Sei sulla pagina 1di 22

ANNUAL

REVIEWS

Ann.

Rev. Fluid Mech.1978. 10: 267-88

Copyright 1978 by

Further

Quick links to online content

Annual Reviews Inc. All rights reserved

TURBULENCE AND MIXING

=-=8124

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

IN STABLY STRATIFIED WATERS


Frederick S. Sherman
Department of Mechanical Engineering, University of California, Berkeley,
California 94720

J org Imberger
Department of Civil Engineering, University of California, Berkeley, California 94720

Gilles M. Corcos
Department of Mechanical Engineering, University of California, Berkeley,
California 94720

INTRODUCTION
Not only is mixing of stably stratified fluids an important element of many problems
of oceanography, limnology, meteorology, and environmental engineering, but it
also stands out from the infinite variety of turbulent flow phenomena as one that
may be understood in fair detail, even in our lifetimes. At least, this is what we
thought when we promised, two or three years ago, to write this survey!
Stable stratification seems to suppress the turbulence enough so that interesting
events are somewhat spread out in space and time. One has a chance to focus on
them, almost one at a time, and to get to know them by name. Thus we speak of
"Kelvin-Helmholtz billows," "salt fingers," "lateral intrusions," and several other
identifiable processes or events, instead of

j ust

"eddies." Because the bursts of

turbulence are unusually intermittent, and conditions have a chance to settle down
a bit between them, there seems to be more than usual scope for stability theory
and numerical models of complex but laminar flows to be used as aids to
understanding.
Most particularly, there seem to be unique opportunities to identify and trace
to its source the particular energy supply tapped by the small-scale motions that
do the mixing in various circumstances. This provides the organizational theme
for our review, and we discuss mixing motions in three categories:

1.

those deriving their kinetic energy from the gravitational potential energy of a
water column that is in some sense top-heavy;

2.

those that convert to smaller scales the kinetic energy associated with large-scale
shear flows or internal waves;

0066-4189/78/0115-0267$01.00

267

268

SHERMAN, 1M BERGER & CORCOS

3. those that are driven by some fairly obvious external source of small-scale
agitation. Here we think particularly of the wind mixing of surface waters.

Readers unfamiliar with this general subject can get an excellent background
from the book by Turner (1973) and from two articles in recent volumes in this
series by Turner (1974) and Maxworthy & Browand (1975).

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

DOUBLY-DIFFUSIVE CONVECTION

Density variations in saline waters are determined mostly by temperature and salinity
variations, according to the equation of state
dp/ p= - rxdT+{JdS.
In the ocean and in saline lakes and reservoirs, the two terms on the right often
nearly cancel because their sum is small compared to their individual values. Then,
even though p may increase with depth, as required for static stability, the individual
contribution of T or S to the density distribution may be top-heavy. Since the
thermal diffusivity, KT, greatly exceeds the diffusivity of salt in water, Ks, the potential
energy of the top-heavy component can be released, and convective mixing can
proceed. The same can happen in a double solution, for example of salt and sugar
in water, if one component diffuses more readily than the other.

Linear Theory of Instability of Steady, Purely Diffusive Transport


Turner (1974) reviews this topic adequately; the stability boundary for normal-mode
disturbances that vanish on stress-free, conducting horizontal boundaries at z = 0
and z = h are defined by the two lines

RT=

a+T

27n4

- R - ( T)
a+1

(1)

1 +- (1+r)
a

and
RT = Rs/r+27n4/4.

(2)

RT and Rs are thermal and saline Rayleigh numbers:


RT= rxl'J. T(gh3 /VKT),
Rs = {JI'J.S(gh3/v/(T)'

The other parameters are the ratio of diffusivities,


number a V/KT'

r =

Ks/KT, and the Prandtl

Formal Nonlinear Theories


The evolution of normal modes to finite amplitudes has been studied for both
instabilities. This work (Veronis 1968, Straus 1972, Huppert & Moore 1976) has
been beautifully done, and can be highly recommended to students of stability
theory and numerical analysis. Two-dimensional flow is assumed and variations
in space are resolved either by truncated Fourier series or by finite differences.

MIXING IN STABLY STRATIFIED WATERS

269

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

In either approach, limits on computer storage and run time restrict the range of
Rs and RT, within which accurate simulations can be made:
TEMPERATURE DESTABILIZATION
Veronis (1968) and Huppert & Moore (1976)
studied the doubly diffusive regime, looking for transitions from oscillatory modes
to steady modes of convection at a fixed horizontal wave number. They also
sought possible subcritical instabilities, through which finite disturbances can grow,
although infinitessimal ones would die. They found all these things, but their results
on subcritical instabilities give only a tantalizing hint of what might happen at
higher values of RT and Rs. Many qualitative features of their results have been
observed in laboratory tests, such as those of Caldwell (1974), although the
boundary conditions that were used cannot be strictly duplicated. Unfortunately
when we observe heat-driven convection between parallel horizontal interfaces in
the laboratory, typical Rayleigh numbers are enormously larger than those in
these studies, and the observed flow does not retain recognizable remnants of these
regular convection patterns.

Straus made a similar analysis of salt-fingering. His


analysis is specifically aimed at fluids for which r 1 and (J = 0(1), so that the
only nonlinear effect he needs to retain is the advective steepening of salinity
gradients. RT and Rs/r are the only remaining parameters of his problem, except
for the dimensionless horizontal wave number a.
Straus's computations show that the finite-amplitude convection patterns are
very bland in the region close to the stability limit, but that thin boundary layers
and slender jets of rising and falling fluid appear at higher values of Rs/r RT.
Unfortunately, the boundary layers could not be accurately resolved when Rslr RT
exceeded about 2 to 8, the lower value for higher 1 RT I, and vice versa
The amplitude of convection was measured by the value of a saline Nusselt
number, Ns h(oS/oz)o/I'!S. Ns depends on RT, Rs/r and a. The values of a that
maximized Ns at RT = 4 X 10-4 are shown in Figure 1.
Straus also assessed the stability of the two-dimensional motions against infinites
simal perturbations. The most threatening perturbation was found to be one that
would convert rolls into rectangular cells, but there was always (within his range
of RT and Rs/r RT) a range of a for which the rolls were stable. The value of a
that gave the greatest margin of stability is shown in Figure 1. In the laboratory,
RT and Rs/r RT are both likely to exceed, by a large margin, the values studied
here, and fingers are observed to have irregular planforms, more nearly square than
anything else (Turner 1974, Williams 1975). This remains unexplained, although
it is again obvious that the boundary conditions used by Straus have never been
matched in the laboratory.
Straus provided a fitting formula for the maximum values of Ns found at each
RT and Rs/r, suggesting that the flow that maximizes salt transport, if stable, is
most likely to be observed. The formula,

SALINITY DESTABILIZATION

Ns

1 +O.l42(Rs/r)113(1- rRT/Rs)4/3

gives an excellent fit if RT :c 104.

(3)

270

SHERMAN, IMBERGER & CORCOS

3.0

2.5

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

2.0
a
1.5

1.0

0.5

2
RS IrRT

Figure 1

Horizontal wave numbers for salt fingers: (a) fastest-growing disturbance in linear

range; (b) most stable two-dimensional fingers (Straus 1972); (c) two-dimensional fingers
that maximize N, (Straus 1972).

