J. Pluid Meeh. (1974), vol. 66, part 2, p p .

223-239 223
Printed in Great Britain

Experimental study of internal waves over a slope

By D A V I D CACCHIONE
Office of Naval Research, 495 Summer Street, Boston, Massachusetts 02210

AND CARL WUNSCH

Department of Earth and Planetary Sciences, Massachusetts Institute of
Technology, Cambridge

(Received 35 January 1974)

Internal waves of the fundamental mode propagating into a shoaling region have
been studied experimentally in a continuously stratified fluid. The waves divide
into three classes depending upon the ratio of the bottom slope y to the
wave-characteristic slope c. For y / c < 1, the amplitude and wavenumber changes
of the waves over the slope are in reasonable accord wihh a simple inviscid linear
theory, prior to wave breakdown near the intersection of the slope and surface.
Considerable mixing occurs in this corner region. When y / c = 1, a striking
instability of the bottom boundary layer is observed and the waves are heavily
damped. When y / c > 1, the waves are inhomogeneous and have complex spatial
dependence.

1. Introduction
Considerable effort has been made towards a theoretical understanding of the
interaction of internal waves and topographic features. Major attention has been
given to the linear problem of internal wave propagation over regions of variable
bottom topography (Sandstrom 1966; Magaard 1962; Wunsch 1969; Keller &
Mow 1969; Baines 1971; and others). The problem not only presents some unusual
mathematical difficulties, but in addition has geophysical applications of con-
siderable interest. For example, the linear theories (e.g. Sandstrom 1966; Wunsch
1969) all predict that if the bottom slope is sufficiently small (defined explicitly
below) internal wave energy will accumulate a t the intersection of the sea bottom
with the top of the stratified layer. Continental margin areas often represent such
a situation, and there may be a significant effect on the mean oceanic conditions
in these boundary areas where incident internal wave energy is dissipated. The
linear theories also predict that, near a particular frequency, the motion induced
along the sea floor becomes very large, and there is a possibility, discussed by
Cacchione & Southard (1974),that sediment may be moved by shoaling internal
waves.
The details of the boundary layer induced along the bottom have been
examined by Wunsch (1969), Wunsch & IIendry (1972), and in a somewhat
indirect fashion by Hart (1971). I n addition to these linear problems, it has been
shown by Wunsch (1971)and Hogg (1971)that Reynolds stress effects of shoaling
324 D.Cacchione and C. Wunsch

FIGURE
1. Geometry and co-ordinate system of the theoretical solutions. The theory does
not take explicit account of the ‘transition corner’ occurring at x = zo.

internal waves may be able to deform the mean density field, and even drive
long-shore (‘long-slope’) mean currents. These effects have never been demon-
strated in the ocean.
I n this paper we describe the results of a series of laboratory experiments
intended to test some of the simpler ideas that emerge from the linear theories for
internal wave propagation over a sloping bottom. A previous crude experiment
related to this work was described by Wunsch (1969). The present experimental
results show some especially striking features in the bottom boundary layer.

2. Theoretical background
The analytic theory of internal waves propagating over a linear sloping bottom
described by Wunsch (1969), hereafter denoted as I, is formally the simplest of
the solutions alluded to above. For this reason the experiments were designed
with this theory as a guide, though the work presented here is primarily experi-
mental, and we have not attempted an extension of the theory to account for the
physical details.
We confine our attention to small amplitude two-dimensional waves propa-
gating in a fluid having constant buoyancy (Brunt-VaisSla) frequency N over
a planar bottom slope as shown in figure 1. The vertical co-ordinate z is positive
upwards; the horizontal co-ordinate x increases in the direction of increasing
water depth. (We shall hereafter refer to the point x = z = 0 as the ‘corner’.)
The water depth is z = - yx. Let c be the wave frequency and define a stream
function such that u = -a@laz, w = a@px,
where u and w are in the conventional x and z directions. The fluid surface is
assumed to be rigid. Wunsch (1969)shows that, if we definec2= a 2 / ( N 2- a2),then

