Chapter 4: Ocean waves

4.1 Wave concepts

When we think about waves usually we have a picture of undulations on the surface of
the sea or lakes. They usually progress from a region of formation to a coast where they are
dissipated or partly reflected (see Fig 1).

Figure 1: Ideal sine or cosine waves.

Assuming that the waves at the sea surface are simple cosine waves.
Wave height (H) and amplitude (A): the vertical distance from trough to crest is called
the “height” of a wave, which is twice of the “amplitude”. The amplitude is the maximum
displacement of the vertical oscillatory motion of the surface above or below the mean water
level.
Wave length (L): The distance from crest to crest or from trough to trough. It is the
distance that a wave travels in one period (T) of time.
Wave number (k): The number of waves that exist for a specified distance. The radian
wave number 2π L
is often used.
Wave period (T): The time between two successive crests or troughs passing a fixed point.
Wave frequency (ω): The number of wave oscillations that take place in a specified time,
say one second. The radian frequency 2π T
is often used.
Wave phase speed (C): The speed at which a particular part of the wave passes a fixed
point, which is C = L/T = ω/k.

i
 Group speed; dispersion: In real ocean, waves are not pure sine waves but are a sum of
sine waves with a range of wavelengths, corresponding periods and amplitudes. The group
speed, Cg , at which the group travels is ∂ω
∂k
. Fig 2 example. [P. & P. Fig 12.5). [see P. & P.
for derivation].
Wave energy travels at the group speed, rather than phase speed. If the group speed
changes with the change of frequency, we say the system is “dispersive”, because the energy
in the system will “disperse” over time. That is, waves at different frequencies travel at
different speed. In contrast, if Cg is the same at all frequencies, the system is called non-
dispersive.
Wave energy density. Total energy per unit area of sea surface is defined as E =
(ρgA2 )/2 = (ρgH 2)/8 (joules m−2 ).
Steepness. The ratio of H/L is called steepness.

Figure 2:

4.2 Surface gravity waves: A homogeneous, inviscid ocean

The equations of motion we derived earlier can help us to understand two important
aspects of the ocean circulation:
(i) the transient response of the ocean to changes in the surface forcing;
(ii) the mean ocean circulation driven by the surface winds and buoyancy effects.
We will discuss both of these aspects in this course, beginning with the transient ocean
response–waves.
We will first discuss how the ocean will respond when there is a change in the forcing
field. To this end a valuable insight into the adjustment of the ocean to changes in forcing
comes from a study of linear ocean waves: Recall from the previous class that the Rossby
and Ekman numbers (Ro and Ex ) are small in the ocean interior, and the interior ocean can
be viewed as “linear” and inviscid. Since Coriolis forcing is very important to large scale
motion, we wish to isolate its effect. One approach to isolate this effect is to “eliminate” it
from the equations first. Then see what it does to the ocean by adding it in.
The class will focus on the concepts and physics rather than detailed math, therefore we

ii
only briefly go through the mathematical procedures during our discussion. First, we will
obtain the dispersion relation, and see what types of waves are available in our simplified
system.
The equations of motion in a homogeneous, inviscid ocean by ignoring the Coriolis force
are:
dU 1 ∂P
=− , (1a)
dt ρ0 ∂x
dV 1 ∂P
=− , (1b)
dt ρ0 ∂y
dW 1 ∂P
=− − g. (1c)
dt ρ0 ∂z
∂U ∂V ∂W
+ + = 0. (1d)
∂x ∂y ∂z
Let’s assume the equilibrium state of the ocean is at the state of rest. That is:
U0 = V0 = W0 = 0,
P0 = P0 (z),
ρ = ρ0 (constant–homogeneous).
For small perturbations u, v, w, p about the resting state, we have:
U = U0 + u = u,
V = V0 + v = v,
W = W0 + w = w,
P = P0 (z) + p,
ρ = ρ0 .
For the background state, the above equations of motion become:
1 ∂P0
− − g = 0. (2)
ρ0 ∂z

For the perturbations, to the first order (ignore the higher order terms), in a non-rotating
frame (f = 0), we have:
∂u 1 ∂p
=− , (3a)
∂t ρ0 ∂x
∂v 1 ∂p
=− , (3b)
∂t ρ0 ∂y
∂w 1 ∂p
=− . (3c)
∂t ρ0 ∂z
∂u ∂v ∂w
+ + = 0. (3d)
∂x ∂y ∂z
From the above equations, we write a single equation in p alone:

iii
 ∂2p ∂2p ∂2p
∂x2
+ ∂y 2
+ ∂z 2
= 0.