In Figure I we see that the values of a that maximize Ns or the stability of


the two-dimensional solutions follow the same trend as those that maximize the
growth rate of infinitessimal disturbances.

The Parallel-Flow Model o.f Salt Fingers


Several authors (see Turner 1974) have made good use of the following simple model
of salt fingers, in which the velocity is supposed to be exactly vertical and the flow
steady. Suppose that T (x,y, z) = <T )+ T(x,y), S(x, y, z) = <S) +S'(x,y), where < n
and <S), the horizontally averaged values of T and S, are linear functions of z.
Then the Boussinesq equations reduce to
dpjdz = pog(a< n -{J<S),

(4)

VV2W+g(rxT'

(5)

KTV2T -wd<T )jdz = 0,

(6)

KSV2S'

(7)

where
"12

oZjox2+0Zjoy 2.

To obtain fingerlike motions, one has to postulate the form, but not the amplitude,
of the functions w, T, and S'.

MIXING IN STABLY STRATIFIED WATERS


w

T'

S'

271

(8)

cos (kx) cos (my).

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

Without boundary conditions at the top and bottom, neithcr planform nor
amplitude of the motion is determined. The horizontal wavc numbers in (8) are
restricted to obey

which exhibits the same scaling found at large RT in the linear stability theory.
We note that this model postulates a perfect balance between viscous and
buoyancy forces, with W, T, and S all increasing together as the intensity of
finger convection increases. With high convection rates, it is easy to imagine that this
delicate balance can be upset and that various instabilities could arise. Turner &
Chen (1974) observed and sketched one, in which periodically spaced lumps pass
down the fingers, like an orange down the neck of an ostrich. This does not seem
to have been analyzed, but it may have something in common with roll waves on
a slope. Alternatively, we can imagine an instability affecting the fingers collectively.

Collective Instability of Salt Fingers


The growth of salt fingers at an interface may continue until they fill the whole
experimental container, but more frequently it is interrupted by a collective swaying
of groups of fingers, which grows until the fingers are sheared off. This effect spreads
horizontally, so that fingering regions are usually bounded by regions of larger-scale,
incoherent, convectively driven motions. This happens under conditions for which
we have no accurate theoretical model of the ordered fingering motions, and
evidently involves a type of instability that was not considered in Straus's analysis.
There is thus no way to set up a formally exact search for a new instability boundary,
but Stern (1969) has constructed a theoretical model that has been very useful.
For a base flow, he adopts a field of infinitely long salt fingers, characterized by
the Brunt-Viiisiilii frequency N2 g(rx d<T )/dzflux F = - <w(rxT'
and < ) denotes the result of averaging over a horizontal plane.
The disturbance is an internal wave, which has nearly horizontal wave fronts
(making a small angle (j to the horizon). This has infinitessimal amplitude and
causes a nearly horizontal shear flow, which oscillates with the low frequency
N sin (J.
The convective motions in the salt fingers have two effects on the internal wave.
When averaged along the wave front, over distances large compared to finger dia
meters, they contribute an inertial "Reynolds stress" term to the momentum budget
of the wave. This effect is neglected. More importantly, they modify the time and
space variation of the body force in that budget. This is because the shear of the
internal wave contorts the salt fingers so that the component of buoyancy flux
normal to the wave fronts varies with distance from the fronts. This periodic re
distribution of density, over and above that caused by an internal wave in a quiescent
fluid, can cause the wave to amplify.
=

272

SHERMAN, IMDERGER & CORCOS

Neglecting KT/V and KS/V in calculations of modified body force, and assuming
that the magnitude of F is not changed by wave-induced tipping of the fingers,
Stern finds an equation for the amplification rate of the wavc, Q, in terms of its
wave number /11. This is
Q( Q2 +vm2Q+ N2 sin2 e) +m2gF sin2 e cos2 e

O.

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

The critical condition, at which the destabilizing effect orthe salt fingers just balances
viscous damping of the wave motions, is F = vN 2 /g cos2 e. Instability results if
F > vN2/g, for any wave number m.

Salt Fingers That Maximize Buoyancy Flux


We recall that Straus (1972) attributed special significance to those two-dimensional
salt-finger patterns that maximized the saline Nusselt number and were most stable
against perturbations ofplanform. Stern (1976) returns to these themes, but in quite
a different way. He finds that it is possible to remove the indeterminacy of the
parallel-flow model (4)-(7) by supposing that thc fingers will transport buoyancy
as fast as they can without becoming collectively unstable.
A simplified version of Stern's calculation will give the general idea. In a situation
in which only a. T and fiS across a fingering layer are given, but the simple
profiles (g) are assumed, the parallel-flow model contains seven undetermined
parameters, W, T, S, d< T)/dz, d<S)/dz, k, and 111. Only three restrictions, Equations
(5)-(7), arise from the conservation equations. We need four more. One comes if
we assume a square planform. Another is provided by the stability limit Fg vN2.
Using these conditions with (5)-(7), we can obtain the expression
=

F3

where r

(fiS/2)4(gKT)(r-l)3(r-r),
=

a.T/fiS.

If we maximize this with respect to

r,

we get

when r = (1 + 3r)/4. A final maximization, with respect to fiS, is obtained if S is as


large as is physically possible for given S, namely when I S I S/2. This evaluation
slants the whole theory strongly towards the case T 1.
For convenience in writing the final prediction, we introduce the reference length
L (VKT/gf3S)'/3, the velocity V (gKT)1/3, and stability ratio R a. T/fJS .
Then we have:
=

Saline buoyancy flux: Fs


Ratio offluxes:

F T/Fs

Finger breadth:

hT
b

=
=

O.099V(fJS)4/3(1- r)'/3

(1 + 3r)/4

Maximum velocity in fingers:


Finger length:

T/(d< T)/dz)

nk

O.794V(fiS)1/3(I-r)'/3
=

2 J.3RfIT2/3(1

r) - 1/3( 1 + 3r)-1

2.64LO'1/6(1-T) )/3
-

The quantitative laboratory measurements, mostly due to Linden (1973), confirm


most oC Stern's predictions remarkably well. The indicated depcndencc of all

MIXING IN STABLY STRATIFIED WATERS

2 73

quantities on {38.S is certainly correct, and the constant multipliers agree with the
data about as well as one data set agrees with another. For example, Turner
(196 7) found FT/Fs = 0.56 for T 1 / 80 and Linden (1 973) found FT/Fs = 0.1 for
T = 1 /240, while Stern predicts F T /Fs
0.25 for both systems. Linden (1973) and
Turner (196 7) would replace the -0.099 in Fs by the weak function of R,
-O.l2R 1/3. On the other hand, agreement with salt-sugar data is not good (Stern
1 976).
We note in closing that Stern's 1976 predictions of Fs are, for salt water, about
24 times greater than those of Straus (1972). The feeding of salt and heat into
fingers between convecting layers must be a great deal more effective than molecular
diffusion through boundary layers on rigid top and bottom walls.