The inviscid linear solutions to (1)fall into three classes: (i) y < c, ‘subcritical’
or ‘transmissive ’ slope, representing waves with phase and group velocity
 Experimental study of internal waves over a slope 225
everywhere in the direction of shoaling water; (ii) y > c, ‘supercritical’ or
‘reflective’ slope, for which all incident wave energy is reflected back into the
interior; (iii)y = c, ‘critical’ slope, for which the direction of energy propagation
is parallel to the bottom slope. The reflexion properties of the slope can be
inferred most directly from the reflexion conditions deduced by Phillips ( 1 966).
Note that c is the slope of the characteristics of the hyperbolic equation (1) along
which wave energy propagates.
For class (i)the solutions given in I for waves travelling towards the corner are
I++@, x ) = A{exp [ - iqln (cx- z ) ] - exp [ - ipln (cx +z ) ] } ,
p=Znn/lnA, A = ( c + y ) / ( c - y ) , n = 1 , 2,....
A strong amplification is predicted along the bottom slope, and the wave energy
1 (2)

propagates into the corner. Greenspan (1969) demonstrates that, as long as linear
theory is valid, the group velocity of the incident waves decreases with the water
depth in such a way that the travel time into the corner is infinite. As far as the
class (i)linear solutions are concerned, the container is ‘open’ (cf.Beardsley 1970).
I n practice of course, nonlinearities or viscous effects must become important to
prevent the infinity in the energy density in the corner. It was also shown in I
that, for sufficiently small values of y/c and far from the corner ( x % zlc), the
horizontal wavelength implicit in (2)is asymptotically the same as the wavelength
predicted over a flat bottom (Lamb 1932, p. 379). Keller &Mow (1969) obtained
the same result from a geometrical optics argument.
When y > c the solutions in (2) are formally valid but have a singularity present
in the interior of the fluid along the line z = -cx emanating from the corner.
Solutions in which this singularity was removed are also presented in I. Energy
impinging on the slope is reflected backwards into deeper water in this case. (See
also the discussion by Robinson 1970.)
Equation (2) is singular as y -+ c; infinite velocities are predicted along the
slope. When y = c the singularity a t z = - y x cannot be removed within the
inviscid small amplitude theory.
The Schmidt number for the salt solutions used in the experiments is about 103,
and we shall assume that for discussion purposes the boundary layers for infinite
Schmidt number given in I are adequate. I n I it is shown that for y #= c the
boundary layer along the bottom slope is essentially a Stokes oscillating boundary
layer for which the non-dimensional boundary-layer thickness is

( R is a Reynolds number defined as R = NL2/v,where v is molecular viscosity,

L is the local water depth and y = tana.) The dimensional thickness was about
0-3cm in the experiment. When u 2 / N 2- sin2&< O(R-*),i.e. y M c, the boundary
layer formally has non-dimensional thickness R-4. But as we shall see below, the
experimental results show that for y M c the boundary layer exhibits a striking
instability. At the order of approximation of I , no dissipation of the interior
wave field takes place.
Phillips (1970) and Wunsch (1970) have shown that an additional buoyancy
15 FLM 66
226 D.Cacchione and C. Wunsch
To motor ‘Corner:

4 cm

H’inge ‘Transition corner’

FIGURE2 . The wave tank, showing the position of a 15” slope. The paddle
was connected through linkages to a driving motor.

boundary layer exists along the slopingboundary in a stratified fluid. A slow mean
flow is driven up the slope over a boundary-layer thickness of order
( v K ) ~ / sin
( Na)+,
where v and K are the diffusivities of momentum and salt, respectively. I n the
geometry shown in figure 1 and with v = 10-2cm2/s and K = 10-3cm2/s, this
boundary-layer thickness is of order 0.05 cm and may be ignored in the discussion
of the wave-driven boundary layer.

3. Experimental arrangement
3.1. Geometry
The basic geometry of the experimental set-up is shown in figure 2, and differs
from the geometry of the above theoretical discussion in two major ways. First,
a free surface was used over the slope, instead of a rigid upper lid. Phillips (1966)
shows that a t small wavelengths this difference has negligible effects on the wave
motion. I n the experiments, vertical motion of the free surface was almost
undetectable and is not thought to have influenced significantly the internal
motion over the sloping bottom. Of course, the structure of the upper surface
boundary layer (a Stokes layer) is slightly different for a rigid top, but this
change is of no consequence in the results.
Second, the wave motion was generated by a paddle in a region of constant
depth and a finite distance from the sloping bottom, leading to a ‘transition
corner’ region, depicted in figure 2, at a distance x,,from the origin a t the apex.
Robinson (1970) has considered the effects of such a corner in detail, by discussing
the diffractive contributions to the ordinary ray solution (whichis asymptotically
equal to equation (2) as the distance from the transition corner increases).
His conclusion is that the ray solution is quantitatively correct as one departs
from the ray (2-cx = constant) emanating from the corner itself. I n the
immediate vicinity of this ray, he shows that linear theory breaks down, non-
linearity smooths out the singularity in the linear solution and, in addition,
diffractive effects occur a t the corner. Anticipating the results of the experiments,
we did not detect any singularities associated with this corner.

