ATMOSPHERIC

AND OCEANIC
FLUID DYNAMICS
Supplementary Material for 2nd Edition
Georey K. Vallis
ii
Contents
Preface vii
Part I GEOPHYSICAL FLUID DYNAMICS 1
1 Wave Fundamentals and Rossby Waves 3
1.1 Wave Fundamentals 4
1.1.1 Denitions and kinematics 4
1.1.2 Wave propagation and phase speed 5
1.1.3 The dispersion relation 6
1.2 Group Velocity 7
1.2.1 Superposition of two waves 9
1.2.2

Superposition of many waves 10
1.2.3

The method of stationary phase 12
1.3 Ray Theory 13
1.3.1 Ray theory in practice 15
1.4 Rossby Waves 16
1.4.1 Waves in a single layer 16
1.4.2 The mechanism of Rossby waves 18
1.4.3 Rossby waves in two layers 19
1.5 * Rossby Waves in Stratied Quasi-Geostrophic Flow 21
1.5.1 Setting up the problem 21
1.5.2 Wave motion 22
1.6 Energy Flux of Rossby Waves 23
1.6.1

Rossby wave reection 26
1.7 Rossby-gravity Waves: an Introduction 31
iii
iv Contents
1.7.1 Wave properties 34
1.7.2 Planetary geostrophic Rossby waves 36
1.8

The Group Velocity Property 38
1.8.1 Group velocity in homogeneous media 38
1.8.2

Group velocity property: a general derivation 39
1.8.3 Group velocity property for Rossby waves 41
4 Gravity Waves 55
4.1 Surface gravity waves 56
4.1.1 Boundary conditions 56
4.1.2 Wave solutions 57
4.1.3 Properties of the solution 58
4.2 Internal Gravity waves in a Non-Rotating Boussinesq uid 62
4.3 Energetics of Poincar Waves 64
4.3.1 One-dimensional problem 64
4.3.2 Two-dimensional Poincar waves 66
4.4 Waves on Fluid Interfaces 67
4.4.1 Equations of motion 68
4.4.2 Dispersion relation 69
4.5 Internal waves in a Continuously Stratied Boussinesq uid 70
4.6 Properties of Internal Waves 72
4.6.1 A few interesting properties 72
4.6.2 Group velocity and phase speed 74
4.6.3 Energetics of internal waves 76
4.7 Internal Wave Reection 78
4.7.1 Properties of internal wave reection 79
4.8 Internal Waves in a Fluid with Varying Stratication 81
4.8.1 An alternative derivation 83
4.8.2 An atmospheric case 84
4.8.3 An atmospheric waveguide 84
4.9 Internal Waves in a Rotating Frame 85
4.9.1 Equations of motion 85
4.9.2 Dispersion Relation 86
4.9.3 Polarization relations 88
4.9.4 Geostrophic motion and vortical modes 88
4.10 Generation of Internal Waves 91
4.10.1 The problem and its solution 91
4.10.2 Energy Propagation 92
4.10.3 Lee waves and ow over topography 94
4.10.4 Gravity waves in the atmosphere 94
4.11

Acoustic-Gravity Waves in an Ideal Gas 94
4.11.1 Interpretation 95
4.12 The Moving Flame Eect 98
4.13 Breaking of Internal Waves 98
4.13.1 Relation to diapycnal diusivity 98
4.13.2 The Garrett-Munk spectrum 98
Contents v
References 101
vi CONTENTS
Preface
September 27, 2012
This is some additional material related to the book Atmospheric and Oceanic Fluid
Dynamics (AOFD). Eventually the material will be incorporated into a second edition of
that book, but that is a couple of years away.
The major items to be added may include:
(i) The material on waves will be consolidated, and most of it will be moved out of Part
I into Part II. Part II will begin with a chapter on wave basics and Rossby waves.
(ii) A chapter on gravity waves including some material on their importance to the
general circulation.
(iii) A chapter on linear dynamics at low latitudes (equatorial waves and the Matsuno
Gill problem).
(iv) A chapter on the tropical atmosphere, if it can be made coherent.
(v) A chapter on the equatorial ocean and El Nio (probably two chapters in total, one
being the chapter already posted above).
(vi) Up to a chapter on stratospheric dynamics.
(vii) Tentatively, a chapter on dynamical regimes of planetary atmospheres. This might
have to wait until a third edition.
Some of these items are present in this document. Others remain to be started. A number
of corrections will be made throughout the existing book, and some other sections will
be shortened, claried or omitted.
In general I will post new items to the web when there is something reasonably
substantial to be read, typically half a chapter or so of new material. The material will
rst be posted when it is readable, but before it is complete or nalized. (There is no point
in asking for comments on material that is nished.) I would appreciate any comments
you, the reader, may have whether major or minor. Suggestions are also welcome on
vii
viii Preface
material to include or omit. There is no need, however, to comment on typos in the text
these will be cleaned up in the nal version. However, please do point out typos in
equations and, perhaps even more importantly, thinkos, which are sort of typos in the
brain.
Student Edition
As the second edition of the book will perforce be rather long (perhaps close to 1000
pages), it may not be appropriate for graduate students who do not plan a career in
dynamics. Thus, the publisher (CUP) and I are considering a shorter student edition,
which would have the advanced or more arcane material omitted and some of the expla-
nations simplied. The resulting would likely be about 500 pages. Please let me know if
you have any comments on this.
Problem Sets
One omission in the rst edition is numerically-oriented problems that graphically illus-
trate some phenomena using Matlab or Python or similar. If you have any such problems
or would like to develop some that could be linked to this book, please let me know. Ad-
ditional problems of a conventional nature would also be welcome. Again, please contact
me.
Thank you!
Geoff Vallis
gkv@princeton.edu
Part I
GEOPHYSICAL FLUID DYNAMICS
1
Catch a wave and youre sitting on top of the world.
The Beach Boys
CHAPTER
ONE
Wave Fundamentals and Rossby Waves
In this chapter we provide an introduction to wave motion and a description of perhaps
the most important kind of wave motion occurring at large scales in the ocean and at-
mosphere, namely Rossby waves.
1
The chapter has three main parts to it. In the rst,
we provide a brief discussion of wave kinematics and dynamics, introducing such basic
concepts as phase speed and group velocity. The second part, beginning with section
3.4, is a discussion of the dynamics of Rossby waves; this may be considered to be the
natural follow-on from the previous chapter. Finally, in section 3.8, we return to group
velocity in a somewhat more general way. Wave kinematics is a somewhat formal topic,
yet closely tied to wave dynamics: kinematics without a dynamical example is jejune and
dry, yet understanding wave dynamics of any sort is hardly possible without appreciating
at least some of the formal structure of waves. Readers should ip pages back and forth
as necessary.
Those readers who already have a knowledge of wave motion, or those who wish to
cut to the chase quickly, may wish to skip the rst few sections and begin at section 3.4.
Other readers may wish to skip the sections on Rossby waves altogether and, after ab-
sorbing the sections on the wave theory move on to chapter 4 on gravity waves, returning
to Rossby waves (or not) later on. The Rossby wave and gravity wave discussions are
largely independent of each other, although they both require that the reader is familiar
with the basic ideas of wave analysis such as group velocity and phase speed. Close to
the equator Rossby waves and gravity waves become more intertwined and we deal with
the ensuing waves in chapter ??.
3
4 Chapter 1. Wave Fundamentals and Rossby Waves
1.1 FUNDAMENTALS AND FORMALITIES
1.1.1 Denitions and kinematics
What is a wave? Rather like turbulence, a wave is more easily recognized than dened.
Perhaps a little loosely, a wave may be considered to be a propagating disturbance that
has a characteristic relationship between its frequency and size; more formally, a wave
is a disturbance that satises a dispersion relation. In order to see what this means,
and what a dispersion relation is, suppose that disturbance, (x, t) (where might be
velocity, streamfunction, pressure, etc), satises some equation
L() = 0, (1.1)
where L is a linear operator, typically a polynomial in time and space derivatives; an
example is L() =
2
/t +/x. We will mainly deal with linear waves for which
the operator L is linear. Nonlinear waves certainly exist, but the curious reader must look
elsewhere to learn about them.
2
If (3.1) has constant coefcients (if is constant in this
example) then solutions may often be found as a superposition of plane waves, each of
which satisfy
= Re ` e
i(x,t)
= Re ` e
i(kxt)
. (1.2)
where ` is a constant, is the phase, k is the vector wavenumber (k, l, m), and is the
wave frequency. We also often write the wave vector as k = (k
x
, k
y
, k
z
).
Earlier, we said that waves are characterized by having a particular relationship be-
tween the frequency and wavevector known as the dispersion relation. This is an equation
of the form
= (k) (1.3)
where (k) [meaning (k, l, m)] is some function determined by the form of L and so
depends on the particular type of wave the function is different for sound waves, light
waves and the Rossby waves and gravity waves we will encounter in this book (peak
ahead to (3.56) are (4.42), and there is more discussion in section 3.1.3). Unless it is
necessary to explicitly distinguish the function from the frequency , we will often
write = (k).
If the medium in which the waves are propagating is inhomogeneous, then (3.1)
will probably not have constant coefcients (for example, may vary meridionally).
Nevertheless, if the medium is slowly varying, wave solutions may often still be found
although we do not prove it here with the general form
= Rea(x, t) e
i(x,t)
, (1.4)
where a(x, t) varies slowly compared to the variation of the phase, . The frequency and
wavenumber are then dened by
k ,

t
, (1.5)
which in turn implies the formal relation between k and :
k
t
+= 0. (1.6)
1.1 Fundamentals and Formalities 5
1.1.2 Wave propagation and phase speed
An almost universal property of waves is that they propagate through space with some
velocity (which might be zero). Waves in uids may carry energy and momentum but
not normally, at least to a rst approximation, uid parcels themselves. Further, it turns
out that the speed at which properties like energy are transported (the group speed) may
be different from the speed at which the wave crests themselves move (the phase speed).
Lets try to understand this beginning with the phase speed.
Phase speed
Let us consider the propagation of monochromatic plane waves, for that is all that is
needed to introduce the phase speed. Given (3.2) a wave will propagate in the direction
of k (Fig. 3.1). At a given instant and location we can align our coordinate axis along
this direction, and we write k x = Kx

, where x

increases in the direction of k and

K
2
= |k|
2
is the magnitude of the wavenumber. With this, we can write (3.2) as
= Re ` e
i(Kx

t)
= Re ` e
iK(x

ct)
, (1.7)
where c = /K. From this equation it is evident that the phase of the wave propagates
at the speed c in the direction of k, and we dene the phase speed by
c
p


K
. (1.8)
The wavelength of the wave, , is the distance between two wavecrests that is, the
distance between two locations along the line of travel whose phase differs by 2 and
evidently this is given by
=
2
K
. (1.9)
In (for simplicity) a two-dimensional wave, and referring to Fig. 3.1(a), the wavelength
and wave vectors in the x- and y-directions are given by,

x
=

cos
,
y
=

sin
, k
x
= Kcos , k
y
= Ksin. (1.10)
In general, lines of constant phase intersect both the coordinate axes and propagate along
them. The speed of propagation along these axes is given by
c
x
p
= c
p
l
x
l
=
c
p
cos
= c
p
K
k
x
=

k
x
, c
y
p
= c
p
l
y
l
=
c
p
sin
= c
p
K
k
y
=

k
y
, (1.11)
using (3.8) and (3.10). The speed of phase propagation along any one of the axis is
in general larger than the phase speed in the primary direction of the wave. The phase
speeds are clearly not components of a vector: for example, c
x
p
c
p
cos . Analogously,
the wavevector k is a true vector, whereas the wavelength is not.
To summarize, the phase speed and its components are given by
c
p
=