= =

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

Laboratory Experiments
Since the review of Turner (1974) reports of four major laboratory experiments on
double diffusion have been published.
Caldwell (1974) made some very delicate heat-flux measurements in salt water in
which a stable salinity gradient had been established by Soret diffusion. These
verified the stability boundary predicted by linear theory and some of the qualitative
features predicted by Veronis (1 968) and by Huppert & Moore (1976). However, the
test conditions fell in the range 0.043 < Rs/RT < DADS, so that the stability boundary
is crossed in the statically unstable region, and this is not an example of a system
destabilized by differential diffusion.
Crapper (1975) repeated and refined Turner's (1 965) measurements of heat flux
across a salt-stabilized interface. His data reduce the scatter of Turner's, but lead
to no new conclusions.
Mannorino & Caldwell (1 976) also measured fluxes across a salt-stabilized
interface, extending the range of stability ratio {38.S/rt8. T to about 12, and operating
over a wide range of rt8. T. The objective was to test the validity of Turner's empirical
laws at values of rt8. T closer to those found typically in the ocean (which are at
least two orders of magnitude smaller than those previously employed in the
lahoratory). Great care was taken to obtain conditions as nearly steady as possible.
At high stability ratios and low rt8. T these new data seem to depart significantly
from the trends established by Turner (1965) and Crapper (1975). For example, at
{J8.S /CJ8. T 7.5, the new values of normalized heat flux are about twice the old, and
the ratio of salt flux to heat flux appears to rise significantly, again about two fold
at the lowest values of CJ T, over the previously accepted value of 0.15.
These discrepancies merit close study, since it was thought from the success with
which Huppert & Turner (1972) applied laboratory correlations to calculate the
temperature distribution in Lake Vanda that those correlations could not possibly
be wrong by a factor of two. The rt8. T values of Lake Vanda are still somewhat
lower than those reached by Marmorino and Caldwell.
Linden (1976) reports a very fruitful study of layer formation in the salt-sugar
system, particularly when the initial state of the system contained nearly com
pensating uniform gradients of sugar (stabilizing) and salt (destabilizing) concentra
tions, so that it was poised, so to speak, just below the stability boundary for the

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

274

SHERMAN, IMBERGER & COR COS

double-diffusion regime. Then a very slight "cooling" at the top (actually adding of
salt at the top) triggered the rapid formation of a whole series of convecting layers
of nearly equal thickness. On the contrary, when layer formation is caused entirely
by a destabilizing flux from a boundary, layers form more slowly and are thicker
when they are closer to the source.
Linden also extended Turner's theory of the formation of layers (see Turner
1974) to account for pre-existing destabilizing gradients, and presented a number
of conclusions that should be very helpful to those who study such layered systems
in the field.

Migration of Interfaces and Merging of Layers


In several of the newer studies of interfaces, (Turner & Chen 1974, Linden 1976,
Marmorino & Caldwell 1976) observations of interface migration have been made,
and some explanations of the phenomenon suggested. Huppert (1971) presented
an elegant but evidently rather incomplete analysis of the stability of a system of
layers, in which migration was deemed to be less important than destabilization
of a stationary layer by gradual changes of iJ( T and f3S. This theory will
undoubtedly be reworked, but one of its main conclusions seems unlikely to be
shaken: systems of layers with high stability ratios (probably f3S/iJ( T > 2 is large
enough in salt water) tend to be long-lived and stationary.

Field Observations of Doubly Diffusive Phenomena


In recent years a number of geographical locations have become famous (among
enthusiasts for this sort of thing) for their layered waters. There no longer seems
to be any substantial doubt that the prominent and very regular layers, separated
by very thin interfaces, that are found in Lake Kivu (Newman 1976), in the Red Sea
brine pools (Voorhis & Dobson 1975), in ice-bound Lake Vanda (Huppert &
Turner 1972), under Arctic ice islands (Neshyba & Neal 1971), and in the Weddell
Sea (Foster & Carmack 1976) are formed by heat-destabilized double-diffusive con
vection. On the other hand, the layered fine structure so prominent under the
Mediterranean outflow, in the Tyrrhenian Sea, and in parts of the Caribbean Sea are
demonstrably the result of salt fingering. Williams (1974, 1975) has actually photo
graphed salt fingers in the interfaces of each of these regions, while Magnell (1976)
has observed their conductivity signal with a specially designed sensor, towed
beneath the Mediterranean outflow.
MIXING DRIVEN BY SHEAR OR WAVES

Large-scale shears are common in stratified fluids, since any horizontal pressure
gradient will interact with vertical density gradients to generate vorticity.
Gravitational stability inhibits the diffusion of this vorticity, and wherever we find
relatively sharp density contrasts, as might be set up by double-diffusive phenomena
or by the interleaving of distinct water masses, we also find concentrated vorticity.
This sets the stage for possible dynamical instabilities of the Kelvin-Helmholtz
type, which have been well reviewed in Turner (1973). Our understanding of these

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

MIXING IN STABLY STRATIFIED WATERS

275

instabilities and of the mixing thcy can produce has been rounded out nicely in the
last few years, as we shall see below. In his book, however, Turner almost buries
the idea that Kelvin-Helmholtz instabilities can by themselves account for much
vertical mixing. His simple energy argument, showing that they can only disperse
an initial discontinuity to a modest degree, cannot really be refuted. Indeed, more
detailed studies only reveal the degree to which this simple argument (Turner
1973, pp. 322, 323) overestimates the thickness of the mixed layer after Kelvin
Helmholtz waves have come and gone. Nevertheless, these studies involve some
good fluid mechanics and are worth a brief review.
We must also say something about another timeless idea, the concept of breaking
internal waves. We are particularly interested in waves that, while propagating
in a smoothly stratified region, break not because they are driven to huge amplitudes
by some obvious external agency, but because of subtle interactions between waves
that may be prevalent under "average" oceanic conditions. We limit our review
to one particularly significant recent paper by McEwan & Robinson (1975).
Kelvin-Helmholtz Instabilities
These have been dealt with in other recent reviews, for example Maxworthy &
Browand (1975) and Sherman (1976), so this discussion is brief. Turner's energy
argument is so simple that we start by repeating it. Suppose that an unspecified
mixing event transforms initial discontinuities U and I'lp into constant gradient
regions of thickness bu and bp, both end states being stationary. Mass and
horizontal momentum have been conserved. The process has decreased the kinetic
energy of the system by an amount PaU2bu/12 and increased the potential energy
by gl'lpbj24. If no mechanical energy is lost by viscous dissipation, we must get
bN)u 2PoU2/gl'lp. If fJp and fJu are roughly equal, we get a firm upper limit for
bp after the mixing event. If bu is for some reason substantially greater than bp
so that kinetic energy is drained from a relatively deep layer, we can get a larger
bp, but this circumstance seems unlikely.
Both controlled laboratory experiments (Thorpe 1973, Koop 1976) and numerical
experiments (Patnaik, Sherman & Corcos 1976) have established that in the Kelvin
Helmholtz instability at least 70%, and sometimes more than 90% of the kinetic
energy given up by the shear flow is simply lost to viscosity. As a good rule of
thumb, we can estimate bp = 0.3PaU2/gl'lp.
The detailed story of this process is quite fascinating. If the initial values of
fJp and bu, which will never be quite zero, are much smaller than the estimate just
given, the interface will first be wrinkled by those wavelike disturbances that grow
most quickly according to linear theory. These waves roll up in a fashion made
unforgettable by Thorpc's beautiful pictures of flow in a tilting tube. The reSUlting
array of more or less concentrated vortices is in turn unstable to a subharmonic
wave that causes alternate vortices to orbit and amalgamate, and this process
continues until the height of the "billows" becomes of order PaU2jgl'lp. The process
is essentially two-dimensional and inviscid, and entrains a substantial amount of
fluid that was outside the initial layer, into the vortex cores. When buoyancy prevents
further pairing of vortices, the vortex cores are typically quite top-heavy, there
=