3.2. Apparatus
A glass rectangular wave tank 4.9 m long, 21 cm wide and 38 cm deep was filled
with a salt-stratified water solution using a technique similar to that described
by Oster (1965). The range of density gradients selected corresponded to
 Experimental study of internal waves over a slope 227

N(s-')
(Specify gravity - 1.0) x
FIGURE 3. Specific gravity and buoyancy frequency N in a typical experimental run.
0, specific-gravity measurements : A,buoyancy-frequency values N computed from the
specific-gravity profile. 'Tails' at top and bottom are due t o boundary conditions.

buoyancy frequencies N between 0.6s-1 and 1.0s-l. Density profiles like that
shown in figure 3 were obtained by determining the index of refraction of samples
of small volume ( < 1ml) with a hand-held refractometer. The index of refraction
was converted directly to density by applying a calibration technique that
included a correction for local temperature measured with a thermistor.
Nonlinear tails in the density gradients normally formed near the bottom and
below the free surface (cf. Mowbray 1967). Evaporative cooling, when the
flexible plastic cover was removed from the slope section during the measure-
ments, resulted in a more homogeneous layer at the free surface. Evaporation was
reduced significantly (to less than 1mm/day) when a plastic cover was in place.
A paddle capable of generating the first-mode internal wave was placed a t the
far end of the tank (figure 2 ) . The paddle was hinged at its mid-point and driven
by a variable-amplitude scotch yoke connected to a variable-speed drive on a
a h.p. motor. Frequencies could be selected in a continuous band over the entire
range of internal wave frequencies (up to 1-5s-l). Rubber gaskets sealed the
edges of the paddle. That the fundamental mode was indeed being generated was
checked by streak photographs and through the dispersion relation. Extraneous
modes appear to have represented only a few per cent of the energy emanating
from the paddle.
Time series of conductivity fluctuations at selected points in the tank were
used to measure the internal wave motion. Over the range of salinity values used
15-2
228 D. Cacchione and C. Wunsch
during this study, salinity is approximately proportional to conductivity. The
temperature variations in the water were small (usually less than 0.3 "C over the
total depth) except in the upper 2 cm, so that the total change in conductivity at
a point below the upper few centimetres is a valid estimate of the totaI change of
salinity. The conductivity probes and associated circuitry used to make the
measurements are described by Cacchione ( I 970).
Six data channels were sampled sequentially once per second and recorded on
magnetic tape. Normally, voltage signals from four conductivity probes, the
potentiometer circuit on the wave maker and a stable constant reference voltage
were recorded. In situ calibration of each probe was accomplished during each
experiment by raising and lowering the probe over several 1 mm depth incre-
ments and logging voltage vs. position.
Periodogram estimates were computed from the conductivity and wave-
maker records. The starting point on each data record for periodogram compu-
tation was selected when it had been determined that the wave motion was well
established. Calibration constants were then applied to the periodogram ampli-
tudes a, to yield the wave amplitude in cm at each probe location.
The periodogram estimates were usually normalized by similar estimates a,
obtained from conductivity data taken over the flat-bottom regions during the
same experimental run and at the same depth by the same probe. The results are
in the form amla,,and are a measure of the change in wave amplitude from the
flat bottom to the slope taken by each probe. This normalization conveniently
removes certain calibration constants, and provides a non-dimensionalexpression
that can be compared with an equivalent theoretical estimate.
The phase values I9 for closely spaced adjacent probes at the same depth were
used to determine experimental values of the horizontal wavenumber FcE from
the expression ICE z A0/Ax for Ax < A ,
where A0 = 19i+l - Oi, Ax = xi+l - xi and h is the horizontal wavenumber.
(idenotes an arbitrary probe location in the tank, and i + 1 that of the adjacent
probe.) Record durations of at least 50 wave periods, often up to 100, were
obtained for quasi-steady conditions.

4. Results
The results are separated into three basic categories: (i) one that describes the
interior wave field (wavenumber and wave amplitude); (ii)another that describes
the activity a t the boundaries; and (iii) a discussion of wave breakdown. Con-
siderably more detail about the results may be found in Cacchione (1970).
Since N and y were constant during each experiment, the ratio y / c could be
altered by changing g. I n this way, all three hydrodynamic regimes y/c < 1,
y/c > 1 and y/c = 1were investigated during each experiment.

4.1. Interior wave jield : y < c

Streak photographs of neutrally buoyant particles confirmed that the funda-
mental mode was indeed the dominant motion. The predicted bottom intensifica-
tion could also be clearly seen in the photographs.
 Experimental study of internal waves over a slope 229

0
” 3

2
1

0 1 2 3 4 5 6 7 8
102~/[22- (z/c)2]

FIGURE 4. The experimental wavenumber plotted vs. its theoretical value for y < c. I n
both cases shown, a, = 0.2 cm. 0 , ‘low’ frequency waves in which u / N N 0.07; 0 , ‘high’
frequency waves with u / N 1: 0.13. However, in both cases different runs had somewhat
different frequencies. The corner is a t the upper right. Data were taken a t various values
of 2.