K
, c
x
p
=

k
x
, c
y
p
=

k
y
. (1.12)
6 Chapter 1. Wave Fundamentals and Rossby Waves
Fig. 1.1 The propagation of a two-dimensional wave. (a) Two lines of constant phase
(e.g., two wavecrests) at a time t
1
. The wave is propagating in the direction k with
wavelength . (b) The same line of constant phase at two successive times. The phase
speed is the speed of advancement of the wavecrest in the direction of travel, and so
c
p
= l/(t
2
t
1
). The phase speed in the x-direction is the speed of propagation of the
wavecrest along the x-axis, and c
x
p
= l
x
/(t
2
t
1
) = c
p
/ cos .
Phase velocity
Although it is not particularly useful, there is a way of dening a phase speed so that is a
true vector, and which might then be called phase velocity. We dene the phase velocity
to be the phase speed in the direction in which the wave crests are propagating; that is
c
p


K
k
|K|
, (1.13)
where k/|K| is the unit vector in the direction of wave-crest propagation. The compo-
nents of the phase velocity in the the x- and y-directions are then given by
c
x
p
= c
p
cos , c
y
p
= c
p
sin. (1.14)
Dened this way, the quantity given by (3.14) is indeed a true vector velocity. However,
the components in the x- and y-directions are manifestly not the speed at which wave
crests propagate in those directions. It is therefore a misnomer to call these quantities
phase speeds, although it is helpful to ascribe a direction to the phase speed and so the
quantity given by (3.14) can be useful.
1.1.3 The dispersion relation
Much of above description is mostly kinematic and a little abstract, applying to almost
any disturbance that has a wavevector and a frequency. The particular dynamics of a
wave are determined by the relationship between the wavevector and the frequency;
that is, by the dispersion relation. Once the dispersion relation is known a great many
of the properties of the wave follow in a more-or-less straightforward manner, as we will
1.2 Group Velocity 7
see. Picking up from (3.3), the dispersion relation is a functional relationship between
the frequency and the wavevector of the general form
= (k). (1.15)
Perhaps the simplest example of a linear operator that gives rise to waves is the one-
dimensional equation

t
+c

x
= 0. (1.16)
Substituting a trial solution of the form = ReAe
i(kxt)
, where Re denotes the real
part, we obtain (i+cik)A = 0, giving the dispersion relation
= ck. (1.17)
The phase speed of this wave is c
p
= /k = c.
A few other examples of governing equations, dispersion relations and phase speeds
are:

t
+c = 0, = c k, c
p
= |c| cos , c
x
p
=
c k
k
, c
y
p
=
c k
l
(1.18a)

t
2
c
2

2
= 0,
2
= c
2
K
2
, c
p
= c, c
x
p
=
cK
k
, c
y
p
=
cK
l
, (1.18b)

2
+

x
= 0, =
k
K
2
, c
p
=

K
, c
x
p
=

K
2
, c
y
p
=
k/l
K
2
. (1.18c)
where K
2
= k
2
+l
2
and is the angle between c and k.
A wave is said to be nondispersive or dispersionless if the phase speed is independent
of the wavelength. This condition is clearly satised for the simple example (3.16) but
is manifestly not satised for (3.18c), and these waves (Rossby waves, in fact) are dis-
persive. Waves of different wavelengths then travel at different speeds so that a group
of waves will spread out or disperse (hence the name), even if the medium is homo-
geneous. When a wave is dispersive there is another characteristic speed at which the
waves propagate, known as the group velocity, and we come to this in the next section.
Most media are, of course, inhomogeneous, but if the medium varies sufciently
slowly and in particular if the variations are slow compared to the wavelength we
may still have a local dispersion relation between frequency and wavevector,
= (k; x, t). (1.19)
Although is a function of k, x and t the semi-colon in (3.19) is used to suggest that x
and t are slowly varying parameters of a somewhat different nature than k. Well pick
up our discussion of this in section 3.3, but before that we must introduced the group
velocity.
1.2 GROUP VELOCITY
Information and energy travel clearly cannot travel at the phase speed, for as the direc-
tion of propagation of the phase line tends to a direction parallel to the y-axis, the phase
8 Chapter 1. Wave Fundamentals and Rossby Waves
Wave Fundamentals
A wave is a propagating disturbance that has a characteristic relationship between
its frequency and size, known as the dispersion relation. Waves typically arise as
solutions to a linear problem of the form
L() = 0, (W.1)
where L is (commonly) a linear operator in space and time. Two examples are

t
2
c
2

2
= 0 and

t

2
+

x
= 0. (W.2)
The rst example is so common in all areas of physics it is sometimes called the wave
equation. The second example gives rise to Rossby waves.
Solutions to the governing equation are often sought in the form of plane waves that
have the form
= ReAe
i(kxt)
, (W.3)
where A is the wave amplitude, k the wavevector, k = (k, l, m), and is the frequency.
The dispersion relation is a functional relationship between the frequency and
wavevector of the form = (k) where is a function. It arises from substitut-
ing a trial solution like (W.3) into the governing equation (W.1). For the examples
of (W.2) we obtain = c
2
K
2
and = k/K
2
and K
2
= k
2
+ l
2
+ m
2
or, in two
dimensions, K
2
= k
2
+l
2
.
The phase speed is the speed at which the wave crests move. In the direction of
propagation and in the x, y and z directions group speed is given by, respectively,
c
p
=

K
, c
x
p
=

k
, c
y
p
=

l
, c
z
p
=

m
. (W.4)
where K = 2/ where is the wavelength. Phase speed is not a vector.
The group velocity is the velocity at which a wave packet or wave group moves. It is
a vector and is given by
c
g
=

k
, c
x
g
=

k
, c
y
g
=

l
, c
z
g
=

m
. (W.5)
Energy and various other physical quantities are also transported at the group velocity.
If the medium is inhomogeneous but only slowly varying in space and time, then
approximate solutions may sometimes be found in the form
= ReA(x, t) e
i(x,t)
, (W.6)
where the amplitude A is also slowly varying and the local wavenumber and frequency
are related to the phase, , by k = and = /t . The dispersion relation is
then a local one of the form = (k; x, t).
1.2 Group Velocity 9
Fig. 1.2 Superposition of two sinusoidal waves with wavenumbers k and k + k, pro-
ducing a wave (solid line) that is modulated by a slowly varying wave envelope or wave
packet (dashed line). The envelope moves at the group velocity, c
g
= /k and the
phase of the wave moves at the group speed c
p
= /k.
speed in the x-direction tends to innity! Rather, it turns out that most quantities of
interest, including energy, propagate at the group velocity, a quantity of enormous impor-
tance in wave theory. Rather roughly, this is the velocity at which a packet or a group of
waves will travel, whereas the individual wave crests travel at the phase speed. To intro-
duce the idea we will consider consider the superposition of plane waves, noting that a
monochromatic plane wave already lls space uniformly so that there can be no propa-
gation of energy from place to place. We will restrict attention to waves propagating in
one direction, but the argument may be extended to two or three dimensions.
1.2.1 Superposition of two waves
Consider the linear superposition of two waves. Limiting attention to the one-dimensional
case for simplicity, consider a disturbance represented by
= Re ` (e
i(k
1
x
1
t)
+e
i(k
2
x
2
t)
). (1.20)
Let us further suppose that the two waves have similar wavenumbers and frequency, and,
in particular, that k
1
= k + k and k
2
= k k, and
1
= + and
2
= .
With this, (3.20) becomes
= Re ` e
i(kxt)
[e
i(kxt)
+e
i(kxt)
]
= 2Re ` e
i(kxt)
cos(kx t).
(1.21)
The resulting disturbance, illustrated in Fig. 3.2 has two aspects: a rapidly varying com-
ponent, with wavenumber k and frequency , and a more slowly varying envelope, with
wavenumber k and frequency . The envelope modulates the fast oscillation, and
moves with velocity /k; in the limit k 0 and 0 this is the group velocity,
c
g
= /k. Group velocity is equal to the phase speed, /k, only when the frequency is
a linear function of wavenumber. The energy in the disturbance must move at the group
velocity note that the node of the envelope moves at the speed of the envelope and no
10 Chapter 1. Wave Fundamentals and Rossby Waves
energy can cross the node. These concepts generalize to more than one dimension, and if
the wavenumber is the three-dimensional vector k = (k, l, m) then the three-dimensional
envelope propagates at the group velocity given by
c
g
=

k

k
,

l
,

m

. (1.22)
1.2.2

Superposition of many waves
Now consider a generalization of the above arguments to the case in which many waves
are excited. In a homogeneous medium, nearly all wave patterns can be represented as a
superposition of an innite number of plane waves; mathematically the problemis solved
by evaluating a Fourier integral. For mathematical simplicity well continue to treat only
the one-dimensional case but the three dimensional generalization is straightforward.
A superposition of plane waves, each satisfying some dispersion relation, can be rep-
resented by the Fourier integral
(x, t) =

`
A(k) e
i(kxt)
dk. (1.23a)
The function
`
A(k) is given by the initial conditions:
`
A(k) =
1
2

(x, 0) e
ikx
dx. (1.23b)
As an aside, note that if the waves are dispersionless and = ck where c is a constant,
then
(x, t) =

`
A(k) e
ik(xct)
dk = (x ct, 0), (1.24)
by comparison with (3.23a) at t = 0. That is, the initial condition simply translates at a
speed c, with no change in structure.
Although the above procedure is quite general it doesnt get us very far: it doesnt
provide us with any physical intuition, and the integrals themselves may be hard to eval-
uate. A physically more revealing case is to consider the case for which the disturbance
is a wave packet essentially a nearly plane wave or superposition of waves conned to
a nite region of space. We will consider a case with the initial condition
(x, 0) = a(x) e
ik
0
x
(1.25)
where a(x), rather like the envelope in Fig. 3.3, modulates the amplitude of the wave on
a scale much longer than that of the wavelength 2/k
0
, and more slowly than the wave
period. That is,
1
a
a
x
k
0
,
1
a
a
t
k
0
c, (1.26a,b)
and the disturbance is essentially a slowly modulated plane wave. We suppose that a(x)
is peaked around some value x
0
and is very small if |x x
0
| k
1
0
; that is, a(x) is small
if we are several wavelengths of the plane wave away from the peak. We would like to
know how such a packet evolves.
1.2 Group Velocity 11
a(x)

0
= 2/k
0
L
x
L
x

0
c
p
c
g
Fig. 1.3 A generic wave packet. The envelope, a(x), has a scale L
x
that is much larger
than the wavelength,
0
, of the wave embedded within in. The envelope moves at the
group velocity, c
g
, and the phase of the waves at the phase speed, c
p
.
We can express the envelope as a Fourier integral by rst noting that that we can
write the initial conditions as a Fourier integral,
(x, 0) =

`
A(k) e
ikx
dk where
`
A(k) =
1
2

(x, 0) e
ikx
dx, (1.27a,b)
so that, using (3.25),
`
A(k) =
1
2

a(x) e
i(k
0
k)x
dx and a(x) =

`
A(k) e
i(kk
0
)x
dk. (1.28a,b)
We still havent made much progress beyond (3.23). To do so, we note rst that a(x) is
conned in space, so that to a good approximation the limits of the integral in (3.28a)
can be made nite, L say, provided L k
1
0
. We then note that when (k
0
k)x is large
the integrand in (3.28a) oscillates rapidly; successive intervals in x therefore cancel each
other and make a small net contribution to the integral. Thus, the integral is dominated
by values of k near k
0
, and
`
A(k) is peaked near k
0
. (Note that the nite spatial extent of
a(x) is crucial for this argument.)
We can now evaluate how the wave packet evolves. Beginning with (3.23a) we have
(x, t) =