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

276

SHERMAN, IMBERGER & CORCOS

having been little molecular mixing as yet. From this point on, the system of
vortices collapses and the layer begins to become horizontally homogeneous again.
In the laboratory, Thorpe (1973) and, most particularly, Koop (1976) show that this
collapse is accompanied by a great deal of three-dimensional, smaller-scale motion.
Both suspect convective instabilities in the top-heavy cores. These small-scale
motions are needed to promote permanent mixing, but are apparently only partially
effective, as the free convection allows much of the fluid that has been hoisted up
by the Kelvin-Helmholtz mechanism to fall back down without mixing. These
observations are for salt-stratified water, and the situation in the atmosphere may
develop differently at this point because ofthe more comparable diffusivities for heat
and momentum.
Even the numerical simulations of Patnaik et al (1976), which show no onset of
small-scale motions, reveal a sort of rebounding from the state in which the vortices
attain their greatest height, and a noticeable surrender of potential energy at this
stage.
Koop's experiments reveal a host of other interesting details, especially about a
switching of preferred modes in the finite-amplitude instability of flows in which
<5u initially exceeds <5 p by a considerable margin. This is a problem that has'so far
defied theoreticians, and is a bit reminiscent of the mode-switching phenomena
successfully analyzed by Veronis (1968) and by Huppert & Moore (1976) in heat
destabilized, doubly diffusive convection.
In our present context, three main conclusions may be drawn about Kelvin
Helmholtz instabilities:
1. They require a pre-existing sharp interface to get started, and hence cannot be
invoked as a first cause of layered structures in stably stratified regions.
2. They convert mean flow kinetic energy to potential energy rather inefficiently.
3. They are self-limiting, so that an interface, once thickened by a Kelvin-Helmholtz
episode, will not support subsequent instabilities of the same kind unless it
becomes more strongly sheared, or unless it is followed by some process that
sharpens the interface again.
We do not mean to suggest by this that we should reject all the enthusiastic
estimates that have been offered concerning the importance of these waves in
oceanic mixing, but to propose that we now need to study the possible symbiotic
relationships between Kelvin-Helmholtz waves and other mixing mechanisms.

The Inherent Instability of Long Internal Waves in a Continuously


Stratified Fluid
The description of the origin of molecular mixing involves invariably the study of
one or several sequences of dynamic processes or events, each of which succeeds
in either creating or altering scales of motion and in providing an opportunity for
a new type of instability. The final molecular mixing is likely to be intimately
associated with strong deformations, the proximate origin of which might be small
scale shear or gravitational instability. It is thus perhaps too restrictive, in view of
the number of possible separate events that follow each other, to offer, as Orlanski

MIXING IN STARLY STRATIFIED WATERS

277

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

(1972) did, a criterion for the final degeneration of internal waves. This criterion

(that for static instability, i.e. fl..p' g < 0) may be accurate (though it is difficult to
prove it is either necessary or sufficient), but it should probably not be applied
directly to test a type of motion that may provide only the first event of a sequence.
Early work on the departure of internal waves from iinear propagation theory
(Ball 1964, Thorpe 1966, Davis & Acrivos 1967, Hasselman 1967, Phillips 1968,
Martin, Simmons & Wunsch 1972, McEwan 1971) had focused on the possibility
of interaction of internal gravity waves within continuously stratified fluids or at an
interface and interaction of surface waves with internal gravity waves. Two types of
interactions were singled out. These were weak resonant interactions of second
order and strong interactions. These are discussed by Phillips (1966, Chapter 5) and
adequately summarized in Turner's book (1973).
Resonant interactions require two waves, which separately satisfy the dispersion
relation dictated by the stratification, to be such that the sum or the difference of
their wave numbers be related to the sum or the difference of their frequencies by
the same dispersion relation. Then a third wave is able to extract energy from one
or both of the two primary waves. Strong interactions are far less selective. An
isolated internal wave of a given wave number and frequency is an exact solution
of the full equations of motion because for it, the nonlinear convective terms vanish
identically. Hence it has been thought that there is, in principle, no restriction on
its amplitude as there is for surface waves. Let us leave aside for the moment the
question of its inherent stability. Consider an arbitrary pair of such waves. They
will only exceptionally be part of a resonant triad. In general their sum and difference
frequencies will not bear any special relation to their sum and difference wave
numbers. Nevertheless, the nonlinear term in the equation of motion will generate
a third wave with amplitude quadratically related to that of the forcing waves.
Since these may be large, the amplitude of the forced mode may be appreciable
and the transfer of energy is rapid.
More recently, McEwan & Robinson (1975) have studied an intriguing mechanism
for the growth of perturbations to a single wave of only modest amplitude. The
mechanism is not very selective in the sense that a given forcing wave may give
rise to waves within a rather large range of wave numbers, although the magnitude
of the disturbance wave number is bounded above for a given primary wave
amplitude and the directions of the disturbance wave vector are also determined
by the forcing wave. Furthermore, both their laboratory observations and the
formulation of their analysis make it clear that the mechanism operates to foster
the growth of waves considerably shorter in length than those that spawn them,
so that the forcing wave provides a time-dependent environment that only gradually
varies in space.
.
Some simple considerations give a qualitative idea of the nature of this instability.
Imagine a progressive internal wave of long wavelength and of arbitrary wave
number vector orientation ko = iko + jl110 in a continuously stratified fluid. The
particle velocity components are given in terms of the angle rjJ made by isopycnic
lines with the horizontal

2 78

SHERMAN, IMBERGER & COR COS

if ,p is small enough, while

,p(x, z, t)

,pM exp i(kox+moz+wot).

The vorticity is
(=

i,pMWO(l +15) exp i(kox+moz+ wot),

where 10 mo/ko, which for a small neighborhood of a point such as the origin is
approximately a time-periodic solid-body rotation proportional to ,pM and to
(1 + tan2 00), where 00 is the angle made by the wave number vector with the horizon.
A second short wave would propagate with respect to this rotating medium. We
use orthogonal coordinates such that x is along an isopycnic of the first wave. Fluid
particles will, on account of the second wave, move along a path s normal to the
vector k (the wave number of the second wave in relative coordinates). See
Figure 2.
Denote by s the distance of the particle along s from its equilibrium position.
Since the basic stratification Vpm i s tilted by an angle ,p, the buoyancy force on
the particle is p'g sN2 cos e and the component of this vertical force along sis
==

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

sN2 cos B cos (B+,p) = -sN2 cos1 B(lsin-cos),

where I m/k
therefore
=

tan B. The approximate equation for the particle displacement is

s-sN2 cos2 B[I sin ,pIt) - cos ,p(t)] =

0,

but ,p itself is a periodic function of time, i.e. ,pM cos wot. For small ,p, sin,p ,p
and cos,p 1 and the equation becomes
s- N2

cos2 B(l,pM cos wot-l)s

o.