Wavenumber. An experimental wavenumber IC, was computed from the

measured phase differences between adjacent probes. I n figure 4 we plot lcJq
[q is defined in ( 2 ) ]against its theoretical value x/[xz- ( z / c ) ~over
] the slope, where
the expression is averaged for the position of the two probes. Measurements are
shown only in the region well away from the wave breakdown area and outside
the bottom boundary layer, where the linear inviscid theory is presumably
applicable.
The theoretical curve in figure 4 is a straight line with a slope of unity. Typical
experiments with different frequencies and amplitudes are shown. The upper
right port’ion of the diagram corresponds to the corner region (x,x = 0 ) of the
slope. The plot shows that the wavenumber decreases with increasing distance
from the corner. The agreement of the results with the theoretical curve is best
for waves over the lower section of the slope, well away from the corner.
Apparently, the transition corner region has little or no effect on the wave
motion. Towards the corner, the data points depart significantly from the
theoretical curve for the lower frequency waves where the predicted increase in
wavenumber does not take place. There is also some evidence in other data of an
amplitude dependence in the wavenumber variation. The experimentally deter-
mined value of the wave steepness e = a,km at the positions nearest the corner
is approximately 0.15; nonlinear effects are apparently already beginning to
have a significant effect on the motion.
Wave amplitude I n figure 5 we show the measured 9s. theoretical wave ampli-
tude squared. We chose this form of representation since the measured wave
displacement is a complicated function of position. Lower frequency waves, with
u / N E 0.07, show earlier departures from the theoretical curves than do the high
frequency waves, with (TINci 0.13. Actual amplitudes are much lower than those
230 D . Cacchione and C. Wunsch

60 -
0

50 -

40-

N
h

5B 30 -
v

20 -

0 10 20 30 40 50
Theoretical value
FIGURE 5 . The measured square of the amplitude ratio as.its theoretical value. 0,results
for ‘high’ frequencies, u ~ N1: 0.13, which are in quite good agreement with the theory
even a t very large values; El, results for lower frequencies, a / N 1: 0.07, which fail to
attain the predicted high values, principally because of the strong dissipation along the
slope. Data were taken a t various values of x.

predicted on the middle and upper sections of the slope for the lower frequency
waves. At these slope sections it was also found that the measured values of
steepness are considerably less than those estimated from inviscid t>heory.The
larger input waves (a,, = 0.4cm) show a tendency to peak earlier than predicted.
Departure from theoretical estimates is most noticeable for the very low
frequencies. Possibly a transformation from the sinusoidal wave form to a
different wave structure (such as a solitary wave) that is heavily damped along
the slope accounts for the nature of t,he experimental data for these kinds of
waves (see below). I n any event, the amplitude of these lower frequency waves is
not as large as predicted, and there seems to be evidence of strong dissipation
along the slope. The waves were observed to surge as they approach the corner.

4.2. Boundary activity and breaking : y < c

These results are primarily qualitative. Observations of the oscillatory flow in
the boundary layer and breaking zone were made with shadowgraph images, the
motion of neutrally buoyant floats and dye streaks. One obvious major feature
of this flow was its intensification near the bottom along the slope as compared
with the flow along the horizontal bottom. The maximum velocities generally
increased a t positions further upslope, with a sharp growth in the velocities near
 Experimental study of internal waves over a slope 231

UlN a, = 0.2 a, = 0.4 cni

0.74 13.1 14.0
0.69 12.5 12.9
0.61 10.6 11.1
0.58 9.9 10-5
0.56 9.0 9.5
0.54 8.2 8.6
0.52 7.7 8.2
TABLE1. Estimated position of breaking L, measured along the
slope from the corner x = z = 0.