`
A(k) exp{i(kx (k)t)} dk (1.29a)

`
A(k) exp

i[k
0
x (k
0
)t] +i(k k
0
)x i(k k
0
)

k

k=k
0
t

dk (1.29b)
having expanded (k) in a Taylor series about k and kept only the rst two terms, noting
that the wavenumber band is limited. We therefore have
(x, t) = exp{i[k
0
x (k
0
)t]}

`
A(k) exp

i(k k
0
)

x

k

k=k
0
t

dk (1.30a)
= exp{i[k
0
x (k
0
)t]} a

x c
g
t

(1.30b)
12 Chapter 1. Wave Fundamentals and Rossby Waves
where c
g
= /k evaluated at k = k
0
. That is to say, the envelope a(x) moves at the
group velocity.
The group velocity has a meaning beyond that implied by the derivation above: there
is no need to restrict attention to narrow band processes, and it turns out to be a quite
general property of waves that energy (and certain other quadratic properties) propagate
at the group velocity. This is to be expected, at least in the presence of coherent wave
packets, because there is no energy outside of the wave envelope so the energy must
propagate with the envelope.
1.2.3

The method of stationary phase
We will now relax the assumption that wavenumbers are conned to a narrow band and
look for solutions at large t; that is, we will be seeking a description of waves far from
their source. Consider a disturbance of the general form
(x, t) =

`
A(k) e
i[kx(k)t]
dk =

`
A(k) e
i(k;x,t)t
dk (1.31)
where (k : x, t) kx/t (k). (Here we regard as a function of k with parameters
x and t; we will sometimes just write (k) with

(k) = /k.) Now, a standard result

in mathematics (known as the RiemannLebesgue lemma) states that
I = lim
t

f (k) e
ikt
dk = 0 (1.32)
provided that f (k) is integrable and

f (k) dk is nite. Intuitively, as t increases the

oscillations in the integral increase and become much faster than any variation in A(k);
successive oscillations thus cancel and the integral becomes very small (Fig. 3.4).
Looking at (3.31), the integral will be small if is everywhere varying with k. How-
ever, if there is a region where does not vary with k that is, if there is a region where
the phase is stationary and /k = 0 then there will be a contribution to the integral
from that region. Thus, for large t, an observer will predominantly see waves for which

(k) = 0 and so, using the denition of , for which

x
t
=

k
. (1.33)
In other words, at some space-time location (x, t) the waves that dominate are those
whose group velocity /k is x/t. In the example plotted in Fig. 3.4, = /k so that
the wavenumber that dominates, k
0
say, is given by solving /k
2
0
= x/t, which for x/t = 1
and = 400 gives k
0
= 20.
We may actually approximately calculate the contribution to (x, t) from waves mov-
ing with the group velocity. Let us expand (k) around the point, k
0
, where

(k) = 0.
We obtain
(x, t) =

`
A(k) exp

it

(k
0
) +(k k
0
)

(k
0
) +
1
2
(k k
0
)
2

(k
0
) . . .

dk (1.34)
The higher order terms are small because k k
0
is presumed small (for if it is large the
integral vanishes), and the term involving

(k
0
) is zero. The integral becomes
(x, t) =
`
A(k
0
)e
i(k
0
)

exp

it
1
2
(k k
0
)
2

(k
0
)

dk. (1.35)
1.3 Ray Theory 13
0 5 10 15 20 25 30 35 40 45 50
1
0.5
0
0.5
1
0 5 10 15 20 25 30 35 40 45 50
1
0.5
0
0.5
1
Wavenumber, k
Wavenumber, k
k
0
t = 1
t = 12
k
0

A
(
k
)
e
i
t

(
k
;
x
,
t
)

A
(
k
)
e
i
t

(
k
;
x
,
t
)

A(k)

A(k)
Fig. 1.4 The integrand of (3.31), namely the function that when integrated over wave-
number gives the wave amplitude at a particular x and t. The example shown is for a
Rossby wave with = /k, with = 400 and x/t = 1, and hence k
0
= 20, for two
times t = 1 and t = 12. (The envelope,
`
A(k), is somewhat arbitrary.) At the later time
the oscillations are much more rapid in k, so that the contribution is more peaked
from wavenumbers near to k
0
.
We therefore have to evaluate a Gaussian, and because

e
cx
2
dx =

/c we obtain
(x, t)
`
A(k
0
)e
i(k
0
)

2/(it

(k
0
))

1/2
=
`
A(k
0
)e
i(k
0
x(k
0
)t)

2i/(t

(k
0
))

1/2
.
(1.36)
The solution is therefore a plane wave, with wavenumber k
0
and frequency (k
0
), slowly
modulated by an envelope determined by the form of (k
0
; x, t), where k
0
is the wave-
number such that x/t = c
g
= /k|
k=k
0
1.3 RAY THEORY
Most waves propagate in a media that is inhomogeneous. In the Earths atmosphere
and ocean the stratication varies with altitude and the Coriolis parameter varies with
latitude. In these cases it can be hard to obtain the solution of a wave problem by
Fourier methods, even approximately. Nonetheless, the ideas of signals propagating t
the group velocity is a very robust one, and it turns out that we can often obtain much
14 Chapter 1. Wave Fundamentals and Rossby Waves
of the information we want and in particular the trajectory of a wave using an
approximate recipe known as ray theory, using the word theory a little generously.
3
In an inhomogeneous medium let us suppose that the solution to a particular wave
problem is of the form
(x, t) = a(x, t) e
i(x,t)
, (1.37)
where a is the wave amplitude and the phase, and a varies slowly in a sense we will
make more precise shortly. The local wavenumber and frequency are dened by
k ,

t
. (1.38)
We suppose that the amplitude a varies slowly over a wavelength and a period; that is
|a|/|a| is small over the length 1/k and the period 1/ or
|a/x|
a
|k|,
|a/t |
a
(1.39)
We will assume that the wavenumber and frequency as dened by (3.38) are the same as
those that would arise if the medium were homogeneous and a were a constant. Thus,
we may obtain a dispersion relation from the governing equation by keeping the spatially
(and possibly temporally) varying parameters xed and obtain
= (k; x, t), (1.40)
and then allow x and t to vary, albeit slowly. (This procedure may be formalized using a
two-scale approximation, or equivalently using WKB methods.)
Let us now consider how the wavevector and frequency might change with position
and time. We recall from (3.6) that the wavenumber and frequency are related by
k
i
t
+

x
i
= 0, (1.41)
where we use a subscript notation for vectors, and in what follows repeated indices will
be summed. Using (3.41) and (3.40) gives
k
i
t
+

x
i
+

k
j
k
j
x
i
= 0. (1.42)
This may be written
k
t
+c
g
k = (1.43)
where
c
g
=

k
=

k
,

l
,

m

, (1.44)
is, again, the group velocity, sometimes written as c
g
=
k
or, in subscript notation,
as c
gi
= /k
i
. The left-hand side of (3.23a) is similar to an advective derivative, but
with the velocity being a group velocity not a uid velocity. Evidently, if the dispersion
relation for frequency is not an explicit function of space the wavevector is propagated
at the group velocity.
1.3 Ray Theory 15
The frequency is, in general, a function of space, wavenumber and time, and from
the dispersion relation, (3.40), its variation is governed by

t
=

t
+

k
i
k
i
t
=

t


k
i

x
i
(1.45)
using (3.41). Thus, using the denition of group velocity, we may write

t
+c
g
=

t
. (1.46)
As with (3.43) the left-hand side is like an advective derivative, but with the velocity
being a group velocity. If the dispersion relation is not a function of time, the frequency
propagates at the group velocity.
Motivated by (3.43) and (3.46) we dene a ray as the trajectory traced by the group
velocity. Noting and if the frequency is not an explicit function of space or time, then
both the wavevector and the frequency are constant along a ray.
1.3.1 Ray theory in practice
What use is ray theory? The idea is that we use (3.43) and (3.46) to track a group of
waves from one location to another without solving the full wave equations of motion.
Indeed, it turns out that we can sometimes solve problems using ordinary differential
equations (ODEs) rather than partial differential equations (PDEs).
Suppose that the initial conditions consist of a group of waves at a position x
0
, for
which the amplitude and wavenumber vary only slowly with position. We also suppose
that we know the dispersion relation for the waves at hand; that is, we know the func-
tional form of (k; x, t). Let us dene the total derivate following the group velocity
as
d
dt


t
+c
g
, (1.47)
so that (3.43) and (3.46) may be written as
dk
dt
= , (1.48a)
d
dt
=

t
. (1.48b)
These are ordinary differential equations for wavevector and frequency, solvable pro-
vided we know the right-hand sides; that is, provided we know the space and time loca-
tion at which the dispersion relation [i.e., (k; x, t)] is to be evaluated. But the location
is known because it is moving with the group velocity and so
dx
dt
= c
g
. (1.48c)
where c
g
= /k|
x,t
(i.e., c
gi
= /k
i
|
x,t
).
The set (3.48) is a triplet of ordinary differential equations for the wavevector, fre-
quency and position of a wave group. They may be solved, albeit sometimes numerically,
16 Chapter 1. Wave Fundamentals and Rossby Waves
Fig. 1.5 Schema of the trajec-
tory of two wavepackets, each
moving with a dierent group
velocity, as might be calculated
using ray theory. If the wave
packets collide ray theory must
fail.
T
i
m
e
x
Wave packet collision.
Ray theory fails.
Trajectory 1
Trajectory 2
c
g1
c
g2
to give the trajectory of a wave packet or collection of wave packets, as schematically il-
lustrated in Fig. 3.5. Of course, if the medium or the wavepacket amplitude is not slowly
varying ray theory will fail, and this will perforce happen if two wave packets collide.
The amplitude of the wave packet is not given by ray theory. However, given our
earlier discussions, it should come as no surprise to nd that a quantity related to the
amplitude of a wave packet specically, the wave activity, which is quadratic in the
wave amplitude is also propagated at the group velocity, but we leave our discussion
of that for later sections. We now turn our attention to a specic form of wave Rossby
waves but the reader whose interest is more in the general properties of waves may
skip forward to section 3.8
1.4 ROSSBY WAVES
We now shift gears and consider in some detail a particular type of waves, namely Rossby
waves in a quasi-geostrophic system. These waves are perhaps the most important large-
scale wave in the atmosphere and ocean (although gravity waves, discussed in the next
chapter, are arguably as important in some contexts).
4
1.4.1 Waves in a single layer
Consider ow of a single homogeneous layer on a at-bottomed -plane. The unforced,
inviscid equation of motion is
D
Dt
( +f k
2
d
) = 0, (1.49)
where =
2
is the vorticity and is the streamfunction and k
d
= 1/L
d
is the inverse
radius of deformation. (Note that denitions of k
d
and L
d
vary, even )
1.4 Rossby Waves 17
Innite deformation radius
If the scale of motion is much less than the deformation scale then on the -plane the
equation of motion is governed by
D
Dt
( +y) = 0. (1.50)
Expanding the material derivative gives

t
+u +v = 0 or

t
+J(, ) +

x
= 0. (1.51)
We now linearize this equation; that is, we suppose that the ow consists of a time-
independent component (the basic state) plus a perturbation, with the perturbation
being small compared with the mean ow. Such a mean ow must satisfy the time-
independent equation of motion, and purely zonal ow will do this. For simplicity we
choose a ow with no meridional dependence and let
= +

(x, y, t), (1.52)

where = Uy and |

| ||. (The symbol U represents the zonal ow of the basic

state, not a magnitude for scaling purposes.) Substitute (3.52) into (3.51) and neglect
the nonlinear terms involving products of

to give

t
+J(,

) +

x
= 0 or

t

+U

x
+

x
= 0. (1.53a,b)
Solutions to this equation may be found in the form of a plane wave,