This is the Mathieu equation. It gives rise under some restrictions to parametrically
excited (unstable) solutions. The effective component of gravity must be modulated
in such a way as to heIp rather than hinder the oscillation. This occurs when the
forcing frequency is nearly double the frequency of the growing disturbance.1 This
maximizes the work done by the forcing wave on the disturbance.2 As a consequence,
the wave-number vector of the disturbance is found at a shallower angle to the
horizon than that of the forcing wave. If the wave were not dissipative there would
be no lower bound to the amplitude ,pM of the primary wave required for the
instability to occur within the right frequency range. McEwan and Robinson take
account of thc effect of viscous forces on the forced wave and thus obtain a
minimum amplitude for the primary wave, which is an increasing function of the
perturbation wave number.
In an experiment that must surely be held up as an example for its elegance and
the care with which it is interpreted, McEwan and Robinson verify the conclusions
1 Since one of the modes forced by quadratic interaction will then have nearly equal and
opposite wavenumber and hence equal frequency, this is equivalent to a high wavenum her
resonant triad interaction. (q.v. Phillips, 1966).
2 Compare with the strategy llsed by a chiJd to propel himself on a swing.

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

MIXING IN STABLY STRATIFIED WATERS

Figure 2

279

Interacting internal waves. The forcing wave causes a periodic rotation (t) of the

disturbance wave vector k. By continuity, the path of a fluid particle is along perpendicular
lines such as

s.

of their theory by using a cylindrical container, i.e. for standing waves. Since, within
can only be discrete wave
numbers for a given frequency. Of these the most unstable member can be identified
and its minimum forcing amplitude can be calculated. In the experiment most of the
dissipation occurs near the walls, whereas in oceans, viscous losses during propaga
tion would be the relevant source of dissipation. The quantities measured are the
minimum amplitude cPM of the primary wave for the disturbance to be barely visible,
the characteristic angle of the disturbance (i.e. the direction of its group velocity),
and the phase difference between the nodes of cP and the nodes of perturbation
density. For all three the agreement is truly impressive, which indicates not only
that parametric instability occurs but also that McEwan and Robinson's mathe
matical analysis, which relies in several instances on the smallness of the primary
wave amplitude, is nevertheless quite accurate. For the experiment, the value of
cPM required to sustain a visible disturbance was only 7, and in the ocean, with
much smaller values of disturbance wave numbers, the critical primary wave
amplitude required would be smaller.
The outcome of a well-developed parametric instability in the laboratory seems
to be the irreversible intensification of small-scale density gradients called
"traumata," i.e. strong local layering, and such accidents are reported to have
occurred spontaneously in previous experiments on nonresonantly interacting
internal waves.
a container, free wave mode are eigenfunctions, there

APPLICATION TO THE OCEANS

McEwan and Robinson argue that the instability,


whose occurrence has been verified in experiments, i.e. with one or two standing
waves or with fluid in solid-body periodic rotation acting on necessarily standing
disturbances, should commonly occur in the ocean. There, two main differences
need to be considered. First, the waves that can serve as forcing functions are
spectrally distributed' rather than truly isolated. Thus the combined effect of such
waves has limited coherence in any coordinate system. Second, even for a narrow
forcing wave band, since the group velocity of the disturbances is in general quite
different from that of a forcing wave, a packet of long waves will in general outrun a
disturbance that it has started to amplify. McEwan and Robinson in effect believe

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

280

SHERMAN, IMBERGER & COR COS

that both the coherence that would result if the spectrum is semidiscrete and the
interaction time that is available if the forcing wave packets are not too short are
sufficient to permit the parametric instahility to develop_ They use the representation
by Garrett & Munk ( 1 972) of the internal wave spectrum of the oceans to compute
a mean square isopycnic slope and deduce that it is sufficient to excite unstable
internal waves, the vertical wave length of which could be as small as ten or twenty
centimeters, particularly where the value of the Brunt-Viiisiilii frequency is highest.
It may be possible to treat this matter more thoroughly by replacing the deter
ministic formalism of the above paper by an admittedly more involved analysis
that would take into account both the random nature of the forcing function and the
different propagating properties of forcing function and unstable waves, perhaps in
the spirit of Phillips' paper ( 1 957) on the generation of surface waves by random
wind-pressure perturbations_
It also seems possible to evaluate an energy transfer function based on the
parametric instability mechanism, so long as the mechanism operates as a cascade,
i.e. so long as traumata and other finite-amplitude manifestations of the instability
are confined to high wave numbers. Such a transfer function might allow the
computation of the spectrum of internal waves in those spectral regions that are not
directly supplied by external forces. This might help assess the ultimate importance
of the proposed instability.
In conclusion, it would seem that strong circumstantial evidence has been found
for a new mechanism that is capable, under ubiquitous circumstances, of creating
density inhomogeneities and small-scale shear and transferring energy from a broad
range of large-scale internal waves, directly or indirectly, to much smaller scales,
thus setting the stage for ultimate mixing.
MIXING DUE TO EXTERNAL ENERGY INPUTS

Consider a stationary stably stratified fluid that is being stirred at either its upper
or lower boundary. The stirring mechanism may be due to a wind or mechanical
stress, a negative buoyancy flux, or mechanical stirring with an oscillating grid.
Near the source of energy the stabilizing gradient is broken down and the fluid
becomes mixed. This "mixed" region propagates into the remainder of the fluid with
continued stirring as some of the turbulent kinetic energy, introduced by the stirring
mechanism, is exchanged for mean potential energy.
This problem originated with the study of the meteorological boundary layer,
the ocean mixed layer, entrainment at the edge of density currents, jets and plumes,
and natural convection from heated surfaces.
The propagation of such a mixed region into a quiescent stratified or unstratified
body of fluid has been studied in essentially three ways; with experiments, with
integral energy models, and with the application of turbulence-closure schemes
originally developed for boundary-layer predictions.
The latter avenue was initiated by Munk & Anderson (1948), who postulated
variable eddy transport coefficients dependent on the local gradient Richardson
number. This approach has been continued by Mellor & Durbin ( 1 975). An intensive

MIXING IN STABLY STRATIFIED WATERS

281

and promising effort to devise second-order closure models for turbulent flows with
buoyancy effects is being made at the present time and is exemplified by the paper
of Zeman & Lumley (1977). This line of work, however, is so complex as to deserve
a review of its own. We shall therefore concentrate on experimental results and their
interpretation in terms of simple integral models.
A

General Integral Model

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

The few available field and laboratory observations (Niiler 1975, Gregg 1976) suggest
that the mixed layer consists of three distinct zones:
1. A comparatively thin surface layer near the stirring agent. The dominant activity
here is the production of turbulent kinetic energy which is then exported to the
fluid below.
2. A uniform central layer in which the energy exported from the surface layer is
used to homogenize the fluid.
3. A thin front separating the turbulent interior from the quiescent fluid below. Here
the remainder of the eddy energy from the surface, plus any that may be locally
generated by shear, less that which is locally dissipated or radiated downward
by internal waves, is expended to entrain quiescent fluid into the central layer
above.
Following Niiler (1975), we consider the turbulent kinetic energy budget for a
column of fluid extending from the agitated surface to a constant level in the
quiescent fluid. The work done against gravity by the turbulent velocity fluctuations
simply increases the mean potential energy of this column, according to the equation
(9)
Here V is the potential energy, AK is the kinematic flux of turbulent kinetic energy
imposed at the top surface, As is the rate of production of turbulent kinetic energy
by the shear across the bottom front, AT is the temporal rate of increase of
turbulent kinetic energy in the column, AI, is the leakage of energy by radiation of
internal waves into the quiescent fluid, and AD is total rate of energy dissipation
by viscosity in the column.
The layer characteristics that one desires to predict are its mean temperature and
velocity, T and v, and its depth h. These depend on the externally given quantities
u .. (friction velocity of the applied wind stress), H (rate of warming at the top
surface), N (the buoyancy frequency of the quiescent fluid just below the front),
g, (I., v, and 1\"[". In what follows, T is represented by T, the temperature jump
across the front, which is positive when the mixed layer is warmer than the water
below. If we can express the A's in terms of these parameters, and add to (9) a
heat-balance equation and a mean momentum equation, the model will be
determinate.
First, we set 2As = C sv v dh/dt, where Cs is a shape factor to be estimated by
comparison with model experiments.