the zone of breaking. The oscillatory flow remained laminar in the boundary
layer up to the zone of breaking.
The breaking of waves on the relatively steep slopes (30and 45")is characterized
by considerable turbulence and vertical motion below the free surface and by the
generation of spatially irregular fine-structure in the density field. This kind of
breaking is also observed for relatively high frequency waves on the smaller
slopes (7 and 15"). I n contrast, the lower frequency waves (i.e. those waves
with frequencies closer to the condition y = c ) break much less violently, instead
propagating upslope as distinctive bore-like features or 'surges ' that eventually
collapse and form a thin fluid stream along the slope, similar to the swash and
backwash associated with the breaking of surface waves on a beach (see figure 6,
plate I). This phenomenon might also be related to the internal surf and swash
described by Emery & Gunnerson ( I973). Over larger slopes, the vortex break-
down is more violent.
For higher frequency waves, shortly after the first arrival of the waves near the
top of the smaller slopes (a = 7 and 15"), a small vortex forms along the slope
beneath the crest of the advancing wave form. Figure 7 (plate 2) illustrates the
onset of this instability beneath the wave crest. I n the figure, the maximum
amplitude of the incoming wave a t the entrance to the slope region is approxi-
mately 0.4 cm. A similar phenomenon was noted by Thorpe (1966) (reproduced
by Turner 1973, pl. XII). As the wave breakdown proceeds, one can see the
protrusion of thin horizontal layers back into the interior region over the slope.
The protrusions lose their intensity (as measured by photographic contrast) over
very short horizontal distances (5-10 cm) and are confined to the upper vertical
levels, where the breaking occurs. The protrusions, thin tongues of mixed fluid
that are weakly convected back into the interior, are caused by mixing in the
zone of breaking.
Table 1presents a summary of shadowgraph observations of the positions L, of
breaking on slopes of 30". L, is measured from the corner along the slope. The
accuracy of the distance L, is low, owing to the difficulty in defining the point of
breaking; however, the relative order of the values of L, for the waves of different
frequencies shown in the table is reliable. The general trend indicates that for a
given N the higher frequency waves break first, i.e. in deeper water.
232 D. Cacchione and C. Wunsch
0

A B
8. Detail of the reflexion characteristics for y > c . Note the concentration of
FIGURE
energy between the line z = - c z and the slope (triangular area AOB).

4.3. Interior wave j e l d : y > c and y = c

For y > c (supercritical),two features are suggested by the theoretical reflexion
properties: (i) the presence of back-reflected wave energy; (ii) concentration of
this back-reflected energy in a zone bounded by the bottom slope and the charac-
teristic that intersects the corner. Wave characteristics which are incident to the
slope from above, i.e. those with a downward-pointing vertical energy com-
ponent, upon reflexion from the slope will remain below the characteristic (defined
by z = - cx) that intersects the corner (figure 8). The energy flux associated with
these characteristics enters the slope region in a relatively broad band and leaves
in a confined triangular zone, the triangular area AOB in figure 8. There is a
smaller influx of energy which, upon reflexion from the flat bottom, propagates
upslope within the narrow bottom zone (for example, the path of the charac-
teristic (b) in triangle AOB). The characteristics associated with this latter
energy flux have upward-pointing vertical components as they enter their slope
regions; upon reflexion they preserve the sense of the upward-directed vertical
components. I n both cases (characteristics ( a ) and (b) in figure 8 ) , once the
characteristics reflect from the slope, the horizontal direction of energy propa-
gation is away from the corner. The damping was sufficientlygreat that we could
find no evidence for reflexion of energy from the paddle; some was undoubtedly
present.
The theory of I selects n = - 2 as the lowest non-singular inviscid mode
permissible over the slope region for each of the experimental conditions tested.
We had no means of generating this mode and hence could not make a direct
experimental test of the reflective-mode solutions. For reference however, we
plot the theoretical behaviour of the n = - 1solution (though it is singular along
z = - cx) and the n = - 2 solution in this regime.
I n the experimental runs, the frequency of the wave a t critical frequency was
within 2 % of the required value.
Wavenumber. Figure 9 illustrates the complicated behaviour of the experi-
mentally determined wavenumbers for y > c. The calculated position of the
critical characteristic (determined by x = - ex) for the experimental groups is
shown along the horizontal axis. The theoretical curves were computed for two
different values of the mode number.
There is a qualitative indication that the curves for n = - 1 follow the same
trend as the experimental results. The experimental values near the calculated
singular lines are large, as expected; however, there is a significant shift in the
 Experimental stu,dy of internal waves over a slope 233
22
20

15

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
U
.
1X

9. Spatial variation of the wavenumber for y >

FIGURE c. a, = 0.2 em; a/N = 0.05. Also
shown are the theoretical results for an n = - 1 wave (which is singular) and an n = - 2
wave (which is continuous). The wave generated corresponds to n = - 1 .