= Re ` e
i(kx+lyt)
, (1.54)
where Re indicates the real part of the function (and this will sometimes be omitted
if no ambiguity is so-caused). Solutions of the form (3.54) are valid in a domain with
doubly-periodic boundary conditions; solutions in a channel can be obtained using a
meridional variation of sinly, with no essential changes to the dynamics. The amplitude
of the oscillation is given by ` and the phase by kx + ly t, where k and l are the x-
and y-wavenumbers and is the frequency of the oscillation.
Substituting (3.54) into (3.53) yields
[(+Uk)(K
2
) +k] ` = 0, (1.55)
where K
2
= k
2
+l
2
. For non-trivial solutions this implies
= Uk
k
K
2
. (1.56)
This is the dispersion relation for Rossby waves. The phase speed, c
p
, and group velocity,
c
g
, in the x-direction are
c
x
p


k
= U

K
2
, c
x
g


k
= U +
(k
2
l
2
)
(k
2
+l
2
)
2
. (1.57a,b)
18 Chapter 1. Wave Fundamentals and Rossby Waves
The velocity U provides a uniform translation, and Doppler shifts the frequency. The
phase speed in the absence of a mean ow is westwards, with waves of longer wave-
lengths travelling more quickly, and the eastward current speed required to hold the
waves of a particular wavenumber stationary (i.e., c
x
p
= 0) is U = /K
2
. We discuss the
meaning of the group velocity in the appendix.
Finite deformation radius
For a nite deformation radius the basic state = Uy is still a solution of the original
equations of motion, but the potential vorticity corresponding to this state is q = Uy/L
2
d
+
y and its gradient is q = ( +U/L
2
d
)j. The linearized equation of motion is thus

t
+U

x

(
2

/L
2
d
) +( +U/L
2
d
)

x
= 0. (1.58)
Substituting

= ` e
i(kx+lyt)
we obtain the dispersion relation,
=
k(UK
2
)
K
2
+1/L
2
d
= Uk k
+U/L
2
d
K
2
+1/L
2
d
. (1.59)
The corresponding x-components of phase speed and group velocity are
c
x
p
= U
+Uk
2
d
K
2
+k
2
d
=
UK
2

K
2
+k
2
d
, c
x
g
= U +
( +Uk
2
d
)(k
2
l
2
k
2
d
)
(k
2
+l
2
+k
2
d
)
2
, (1.60a,b)
where k
d
= 1/L
d
. The uniformvelocity eld nowno longer provides just a simple Doppler
shift of the frequency, nor a uniform addition to the phase speed. From (3.60a) the waves
are stationary when K
2
= K
2
s
/U; that is, the current speed required to hold waves
of a particular wavenumber stationary is U = /K
2
. However, this is not simply the
magnitude of the phase speed of waves of that wavenumber in the absence of a current
this is given by
c
x
p
=

K
2
s
+k
2
d
=
U
1 +k
2
d
/K
2
s
. (1.61)
Why is there a difference? It is because the current does not just provide a uniform trans-
lation, but, if L
d
is non-zero, it also modies the basic potential vorticity gradient. The
basic state height eld
0
is sloping, that is
0
= (f
0
/g)Uy, and the ambient potential
vorticity eld increases with y, that is q = ( + U/L
2
d
)y. Thus, the basic state denes a
preferred frame of reference, and the problem is not Galilean invariant.
5
We also note
that, from (3.60b), the group velocity is negative (westward) if the x-wavenumber is suf-
ciently small, compared to the y-wavenumber or the deformation wavenumber. That
is, said a little loosely, long waves move information westward and short waves move in-
formation eastward, and this is a common property of Rossby waves The x-component of
the phase speed, on the other hand, is always westward relative to the mean ow.
1.4.2 The mechanism of Rossby waves
The fundamental mechanism underlying Rossby waves is easily understood. Consider a
material line of stationary uid parcels along a line of constant latitude, and suppose that
1.4 Rossby Waves 19
Fig. 1.6 The mechanism of a two-dimensional (xy) Rossby wave. An initial distur-
bance displaces a material line at constant latitude (the straight horizontal line) to the
solid line marked (t = 0). Conservation of potential vorticity, y + , leads to the
production of relative vorticity, as shown for two parcels. The associated velocity eld
(arrows on the circles) then advects the uid parcels, and the material line evolves into
the dashed line. The phase of the wave has propagated westwards.
some disturbance causes their displacement to the line marked (t = 0) in Fig. 3.6. In the
displacement, the potential vorticity of the uid parcels is conserved, and in the simplest
case of barotropic ow on the -plane the potential vorticity is the absolute vorticity,
y + . Thus, in either hemisphere, a northward displacement leads to the production
of negative relative vorticity and a southward displacement leads to the production of
positive relative vorticity. The relative vorticity gives rise to a velocity eld which, in
turn, advects the parcels in material line in the manner shown, and the wave propagates
westwards.
In more complicated situations, such as ow in two layers, considered below, or in
a continuously stratied uid, the mechanism is essentially the same. A displaced uid
parcel carries with it its potential vorticity and, in the presence of a potential vorticity
gradient in the basic state, a potential vorticity anomaly is produced. The potential
vorticity anomaly produces a velocity eld (an example of potential vorticity inversion)
which further displaces the uid parcels, leading to the formation of a Rossby wave. The
vital ingredient is a basic state potential vorticity gradient, such as that provided by the
change of the Coriolis parameter with latitude.
1.4.3 Rossby waves in two layers
Now consider the dynamics of the two-layer model, linearized about a state of rest. The
two, coupled, linear equations describing the motion in each layer are

1
+F
1
(

1
)

1
x
= 0, (1.62a)

2
+F
2
(

2
)

2
x
= 0, (1.62b)
where F
1
= f
2
0
/g

H
1
and F
2
= f
2
0
/g

H
2
. By inspection (3.62) may be transformed into two
uncoupled equations: the rst is obtained by multiplying (3.62a) by F
2
and (3.62b) by
20 Chapter 1. Wave Fundamentals and Rossby Waves
F
1
and adding, and the second is the difference of (3.62a) and (3.62b). Then, dening
=
F
1

2
+F
2

1
F
1
+F
2
, =
1
2
(

2
), (1.63a,b)
(think for temperature), (3.62) become

2
+

x
= 0, (1.64a)

(
2
k
2
d
)

x
= 0, (1.64b)
where now k
d
= (F
1
+ F
2
)
1/2
. The internal radius of deformation for this problem is the
inverse of this, namely
L
d
= k
1
d
=
1
f
0

H
1
H
2
H
1
+H
2

1/2
. (1.65)
The variables and are the normal modes for the two-layer model, as they oscillate
independently of each other. [For the continuous equations the analogous modes are the
eigenfunctions of
z
[(f
2
0
/N
2
)
z
] =
2
.] The equation for , the barotropic mode, is
identical to that of the single-layer, rigid-lid model, namely (3.53) with U = 0, and its
dispersion relation is just
=
k
K
2
. (1.66)
The barotropic mode corresponds to synchronous, depth-independent, motion in the two
layers, with no undulations in the dividing interface.
The displacement of the interface is given by 2f
0
/g

and so proportional to the

amplitude of , the baroclinic mode. The dispersion relation for the baroclinic mode is
=
k
K
2
+k
2
d
. (1.67)
The mass transport associated with this mode is identically zero, since from (3.63) we
have

1
= +
2F
1

F
1
+F
2
,
2
=
2F
2

F
1
+F
2
, (1.68a,b)
and this implies
H
1

1
+H
2

2
= (H
1
+H
2
). (1.69)
The left-hand side is proportional to the total mass transport, which is evidently associ-
ated with the barotropic mode.
The dispersion relation and associated group and phase velocities are plotted in
Fig. 3.7. The x-component of the phase speed, /k, is negative (westwards) for both
baroclinic and barotropic Rossby waves. The group velocity of the barotropic waves
is always positive (eastwards), but the group velocity of long baroclinic waves may be
negative (westwards). For very short waves, k
2
k
2
d
, the baroclinic and barotropic ve-
locities coincide and their phase and group velocities are equal and opposite. With a
deformation radius of 50 km, typical for the mid-latitude ocean, then a non-dimensional
frequency of unity in the gure corresponds to a dimensional frequency of 5 10
7
s
1
or a period of about 100 days. In an atmosphere with a deformation radius of 1000km
1.5 * Rossby Waves in Stratied Quasi-Geostrophic Flow 21
Fig. 1.7 Left: the dispersion relation for barotropic (
t
, solid line) and baroclinic
(
c
, dashed line) Rossby waves in the two-layer model, calculated using (3.66) and
(3.67) with k
y
= 0, plotted for both positive and negative zonal wavenumbers and
frequencies. The wavenumber is non-dimensionalized by k
d
, and the frequency is
non-dimensionalized by /k
d
. Right: the corresponding zonal group and phase ve-
locities, c
g
= /k
x
and c
p
= /k
x
, with superscript t or c for the barotropic or
baroclinic mode, respectively. The velocities are non-dimensionalized by /k
2
d
.
a non-dimensional frequency of unity corresponds to 1 10
5
s
1
or a period of about 7
days. Non-dimensional velocities of unity correspond to respective dimensional velocities
of about 0.25ms
1
(ocean) and 10ms
1
(atmosphere).
The deformation radius only affects the baroclinic mode. For scales much smaller
than the deformation radius, K
2
k
2
d
, we see from (3.64b) that the baroclinic mode
obeys the same equation as the barotropic mode so that

2
+

x
= 0. (1.70)
Using this and (3.64a) implies that

i
+

i
x
= 0, i = 1, 2. (1.71)
That is to say, the two layers themselves are uncoupled from each other. At the other
extreme, for very long baroclinic waves the relative vorticity is unimportant.
1.5 * ROSSBY WAVES IN STRATIFIED QUASI-GEOSTROPHIC FLOW
1.5.1 Setting up the problem
Let us now consider the dynamics of linear waves in stratied quasi-geostrophic ow
on a -plane, with a resting basic state. (In chapter ?? we explore the role of Rossby
waves in a more realistic setting.) The interior ow is governed by the potential vorticity
equation, (??), and linearizing this about a state of rest gives

+
1
` (z)

` (z)F(z)

x
= 0, (1.72)
22 Chapter 1. Wave Fundamentals and Rossby Waves
where ` is the density prole of the basic state and F(z) = f
2
0
/N
2
. (F is the square of the
inverse Prandtl ratio, N/f
0
.) In the Boussinesq approximation ` =
0
, i.e., a constant.
The vertical boundary conditions are determined by the thermodynamic equation, (??).
If the boundaries are at, rigid, slippery surfaces then w = 0 at the boundaries and if
there is no surface buoyancy gradient the linearized thermodynamic equation is

= 0. (1.73)
We apply this at the ground and, with somewhat less justication, at the tropopause
the higher static stability of the stratosphere inhibits vertical motion. If the ground is
not at or if friction provides a vertical velocity by way of an Ekman layer, the boundary
condition must be correspondingly modied, but we will stay with the simplest case here
and apply (3.73) at z = 0 and z = H.
1.5.2 Wave motion
As in the single-layer case, we seek solutions of the form

= Re ` (z) e
i(kx+lyt)
, (1.74)
where ` (z) will determine the vertical structure of the waves. The case of a sphere is
more complicated but introduces no truly new physical phenomena.
Substituting (3.74) into (3.72) gives