Boundary conditions:

• At the ocean bottom z = −D, we have w = 0. [Assume flat ocean bottom where
D=constant].

• At surface, z = η, we require any fluid parcels at the surface to remain at the surface
(material boundary condition). For such fluid parcels, z − η = 0 all the time. So
d(z − η)/dt = 0, and
dz ∂η
w= dt
= ∂t
(ignore nonlinear terms).

• At z = η, P = P0 + p = 0, (zero pressure at the ocean surface, ignore the surface

atmospheric pressure). Then p = −P0 = ρ0 gη.
Since η << D so we can take z = η = 0 so the condition is:
At z = 0, p = ρ0 gη.

Now, the ocean is bounded in z direction but open in x and y directions. So we can
assume free wave form in x and y, but leave the vertical structure undetermined.
Assume a traveling wave-like solutions of the form:
η = η 0 cos(kx + ly − ωt),
p = pi (z)cos(kx + ly − ωt),
Substituting these solutions to the single equation in p alone, we have:
∂ 2 pi
∂z 2
− (k 2 + l2 )pi = 0.
Thus,
pi (z) = Asinh(κz) + Bcosh(κz),
where κ2 = k 2 + l2 .
∂p
By applying bottom boundary condition (at z = −D, w = 0 and thus ∂z = 0 from
equation (3c)), and surface boundary condition p = ρ0 gη at z = 0, we obtain,

gρ0 η0 cos(kx + ly − ωt)coshκ(z + D)

p= . (4)
cosh(κD)
From equation (3c) we have

κgη0 sin(kx + ly − ωt)sinhκ(z + D)

w= . (5)
ωcosh(κD)
∂η
Now let equations (4) and (5) satisfy boundary condition we derived above w = ∂t
at
z = 0, we obtain the dispersion relation:

ω 2 = gκtanh(κD). (6)
See Fig 3 for the dispersion relation diagram.

iv
 Figure 3: Dispersion relation for surface gravity waves.

• Long waves. For long waves such that L − − κ1 >> D and thus κD << 1===>
tanh(κD) = κD. Equation (6) becomes:
ω 2 = gDκ2 .
∂ω
√
Thus, the wave speed C 2 = gD. Cg = ∂κ
= ± gD, which does not change with
frequency, and thus is non-dispersive.
Long surface gravity waves are non-dispersive. These waves are called gravity
waves because their restoring force is “gravity”. They can propagate in all directions
of a horizontal plane.
From equations (3) we can see that, for an initial perturbation in p, the ocean will
respond in transient motion by radiating gravity waves in all directions of a horizontal
∂
plane. After the gravity waves radiate out, that is the ∂t terms equal 0, the ocean reaches
a state of rest. After an initial perturbation, the ocean adjusts to an equilibrium state
under gravity in a non-rotating ocean. A good example is we throw a big stone in a
calm pond. The initial perturbation will cause gravity waves to radiate out in circles,
after these waves radiate out, the pond is calm again.
Vertical structure. Since (κD) << 1, then cosh(κD) − 1, and

p = gρ0 η 0 cos(kx + ly − ωt), (7)

∂w
= 0. (8)
∂t
and thus w = constant.
Independent of depth. They feel the bottom of the ocean.

• Short waves. For the short waves with L− κ1 << D and thus (κD) >> 1, the dispersion
relation becomes:

v
 ω 2 = gκ.
ω ∂ω
C= κ
and Cg = ∂κ
is a function of ω and thus they are dispersive.
By simplifying their solutions by the dispersion relation, we have

p = gρ0 η 0 cos(kx + ly − ωt)eκz , (9)

κgη0 sin(kx + ly − ωt)eκz

w= (10)
ω
Short waves are trapped to the upper ocean to a region of depth κ−1 . They do not feel
the ocean bottom.
Frequency range for gravity waves: 0–infinity in a non-rotating system
(f=0), and at low frequency, wave number is low and thus wavelength is
long. As we shall see, in a rotating system these long waves are modified
and thus gravity waves exist only in the period band of seconds–days.
The wave energy propagation is different from water particle movement. In fact, to
the lowest order, the water particles for short waves are circular and for long waves are
elliptical. [Need to solve the solutions to know the details]. If we consider the higher
order effects the orbits of particles are not quite closed; for short waves there is a net
flow in the direction of travel and this net transport is called Stokes drift. However,
this effect is small (a small fraction of the orbital speed).