peak value to the right of the calculated critical point in figure 9. This shift
probably indicates that the singularity is removed by nonlinear processes (as
suggested by Robinson 1970) and, as expected, that the linear inviscid solutions
are not valid here. The measured values of the steepness parameter 8 for this
particular data set are as large as 0-5 a t x/xo = 0.36. Nonlinear effects are probably
significant near these positions.
Figure 10 summarizes the experimental results for the wavenumber for the
critical case y = c. The plot emphasizes the steep rise towards larger values of the
wavenumber for decreasing distance to the corner. I n general, the wavenumber
variations with distance to the corner for this case are similar to the variations
shown in figure 9 for the supercritical case.
I n summary, the results show a complicated relationship between the wave-
number and position over the slope. There is evidence that the wavenumber
increases rapidly as the waves pass over the lower and middle sections of the
slope for the supercritical case. The variations in wavenumber for the critical
case are qualitatively similar to those for the supercritical case. For both cases,
the measured values of 8 are greater than 0.1 for most of the measurements taken
closest to the corner.
Wave amplitude. The results in figure 11 are presented as ratios of the square
of the wave amplitude over the slope to the square of the wave amplitude over
the flat bottom for various frequencies of the input waves. The horizontal axis,
-z/cx, was chosen to facilitate interpretation of the results in terms of the
reflexion properties of the slope. As described earlier, theory indicates that near
the critical characteristic defined by z = - cx and in a zone between this charac-
teristic and the bottom slope (figure 8) there is a possible region of concentrated
234 D . Cacchione and C. Wunsch

0
0 A

4%
10. Spatial variation of the wavenumber for y w c. cr/N = 0.075.
FIGURE
0 , A,a, = 0.2 cm; CI, a, = 0.4 cm.

energy flux. Figure 11 shows that the measured values of the wave amplitude
are highest near - z/cx = 1.0 for each frequency. However, it is not conclusively
shown in this figure that the amplitudes peak exactly a t this position. The
experimental values (u,/ao)2 increase sharply towards the critical characteristic;
beyond this point, the amplitude ratios then decrease towards the slope. The
larger input waves grow more rapidly as the critical characteristic is approached.
Other results show that the amplitude ratios increase with decreasing wave
frequency at equal values of - zlcx.
The theoretical curves for n = - 1 and n = - 2 are also shown in figure 11. The
theory of I selects n = - 2 as the lowest non-singular inviscid mode for the
experimental conditions tested. The curve for n = - 2 does not predict the sharp
increase in measured amplitudes toward the critical characteristic, nor does it
show the decrease in amplitudes at positions between the critical characteristic
and the slope ( - z / c x < 1.0).
Results for the probe measurements of wave amplitude when y z c are shown
in figure 12. The increase in amplitude over the range 0.38 < x/xo < 0.60is similar
to the increase shown in figure 11 for the same range. However, the maximum
 Experimental study of internal waves over a slope 335

A
0 On[ n=-I

n=-2&
I
'.2-.-- _ _ _
I > - J L ---r _--I
,
I I
0 0.2 0.6 1 .o 1.4 1.8
- z/cx
FIGURE11. Spatial variation of the wave amplitude for y > c. u / N = 0.05. 0, 0,
a, = 0.2 cm; A,a, = 0.4 cm. The data were taken at a constant value of - 2 = 6.0 cm.

B
0
$ A
0 8

0.1 0.2 0.5 1.o

4x0

12. Amplitude variation for y w c. u / N = 0.075.

FIGURE
0 , a, = 0.2 cm; A,a, = 0.4 em.
336 D. Cacchione and C . Wunsch
amplitude measured for these waves is considerably less than the corresponding
value for the supercritical case. Larger viscous dissipation along the slope in the
critical case (see below) appears to explain the smaller peak amplitude levels.
I n summary, the data indicate that the wave amplitudes generally increase
sharply during shoaling over the middle sections of the slope. Peak levels of
amplitude were generally found near - z/cx = 1.0. Peak values of amplitudes for
waves of critical frequency are consistently lower than those a t lower frequencies.
I n both cases the wave amplitudes diminish near the slope, but remain larger
than the input values. Measurements of amplitude agree qualitatively with the
reflexion properties of internal waves from a rigid sloping bottom, but cannot
be compared directly to the predictions in I.