K
2
` (z) +
1
`

` F(z)
`
z

k ` (z) = 0. (1.75)
Now, if ` satises
1
`

` F(z)
`
z

= ` , (1.76)
where is a constant, then the equation of motion becomes

K
2
+

` k ` = 0, (1.77)
and the dispersion relation follows, namely
=
k
K
2
+
. (1.78)
Equation (3.76) constitutes an eigenvalue problem for the vertical structure; the bound-
ary conditions, derived from (3.73), are ` /z = 0 at z = 0 and z = H. The resulting
eigenvalues, are proportional to the inverse of the squares of the deformation radii for
the problem and the eigenfunctions are the vertical structure functions.
1.5 * Rossby Waves in Stratied Quasi-Geostrophic Flow 23
A simple example
Consider the case in which F(z) and ` are constant, and in which the domain is conned
between two rigid surfaces at z = 0 and z = H. Then the eigenvalue problem for the
vertical structure is
F

2
`
z
2
= ` (1.79a)
with boundary conditions of
`
z
= 0, at z = 0, H. (1.79b)
There is a sequence of solutions to this, namely
`
n
(z) = cos(nz/H), n = 1, 2. . . (1.80)
with corresponding eigenvalues

n
= n
2
F
2
H
2
= (n)
2

f
0
NH

2
, n = 1, 2. . . . (1.81)
Equation (3.81) may be used to dene the deformation radii for this problem, namely
L
n

1

n
=
NH
nf
0
. (1.82)
The rst deformation radius is the same as the expression obtained by dimensional anal-
ysis, namely NH/f , except for a factor of . (Denitions of the deformation radii both
with and without the factor of are common in the literature, and neither is obviously
more correct. In the latter case, the rst deformation radius in a problem with uniform
stratication is given by NH/f , equal to /

1
.) In addition to these baroclinic modes,
the case with n = 0, that is with ` = 1, is also a solution of (3.79) for any F(z).
Using (3.78) and (3.81) the dispersion relation becomes
=
k
K
2
+(n)
2
(f
0
/NH)
2
, n = 0, 1, 2. . . (1.83)
and, of course, the horizontal wavenumbers k and l are also quantized in a nite domain.
The dynamics of the barotropic mode are independent of height and independent of the
stratication of the basic state, and so these Rossby waves are identical with the Rossby
waves in a homogeneous uid contained between two at rigid surfaces. The structure
of the baroclinic modes, which in general depends on the structure of the stratication,
becomes increasingly complex as the vertical wavenumber n increases. This increasing
complexity naturally leads to a certain delicacy, making it rare that they can be unam-
biguously identied in nature. The eigenproblem for a realistic atmospheric prole is
further complicated because of the lack of a rigid lid at the top of the atmosphere.
6
24 Chapter 1. Wave Fundamentals and Rossby Waves
Rossby Waves
Rossby waves are waves that owe their existence to a gradient of potential vorticity
in the uid. If a uid parcel is displaced, it conserves its potential vorticity and so its
relative vorticity will in general change. The relative vorticity creates a velocity eld
that displaces neighbouring parcels, whose relative vorticity changes and so on.
A common source of a potential vorticity gradient is differential rotation, or the -
effect. In the presence of non-zero the ambient potential vorticity increases north-
ward and the phase of the Rossby waves propagates westward. In general, Rossby
waves propagate pseudo-westwards, meaning to the left of the direction of the poten-
tial vorticity gradient.
In the simple case of a single layer of uid with no mean ow the equation of motion
is

t
(
2
+k
2
d
)+

x
= 0 (RW.1)
with dispersion relation
=
k
k
2
+l
2
+k
2
d
. (RW.2)
The phase speed is always negative, or westward. If there is no mean ow, the com-
ponents of the group velocity are given by
c
x
g
=
(k
2
l
2
k
2
d
)
(k
2
+l
2
+k
2
d
)
2
, c
y
g
=
2kl
(k
2
+l
2
+k
2
d
)
2
, (1.3a,b)
The group velocity is westward if the zonal wavenumber is sufciently small, and
eastward if the zonal wavenumber is sufciently large.
The reection of such Rossby waves at a wall is (like light) specular, meaning that
the group velocity of the reected wave makes the same angle with the wall as the
group velocity of the incident wave. The energy ux of the reected wave is equal
and opposite to that of the incoming wave in the direction normal to the wall.
Rossby waves exist in stratied uids, and have a similar dispersion relation to (RW.3)
with an appropriate redenition of the inverse deformation radius, k
d
.
1.6 ENERGY FLUX OF ROSSBY WAVES
We now consider how energy is uxed in Rossby waves. To keep matters reasonably
simple from an algebraic point of view we will consider waves in a single layer and
without a mean ow, but we will allow for a nite radius of deformation. To remind
ourselves, the dynamics are governed by the evolution of potential vorticity equation
1.6 Energy Flux of Rossby Waves 25
and the linearized evolution equation is

2
k
2
d

x
= 0. (1.84)
The dispersion relation follows in the usual way and is
=
k
K
2
+k
2
d
. (1.85)
which is a simplication of (3.59), and the group velocities are
c
x
g
=
(k
2
l
2
k
2
d
)
(K
2
+k
2
d
)
2
, c
y
g
=
2kl
(K
2
+k
2
d
)
2
, (1.86a,b)
which are simplications of (3.60), and as usual K
2
= k
2
+l
2
.
To obtain an energy equation multiply (3.84) by to obtain, after a couple of lines
of algebra,
1
2

()
2
+k
2
d

t
+i
2

= 0. (1.87)
where i is the unit vector in the x direction. The rst group of terms are the energy itself,
or more strictly the energy density. (An energy density is an energy per unit mass or per
unit volume, depending on the context.) The term ()
2
/2 = (u
2
+v
2
)/2 is the kinetic
energy and k
2
d

2
/2 is the potential energy, proportional to the displacement of the free
surface, squared. The second term is the energy ux, so that we may write
E
t
+ F = 0. (1.88)
where E = ()
2
/2 +k
2
d

2
and F =

/t +i
2

. We havent yet used the fact

that the disturbance has a dispersion relation, and if we do so we may expect, following
the derivations of section 3.2, that the energy moves at the group velocity. Let us now
demonstrate this explicitly.
We assume solution of the form
= A(x) cos(k x t) = A(x) cos (kx +ly t) (1.89)
where A(x) is assumed to vary slowly compared to the nearly plane wave. (Note that k
is the wave vector, to be distinguished from k, the unit vector in the z-direction.) The
kinetic energy in a wave is given by
KE =
A
2
2

2
x
+
2
y

(1.90)
so that, averaged over a wave period,
KE =
A
2
2
(k
2
+l
2
)

2

2/
0
sin
2
(k x t) dt. (1.91)
The time-averaging produces a factor of one half, and applying a similar procedure1 to
the potential energy we obtain
KE =
A
2
4
(k
2
+l
2
), PE =
A
2
4
k
2
d
, (1.92)
26 Chapter 1. Wave Fundamentals and Rossby Waves
so that the average total energy is
E =
A
2
4
(K
2
+k
2
d
), (1.93)
where K
2
= k
2
+l
2
.
The ux, F, is given by
F =

t
+i
2

= A
2
cos
2
(k x t)

ki

, (1.94)
so that evidently the energy ux has a component in the direction of the wavevector,
k, and a component in the x-direction. Averaging over a wave period straightforwardly
gives us additional factors of one half:
F =
A
2
2

k+i

. (1.95)
We now use the dispersion relation = k/(K
2
+k
2
d
) to eliminate the frequency, giving
F =
A
2

k
k
K
2
+k
2
d
i
1
2

, (1.96)
and writing this in component form we obtain
F = i
A
2

k
2
l
2
k
2
d
K
2
+k
2
d

+j
2kl
K
2
+k
2
d
. (1.97)
Comparison of (3.97) with (3.86) and (3.93) reveals that
F = c
g
E (1.98)
so that the energy propagation equation, (3.88), when averaged over a wave, becomes
E
t
+ c
g
E = 0 . (1.99)
It is interesting that the variation of A plays no role in the above manipulations, so
that the derivation appears to go through if the amplitude A(x, t) is in fact a constant and
the wave is a single plane wave. This seems to y in the face of our previous discussion,
in which we noted that the group velocity was the velocity of a wave packet or at least of
a superposition of plane waves. Indeed, the derivative of the frequency with respect to
wavenumber means little if there is only one wavenumber. In fact there is nothing wrong
with the above derivation if A is a constant, and the resolution of the paradox arises by
noting that a plane wave lls all of space and time. In this case there is no convergence
of the energy ux and the energy propagation equation is trivially true.
1.6.1

Rossby wave reection
We now consider how Rossby waves might be reected from a solid boundary. The topic
has particular oceanographic relevance, for the reection of Rossby waves turns out to
one way of interpreting why intense oceanic boundary currents formon the western sides
of ocean basins, not the east. As a preliminary, let us give a useful graphic interpretation
of Rossby wave propagation.
7
1.6 Energy Flux of Rossby Waves 27
l
k
k
c
=

2
+

2
k
2
d

1/2
k
c
c
g
C
W k
O
C =

2
, 0

|

WC| =

2
k
2
d

1/2
Fig. 1.8 The energy propagation diagram for Rossby waves. The wavevectors of a
given frequency all lie in a circle of radius [(/2)
2
k
2
d
]
1/2
, centered at the point C.
The closest distance of the circle to the origin is k
c
, and if the deformation radius is
innite k
c
the circle touches the origin. For a given wavenumber k, the group velocity
is along the line directed from W to C.
The energy propagation diagram
The dispersion relation for Rossby waves, = k/(k
2
+l
2
+k
2
d
), may be rewritten as
(k +/2)
2
+l
2
= (/2)
2
k
2
d
. (1.100)
This equation is the parametric representation of a circle, meaning that the wavevector
(k, l) must lie on a circle centered at the point (/2, 0) and with radius [(/2)
2

k
2
d
]
1/2
, as illustrated in Fig. 3.8. If the deformation radius is zero the circle touches the
origin, and if it is nonzero the distance of the closest point to the circle, k
c
say, is given
by k
c
= /2+[(/2)
2
k
2
d
]
1/2
. For low frequencies, specically if /2k, then
k
c
k
2
d
/. The radius of the circle is a positive real number only when < /2k
d
.
This is the maximum frequency possible, and it occurs when l = 0 and k = k
d
and when
c
x
g
= c
y
g
= 0.
It turns out that the group velocity, and hence the energy ux, can be visualized
graphically from Fig. 3.8. By direct manipulation of the expressions for group velocity
and frequency we nd that
c
x
g
=
2

K
2
+k
2
d

k +

2

, (1.101a)
c
y
g
=
2

K
2
+k
2
d

2
l. (1.101b)
(To check this, it is easiest to begin with the right-hand sides and use the dispersion
relation for .) Now, since the center of the circle of wavevectors is at the position
28 Chapter 1. Wave Fundamentals and Rossby Waves
Fig. 1.9 The reection of a Rossby wave at a
western wall, in physical space. A Rossby wave
with a westward group velocity impinges at an
angle
i
to a wall, inducing a reected wave
moving eastward at an angle
r
. The reection
is specular, with
r
=
i
, and energy conserv-
ing, with |c
gr
| = |c
gi
| see text and Fig. 3.10.

c
gi
c
gr
x
y
(/2, 0), and referring to Fig. 3.8, we have that
c
g
=
2
(K
2
+k
2
d
)
2
R (1.102)
where R =

