greatest
deviation usually occurs in the region above the interface where the
curvature of the density profile is maximum.
Figure 26 shows a plot of the interfacial thickness parameter, 6,
versus the Keulegan number, K. The solid dots represent the data
taken at Station 800 while the open circles are data taken at Sta
tion 1600. The solid line is the line of least square error through
the solid dots. The corresponding correlation coefficient is 0.43,
which for 25 degrees of freedom, is significant at the 0.05 level.
The slope of this line gives
d6
0.0126 cm.
dK
The data sets defined by O
exists between 6 and x. The average interfacial thickness for the
data in this range was 0.44 at Station 800 while at Station 1600 it
was 0.64, thus
= 2.5 x 104
AX
The correlation coefficient between 6 and x for these data sets was
0.59, which given the 28 degrees of freedom is again significant at
0.001 level. Due to the limited number of stations and experiments
it was not possible to determine the functional relationship between
6 and x or between 6 and K, but there is a positive correlation in
both cases.
Mean Velocity
The mean velocity profiles fall into two distinctly different
groups. The primary feature differentiating the groups in the bound
ary layer thickness, 6 defined as the point where the mean velocity
reaches 95 percent of the freestream velocity. The freestream velocity,
U, is, in turn, defined as the average velocity 15 cm or more above the
density interface. The velocity in this region is very nearly uniform.
Figure 27 shows the nondimensionalized velocity profiles for the first
group; the boundary layer thickness for these profiles is between
10.5 cm and 11.5 cm. In the second group, Figure 28, the boundary layer
thickness is only between 1.15 cm and 3.2 cm. The two nondimensional
ized profiles also have different shapes. For comparison, the two
velocity profiles shown in Figures 27 and 28 are drawn together in
Figure 29. In the region (En)/6u > 0.6 the two profiles are identical,
within experimental error. However, below this region the profiles are
distinctly different.
No statistically significant distinction can be made between pro
file groups on the basis of any of the parameters listed in Figure 14.
However, there is a slight tendency for the group two profiles to occur
at lower velocities and at locations closer to the entrance. The small
variations in the boundary layer thickness for the profiles in Figure 27
have no significant correlation with mean velocity, distance downstream
or the Reynolds number based on these two quantities. Therefore, the
flow is believed to be fully developed and in equilibrium. This,
~_
however, is not the case for the group two profiles (Figure 28). The
boundary layer thickness for this set correlates well with the inverse
of the Reynolds number based on the downstream position. The correla
tion coefficient is 0.969 which, given two degrees of freedom, is still
significant at the 0.03 level.
The origin of the vertical coordinate in both Figures 27 and 28 is
defined as the point where the velocity equals 50 percent of the free
stream velocity. This origin is shifted a small amount, n, from the
origin defined by the density interface (i.e. C=0). This small shift
between the density and velocity profiles appears reasonable upon exam
ination of the individual density profiles (Appendix B). A small dis
crepancy of less than 0.2 cm could result from errors in positioning
the conductivity and hot film probes, while larger shifts are usually
associated with the existence of a double interfacial structure. It
appears that the velocity profile, in some way, "locks in" on the upper
most density interface.
The experiments which exhibit the group two velocity profiles also
exhibit relatively smooth density profiles having a single density
interface. On the other hand, the experiments which fall into group one
have density profiles which exhibit either a double interface or a
highly agitated structure (see Appendix B). This correlation between
velocity profiles and density structure extends to the vertical shift
parameter, n. As shown in Figures 27 and 28 the group one velocity pro
files have values of n ranging from 0.38 to 0.80 cm while for the group
two profiles, n falls between 0.05 cm and 0.25 cm.
The mean velocity profile for group one is plotted on lognormal
paper in Figure 30. This illustration points out the existence of a
region above the interface, (c
file has a logarithmic shape.
region.
 n)/6u > 0.15, where the velocity pro
The group two profiles have no such
The Richardson Number
The gradient Richardson number at any point is defined by:
gap N2
Ri = aU aU '2
g p( ( ) ( ),
where N is the BruntVaisala frequency. Given the Vaisala frequency
profile,
N(s) = NMAX sech(2)
where by definition,
NMAX2 = 29 Ap
(P +P )6
and the velocity profile,
U =Uf( ) ,
U
The Richardson number profile can be expressed as:
2g Ap 62 sech (2)
Ri (;) = u
(p +p )26 f( )
where f'(__) is the derivative of f with respect to E1
u u
The gradient Richardson number at the inflection point in the velocity
profile (E=n) then becomes:
S2
2g Ap 62 sech(2)
Ri (g=n) = u
g 2(p + p)6 f' (0)
The gradient Richardson number at =n was calculated for each experi
ment using the above equation and the appropriate values listed in
Figures 14 and 15 (the average value of the 6's listed in Figure 15
was used for 6). The values of f'(0) were obtained for the two
velocity groups from Figures 27 and 28. The results are shown below:
Exp. No. Date Group Ri(=n) f'()
2 5/3/77 1 0.08 8.0
3 10/19/77 1 0.13 8.0
6 10/26/78 1 0.52 8.0
7 10/26/78 1 1.06 8.0
8 11/8/78 1 0.07 8.0
1 7/6/76 2 88.60 0.8
4 7/24/78 2 1.23 0.8
5 10/13/78 2 7.18 0.8
9 12/6/78 2 7.06 0.8
Thus,
Ri (=n) < 1.06 for group one profiles
and
Ri (E=n) > 1.23 for group two profiles.