4.4. Boundary activity: y = c and y > c

With y = c and with steep slopes (30and 45"), intense mixing occurs in a zone
of waves breaking near the corner similar to that described for the subcritical
waves on a 30" slope. For runs with gentler slopes (7 and 15"), each wave, for
both the critical and supercritical cases, forms a surge that usually ends with a
short run-up. This run-up typically covers much shorter distances than run-up
for waves in the subcritical case.
Figure 13 (plate 3) illustrates in a very striking way that the shoaling of
internal waves for critical and supercritical cases produces an instability along
the slope prior to breaking. Shadowgraphs of the instability illustrated in
figure 13 indicate that a line of regularly spaced vortices forms near the bottom.
These vortices grow and decay over each half-cycle as they oscillate along the
slope a t the frequency of the input waves. During downslope particle motion in
the waves, the vortices 'flatten', indicating that they form only during the
upslope movement of the water particles. The circulations of adjacent vortices
have the same sense and the axis of each vortex is oriented approximately normal
to the side walls of the tank. The maximum distance observed between the core
of adjacent vortices was about 2 cm. Unlike the vortices which form and propa-
gate upslope beneath the shoaling subcritical waves prior to violent breaking,
these vortices ( y = c) retain their cyclic growth and decay pattern for the duration
of the input wave motion. Their regular spacing is much shorter than the wave-
length of the input waves, and presumably is related to the boundary-layer scale.
The onset of vortex activity for any particular run normally occurs within
10-20 wave periods after the arrival of the first wave motion at the corner region
of the slope. The initial vortex generally appears somewhere along the middle
section of the slope, and the instability spreads gradually towards the corners.
The initiation of vortex activity was observed as a function of input frequency.
These results indicate that (i) the vortices occur earliest (in time measured from
the start of the wave maker) for y = c, (ii) for small enough c < y , no vortices
appear, and (iii) there is a small range of y < c (subcritical) when vortices do
appear though precise values are hard to specify.
The diameters of the vortices were difficult to measure accurately from the
shadowgraph images because of the constant process of either growth or decay
during each half-cycle. From data taken from various photographs, the size of
 Experimental study of internal waves over a slope 237
-1.0 -0.5 o 0.5 1.0 hp x 103 (g/cc)
I I I I I
-0.5 0 0.5 A V N )
1 " 1 1 ' 1 1 1 ' 1 1

9-

-E
h

10-

11 -

12 -
FIGURE 14. A density profile taken through the streamers of figure 13, showing the micro-
structure that is formed along the slope. Numbers are adjacent to the nearby homogeneous
streamers that are formed by mixing in the vortex cells.

the vortex diameters is largest for c y and generally decreases for lower and
N

higher values of c.
Shortly after the formation of the line of vortices, thin streamers or wisps
appear near the cusp of each vortex and penetrate into the interior of the fluid
(note the thin lighter bands that alternate with the darker bands in figure 13).
Qualitatively, it appears that fluidof slightly different densities fromvertical levels
adjacent to each vortex cell is mixed in the cell and returns to the interior as a
thin streamer a t anew equilibrium level. The result ofthis processis alternatelayers
of thin nearly homogeneous streamers and thicker linearly stratified laminae.
Shadowgraph images of the thin horizontal streamers usually intensify with
time if the wave motion continues. The shadowgraph images of the streamers are
sharply outlined near the vortices, but within short horizontal distances from
the slope (of order 2-4 cm) the images weaken, although they can be detected out
to distances of 20 cm horizontally from the bottom.
Several conductivity probe traverses through short vertical sections (3 and
4 cm) were made in order to detect changes in the linear density-depth profiles
that might be associated with the streamers. Figure 14 shows the measured
density structure for a typical vertical traverse of a probe after the formation of
the streamers. Small-scale irregularities which can be described as 'micro-
structure' are typical of these profiles. Sharp changes in density a t the edges of
the nearly homogeneous streamers are apparent in the shadowgraphs and
conductivity measurements. The density gradients are not homogeneous within
the laminae, but are approximately equal to the overall mean gradient.
The net movements of neutrally buoyant particles in the streamers and in the
intervening laminae were observed for several runs. The typical net motion was
away from the slope in the streamers and towards the slope in the laminae with
magnitude of about 0.2 mm/s.
238 D. Cucchione and C. Wunsch

5. Summary
The linear theory of internal waves of the fundamental mode travelling over
a uniform slope is in quite good agreement with experimental results in the
subcritical case, sufficiently far from the corner region. Breakdown of the waves
occurs through strong dissipation along the slope, and especially a t high
frequencies, through a breaking mechanism. At lower frequencies, the waves
have a tendency to ‘surge’ along the bottom with large viscous losses.
I n the case of a supercritical slope, no theory was actually tested, but the wave
field is clearly inhomogeneouswith nonlinear processes important in the vicinity
of the critical characteristic. These waves tend not to break, but to dissipate
through laminar processes.
At the critical frequency, where the direction of energy flux in the wave is
parallel to the slope, the boundary layer at the bottom is unstable, generating
periodic vortices which mix fluid locally. This mixed fluid propagates into the
interior of the fluid and is apparent as a density ‘microstructure ’. Breaking that
is considerably less energetic than that for the subcritical case also occurs near
the corner when y w c ; its intensity diminishes with y > c.

Supported by the Office of Naval Research under contracts N00014-67-A-0204-

0047, -0048. The glass tank was provided by the Educational Development
Center of Newton, Massachusetts. Dr Seelye Martin made several helpful sug-
gestions concerning the experimental techniques.