WC is the vector directed from W to C, that is from the end of the wavevector
itself to the center of the circle around which all the wavevectors lie.
Eq. (3.102) and Fig. 3.8 allow for a useful visualization of the energy and phase.
The phase propagates in the direction of the wave vector, and for Rossby waves this is
always westward. The group velocity is in the direction of the wave vector to the center
of the circle, and this can be either eastward (if k
2
> l
2
+k
2
d
) or westward (k
2
< l
2
+k
2
d
).
Interestingly, the velocity vector is normal to the wave vector. To see this, consider a
purely westward propagating wave for which l = 0. Then v = /x = ik ` and u =
/y = il ` = 0. We now see how some of these properties can help us understand
the reection of Rossby waves.
[Do we need a gray box summarizing some of the properties of reection? xxx]
Reection at a wall
Consider Rossby waves incident on wall making an angle with the x-axis, and suppose
that somehow these waves are reected back into the uid interior. This is a reasonable
expectation, for the wall cannot normally simply absorb all the wave energy. We rst note
a couple of general properties about reection, namely that the incident and reected
wave will have the same wavenumber component along the wall and their frequencies
must be the same To see these properties, consider the case in which the wall is oriented
meridionally, along the y-axis with = 90. There is no loss of generality in this choice,
because we may simply choose coordinates so that y is parallel to the wall and the -
effect, which differentiates x from y, not enter the argument. The incident and reected
waves are

i
(x, y, t) = A
i
e
i(k
i
x+l
i
y
i
t)
,
r
(x, y, t) = A
r
e
i(k
r
x+l
r
y
r
t)
, (1.103)
1.6 Energy Flux of Rossby Waves 29
with subscripts i and r denoting incident and reected. At the wall, which we take to be
at x = 0, the normal velocity u = /y must be zero so that
A
i
l
i
e
i(l
i
y
i
t)
+A
r
l
r
e
i(l
r
y
r
t)
= 0. (1.104)
For this equation to hold for all y and all time then we must have
l
r
= l
i
,
r
=
i
. (1.105)
This result is independent of the detailed dynamics of the waves, requiring only that the
velocity is determined from a streamfunction. When we consider Rossby-wave dynamics
specically, the x- and y-coordinates are not arbitrary and so the wall cannot be taken
to be aligned with the y-axis; rather, the result means that the projection of the incident
wavevector, k
i
on the wall must equal the projection of the reected wavevector, k
r
. The
magnitude of the wavevector (the wavenumber) is not in general conserved by reection.
Finally, given these results and using (3.104) we see that the incident and reected
amplitudes are related by
A
r
= A
i
. (1.106)
Now lets delve a little deeper into the wave-reection properties.
Generally, when we consider a wave to be incident on a wall, we are supposing that
the group velocity is directed toward the wall. Suppose that a wave of given frequency,
, and wavevector, k
i
, and with westward group velocity is incident on a predominantly
western wall, as in Fig. 3.9. (Similar reasoning, mutatis mutandis, can be applied to a
wave incident on an eastern wall.) Let us suppose that incident wave, k
i
lies at the point
I on the wavenumber circle, and the group velocity is found by drawing a line from I to
the center of the circle, C (so c
gi


IC), and in this case the vector is directed westward.
The projection of the k
i
must be equal to the projection of the reected wave vector,
k
r
, and both wavevectors must lie in the same wavenumber circle, centered at /2,
because the frequencies of the two waves are the same. We may then graphically deter-
mines the wavevector of the reected wave using the construction of Fig. 3.10. Given
the wavevector, the group velocity of the reected wave follows by drawing a line from
the wavevector to the center of the circle (the line

RC). We see from the gure that
the reected group velocity is directed eastward and that it forms the same angle to the
wall as does the incident wave; that is, the reection is specular. Since the amplitude
of the incoming and reected wave are the same, the components of the energy ux
perpendicular to the wall are equal and opposite. Furthermore, we can see from the
gure that the wavenumber of the reected wave has a larger magnitude than that of
the incident wave. For waves reecting of an easter boundary, the reverse is true. Put
simply, at a western boundary incident long waves are reected as short waves, whereas
at an eastern boundary incident short waves are reected as long waves.
Quantitatively solving for the wavenumbers of the reected wave is a little tedious in
the case when the wall is at angle, but easy enough if the wall is a meridional, along the
y-axis. We know the frequency, , and the y-wavenumber, l, so that the x-wavenumber
is may be deduced from the dispersion relation
=
k
i
k
2
i
+l
2
+k
2
d
=
k
r
k
2
r
+l
2
+k
2
d
. (1.107)
30 Chapter 1. Wave Fundamentals and Rossby Waves
l
k C
O
k
i
k
r
I
R
c
gi
c
gr

x
y
Fig. 1.10 Graphical representation of the reection of a Rossby wave at a western
wall, in spectral space. The incident wave has wavevector k
i
, ending at point I. Con-
struct the wavevector circle through point I with radius

(/2)
2
k
2
d
and center
C = (/2, 0); the group velocity vector then lies along

IC and is directed west-
ward. The reected wave has a wavevector k
r
such that its projection on the wall is
equal to that of k
i
, and this xes the point R. The group velocity of the reected wave
then lies along

RC, and it can be seen that c
gr
makes the same angle to the wall as
does c
gi
, except that it is directed eastward. The reection is therefore both specular
and is such that the energy ux directed away from the wall is equal to the energy
ux directed toward the wall.
We obtain
k
i
=

2
+

l
2
+k
2
d

, k
r
=

2

l
2
+k
2
d

. (1.108a,b)
The signs of the square-root terms are chosen for reection at a western boundary, for
which, as we noted, the reected wave has a larger (absolute) wavenumber than the
incident wave. For reection at an eastern boundary, we simply reverse the signs.
Oceanographic relevance
The behaviour of Rossby waves at lateral boundaries is not surprisingly of some oceano-
graphic importance, there being two particularly important examples. One of them con-
cerns the equatorial ocean, and the other the formation of western boundary currents,
common in midlatitudes. We only touch on these topics here, deferring a more extensive
treatments to later chapters.
Suppose that Rossby waves are generated in the middle of the ocean, for example
by the wind or possibly by some uid dynamical instability in the ocean. Shorter waves
1.7 Rossby-gravity Waves: an Introduction 31
will tend to propagate eastward, and be reected back at the eastern boundary as long
waves, and long waves will tend to propagate westward, being reected back as short
waves. The reection at the western boundary is believed to be particularly important
in the dynamics of El Nio. although the situation is further complicated because the
reection may also generate eastward moving equatorial Kelvin waves, which we discuss
more in the next chapter.
In mid-latitudes the reection at a western boundary generates Rossby waves that
have a short zonal length scale (the meridional scale is the same as the incident wave
if the wall is meridional), which means that their meridional velocity is large. Now, if
the zonal wavenumber is much larger than both the meridional wavenumber l and the
inverse deformation radius k
d
then, using either (3.57) or (??) the group velocity in the
x-direction is given by cg
x
= U +/k
2
, where U is the zonal mean ow. If the mean ow
is westward, so that U is negative, then very short waves will be unable to escape from
the boundary; specically, if k >

/ U then the waves will be trapped in a western
boundary layer. [More here ?? xxx]
1.7 ROSSBY-GRAVITY WAVES: AN INTRODUCTION
We now consider Rossby waves and shallow water gravity waves together, keeping f
constant except where differentiated, following Hendershott (5.11). The equations of
motion are the shallow water equations in Cartesian coordinates in a rotating frame of
reference, namely
u
t
f v = g

x
,
v
t
+f u = g

y
, (1.109a,b)

t
+c
2

u
x
+
v
y

= 0 (1.109c)
where, in terms of possibly more familiar shallow water variables, = g

and c
2
= g

H,
where is the kinematic pressure, c will be a wave speed, is the free surface height, H
is the reference depth of the uid and g

is the reduced gravity.

After some manipulation (described more fully in section ??) we may obtain, without
additional approximation, a single equation for v, namely
1
c
2

3
v
t
3
+
f
2
c
2
v
t


t

2
v
v
x
= 0. (1.110)
In this equation the Coriolis parameter is given by the -plane expression f = f
0
+ y;
thus, the equation has a non-constant coefcient, entailing considerable algebraic dif-
culties. We will address some of these difculties in chapter ??, but for now we take a
simpler approach: we assume that f is constant except where differentiated, an approx-
imation that is reasonable in mid-latitudes provided we are concerned with sufciently
small variations in latitude. Equation (3.110) then has constant coefcients and we may
look for plane wave solutions of the form v = ` v exp[i(k x t)], whence

2
f
2
0
c
2
(k
2
+l
2
)
k

= 0. (1.111a)
32 Chapter 1. Wave Fundamentals and Rossby Waves
Rossby waves
l
k
k
O
C =

2
, 0

C
k
c
R
R =

2
+

2
f
2
0
c
2

1/2

f
2
0
c
2

1/2

k
c
=

2
+R
f
2
0
c
2
l
k
k
O
C
R
R =

2
+

2
f
2
0
c
2

1/2

2
f
2
0
c
2

1/2

C =

2
, 0

k
c
=

2
+R R
Gravity waves
Fig. 1.11 Wave propagation diagrams for Rossby-gravity waves, obtained using
(3.111). The top gure shows the diagram in the low frequency, Rossby wave limit,
and the bottom gure shows the high frequency, gravity wave limit. In each case the
the locus of wavenumbers for a given frequency is a circle centered at C = (/2, 0)
with a radius R given by (3.117), but the approximate expressions dier signicantly
at high and low frequency.
1.7 Rossby-gravity Waves: an Introduction 33
Gravity waves
l = 0
l = 1
l = 2

Wavenumber, kc/f
0
F
r
e
q
u
e
n
c
y
,

/
f
0
Planetary waves, l = 0, 1, 2
Fig. 1.12 Dispersion relation for Rossby-gravity waves, obtained from (3.123) with

= 0.2 for three values of l. There a frequency gap between the Rossby or planetary
waves and the gravity waves. For the stratied mid-latitude atmosphere or ocean the
frequency gap is in fact much larger.
or, written differently,

k +

2

2
+l
2
=

2
+

2
f
2
0
c
2
. (1.111b)
This equation may be compared to (3.100): noting that k
2
d
= f
2
0
/g

H = f
2
0
/c
2
, the two
equations are identical except for the appearance of a term involving the frequency on
the right-hand side in (3.117). The wave propagation diagram is illustrated in Fig. ??.
The wave vectors a a given frequency all lie on a circle centered at (/2, 0) and with
radius r given by
R =

2
+

2
f
2
0
c
2

1/2
, (1.112)
and the radius must be positive in order for the waves to exist. The wave propagation
diagram is illustrated in Fig. 3.11. In the low frequency case the diagram is essentially
the same as that shown in Fig. 3.8, but is quantitatively signicantly different in the high
frequency case. These limiting cases are discussed further in section 3.7.1 below.
To plot the full dispersion relation it is useful to nondimensionalize using the follow-
ing scales for time (T), distance (L) and velocity (U)
T = f
1
0
, L = L
d
= k
1
d
= c/f
0
, U = L/T = c, (1.113a,b)
34 Chapter 1. Wave Fundamentals and Rossby Waves
so that, denoting nondimensional quantities with a hat,
= f
0
, (k, l) = (

k,

l)k
d
, =

f
2
0
c
=

f
0
L
d
=

f
0
k
d
. (1.114)
The dispersion relation (3.111) may then be written as

2
1 (

k
2
+

l
2
)

k

= 0 (1.115)
This is a cubic equation in , as might be expected given the governing equations
(3.109). We may expect that two of the roots correspond to gravity waves and the third
to Rossby waves. The only parameter in the dispersion relation is

= c/f
2
0
= L
d
/f
0
.
In the atmosphere a representative value for L
d
is 1000 km, whence

= 0.1. In the
ocean L
d
100km, whence

= 0.01. If we allow ourselves to consider external Rossby
waves (which are of some oceanographic relevance) then c =

gH = 200ms
1
and
L
d
= 2000km, whence

= 0.2.
To actually obtain a solution we regard the equation as a quadratic in k and solve in
terms of the frequency, giving

k =

2

1
2

2

2
+4(
2

l
2
1)