Shear Stress Profiles
The total shear stress in a turbulent flow is the sum of the
viscous and the Reynolds stresses defined as
T = + T = puv'
T v t 
where IT is the total stress, T is the viscous stress, and Tt
is the Reynolds stress.
Figures 31 and 32 show the total shear stress profile correspond
ing to the group one and group two velocity profiles respectively. The
stress has been normalized by Ti, the total stress at the velocity
interface, E=n. Both these profiles exhibit a marked increase in the
neighborhood of the interface, dropping off rapidly on either side.
The fact that the total shear stress becomes very small just below the
density interface is in agreement with the observation of little or no
motion in the bottom layer even after the upper layer has been moving
for three to six hours. The rapid decrease in the stress profile above
the interface, however, is unexpected since equations 2.2 and 2.4 pre
dict a linear stress profile between the freesurface and the density
interface. In an attempt to explain this unexpected behavior, the theo
retical shear stress profile was calculated from equations 2.1 through
2.3. These equations account for the additional effects of side wall
friction, mean advective accelerations and mean density gradients. The
latter two result from boundary layer growth and turbulent diffusion at
the interface, setup of the density interface and setdown of the
freesurface. Details of the solution techniques are presented in
Appendix E. Figure 33 shows the computed shear stress profiles under
group one conditions and the magnitude of the five parameters which
determine the shape of the stress profiles. These profiles more closely
represent the data (see Figure 31) than does the simple linear model.
However, this solution is still not completely satisfactory. First,
the calculated stress is only 40 to 50 percent of the measured stress
at the interface. In addition, except for profiles one and two,
the magnitude of the parameters are unreasonably large. A possible
explanation for why this analysis is unable to predict the shear stress
profile is given in Chapter V.
The interfacial friction factor, defined by
8 Ti
f. 1
1 p2
is plotted versus Reynolds number in Figure 34 along with the line
corresponding to turbulent flow over a hydraulically smooth rigid
surface. This figure shows that, at the same Reynolds number, the shear
stress at the interface is generally less than or equal to the stress
at a smooth rigid surface. The data points, except for one, cluster
around a line with a 2 slope which implies the interfacial shear stress
is independent of the mean flow velocity. No significant correlation is
found between the interfacial friction factor and the data groups.
Figures 35 and 36 show the ratio of viscous stress to total stress
for the group one and two velocity profiles. There is again an obvious
difference between groups. For group one, the viscous stress drops off
in an exponential manner above the interface; at the interface, the
viscous stress is only 60 to 70 percent of the total stress. For group
two, the viscous stress remains large above the interface until
(Sn/6 ) > 1.0 and then drops off rapidly. For (cn/ u) < 1.0 the
viscous stress accounts for nearly the entire stress and in some cases
balances negative Reynolds stresses. Again for the purpose of compari
son, the curves from Figures 35 and 36 are shown together in Figure 37.
Energy Spectra
Energy spectra of the horizontal and vertical velocity fluctua
tions were measured during four of the nine experiments at several
positions above the interface. These plots, along with pertinent facts
on the measurement technique, are given in Appendix D. The spectra mea
sured well above the interface generally have a smoothly varying shape
indicating broad band phenomena denoting a turbulent flow. The energy
content of the high frequency fluctuations appears to be about the
same for the horizontal and vertical components indicating local
isotropy. However, at lower frequencies the energy in the vertical
fluctuations drops below that for the horizontal fluctuations due to
the presence of the stable density interface.
The spectra measured close to or in the density interface exhibit
a more erratic behavior and a rather sudden change in the slope of the
energy curve between one and two Hertz; this corresponds roughly to
the maximum Vaisala frequency of the interface. There is often a marked
rise in the energy level in this frequency band, especially in the ver
tical velocity component and a rapid decrease in the energy at higher
frequencies.
This characteristic of the energy spectra indicates the presence
of internal wave energy. The fact that the spectra drops off rapidly
above the Vaisala frequency is indicative of the fact that disturbances
with frequencies above the Vaisala frequency are damped rapidly. The
density interface prevents the transfer of high frequency energy from
one side of the interface to the other.
Finally, the spectra measured on December 6, 1978, at E = 0.37 cm
and on July 6, 1976, at = 0.46 cm correspond to group two velocity
profiles. These spectra indicate the presence of internal waves.