REFERENCES
BAINES,P. G. 1971 The reflexion of internal/inertial waves from bumpy surfaces. Part 2.
Split reflexion and diffraction. J . Fluid Mech. 49, 113-131.
BEARDSLEY, R. C. 1970 An experimental study of inertial waves in B closed cone. Studies
i n AppZ. Math. 49, 187-196.
CACCHIONE, D. 1970 Experimental study of internal waves on a slope. Ph.D. thesis, p. 226,
M.I.T.IW.H.0 .I.
CACCHIONE,D. & SOUTHARD,J. B. 1974 Incipient sediment movement by shoaling
internal gravity waves. J . Geophys. Res. 79, 2237-2242.
EMERY, K. 0. & GUNNERSON,C. G. 1973 Internal swash and surf. Proc. Nut. Acad. Xci.
70, 2379-2380.
GREENSPAN,H. P. 1969 On the inviscid theory of rotating fluids. Studies in A p p l . Math.
48, 19-28.
HART,
J. E. 197 1 A possible mechanism for boundary layer mixing and layer formation in
a stratified fluid. J . Phys. Oceanog. 1, 255-262.
HOGG, N. G. 1971 Long shore current generation by obliquely incident internal waves.
Geophys. Pluid Dyn. 2, 361-376.
HURLEY, P. M. 1970 Internal waves in a wedge-shaped region. J . Fluid Mech. 43,97-120.
KELLER,J. B. & Mow, V. C. 1969 Internal wave propagation in an inhomogeneous fluid
of non-uniform depth. J . Fluid Mech. 38, 365-374.
LAMB,H. 1932 Hydrodynamics. Cambridge University Press.
MAGAARD,L. 1962 Zur berechung interner Wellen im meeresraumen mit nichtebenen
Boden dei einer spezielleii Dichteverteilung. KieZ Meeresforsch. 18, 161-183.
MOWBRAY, D. E. 1967 The use of schliereii and shadowgraph techniques in the study of
flow patterns in density stratified fluids. J . Fluid Mech. 27, 595-608.
 Experimental study of internal waves over a slope 239
OSTER,G. 1965 Density gradients. Sci. Am. 213, 70-76.
PHILLIPS,0. M. 1966 Dynamics of the Upper Ocean. Cambridge University Press.
PHILLIPS,0. M. 1970 On flows induced by diffusion in a stably stratified fluid. Deep-sea
Res. 17, 435-443.
ROBINSON, R. M. 1970 The effects of a corner on a propagating internal gravity wave.
J . Fluid Mech. 42, 257-267.
SANDSTROM, H. 1966 The importance of topography in generation and properties of
internal waves. Ph.D. thesis, University of California, San Diego.
THORPE,S. A. 1966 Internal gravity waves. Ph.D. thesis, University of Cambridge.
TURNER, J. S. 1973 Buoyancy Effects i n Fluids. Cambridge University Press.
WUNSCH, C. 1969 Progressive internal waves on slopes. J . Fluid Mech. 35, 131-144.
WUNSCH,C. 1970 An oceanic boundary mixing. Deep-sea Res. 17, 293-301.
WUNSCH, C. 1971 Note on some Reynolds stress effects of internal waves on slopes. Deep-
Sea Res. 16, 583-591.
WUNSCH, C. & HENDRY, R. 1972 Array measurembnts of the bottom boundary layer and
the internal wave field on the continental slope. Geophys. Fluid Dyn. 4, 101-145.
Journ.cr.1of Fluid Mechanics, JTol. 66, part 2 Plate I

FIGURE 6. Sequence of shadowgraph pictures showing breakdown of a low frequency wave

with y < c, a t various times. Note the formation of a bore-like phenomenon along the
slopc. I n this case a, = 0.4 cm; (r = 0.4s-l. Distance between heavy dark lines along slope
is 2 em; ixndisturbed free surface is a t the intersection of the dark and light regions a t thc
top of the photographs.
CSCCHIONE AND WUNSCH (Facing $3. 240)
eJournnl of Fluid Mechanics, 6'01. 66, past 2 Plate 2

FIUURE 7. Sequence of shadowgraph pictures showing breakdown of a high frequency wave

with y < c., at various times. A multiple set of crests forms along the slope. In this case
uo = 0.4em; cr = 0.75 s-l.

CACCHIONE ANI) WUNSCH

.Journal of Fluid iVwhnnics, I r d . 66, purt 2 Plate 3

KIOURI? 13. Shadowgraph picture of the instability in the bottom boundary layer i n the
critical case for two phases of the incident waves. The thin streamers (lightcr znncs in thc
shadowgraph) represent fluid t h a t htbs been mixed in the boiindary laycr and is periet,rating
into the interior of the fluid. The long vertical object on the left in the pictures is a coil-
ductivity prohc. Tho spacing between heavy black lines along the slope is 3 cm. a , = 0.3crn;
0- = 0.48 s-1.
CACCHIONE AND WUNSCH