1/2
. (1.116)
The solutions are plotted in Fig. 3.12, with

= 0.2, and we see that the waves fall into
two groups, labelled gravity waves and planetary waves in the gure. The gap between
the two groups of waves is in fact far larger if a smaller (and generally more relevant)
value of

is used. To interpret all this let us consider some limiting cases.
1.7.1 Wave properties
We now consider a few special cases of the dispersion relation.
(i) Constant Coriolis parameter
If = 0 then the dispersion relation becomes

2
f
2
0
(k
2
+l
2
)c
2

= 0, (1.117)
with the roots = 0 and
2
= f
2
0
+ c
2
(k
2
+ l
2
). The root = 0 corresponds to
geostrophic motion (and, since = 0, Rossby waves are absent), with the other
root corresponding to Poincar waves, considered at length in chapter ??, for which

2
> f
2
0
.
(ii) High frequency waves
If we take the limit of f
0
then (3.111a) gives

2
c
2
(k
2
+l
2
)
k

= 0. (1.118)
To be physically realistic we should also now eliminate the term, because if f
0
then, from geometric considerations on a sphere, k
2
k/. Thus, the dispersion
relation is simply
2
= c
2
(k
2
+l
2
). These waves are just gravity waves uninuenced
by rotation, and are a special case of Poincar waves.
1.7 Rossby-gravity Waves: an Introduction 35
Rossby-Gravity Waves
Generically speaking, Rossby-gravity waves are waves that arise under the com-
bined effects of a potential vorticity gradient and stratication. The simplest set-
ting in which they occur is in the linearized shallow water equations which may be
written as a single equation for v, namely
1
c
2

3
v
t
3
+
f
2
c
2
v
t


t

2
v
v
x
= 0. (RG.1)
If we take both f and to be constants then the equation above admits of plane-
wave solutions with dispersion relation

kc
2

= f
2
0
+c
2
(k
2
+l
2
). (RG.2)
In Earths atmosphere and ocean is common for there to be a frequency separation
between the two classes of solution. To a good approximation, high frequency,
gravity waves then satisfy

2
= f
2
0
+c
2
(k
2
+l
2
), (RG.3)
and low frequency, Rossby waves satisfy
=
kc
2
f
2
0
+c
2
(k
2
+l
2
)
=
k
k
2
d
+k
2
+l
2
(RG.4)
where k
2
d
= f
2
0
/c
2
.
Rossby-gravity waves also exist in the stratied equations. Solutions may be found
be decomposing the vertical structure into a series of orthogonal modes, and a
sequence of shallow water equations for each mode results, with a different c for
each mode. Solutions may also be found if f is allowed to vary in (RG.1), at the
price algebraic complexity, as discussed in chapter ??.
(iii) Low frequency waves
Consider the limit of f
0
. The dispersion relation reduces to
=
k
k
2
+l
2
+k
2
d
. (1.119)
This is just the dispersion relation for quasi-geostrophic waves as previously ob-
tained see (3.59) or (3.85) In this limit, the requirement that the radius of the
circle be positive becomes

2
<

2
4k
2
d
. (1.120)
That is to say, the Rossby waves have a maximum frequency, and directly from
36 Chapter 1. Wave Fundamentals and Rossby Waves
Fig. 1.13 The locus of points on
planetary-geostrophic Rossby waves.
Waves of a given frequency all have the
same x-wavenumber, given by xxx
l
k
O
k
1
k
2
k
3
Locus of wavenumbers of
given frequency
k =
f
2
0
c
2

(3.120) this occurs when k = k

d
and l = 0.
The frequency gap
The maximum frequency of Rossby waves is usually much less than the frequency of
the Poincar waves: the lowest frequency of the Poincare waves is f
0
and the highest
frequency of the Rossby waves is /2k
d
. Thus,
Low gravity wave frequency
High Rossby wave frequency
=
f
0
/2k
d
=
f
2
0
2c
. (1.121)
If f
0
= 10
4
s
1
, = 10
11
m
1
s
1
and k
d
= 1/100km
1
(a representative oceanic
baroclinic deformation radius) then f
0
/(/2k
d
) = 200. If L
d
= 1000km (an atmospheric
baroclinic radius) then the ratio is 20. If use a barotropic deformation radius of L
d
=
2000km then the ratio is 10. Evidently, for most midlatitude applications there is a large
gap between the Rossby wave frequency and the gravity wave frequency.
Because of this frequency gap, to a good approximation Fig. 3.12 may be obtained by
separately plotting (3.117) for the gravity waves, and (3.120) for the Rossby or planetary
waves. The differences between these and the exact results become smaller as

gets
smaller, and are less than the thickness of a line on the plot shown.
1.7.2 Planetary geostrophic Rossby waves
A good approximation for the large-scale ocean circulation involves ignoring the time-
derivatives and nonlinear terms in the momentum equation, allowing evolution only to
occur in the thermodynamic equation. This is the planetary-geostrophic approximation,
introduced in section ?? ??, and it is interesting to see to what extent that system sup-
ports Rossby waves.
8
It is easiest just to begin with the linear shallow water equations
1.7 Rossby-gravity Waves: an Introduction 37
themselves, and omitting time derivatives in the momentum equation gives
f v =

x
, f u =

y
, (1.122a,b)

t
+c
2

u
x
+
v
y

= 0. (1.122c)
From these equations we straightforwardly obtain

t

c
2

f
2

x
= 0. (1.123)
Again we will treat both f and as constants so that we may look for solutions in the
form =
`
exp[i(k x t)]. The ensuing dispersion relation is
=
c
2

f
2
0
k =
k
k
2
d
(1.124)
which is a limiting case of (3.120) with k
2
, l
2
k
2
d
. The waves are a form of Rossby
waves with phase and group speeds given by
c
p
=
c
2

f
2
0
, c
x
g
=
c
2

f
2
0
. (1.125)
That is, the waves are non-dispersive and propagate westward. Eq. (3.123) has the
general solution = G(x + c
2
/f
2
t), where G is any function, so an initial disturbance
will just propagate westward at a speed given by (3.125), without any change in form.
Note nally that the locus of wavenumbers in kl space is no longer a circle, as it
is for the usual Rossby waves. Rather, since the frequency does not depend on the y-
wavenumber, the locus is a straight line, parallel to the y-axis, as in Fig. 3.13. Waves of a
given frequency all have the same x-wavenumber, given by k = f
2
0
/(c
2
) = k
2
d
/,
as shown in Fig. 3.13.
Physical mechanism
Because the waves are a form of Rossby wave their physical mechanism is related to
that discussed in section 3.4.2, but with an important difference: relative vorticity is no
longer important, but the ow divergence is. Thus, consider ow round a region of high
pressure, as illustrated in Fig. 3.14. If the pressure is circularly symmetric as shown, the
ow to the south of H in the left-hand sketch, and to the south of L in the right-hand
sketch, is larger than that to the north. Hence, in the left sketch the ow converges at
W and diverges at E, and the ow pattern moves westward. In the ow depicted in the
right sketch the low pressure propagates westward in a similar fashion.
38 Chapter 1. Wave Fundamentals and Rossby Waves
H L W E W E
convergence
at W
divergence
at E
divergence
at W
convergence
at E
N
S
W E
Fig. 1.14 The westward propagation of planetary-geostrophic Rossby waves. The cir-
cular lines are isobars centered around high and low pressure centres. Because of the
variation of the Coriolis force, the mass ux between two isobars is greater to the
south of a pressure center than it is to the north. Hence, in the left-hand sketch there
is convergence to the west of the high pressure and the pattern propagates westward.
Similarly, if the pressure centre is a low, as in the right-hand sketch, there is diver-
gence to the west of the pressure centre and the pattern still propagates westward.
1.8

THE GROUP VELOCITY PROPERTY
In the last section of this chapter we return to a more general discussion of group velocity.
Our goal is to show that the group velocity arises in fairly general ways, not just from
methods stemming from Fourier analysis or from ray theory. In a purely logical sense this
discussion does follow most naturally from the end of the section on ray theory (section
3.3), but for most humans it is helpful to have had a concrete introduction to at least one
nontrivial formof waves before considering more abstract material. We rst give a simple
and direct derivation of group velocity that valid in the simple but important special case
of a homogeneous medium.,
9
Then, in section 3.8, we give a rather general derivation
of the group velocity property, namely that conserved quantities that are quadratic in the
wave amplitude quantities known as wave activities are transported at the group
velocity.
1.8.1 Group velocity in homogeneous media
Consider,
10
waves propagating in a homogeneous medium in which the wave equation
is a polynomial of the general form
L() =

t
,

x

(x, t) = 0. (1.126)
where is a polynomial operator in the space and time derivatives. For algebraic sim-
plicity we again restrict attention to waves in one dimension, and a simple example is
L() = (
xx
)/t + /x or = (
xx
)/t + /x. We will seek a solution of the
form [c.f., (3.4)]
(x, t) = A(x, t) e
i(x,t)
, (1.127)
1.8

The Group Velocity Property 39
where is the phase of the disturbance and A(x, t) is the slowly varying amplitude, so
that the solution has the form of a wave packet. The phase is such that k = /x
and = /t , and the slowly varying nature of the envelope A(x, t) is formalized by
demanding that
1
A
A
x
k,
1
A
A
t
, (1.128)
The space and time derivatives of are then given by

x
=

A
x
+iA

e
i
=

A
x
+iAk

e
i
, (1.129a)

t
=

A
t
+iA

e
i
=

A
t
iA

e
i
, (1.129b)
so that the wave equation becomes
=

t
i,

x
+ik

A = 0. (1.130)
Noting that the space and time derivative of A are small compared to k and we
expand the polynomial in a Taylor series about (, k) to obtain
(i, ik)A+

(i)
A
t
+

(ik)
A
x
= 0. (1.131)
The rst terms is nothing but the linear dispersion relation; that is (i, ik)A = 0 is the
dispersion relation for plane waves. Taking this to be satised, (3.131) gives
A
t

/k
/
A
x
=
A
t
+

k
A
x
= 0. (1.132)
That is, the envelope moves at the group velocity /k.
1.8.2

Group velocity property: a general derivation
In our discussion of Rossby waves in section 3.6 in (3.99) we showed that the energy of
the waves is conserved in the sense that
E
t
+ F = 0, (1.133)
where E is the energy density of the waves and F is its ux. In (3.99) we further showed
that, when averaged over a wavelength and a period, the average ux was related to the
energy by F = c
g
E. This property is called the group velocity property and it is a very
general property, not restricted to Rossby waves or even to energy. Rather, it is a property
of almost any conserved quantity that is quadratic in the wave amplitude, which is he
dening property of a wave activity, and we now demonstrate this in a rather general
way.
11
It is a useful property, because if we can use observations to deduce c
g
then we
can determine how wave activity density propagates.
Our derivation holds generally for waves and wave activities that satisfy the following
three assumptions.
40 Chapter 1. Wave Fundamentals and Rossby Waves
(i) The wave activity, A, and ux, F, obey the general conservation relation
A
t
+ F = 0. (1.134)
(ii) Both the wave activity and the ux are quadratic functions of the wave amplitude.
(iii) The waves themselves are of the general form
= ` e
i(x,t)
+c.c., = k x t, = (k), (1.135a,b,c)
where (3.135c) is the dispersion relation, and is any wave eld. We will carry out
the derivation in case in which ` is a constant, but the derivation may be extended
to the case in which it varies slowly over a wavelength.
Given assumption (ii), the wave activity must have the general form
A = b +ae
2i(kxt)
+a

e
2i(kxt)
, (1.136a)
where the asterisk,

, denotes complex conjugacy, and b is a real constant and a is a
complex constant. For example, suppose that A =
2
and = c e
i(kxt)
+c

e
i(kxt)
,
then we nd that (3.136a) is satised with a = c
2
and b = 2cc

. Similarly, the ux has

the general form
F = g +f e
2i(kxt)
+f

e
2i(kxt)
. (1.136b)
where g is a real constant vector and f is a complex constant vector. The mean activity
and mean ux are obtained by averaging over a cycle; the oscillating terms vanish on
integration and therefore the wave activity and ux are given by
A = b, F = g, (1.137)
where the overbar denotes the mean.
Now formally consider a wave with a slightly different phase, + i , where is
small compared with . Thus, we formally replace k by k+i k and by +i where,
to satisfy the dispersion relation, we have
+i = (k +i k) (k) +i k

k
, (1.138)
and therefore
= k

k
= k c
g
, (1.139)
where c
g
/k is the group velocity.
The new wave has the general form