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

282

SHERMAN, IMBERGER & CORCOS

The quantity AK -rxHgh/2pocp -AD = Al/2 represents the rate of energy introduc
tion at the top surface minus that dissipated within the layer. The term involving H
becomes available to help mix the layer when H is negative (surface cooling). Part
of this energy is available at the front to promote entrainment. As suggested by
Zeman & Tennekes (1977), the dissipation length scale is nearly proportional to
h, so AD should be proportional to the power at the surface, u +u}, where
u} = -aHgh/cppo. Hence, we write Al = cKl, where the turbulent velocity scale
q = ( u} +1]3U;) 1/3. Zeman & Tennekes added the energies to define a velocity scale;
we follow Kraus & Turner (1967) in adding the powers.
With this velocity scale, we can follow Zilitinkevich (1975) by setting
2AT = C Tq2 dh/dt
and adopting parameterization of the leakage,
2AL = CLA2N2h,

where A is a typical amplitude of the internal waves radiating from the base of the
front. Zeman & Tennekes estimated A = q/ N when I1T is small, and A = q2/grxl1T
for severe fronts. An expression that interpolates between these extremes is
A =

(q/N)/(l+grxI1T /qN),

which leads to
2AL = -CLRs3l/(Rs+Ri)2.
Substituting for the A's in Equation (9) and noting that
2dV/dt

rxI1Tghdh/dt+rxHgh/ cpPo,

we obtain the entrainment law


1 dh

CK-CLRs3/(Rs+Ri)2

q dt

Ri(1+CT/Ri-CsF)

(10)

This generalization of existing models contains five coefficients; CK, C T, CS, C Land
'1, which may depend on Ri, Rs, (Y, al1 T, F, Re, and u*/u f, where Ri = al1 Tgh/q2,
Rs Nh/q, Re qh/v, (Y = v/I(, and F = vv/aI1Tgh.
=

Laboratory Experiments
Let us now turn to the determination of the coefficients from experimental results.
Unfortunately, existing data consists almost entirely of measurements of mean
density (temperature or salinity) profiles. Little or no information is available on the
properties of the turbulence in the layer or even the distribution of the mean velocity
within the layer that would allow direct evaluation of the parameter q. The experi
ments previously carried out fall into one of the following limiting categories:
(a) propagation of turbulent fronts into stagnant homogeneous fluid; (b) stirring
with an imposed shear stress; (c) wind stress over water; (d) natural convective
penetration; (e) gravity currents; (f) grid-generated turbulence.
Table 1 summarizes the results from experiments of category (a) to (e), but not

MIXING IN STABLY STRATIFIED WATERS


Table 1

2 83

Experiments that supply coefficients for the model


Relation between

Author
Tennekes & Lu mley

(1972)

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

Willis & Deardorff

Ri

Rs

0.023(CT+1O)

14.8

10 3

0.013(CT+8.5)

3.3

102

103

0.3
3.3

30.8

102

103

0
0

4.0

300
1.3

7.7
Ellison & Turner

102
103

(1959)

1
101

CK 2.4
CK-6.0

Cs 0.2

92

(1977)

Lofquist (1960)

CT = 3.3'7CK

103

(1969)

Azad

0
x

8.5

Kantha, Phillips &

coefficients from best fit

Kato & P hillips

9.7

10

(1974)

Re

Cs

very

0.3

0.1

x 10
10- 2
6 x 10-1
103

small !

'"

CfJ

'13CK

Cs

2.2

0.3 using Lofquist's


value of CD

2.5

(fl. These were omitted because of the great difficulty in estimating the relevant q
and because of the excellent discussion of the results in Turner (1973). The right
hand column of Table I summarizes the constraints put on the coefficients of
Equation (10) in order to satisfy the experimental results. However, there was
considerable uncertainty in estimating the mean velocities in the experiments of
Kato & Phillips (1969) and Kantha, Phillips & Azad (1977), but a value of Cs = OJ
estimated from the experiment (8 = 14) of Ellison & Turner (1959) appears to
apply. The entrainment predicted by Kato and Phillips was considerably smaller,
for the same value of Ri, than that recently found by Kantha, Phillips and Azad
in the same experimental apparatus. The radiation losses contribute to the
differences, but the value of C L determined from Willis & Deardorff (1974) is much
too small to makc the leakage term significant in the experiments of Kato &
Phillips (1969). Thus these two experiments imply different values of Cs, but it
must he remembered the estimates of mean velocity are only very approximate and
further experiments are necessary.
The grid-stirring experiments of Hopfinger & Toly (1976) suggest that (l/q)dh/dt
depends on the Peelet number once Ri > 50. The difference in the functio nal form
from Ri- 1 to Ri- 3/2 is probably not as fundamental as originally thought since the
model equation has a variable slope and describes the functional relationship of

2 84

SHERMAN, 1MBERGER & CORCOS

Table 2

Coefficients chosen by various authors

Author

IJ

Zeman &
Tennekes
(1977)

2.2

CK

Cy

Cs

0.5

3.6

O.SR,

CL

0.024

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

q2
--

grxAT
Niiler (1975)

1.3

1.0

Mahrt & Lenschow


( 1 976)

0.2

Kraus & Turner


(1967)

1 .4

0
I cm
0
fixed
s-stroke
CK'
constant for particular experiment

0.01

CK,(s/h)9/2

Linden (1975 )

Alexander & Kim


( 1 976)

exponential
1 . 1 ( 1 + h/5)

Imberger et al
( 1 977)
Best fit to experi
ments reviewed
in Table 2
-

1 .0

(I + h/20)

( 1 + h/W

1.8

0.38

2.25

0.3

q
N + grx/';. T/q

0.04
--

Kantha, Phillip s , and Azad quite well. However, it would not

-------

be surprising if

C K and C L were functions of molecular diffusivities at large values of Ri

as originally
suggested by Turner (1973), or even weak functions of Ri as postulated by Long
(1975), especially for large values of h.
Assuming the coefficients to be constant, we list in Table 2 the values for best fit
to the data from the e xpe riments detailed in Table 1. For comparison, the values
suggested by previous authors are also given.
Figure 3 displays the right-hand side of Equation (10) as a function of Ri for a
range of values of F and Rs. The chosen values of the coefficients thus allow an
excellent correlation with existing laboratory data, but further experiments, embrac
ing a wider range of the parameter space, are required to estimate the limitations
of the model.
It should be noted that when surface cooling dominates, this class of models

MIXING IN STABLY STRATIFIED WATERS

1 0- 1

3. 3 , Rs

285

F------__
F = 0 , Rs 5

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

1 dh

q dt

10 3

F = O , Rs =

10

10
Ri

FifJure 3

Model entrainment law (see Equation 10).

predicts rapid deepening, slowing at the depth where losses by internal wave radia
tion begin to equal the turbulent energy arriving at the interface. At this level, the
mixed layer cools without appreciable deepening until A T is zero (or even negative
if CT oF 0), whereupon the losses can exceed the incoming kinetic energy since with
A T negative there is then a supply of mean potential energy available. Such a
rejuvenation of the turbulence by an unstable gradient within the mixed layer does
not seem to be unrealistic for very thick mixed layers; however, as yet there is no
experimental data available to support this conclusion.