4.3 Refraction and breaking in shallow water

(i) Refraction. As we just discussed, small

√ amplitude long waves all travel at the same
speed in water of a given depth D (C = gD). But if the bottom depth is changing their
direction of travel may change. As waves move into shallow water their period remains
constant but C decreases and therefore wavelength decreases. [ω = κC. If ω is constant and
C decreases, then κ has to increases and thus wavelength (L = 2π κ
) decreases.] If a series of
parallel-crested waves approaches at an angle to a straight shoreline (Fig 4) over a smooth
sea bottom which shoals gradually, they progressively change direction as the end of the
wave nearer to the shore. As a result, the waves become more parallel to the shore by the
time that they pile up as surf. The change in direction associated with the change of speed is
called refraction. The refraction pattern will be more complicated if the ocean bottom has
complicated topography.
(ii) Breaking. As waves move in-shore and slow down, not only does the wavelength
decrease but also the wave height changes. In a steady state, the energy flux (the product of
the wave energy per square meter times the group speed Cg ) is conserved as the waves move
in-shore (until they break).

vi
 Figure 4: Wave regraction.

Energy flux Ef = ρgH 2 Cg /8=constant. For the long waves, as it moves in-shore, Cg
decreases and wavelength decreases (since frequency does not change), and thus H increases.
Therefore, the waves become steeper and steeper. As the steepness goes to 1/12 in real ocean
(1/7 in theory), the waves break (Fig 5).
4.4 Swells and Tsunamis
Swell: waves generated by persistent wind forcing. Large amplitudes and propagate for
a long distance.
Tsunamis: Long water-waves usually generated by sea bottom movement associated with
earthquakes. Sometimes called “seismic sea wave”; but they can also be caused by Meteorite.
The effects are usually not widespread, but Tsunamis generated by seismic activity on one
side of the Pacific have caused devastation on the other side, as in the case of the tsunami
generated by the 1960 earthquake in Chile which caused serious damage there and also in
Japan nearly 20,000km away. The recent December 2004 tsunami over Asia was devastating.

4.5 Internal gravity waves: density stratified ocean.

In the above, we discussed the waves in a homogeneous, inviscid ocean. We obtain gravity
waves “either surface trapped” (short waves), or “depth independent” whose perturbation
is large at the surface (long waves). We call these waves “surface gravity waves”. In reality,
the ocean is stratified rather than homogeneous. By considering the “stratification” of the
ocean, we obtain “internal gravity waves”.
a) A two-layer model.
The simplest case to represent the stratified ocean is to view the ocean as a two-layer

vii
 Figure 5: Wave breaking.

system. Fig 6. This is “reasonable” because there is a well mixed surface layer in the real
ocean, which is separated from the deep ocean by a sharp pycnocline. Usually, mixed layer
depth H1 = 100m − 200m, and the depth beneath the thermocline H2 = 3800m. The
“interface” in the two layer model represents the “pycnocline”.
This system has two vertical modes: One barotropic mode and one baroclinic mode. The
barotropic mode represents the vertically-averaged motion. The restoring force is gravity “g”.
The baroclinic mode has vertical shear and does not contribute to the vertically-integrated
transport. The restoring force is “reduced gravity”, gg = 0.003, and g ′ = g(ρ2ρ−ρ1 ) = 0.03ms−2 .
′

2
I will skip the derivation and just
p give the phase speeds for the two modes, respectively.
For the barotropic mode C0 = q g(H1 + H2 ).
For the baroclinic mode C1 = g ′ HH11+H
H2
2
.

• For the barotropic mode,

h H2
η
= H1 +H2
≤ 1,
u2
and u1
≤ 1.
Because H2 ≫ H1 , η is close to h and u2 is close toR u1 . This is why we call it “barotropic
0
mode”. It is approximately depth independent. −(H1 +H2 ) udz 6= 0.
For H1 + H2 =4000m, C0 = 200m/s. Very fast. They can cross the Pacific within 2
days. It is surface gravity wave.

viii
 Figure 6: A two-layer model.

Figure 7a: Barotropic mode.

• For the baroclinic mode, h

η
= − gg′ = −300.
So h = −300η. If surface raises 1 cm, η=1cm, then the thermocline will deepen by
3m!! Since surface perturbation is much less than the variation of the “thermocline”,
we call these waves “internal gravity waves”. Velocities for the baroclinic mode have
opposite signs in the two layers. This mode, however, does not contribute to the total
vertically integrated transport.
R0
−(H1 +H2 )
udz = 0.
Its structure is vertically dependent, shear flow. C1 = 2 − 3m/s which is 100 times
slower than the surface gravity wave.

b) A 1 12 -layer, reduced gravity model.