( ' in 0 N0 C co Z3 "IO C)
a 0 Min N 0 C 0. CM 0
II
c) xC9
m O 0C c .o I. I
SD~ m , 0 co 0 n
c\j 
S in N C o o M i I C
x 0o O I. .. oC O I
=)n o CO) C) C\j tn
Ln oj 0 i O 0 0 0L
) rN c\j c rN a) co0 CM1
C) C ) NI Ln o' t0 CO)n Cn
C\j C C) U n O C o O n n m
CO CC O C C C =) C=) C) D
L0 C"J iO in in 1n r
 CC M CD CJ C) Cj Ct Co O E
, Ir" O Ln CD Iu "I
' (n 0 l0 r,
C
 E C I C I in in N N o C) u
u C) C C C C) C CD CD CD
xE 0 0 0 0 O O O O O o
co co co co co D
010 o') CM CIO co U3 00
LU t0 Cr)N I r CO I CO CCO Cj cO COCO o0 co
a > N~i N >i Ni N N i N II
C) s o Ua o z aU 0 ,
____________________.:I
xo
LU
J. 1 I I
r
0
C).
I
CU
Z:
L:
i,
U
E
U
I
C0
u E
So
 S
LO 4
W1
1
r I T T T T
0
CO
0
CMD
00
cr)
10
0
0
o
CO
0
CD
0
or
cn
CD
0
00
LO
CO
O
a
) O; C C) O O j oo
10 O'  .J ,r CO r. CO CO
0 0 N0 0 0 0D 0 0
C.o C0 cO O d CO 1O O 0 a
'o 0, , 0 0 a
c O o o C ; c oC; oC;
S CD 0) C. .mJ C..! C0 C'..!
u co ) to r ;r .O CD
E O O OC C\J O C C'j Cj D OC
o o 0 0 0 0 0 0 0
E (D CD C CD 0 C C C
D CD o o o 0 0 C o
o o oo o
(j 0 O r VJ c 1 t I C' O0 C t
l Jo o o o= 4
1 u I1 ***
oooo o O o o O O Ot
nI
0
10
0D
CD>
oD
C
C)
cr
,i
i
V)
LLI
I
0
LiJ
0
d)
00
Li
0
U
....... ... ..:
.... .. ..... .... .........
t....i.... .....i..
ORTE: 12/6 /78
RUN SGMIN SGMAX scm
o 100 0.16 2.55 0. 46
S101 0.13 2.55 0. 47
+ 400 0.08 2.55 0.44
x 401 0.10 2.55 0. 3
............... .... .. .... :. ... ...... . . . .......... : . .. i_... . .... ............. ....... .... ... ..........
., .
P 1tQ n
0 ... .
... .. .. .. .. . ....... ...... . .. . ... .... .... .... .... .
l : i Ii
0 T
......... .. . ... ....... ...... ... ................. ....... . ........ ................. .......... .
D . . .....
. ... ........ ... . ..... .... ........ ........... . .......
'o I ii i i ii i g I u
.= .,,
0
0
'.2 0.0 0.2 0.u 0.0060
Ll
I
crm
CD
n
O
CE
I
LL
'0.20 0.00 0.20 O.I0O 0.60 080 1.00
NONDIMENSIONRL DENSITY p*
FIGURE 16: NONDIMENSIONAL DENSITY PROFILE 12/6/78
. .......... ...... ..... ...... ....... ............
.:.......... ... ... .. ............... ...... ............... .: ........... .. .. .... ..... ....
I
ni
62
. ..... ..... .............   i . .. .... . .. ...
RUN NMAX bcm
 ..;. .. !... x ................... .... 1 0 0 7 7 2 0 .6 *
S... 100 7.72 0.6 O
.o~ .. .101 8.63 0.147
0
0
0
Cu
(u
a
0
00
IL
J
rO
CD
Co
(I
: i : : i : ,. + : ^< j i .. : i : : : !
I . .. ......
. .. .i ... .... .... .... ... .... .... .... . . . .' .... i.. ...
mp 4. . ...
........ ..... .... ................... ..... .... ..... .. ..... .... .. .. ..... ..........
 ... .. .. ; ........ ..... ;....... .. ; .. ........... ... ...... .... .... ..... .... ...
.... : : ... . ].... T .... .... .. i .......:...... i ....... .. 7 ........ . . i .... i.......... 
..... ...... . ... .. ... ...... ._ .. ... . . .. i....... .. ..._ ..... .... .... . .. ....
~ ~ ~ N '**+0e.hi(r
...+... .... .. :... .. .........: i ..
:. . .... .. :... . .. .. ... .. ... . .. ... .... .... . .... 
............ .~ .i ....
E / i>< 0 :' oci : i : /
X i
X a
.. ..... . ..... ..... ... ..... .... ..... ......... ......... ........ .. ....
Y ^ ' ~ / 0 A + """ ''" " '' '" ' : ~' : ~ ' '
p.. .... ......., ..^ ..... .............. 
. ........ .. ...... ...
+ i
J  6: 1 ^ f   I : 1 : I   I 
0,
IX\ ;:I i. + 400 7.78 0. ...