Field Observations in the Mixed Layer


The difficulty of measuring velocity (both mean and turbulent intensities) in the
ocean and lakes explains why therc is such a scarcity of observations of the mixed
layer behaviour. Thompson (1 976) has recently adapted three of the above limiting
models, (Krause & Turner 1 967, Warren 1 972, and Pollard, Rhines & Thompson
1 973) and compared the predictions for one year of record for Ocean Weather
Station "N." He found that all the models gave much the same answer indicating
that the terms in Equation (9) would be competing. Thompson, following Denman
( 1 973), assumed C s = 1.0 and C KIJ 3 = 2.4.
The experiment of Halpern (1974) indicates that CKIJ 3 should be closer to 6.5 and
the original comparisons made by Turner (1 969) suggest a value of CKIJ 3 of 16.5.
In all these experiments there was great uncertainty about the meteorological forcing
and the estimation of the term V V/q2 . The energy leakage term AL and the temporal

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

2 86

SHERMAN, 1MBERGER & CORCOS

terms AT were very small at all times and could certainly not be used to explain
the differences. Neither Ri nor Rs differed very much between experiments. A much
more likely explanation for the great discrepancy is failure of the one-dimensional
assumption. Gregg (1976) showed how horizontal intrusions may often alter the
vertical structure to a much larger degree than is possible by surface effects.
Particular care must thus be exercised when correlating localized single storms with
predictions from Equation ( 1 0).
Application of the above one-dimensional model to lakes must be treated with
even more caution since, in addition to such interleaving of intrusions, boundary
effects and wind sheltering may cause severe distortion of the thermocline ultimately
leading to internal bore formation (Thorpe et al 1 9 72). Except for very large lakes
(> 10 km in diameter), the shear production term may be neglected ; however,
Imberger et al ( 1 977) found that the effective penetration (CKI73) of the surface
energy for thermoclines 2 0 m deep is typically 0.15, much smaller than values found
in the ocean. Differential deepening due to wind sheltering must be responsible for
this large discrepancy.
It can thus be stated that the model proposed by Niiler (1 975) and Zeman &
Tennekes ( 1 977) which, with minor modification, is expressed by Equation ( 10)
appears to be quite satisfactory to correlate most laboratory data presently available.
However, to apply it to mixed layers in the ocean or in lakes (where the leakage

and temporal terms are essentially zero), we must still face major uncertainties
concerning the values of C s, 17, and CK and the validity of the one-dimensional
assumption underlying the model. Until more detailed observations of both velocity
and density structure, both in the vertical and the horizontal, are obtained in the
field, and laboratory experiments are carried out to verify the turbulence scaling
assumptions, there appears to be little value to be gained in further refining the
models. In lakes the situation is even more complicated and involves the tilting of
the thermocline with subsequent edge effects.
SUMMARY

We have focused attention on a few well-isolated mechanisms and displayed one


scheme for making rough predictions when mechanisms cannot as yet be dis
entangled. One has only to try to reproduce some of these experiments in the
laboratory, or to consult the rapidly growing literature of detailed field studies,
(e.g. Gregg 1 976) to appreciate how much we have simplified or idealized real
world circumstances.
It seems to us that much emphasis now should be placed on processes that
combine vertical mixing and horizontal advection. This is not a new suggestion,
and some important starts were made some years ago by Turner & Chen ( 1974)
and Caccione & Wunsch ( 1 974).

MIXING IN STABLY STRATIFIED WATERS

287

Literature Cited
Alexander, R . c., Kim,1.-W. 1976. Diagnostic
model study of m ixed-layer depths in the
summer North Pacific. J. Phys. Oeeanogr.

6 : 293-98

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

Ball, F. K. 1 964. Energy transfer between


external and internal gravity waves. J.

Fluid M echo 1 9 : 465-78


Caccione, D. , Wunsch, C . 1974. Experimental
study of internal waves over a slope. J.
Fluid M eeh. 66 : 223-40
Caldwell, D. R. 1974. Experimental studies
on the onset of thermohaline convection.

J. Fluid Mech. 64 : 347-67


Crapper, P. F. 1975. Measurements across
a diffusive interface. D eep Sea Res. 22 :
537-45
Davis, R. E., Acrivos, A. 1967. The stability
of oscillatory internal waves. J . Fluid M echo
30 : 723-36
Denman, K. L. 1 973. A time dependent
model of the upper ocean. J. Phys.
Oceanogr. 3 : 173-84
Ellison, T. H., Turner, J. S. 1959. Turbulent
entrainment in stratified flows. J. Fluid
M echo 6 : 423-48
Foster, T. D., Carmack, E. C. 1976. Tempera

ture and salinity structure in the Weddell


Sea. J. Phys. Oceanogr. 6 : 36--44
Garrett, c., Munk, W. 1 972. Space-time
scales of internal waves. Geophys. Fluid
Dyn. 3 : 225-64
Gregg, M. C. 1976, F i ne str u ctu re and micro
structure observations during the passage
of a mild storm. J. Phys. Oceanoyr. 6 :

528-55

Halpern, D. 1 974. Observations of the


deepening of the wind-mixed layer in the
northeastern Pacific Ocean. J. Phys.

Oceanoyr. 4 : 454-66

Hasselman, K. 1967. A criterion for non


linear wave stability. J. Fluid Mech. 30 :

737 -39

Hopfinger, E. J., Toly, J. A. 1 976. Spatially


decaying turbulence and its relation to
mixing across density interfaces. J. Fluid

Mech. 78 : 1 5 5 -77

Huppert, H. E. 1 97 1 . On the stability of a


series of double-diffusive layers. Deep Sea

Res. 1 8 : 1005-21

Huppert, H. E., Moore, D. R. 1976. Non


linear doubk-diffusive convection. J . Fluid

Mech. 78 : 821-54

Huppert, H. E., Turner, J. S. 1 972. Double


diffusive convection and its implications
for the temperature and salinity structure
of the ocean and Lake Vanda. J. Phys.

Oceanoyr. 2 : 456-61

I m berger, J., Patterson, J. P., Hebbert, R. H .


B., Loh, I. C. 1 977. Simulation of the

dynamics of a reservoir of medium size.

J. Hydraul. Div. Proc. ASCE. In press

Kantha, L. H., Phillips, O. M., Azad, R. S.


1977. On turbulent entrainment at a stable
density interface. . J. Fluid Mech. 79 : 753-

68

Kato, H., Phillips, O. M . 1969. On the


penetration of a turbulent layer into a
stratified fluid. J. Fluid Mech. 37 : 643-55
Koop, C. G. 1976. Instability and turbulence
in a stratified shear layer. U niv. Southern

California Dept. Aerosp. Eng. Rep. USCAE


134. J une 1 976
Kraus, E. B., Turner, J. S. 1967. A one

dimensional model of the seasonal thermo


cline : II. The general theory and its con
sequences. T ellus 1 9 : 98-106
Linden, P. F. 1 973. On the structure of salt
fingers. Deep Sea Res. 20 : 325 -40
Linden, P. F. 1975. The deepening of a mixed
layer in a stratified fluid. J. Fluid Mech.