= ( ` + ` ) e
i(kxt)
e
kx+t
+c.c., (1.140)
and, analogously to (3.136), the associated wave activity and ux have the forms:
A

b +b +(a +a) e
2i(kxt)
+(a

+a

) e
2i(kxt)

e
2kx+2t
(1.141a)
F

g +g +(f +f ) e
2i(kxt)
+(f

+f

) e
2i(kxt)

e
2kx+2t
, (1.141b)
1.8

The Group Velocity Property 41
where the quantities are small. If we now demand that A

and F

satisfy assumption
(i), then substituting (3.141) into (3.134) gives, after a little algebra,
(g +g) k = (b +b) (1.142)
and therefore at rst order in quantities, g k = b. Using (3.139) and (3.137) we
obtain
c
g
=
g
b
=
F
A
, (1.143)
and using this the conservation law, (3.134), becomes
A
t
+ (c
g
A) = 0. (1.144)
Thus, for waves satisfying our three assumptions, the ux velocity that is, the propa-
gation velocity of the wave activity is equal to the group velocity.
1.8.3 Group velocity property for Rossby waves
[This subsection may be shifted to a later chapter after we have discussed wave activity
more, and in particular EliassenPalm uxes.]
We now show explicitly that the wave activity for Rossby waves satises the group
velocity property. The Boussinesq quasi-geostrophic equation on the -plane, linearized
around a uniform zonal ow and with constant static stability, is
q

t
+u
q

x
+v

q
y
= 0, (1.145)
where q

= [
2
+(f
2
0
/N
2
)
2
/z
2
]

and, if u is constant, q/y = . Thus we have

t
+u

x

+

z

f
2
0
N
2

x
= 0. (1.146)
Seeking solutions of the form

= Re ` e
i(kx+ly+mzt)
, (1.147)
we nd the dispersion relation,
= uk
k

2
. (1.148)
where
2
= (k
2
+l
2
+m
2
f
2
0
/N
2
), and the group velocity components:
c
y
g
=
2kl

4
, c
z
g
=
2kmf
2
0
/N
2

4
. (1.149)
Also, if u

= Re ` uexp[i(kx +ly +mz t)], and similarly for the other elds, then
` u = Re il ` , ` v = Re ik ` ,
`
b = Reimf
0
` , ` q = Re
2
` ,
(1.150)
42 Chapter 1. Wave Fundamentals and Rossby Waves
The wave activity density is then
A=
1
2
q
2

=

4
4
| `
2
|, (1.151)
where the additional factor of 2 in the denominator arises from the averaging. Using
(3.150) the EP ux, (??), is
F
y
= u

=
1
2
kl| `
2
|, F
z
=
f
0
N
2
v

=
f
2
0
2N
2
km| `
2
|.
(1.152)
Using (3.149), (3.151) and (3.152) we obtain
F = (F
y
, F
z
) = c
g
A . (1.153)
If the properties of the medium are slowly varying, so that a (spatially varying) group
velocity can still be dened, then this is a useful expression to estimate how the wave
activity propagates in the atmosphere and in numerical simulations.
Notes
1 For a review of waves and group velocity, see Lighthill (1965). Throughout this chapter we
also draw on unpublished lecture notes by D. Chapman and P. Malanotte-Rizzoli based on a
lecture course by M. Hendershott (Chapman et al. 1989).
2 See, for example, the book Nonlinear Waves by G. B. Whitham.
3 Another useful reference for ray theory is Lighthill (1978).
4 Rossby waves were probably rst discovered in a theoretical context by ??. He considered
the linear shallow water equations on a sphere, expanding the solution in powers of the sine
of latitude, so obtaining both long gravity waves and Rossby waves. However, the discovery
did not garner very much attention in the meteorological or oceanographic community until
the discovery was reprised by Rossby (1939). Rossby used the beta-plane approximation in
Cartesian co-ordinates, and the simplicity of the presentation along with the meteorological
context now attracted signicant notice.
5 This non-Doppler eect also arises quite generally, even in models in height coordinates. See
White (1977) and problem ??.5.
6 See Chapman & Lindzen (1970).
7 Following Longuet-Higgins (1964).
8 The ensuing waves seem to have been rst noted by ?.
9 Following Pedlosky (2003).
10 Following Pedlosky (2003).
11 The formof this derivation was originally given by Hayes (1977) in the context of wave energy.
See Vanneste & Shepherd (1998) for extensions.
Further reading
Majda, A. J., 2003. Introduction to PDEs and Waves for the Atmosphere and Ocean.
Provides a compact, somewhat mathematical introduction to various equation sets and their
properties, including quasi-geostrophy.
Notes and Problems 43
Problems
1.1 Consider the at-bottomed shallow water potential vorticity equation in the form
D
Dt
+f
h
= 0 (P1.1)
(a) Suppose that deviations of the height eld are small compared to the mean height eld,
and that the Rossby number is small (so || f ). Further consider ow on a -plane such
that f = f
0
+y where |y| f
0
. Show that the evolution equation becomes
D
Dt

+y
f
0

= 0 (P1.2)
where h = H + and || H. Using geostrophic balance in the form f
0
u = g/y,
f
0
v = g/x, obtain an expression for in terms of .
(b) Linearize (P3.2) about a state of rest, and show that the resulting system supports two-
dimensional Rossby waves that are similar to those of the usual two-dimensional barotropic
system. Discuss the limits in which the wavelength is much shorter or much longer than
the deformation radius.
(c) Linearize (P3.2) about a geostrophically balanced state that is translating uniformly east-
wards. Note that this means that:
u = U +u

= (y) +

,
where (y) is in geostrophic balance with U. Obtain an expression for the form of (y).
(d) Obtain the dispersion relation for Rossby waves in this system. Show that their speed
is dierent from that obtained by adding a constant U to the speed of Rossby waves in
part (b), and discuss why this should be so. (That is, why is the problem not Galilean
invariant?)
1.2 Obtain solutions to the two-layer Rossby wave problem by seeking solutions of the form

1
= Re `
1
e
i(kxx+ky yt)
,
2
= Re `
2
e
i(kxx+ky yt)
. (P1.3)
Substitute (P3.3) directly into (3.62) to obtain the dispersion relation, and show that the ensu-
ing two roots correspond to the baroclinic and barotropic modes.
1.3 (Not dicult, but messy.) Obtain the vertical normal modes and the dispersion relationship
of the two-layer quasi-geostrophic problem with a free surface, for which the equations of
motion linearized about a state of rest are

1
+F
1
(
2

1
)

1
x
= 0 (P1.4a)

2
+F
2
(
1

2
) F
ext

2
x
= 0, (P1.4b)
where F
ext
= f
0
/(gH
2
).
1.4 Given the baroclinic dispersion relation, = k
x
/(k
x
2
+ k
2
d
), for what value of k
x
is the
x-component of the group velocity the largest (i.e., the most positive), and what is the corre-
sponding value of the group velocity?
1.5 Show that the non-Doppler eect arises using geometric height as the vertical coordinate,
using the modied quasi-geostrophic set of White (1977). In particular, obtain the disper-
sion relation for stratied quasi-geostrophic ow with a resting basic state. Then obtain the
dispersion relation for the equations linearized about a uniformly translating state, paying
attention to the lower boundary condition, and note the conditions under which the waves are
stationary. Discuss.
44 Chapter 1. Wave Fundamentals and Rossby Waves
1.6 (a) Obtain the dispersion relationship for Rossby waves in the single-layer quasi-geostrophic
potential vorticity equation with linear drag.
(b) Obtain the dispersion relation for Rossby waves in the linearized two-layer potential vor-
ticity equation with linear drag in the lowest layer.
(c) Obtain the dispersion relation for Rossby waves in the continuously stratied quasi-
geostrophic equations, with the eects of linear drag appearing in the thermodynamic
equation for the lower boundary condition. That is, the boundary condition at z = 0 is

t
(
z
) +N
2
w = 0 where w = with being a constant. You may make the Boussinesq
approximation and assume N
2
is constant if you like.
98 Chapter 4. Wave Fundamentals and Rossby Waves
References
Chapman, D. C., Malanotte-Rizzoli, P. & Hendershott, M., 1989. Wave motions in the ocean.
Unpublished. Based on lectures by Myrl Hendershott.
Chapman, S. & Lindzen, R. S., 1970. Atmospheric Tides. Gordon and Breach, 200 pp.
Gill, A. E., 1982. AtmosphereOcean Dynamics. Academic Press, 662 pp.
Hayes, M., 1977. A note on group velocity. Proc. Roy. Soc. Lond. A, 354, 533535.
Lighthill, J., 1978. Waves in Fluids. Cambridge University Press, 504 pp.
Lighthill, M. J., 1965. Group velocity. J. Inst. Math. Appl., 1, 128.
Longuet-Higgins, M. S., 1964. Planetary waves on a rotating sphere, I. Proc. Roy. Soc. Lond.
A, 279, 446473.
Majda, A. J., 2003. Introduction to PDEs and Waves for the Atmosphere and Ocean. Amer-
ican Mathematical Society, 234 pp.
Munk, W. H., 1981. Internal waves and small-scale processes. In B. A. Warren & C. Wunsch,
Eds., Evolution of Physical Oceanography, pp. 264291. The MIT Press.
Pedlosky, J., 2003. Waves in the Ocean and Atmosphere: Introduction to Wave Dynamics.
Springer-Verlag, 260 pp.
Rossby, C.-G., 1939. Relations between variation in the intesity of the zonal circulation and
the displacements of the semi-permanent centers of action. J. Marine Res., 2, 3855.
Vanneste, J. & Shepherd, T. G., 1998. On the group-velocity property for wave-activity
conservation laws. J. Atmos. Sci., 55, 10631068.
White, A. A., 1977. Modied quasi-geostrophic equations using geometric height as verti-
cal co-ordinate. Quart. J. Roy. Meteor. Soc., 103, 383396.
99
Index
Bold face denotes a primary entry or an extended discussion.
Dispersion relation, 4, 6
Rossby waves, 17
Energy ux, 24
Rossby waves, 2426
Frequency, 4
Group velocity, 712
Group velocity property, 3842
Kinematics
of waves, 4
Phase speed, 5, 56
Phase velocity, 5
Plane waves, 4
Ray Theory, 15
Ray theory, 13
Rays, 15
Rossby waves, 1631
barotropic, 16
continuously stratied, 21
dispersion relation, 17
energy ux, 2426
nite deformation radius, 18
group velocity property, 41
mechanism of, 18
planetary geostrophic, 36
reection, 26
two layers, 19
Stationary phase, 12
Wave packet, 10
Wavelength, 5
Waves, 4
barotropic Rossby, 16
frequency, 4
group velocity property, 38
kinematics, 4
Rossby, 16
Rossby dispersion relation, 17
Rossby wave mechanism, 18
Rossby, continuously stratied, 21
Rossby, single-layer, 16
Rossby, two-layer, 19
wavevector, 4
Wavevector, 4
100