.... .... . . ....a .. . 7 7 0 .
x T01 7.52 0.uq
.. .... '...... "....... .. ^ ......^ . . .. .; i ..
. : . . . i ; .. .. .. ... .. .: i . i ... i .. .
....... . .. .. ..... Q:o ..  ....... .... . ................ .... ..... ........ ........!
X.......... ......... ... ..i .i. i .... .. ...... .
..... ... ...... ......... i ....... .. ....... .... .. ..... ..... . . .. ..... .... . ....  ... . ... .... .
.. .. .. .... .... : .. . . . . . .__ ___ . ._ . ._ . .. ._ . .. !. ._ . .. . .. ... .. .. ._ _ . .! .. . .. . L .. .

.20 0.00 0.20 0.110 0.60 80 1.00
DIMENSIONLESS VAISALR FREQUENCY N
FIGURE 17: DIMENSIONLESS VAISALA FREQUENCY 12/6/78
.
.
DATE: 10/13/78
RUN SGMIN SGMRX 6
.I ...... 110 0.o0 2.45 0.56
S 300 0.03 2.45 0.5u
........ + 301 0.01 2.45 0.39
x500 0.05 2. 5 0.44
. ......................... ...
..I. ....... .
.......... ......... ......... i.............:..... .....:.. ......~... .....l....
.. . . .. .. .. .. ...
CD..
Ci l
F~ i i it1
~ Li
Si
CD .... .......... ... ...... ....... ;(!)r' .. ....1
0............... .
tdnh C
. .
0i
............. .... ..... ... ............................. ........... .......... ......
iA
...i.. i..i.. i.... l............. ..... ......... ..............i... l
................. ..... ........... i......... i.......... ...... ...........i........ ..........! ........ ..........i ......... v 1
......... ...... ........ ....... ......... . .....
......... .................. ........ .......
: ...... .. ^ ........ ..: ..........................; ............ :............. :...... ..... .......................... ............ ......... .
...........
.........
LLJ
F,
IL
C:
C:
C
CC
'0.20 0.00 0.20 0.UO 0.60 0.180 J.00
NNOIND ENSIoNRL DENSITY p
FIGURE 18: NONDIMENSIONAL DENSITY PROFILE 10/13/78
(iD
.. . . .. . .
. . . . .
I  ..... .... .... .... ...
... ....... ... ...
..... .... .... ... .... ...... ... . ...
++
...... .... A  i ...
'/ X+i I+
S....... .. ......... i .... .. .. .... i .. .. i ....
i ... .i....... .. 
'
AX. +
. .. .... ......
i.
:x
....:...!...* ...........i
4+
 +
L"N
RUN FNMRXt 1 6
G 110
S300
+ 301
x 500
6.78
6.79
7.93
7.L43
0.56
0.5q
0.39
0.L44
j_i1~i_ *1 &~
A
.. .. ... . .........i .... ~..... .. ..... .........i 
1 ...i ...I ...i ...i ...; .. i. ; ..... i.....i ..........
..
010
f~. sech(2.. )
T ............._............." I ..... ... ^ ".. .^ i......... ......... .......... i
i : ..... '. ...... . l ..... ...... . i ..... : .....   . ... . ..... . . ... : ...... ........ 
.. .. ... ... . .. ... . . .. . . ..
S i .> _ X__ _
0.0 00 .0 0u0 06 .0 10
0 o
Lu
LU
"
0
Ic
ru
0
0
I
o
I.
I
0.20 0.00 0.20 0.10 0.60 0.80 1.00
DIMENSIONLESS VRISRLR FREQUENCYN*
FIGURE 19: DIMENSIONLESS VAISALA FREQUENCY PROFILE 10/13/78
"''
I
.. .. . .. .
'j<
..., i
I ""ii~
.... : ....,.... : ....
I
i
n~TFr 1n/1~/7~
SI D f T E 7 /2LIY/7~ r
RUN ISGMILN SGMRXI &c
1!i ~
.11..>.
''~':''''
i
:....;
i
:!
1
~
2.62
2.62
2.62
2.62
2.62
2.62
.. .. .. . . . ... . .. .
0 +i
oHR
P ~ 1L tan/~ 1 '
................. ................... ......... ........ ...... ........... ................... .......
t
.....
........ .....
CD3 :  : 1 : 1 : 1 1 1 .
.. ....... ........................................ .............. .. ..... ........................ ..............
0.32
0.38
0.39
0.55
0.56
0.48
.................... ........ ..... ... ... . ...... ........ : ........ .. . ...
......... ...
' : i : : 0 : ] : 0 i 0
S100
S101
+ 400
x 401
* 700
" 701
LU
cc
j
a:
cc
0
CD'
'0. 20 0.00 0.20 O. yO 0.60 0.,80
NONOIMENSIONRL DENSITYp*
FIGURE 20: NONDIMENSIONAL DENSITY PROFILE 7/24/78
1.1
00
.. ..... ...... . .............. ....... ..........
;m
I
...i
L
....... ..... ...... ............ ...... .... .. ..... :.... .......... ........... ........ ................... ... .