71 : 385-405

Linden, P. F. 1976. The formation and


destruction of fine-structure by double
diffusive processes. Deep Sea Res. 23 : 895 -

908

Lofquist, L. 1960. Flow and stress near an


interface between stratified fl uids . Phys.

Fluids 3 : 158-74

Long, R. R. 1975. The influence of shear on


mixing across densi ty interfaces. J. Fluid

M echo 70 : 305 -20

Magne\l, B. 1 976. Salt fingers observed in the


Mediterranean outflow region (34N,
I I OW) using a towed sensor. J. Phys.

Oceanogr. 6 : 5 1 1-23

Mahrt, L., Lenschow, D. H. 1976. Growth


dynamics of the convectively mixed layer.

J. Armas. Sci. 33 : 41-5 1

Marmorino, G. 0., Caldwell, D. R. 1 976.


Heat and salt transport through a diffusive
thermohaline interface. Deep Sea Res. 23 :

59-67

Martin, S., Simmons, W., Wunsch, C. 1 972.


The excitation of resonant triads by single
internal waves. J. Fluid M eeh. 53 : 1 7-44
Maxworthy, T., Browand, F. K. 1 975. Experi
ments in rotating and stratified flows :
oceanographic applications. Ann. Rev.

Fluid Mech. 7 : 273-305

McEwan, A. D. 1971. Degeneration of


resonantly-excited
standing
internal
gravity waves. J. Fluid Mech. 50 : 43 1 -48
McEwan, A. D., Robinson, R. M. 1975.
Parametric instability of internal waves.

J. Fluid Mech. 67 : 667-87

Mellor, G. L., Durbin, P. A. 1975. The struc


ture and dynamics of the ocean surface
layer. J. Phys. Oceanogr. 5 : 71 8 -28

Annu. Rev. Fluid Mech. 1978.10:267-288. Downloaded from arjournals.annualreviews.org


by University of California - Santa Barbara on 08/28/09. For personal use only.

288

SHERMAN, IMBERGER & COR COS

Munk, W. H., Anderson, E. A. 1 948. Notes


on a theory of the thermocline. J. Marine
Res. 7 : 276-95
Neshyba, S., Neal, V. T. 1 9 7 1 . Temperature
and conductivity measurements under ice
island T -3. J. Geophys. Res. 76 : 8 107-20
Newman, F. C. 1976. Temperature steps in
Lake Kivu : a bottom heated saline lake.
J. Phys. Oceano(jr. 6 : 1 57-63
Niiler, P. P. 1 975. Deepening of the wind
mixed layer. J. Marine Res. 33 : 405-22
Orlanski, 1. 1 972. On the breaking of standing
internal gravity waves. J. Fluid Mech. 54,
4 : 577-98
Patnaik, P. C, Sherman, F. S., Corcos, G. M.
1 976. A numerical simulation of Kelvin
Helmholtz waves of finite amplitude. J .
Fluid Mecho 73 : 2 1 5-40
Phillips, O. M . 1 957. On the generation of
waves by turbulent wind. J. Fluid Mech.
2 : 4 1 7-45
Phillips, O. M. 1966. The Dynamics of the
Upper Ocean. Cam bridge Univ. Press.
2 6 1 pp.
Phillips, O. M. 1 968. The interactive trapping
of internal gravity waves. J. Fluid M echo

34 : 407-16

Pollard, R . T ., R hines, P. B., Thompson,


R. O. R. Y. 1 973. The deepening of the
wind-mixed layer. Geophys. Fluid Dyn. 3 :
38 1-404
Sherman, F. S. 1976. The dynamics of un
stable free shear layers : effects of buoyancy
and
non-linear
interactions.
Fluid
Dynamics Transactions 8 : 1 41-93
Stern, M. E. 1 969. Collective instability of
salt fingers. J. flUid Mech. 3 5 : 209-1 8
Stern, M. E. 1 976. Maximum buoyancy flux
across a salt finger interface. J. M a r. Res.
34 : 95- 1 1 0
Straus, .T. M . 1 972. Finite amplitude douhly
diffusive convection. J. Fluid Mech. 56 :

353--74

Tennekes, H., Lumley, J. 1. 1972. A First


Course in Turbulence. MIT Press. 300 pp.
Thompson, R. O. R. Y. 1 976. Climatological
numerical models of the surface mixed
layer of the ocean. J. Phys. Oceanogr. 6 :
496-503
Thorpe, S. A. 1966. On wave interactions in a
stratified fluid. J. Fluid Mech. 24 : 737-51
Thorpe, S. A. 1 973. Turbulence in stably
stratified fluids : a review of laboratory

experiments. Boundary-Layer Meteorol. 5 :


95-1 1 9
Thorpe, S . A., H al l, A., Crofts, I . 1 97 2 . Thc
internal surge in Loch Nes s. Nature 237 :
96-98
Turner, J. S. 1 965. Coupled turbulent trans
ports of salt and heat across a sharp density

interface. Int. J. Heat Mass Transfer, 8 :

759-67
Turner, J. S. 1 967. Salt fingers across a density
interface. Deep Sea Res. 1 4 : 599-61 1
Turner, 1 . S. 1969. A note on wind mixing at
the seasonal thermocline. Deep Sea Res.
1 6, Supp!. : 297-300
Turner, J. S. 1 973. Buoyancy Effects in Fluids.
Cambridge Univ. Press
Turner, 1. S. 1 974. Double-diffusive pheno
mena. Ann. Rev. Fluid M echo 6 : 37-56
Turner, J. S., Chen, C F. 1974. Two
dimensional effects in double-diffusive
convection. J. Fluid M echo 63 : 577-92
Veronis, G. 1968. Effect of a stabilizing
gradient of solute on thermal convection.
J. Fluid Mech. 34 : 3 1 5-36
Voorhis, A. D., Dobson, D. 1. 1975. Thermal
convection in the Atlantis I I hot brine
pool. Deep Sea Res. 22 : 1 67-75
Warren, B. 1 972. Insensitivity of subtropical
mode water characteristics to meteoro
logical fluctuations. Deep Sea Res. 1 9 : 1-19
Williams, A. J., III. 1 974. Salt fingers in the
Mediterranean outflow. Science 1 8 5 : 94143
Williams, A. J., III. 1 975. Images of ocean
microstructure. Deep Sea Res. 22 : 8 1 1 -29
Willis, G. E., DeardorlT, 1. W. 1974. A
laboratory model ofthe unstable planetary
boundary layer. J. Atmos. Sci. 3 1 : 1 2971 307
Zeman, 0., Lumley, J. 1. 1977. Buoyancy
effects in entraining turbulent boundary
layers : a second-order closure study. Proc.
Turbulent Shear Flow Symp., A p ril 1977,
Penn. State Univ., pp. 6.2 1 -6.28
Zeman, 0., Tennekes, H. 1 977. Parameteriza
tion of the turbulent energy budget at the
top of the daytime atmospheric boundary
layer. J. A tmos. Sci. 34 : 1 1 1-23
Zilitinkevich, S. S. 1975. Comments on "A
model for the dynamics of the inversion
above a convective boundary layer." J .
Atmos. Sci. 32 : 991-92

Potrebbero piacerti anche