; :
7
! : .. . . ., .
0.08
0.03
0.07
0.06
0.03
0.03
Ix4
......~i
0
0
0.
0
=cr
C
4
LU.
(_
+^
(_
t
i.........1....l... t....l i .i. . . .i. 1
: ; : i : i : I
... .. ;... . .. . .... :. .. .. .
"""""""""""" ""'" "'n.^p'4~~
dI4+
i Ix: i ::, __ L i~ i f i
.. __ A _^ s
xxxa
: : :_ a ., ; : :?<
.... .. .. .. .... .. .. i. .. ... ... .
1^ jpL i0? I i 1
.._i. ....... .........Ji
Xi :i
sech
(6
:4
4: iK i _I!..'i
*1'4(0 __~ ""i"'""'""
+ 0 ~s 3
t ''ao i'_'' IX """i" ___'.. .. ..:.. .. ..~. .. ..;.
*a x e3g
*ax  4
......> w & ****** *** ****** ***i*'*'***** I 
i^ .. DiATE: 7 /2i//78
UN NMRAX cm
i 100 9.00 0.32
.... .... 101 8.17 0.38
i@ +. + o00 8.57 0.39...
401 x 7. 17 0.55
.7.. i 00 7.17 0.5S6
S701 7.8 0. q8L
.... ,i .... i........ i... ...... .... ..... i ......... 8 .
.. .. ... . .. .. .... ... I..... . . . .* .
 "Q "&  X ^  '"   *T' i :* : *:
.... ..... ... ...~~. ... ............. ..... ... .........
e .+ __ __
4 * I ***' * "^ .
D
.. . .... ..
 ..... .... ......... .. ... .... ....
I : : I ....I
O.20 0.00 0.20 0.40 0.60 0.80 1.00
DIMENSIONLESS VRISALR FREQUENCYN*
FIGURE 21: DIMENSIONLESS VAISALA FREQUENCY PROFILE 7/24/78
I
'"~I"'":"""''""~':
"""`"'~~~'~''""
. . ................
)
+.
'''''''''
.... ; ... .i .... ; ....
 : i :
.... : .... ; .... ; ...
i
i
i
r
D TE: 10/19/77
SRUN SGMIN SGMAX &cm
e 100 0.26 2.54 0.142
. 300 0.25 2.54 0.71
0 500 0.26 2.54 0.58
. ..... .......................... .............
4 1
... ... ... ... .. ... ... ... .. ...... ....... ............ .. .. ... .... .. ... . .. . .. . . .... .... ............. .... ..... .i .... ..... .... ..... .
... ......... ........ i ... ........ ......... ......... ... ...... . ................. ......... : ....... .....
... ...... .. ....... : . ... i .... ; ... i ... ; .... i .... .... .... .... .... .... i .... .. ..
: i !i i i 7 .
""!... i ...! .. .. .. .. !.. !... ... ...:" : : : : i l
....... i ......... i ......i ....i .. L .i .. .. ._ .. .....L ........ .. ...i..... .... .....i.....i....i. ... ...I.... i
3 V1
........ ...... ....... .... .... .. .... .. .i.. ....; ....... .... .... ........ ..
. .. . ........... .... 7 ........... .......... ...........1............ ..... ..... ...........
.
i
... .. .
cra
uC
aIc
w
cc
LI
S. . . . . .
. .~i J .. ........ .. ......... .. ... .. .... ........... ................
.......... : ......ii ii i i  y
'0 20 0.00 0.20 0. 0 0.60 0,,80 1.00
NONDIMENSIONRL DENSITY p*
FIGURE 22: NONDIMENSIONAL DENSITY PROFILE 10/19/77
I ......... ........ .... .. .... .....:............................ ......... ....... ........... .................................. ........... .......... .
i
'''''
' ' ' ' ' ' ' ' ' '
CD
=r
C
0
c'J
CDP
LU
I
cr
CD
cc
LU
*1 *.... .. .... 1 .... .. 
 A
ja .......... .......:....:....
......... ... ... ... ........
. ........
N es
. ......... ......... .....
C :. : i : : : i :s:"
: .... ... .. . ... i .. ... .... .. . .. ... .. .. . .. . ... .. .. .... .. ....
. .....^....... ... ... ..... ....
A itz *
'*0.20 I  t00 .20 *  I   I  .00
DATE: 10/19/77 i
RUN NMAX Scm
S1001 7.29 0. L2
S300 6.02 0.71
.+ 500 6.64 0.58 ...!
a
'a. 4
. . . . .~ .. . . .
......... ... ... ....... ........
.... ...........T ... :... ... i. ..; .i
'0.20 0.00 0.20 o.DAO 0.60 0.80 J.00
DIMENSIONLESS VRISRLR FREQUENCY N
FIGURE 23: DIMENSIONLESS VAISALA FREQUENCY PROFILE 10/19/77
i
Y...'? n.__________:_________1________:______
i ~t: i1f
,.......,.........,
>1
j:il
,.... : ....,.... ; ...
: 
I
i
I I ;
h
a
.... : ....,.........
~
. ... ...... ... ... ..:.
i. .;....i
.... : .... ; .... ; .... i
;t.]
 : i :
i : i :
i
LLU
ED
ED
Cc
Cc
LLJ
....... ......... ....... . ........ .... ..... ........ ......... ... ...... .. ..... ... ....... ... .. ......... ..... .. ,
; . . ..  :
A ... ..........
... .... .... .... ... . . .. .... . .... .... .... ..... ... .I ... .... .... .... .... .... ....
........ .............. .......... ... . ... .. .. .. ....... .................... ........... ........ ........... ...... ..........
./ i i.:............ .
.... .... .... ... . ...... ... .... .. .... ......... ....... ..... .... .. .... .... .
.: : . : .
g : .... ...... . ... .. .. ......... .... ..... . .. . . .....
.... .... .... . .... .... ... .... .. . .... ... .... .... .... . . i . "..
S...i ......... ......... i........ ......... i
CD\J
o ;" I""""i": i'l"i""j"i'"l"""'~ 63ai
0~~~~~
'1).20 0.00
0.20 0.10 0.60 0..0 ].00
NONOIMENSIONRL DENSITY P
FIGURE 24: NONDIMENSIONAL DENSITY PROFILE 7/6/76
Cp
cr)
C')
0
0
i "" "DR:DTE 7 /i 6, /7J
. ...... ... ......... ....: . . ... D : E 7 / 6 /c m
RUN SGMIN SGMAX _cm
S 1 0.11 2.51 0.71
......... 3 0.08 2.51 0.66
+ u 0.09 2.51 0. u8
x 0.02 2.51 0.46
1 0 0.00 2.51 0.76
.............. ......... ; i ;
i 110 i
................ .. ..... ...." 06...;~....... ........ .... ..........
:Ii. .......... .i
....:..i... :43. ..../..... .....
... .......
45
... .............. .U. .. ..r. ..... .
RUN NMvX b cm
. 1 6. 17 0.71 ...
.... 3 6.32 0.66
+ 7.56 0.148 ...
x 8 7.68 0. 6
1 0 5.73 0.76
. . ..... ..........
.. ... ....... ....  .... ...... ..... .
. .. . . . .. ,.X . . .. . . . . . . . . . . . + . .* . . . ..: . . . . : . .
. .... .....
S , .*.: ....... ^.i^ .e r p .,.
,i .
)"jec(
*1 1.:. ti.: i11.......i
...... ........ . ....... i ........... ....... ..
.. 4" 2^_X ..
I
r 'I i I S I I
o .. . . .. ., . . .. .. .... .. .. . .. .. .. .. .. .... .... .. .. .. ..... .... . . .. .
o i y  i i:
0.20 I 0.00 0.20 ,0" 0 "I" .60 "e "
uJ
I0
Cco
CD
Cc
U)
Jo
rlo
I
Cc
uJ
0r
0
tN\3
A... .... i
i
0.20 o.oo o.20o o.0 0.60 o.eo
DIMENSIONLESS VRISALR FREQUENCYN*
FIGURE 25: DIMENSIONLESS VAISALA FREQUENCY PROFILE
1.00
7/6/76
' ' ' ' ' ' ' ' '
' ' ' ' '
..:...~a.............. . . .' '!""""'"""
Xa ix
~~i ... . . . . .j~
I
. ..... .... ..~'' II 
....... ...... ...... ......
L ~'.. ..... .......
Z
~1~,
 I 
. . ...
I
S EE
o 00 o
+ 00
I) II
Sco
0 o 0
0 0 0 0 ( 0 S 0
0 00 0
o o
E C
CO
C d
Cif
SLU
CO
cn
LU
I
LU
I)
co L
LU
0
I
ij
2:
LU
U1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
EXP# DATE 8u(c
2 Moy 3,1977 11.(
3 AOct. 19,1977 10.
6 v Oct. 26,1978 10.
7 e Oct. 26,1978 10
8 xNov. 8,1978 I1.
o.1 2 0.4 0.5 0.6 0.7
0.8 0.9 1:0
U
U
NORMALIZED MEAN VELOCITY PROFILE GROUP 1
1.5
1.4
1.3
:m) q(cm) U( s )
0 0.80 3.22
8 0.62 4.10
8 0.38 6.52
.5 0.38 4.76
5 0.65 7.79 x
x
A
A
a I
Group I
I I I I
0.0
0.11
FIGURE 27:
0
0
ci
Group 2
0.1 0.2
0.6 0.7
0.8 0.9
FIGURE 28: NORMALIZED MEAN VELOCITY PROFILE GROUP 2
EXP# DATE Su (cm) rT (cm) u ()
I +July 6,1976 3.20 0.1 2.70
9 0 Dec. 6,1978 1.15 0.05 4.05
4 9 July 24,1978 1.45 0.25 7.10
5 oOct. 13,1978 1.60 0.15 4.50
3.0
0.0
0.2
04
0.6
Group 2
MEAN VELOCITY PROFILE COMPARISON
1.5
1.4
1.3
1.2
I.I
1.0
8u
0.3
0.2
0.0
0.1
FIGURE 29:
0

0 0
ri
S0
J
o
z
_
00
O'
o L
LUJ
'I
o
0
1
~ b
~ 2
 ==: =d
IP O
m X
60U
0.9
0.8
0.7
0.6
0.5
04
0.3
0.I
Group I
0.1 0.2
FIGURE 31: NORMALIZED SHEAR STRESS PROFILE GROUP 1
( cm2
EXP# DATE Ti/Pu sec2
2 May 3,1977 .040
3 Oct. 19,1977 .051
6 v Oct. 26,1978 .050
7 eOct. 26,1978 .041
8 x Nov. 8,1978 .052
Group 2
+0
0.1 0.2 0.3
o TT
Ti
.
NORMALIZED SHEAR STRESS PROFILE GROUP 2
EXP# DATE VI/Pc )
1 0 Dec.6,1978 .0371
9 U July 24,1978 .0585
4 o Oct. 13,1978 .059
5 + July 6,1976 .017
FIGURE 32:
CZ
c
dD
dx
(x103)
dU
(x
(xI03"
dsu
dx
(x103)
d77
dx
(x 103)
D 0.46 0.18 0.18 4.5 0.0
2) 1.20 0.48 1.29 0.0 1.0
) 1.26 0.50 1.36 2.07 12.0
S 1.79 0.72 2.14 6.46 20.0
) 2.90 1.16 3.77 15.6 36.6
GROUP 1
I I"A
0.1 0.5
000
I I I I I
0.6 0.7 0.8
0.9 1.0
T
0.2
FIGURE 33: COMPUTED SHEAR STRESS PROFILES GROUP 1
0.5
0.2
0.1
4.ii
I I
0.1
1
dV
dx
(x I05)
79
I I I I l II
.05
fi *
Hydraulically smooth Slope =2
rigid surface
.01
I I I I I I I I
.0050
104 105
uDe
FIGURE 34: INTERFACIAL FRICTION FACTOR VS. REYNOLDS NUMBER
FIGURE 34: INTERFACIAL FRICTION FACTOR VS. REYNOLDS NUMBER
A
I
AA

\ \

\
A x\ G
I I IA I I 'A
Qo 0.2 0.3 0.,
v
0.5 0.6
roup I
0.7 0.8
0.9 1.0
FIGURE 35: VISCOUS SHEAR STRESS/TOTAL STRESS GROUP 1
1.5
1.4
Ty Viscous Shear Stress
TT Viscous + Reynolds Stress
I.I
1.0
0.9
0.8
0.7
0.6
A
0.3
0.2
0.1
0.1
... .. . I I
VN
0.5
0.4
I I I
3.0r
2.6
2.4
8u
2.0
1.6
1.4( 
1.0
0.8
0.6
0.4
02
U
I
10
\
~\

T, Viscous Shear Stress
TT Viscous + Reynolds Stress
Group 2
0
I I I
0.2 0.4
0.2
0.4
FIGURE 36:
0 \
0\
I
0 +
I I I
I I I I I
0.8 1.0 1.2 1.4 1.6 1.8
O
Tv
^ TT
2.0
VISCOUS SHEAR STRESS/TOTAL STRESS GROUP 2
i ~ ~ ~~~ '
Group 2
Edge of Boundary Layer
Group I
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
VISCOUS SHEAR STRESS/TOTAL STRESS COMPARISON
3.0
2.6
TT
FIGURE 37:
CHAPTER V
CONCLUSIONS
This study has documented the structure of the velocity and
density transition region in a twolayer flow by experimentally mea
suring the time mean velocity and density profiles, the turbulent and
viscous shear stress and the turbulent energy spectra. In addition,
the measured data were used to calculate the interfacial friction factor
and the gradient Richardson number at the inflection point of the veloc
ity profile. From these measurements, the following conclusions are
reached.
1) In a weakly turbulent stratified shear flow, the interfacial
region exists in one of two measurably different states. This bimodal
character is not manifest in the magnitude of the flow properties, but
in the distribution of these properties through the interfacial region.
2) No parameter or dimensionless number derived from global or
overall flow properties accurately predicts the existence of the
two modes or groups. However, the gradient Richardson number at the
The term "global" in this context implies a integral scale property,
for example the mean freestream velocity, the depth or width of the
channel, the total density jump across the density interface, or the
Richardson number based on the above. This is in contrast to the term
"local" which refers to point properties such as the density or the
gradient Richardson number based on the velocity and density gradient
at a point within the flow.
inflection point of the velocity profile is distinctly different for
each group. In group one the Richardson number is less than or equal to
one, while in group two the converse is true.
3) The turbulent energy spectra measured within the boundary layer
above the density interface indicate the presence of internal wave
motions under group two conditions. Under group one conditions, no in
ternal wave subrange is indicated by the energy spectra.
4) The total shear stress in the xy plane is small everywhere ex
cept in a narrow region around the density interface where the stress
reaches a maximum (Figures 31 and 32).
5) Over the range of conditions investigated, the interfacial fric
tion factor is generally less than or equal to the friction factor for
a hydraulically smooth surface at the same Reynolds number. Furthermore,
the friction factor is proportional to Re2 (see Figure 34). Therefore,
the interfacial shear stress is independent of the freestream velocity
when the Reynolds number is between 30,000 and 65,000.
The bimodal character of the flow near the interface is indicated
by the mean velocity and shear stress profiles, the profiles of viscous/
total shear stress, and the gradient Richardson number at the inflection
point of the velocity profile. For each of the above, the experimental
data divide into one of two distinct groups.
With respect to the mean velocity, there are two distinguishing
characteristics. First, the group one profiles have a boundary layer
thickness on the order of 10 cm while group two profiles have a thick
ness of only one to three centimeters. Secondly, when nondimensional
ized, the velocity profiles for the two groups have significantly
different shapes. The group one profiles have a shape characteristic
of a turbulent boundary layer over a rigid wall, possessing a loglinear
distribution over a limited region. The group two profiles do not ex
hibit this characteristic shape. The nondimensionalized shear stress
profiles also bifurcate, each group having a distinctly different shape.
The profiles of viscous/total shear stress point again to the
bimodal character. For group one, the stress is 35 to 40 percent turbu
lent at the interface and the viscous component drops rapidly to less
than 10 percent before reaching the outer edge of the boundary layer
(see Figure 37). The group two data, however, show the shear stress to
be about 90 percent viscous throughout the boundary layer falling off
only outside this region. In addition, the data for group one have
little variance while group two data are widely scattered.
The density profiles also show a character that correlates with the
velocity groups. The group one experiments almost invariably exhibit an
agitated density interface, many times possessing two or three distinct
interfaces while the group two experiments usually have a smoothly chang
ing density profile.
The fact that no global parameter or dimensionless number correlates
significantly with the two groups indicates that the existence of the
two dynamic states is not determined by the global properties of the
flow. Even the interfacial friction factor, which is a measure of the
maximum shear stress at the interface, shows no significant correlation
with the groups. In contrast, the gradient Richardson number at the
inflection point of the velocity profile is found to have a strong
correlation with the two groups. This verifies the fact that the mean
velocity and density distributions through the transition region are
significantly different from one group to the other. Since the gradient
Richardson number is calculated from the velocity and density distribu
tion, it can not be considered as the factor that dictates which dynamic
state will exist, but must be viewed only as an indicator of which state
exists at the time. The significance of the gradient Richardson number
at the inflection point of the velocity profile is brought into focus
when combined with two other observations. First, the velocity inflec
tion point is always at or above the density inflection point. Second,
the density transition region is always thinner than the velocity tran
sition region (at least for this flow regime). These two factors com
bined imply that the gradient Richardson number everywhere above the
velocity inflection point is less than or equal to the Richardson
number at the velocity inflection point. Therefore, the gradient
Richardson number is less than or equal to one everywhere above the in
flection point for group one. For group two, this is not the case. To
maintain turbulence, the gradient Richardson number must be less than
or equal to one (see Turner, 1973).
The author suggests the above facts point to the following conclu
sions. First, the two groups represent, in one case, a well developed
turbulent boundary layer (group one) and, in the second, a gravitation
ally stabilized boundary layer (group two). This conclusion is sup
ported by the fact that the turbulent energy spectra, measured within
the boundary layer above the density interface, indicate the existence
of an internal wave subrange for group two experiments. The energy
spectra measured in the same region (although not at exactly the same
point) show no such subrange for group one experiments. Second, since
no correlation is found between the dynamic modes (groups) and any of
the global flow parameters, the author believes the data groupings
must be related to the initial conditions at the start of each experi
ment. Although there was no conscious or observable difference in the
initiation of each experiment, subtle differences may have been present.
The shear stress profiles for both groups (Figures 31 and 32) have
very unexpected shapes which are difficult to explain. The stress is
small everywhere except in a narrow region around the interface, where
the magnitude increases rapidly. One to three centimeters below the
interface the shear stress goes to zero. This is in agreement with the
observation of little or no motion below the interface, even after the
flow in the upper layer has been in motion for three to six hours.
An attempt was made to solve equations 2.1 through 2.3 in order to
determine the origin of the unexpected stress profile. The equations
were solved (see Appendix E) under the assumption that the velocity and
density profiles were each selfsimilar and that each required only one
length scale for nondimensionalization. This approach failed to repro
duce accurately the measured shear stress profiles (compare Figures 31
and 33). One can only assume that significant terms are omitted
from equations 2.1 through 2.3 (terms which were assumed small, see
Appendix E) or that the selfsimilarity assumption is incorrect.
The author believes the latter is more likely, since nondimensionalizing
the velocity profile by one length scale implies the boundary layer
above the interface grows at the same rate as the boundary layer below