As we have seen above, gravity waves associated with the barotropic mode are very
fast. If we are only interested in the slower baroclinic wave, we wish to eliminate

ix
 Figure 7b: Baroclinic mode.

the barotropic mode. One way to do so is to assume “reduced gravity” condition,

which assumes that the deep ocean (below the surface mixed layer) is motionless. This
assumption eliminates the barotropic mode, which is depth-independent. [NOTE:
Recall that in the “geostrophic method” section, we assume a reference level - the level
of no motion. We argued that this assumption filtered out the barotropic mode.]
Therefore, in a 1 21 -layer system, we only have one baroclinic mode. Fig 8 Schematic
diagram showing the baroclinic mode in a 1 12 -layer system.

Figure 8: The baroclinic mode in the 1 21 -layer model.

As discussed in the 2-layer model, the baroclinic mode has a large-amplitude per-
turbation in the thermocline, and a small-amplitude perturbation at the sea surface.
Therefore, it represents internal waves. The “one active layer” above the “inert deep
ocean” represents the oceanic “surface mixed layer”, and the motion of the mixed layer
bottom represents the movement of the thermocline. In this system, variation of the
layer thickness results primarily
√ from the thermocline motion. The baroclinic mode
speed of gravity is: c = g ′H.

4.6 Effects of rotation, Rossby radius of deformation

x
 In the above, we have considered the adjustment under gravity in a non-rotating ocean.
In this system, after the transient gravity waves radiate out, the ocean reaches a equilibrium
state: the state of rest. We eliminated the rotation effects by setting Coriolis parameter
f = 0.
Now let’s add “f ” back into the equations, and see what it does to the system.
With a uniform rotation (f is assumed to be a constant), the equations of motion for the
unforced, inviscid ocean are:
dU 1 ∂P
− fV = − , (11a)
dt ρ0 ∂x
dV 1 ∂P
+ fU = − , (11b)
dt ρ0 ∂y
dW 1 ∂P
=− − g. (11c)
dt ρ0 ∂z
∂U ∂V ∂W
+ + = 0. (11d)
∂x ∂y ∂z
Assumptions:
(i) f =constant (i.e. the f-plane approximation);
(ii) homogeneous ocean==> ρ = ρ0 =constant;
(iii) hydrostatic approximation ( ∂P
∂z
= −ρ0 g); This approximation eliminates the “short
gravity waves”, leave only the long waves in the system, as discussed in the previous class.
As we recall, the long waves have a structure that is “independent of depth”.
(iv) flat bottom, depth H at where W = 0.
(v) Linearize about a resting basic state (U0 = V0 = W0 = 0)
For small perturbations u, v, w, p about the resting state, we have:
U = U0 + u = u,
V = V0 + v = v,
W = W0 + w = w,
P = P0 (z) + p,
ρ = ρ0 .
To the first order, we have:
∂u 1 ∂p
− fv = − , (12a)
∂t ρ0 ∂x
∂v 1 ∂p
+ fu = − , (12b)
∂t ρ0 ∂y
∂u ∂v ∂w
+ + = 0. (12c)
∂x ∂y ∂z
By applying hydrostatic approximation at z = 0, P = P0 + p = p = ρ0 gη (note: p
is independent of depth), vertically integrate equation (12c) and using surface boundary
condition w = ∂η
∂t
, bottom boundary condition w = 0,

xi
 we obtain:
∂u ∂η
− f v = −g , (13a)
∂t ∂x
∂v ∂η
+ f u = −g , (13b)
∂t ∂y
∂η ∂u ∂v
+ H( + ) = 0. (13c)
∂t ∂x ∂y
Figure 9.

Figure 9: Schematic diagram showing the surface perturbation for the shallow water equa-
tion.

As in the non-rotating case, we combine the above three equations to write a single
equation in η alone.

∂2η
− c2 ∇2 η + f Hζ = 0, (14)
∂t2
∂v
where c2 = gH is phase speed squared, ζ = ∂x − ∂u
∂y
is called relative vorticity, f is
planetary vorticity, and ζ + f is called absolute vorticity.
(a) Non-rotating case (f = 0). Fig 10. Schematics for gravity waves radiation. Let’s
look at the simple, one-dimensional case in x direction. If there is no f , equation (14)
becomes,
∂2η
∂t2
− c2 ∇2 η = 0.
Assume η = η0 cos(kx − ωt), then we obtain from the above equation:
−ω 2 + c2 k 2 = 0, and ω 2 = c2 k 2 , thus

xii
 ω√= ±kc. That is, gravity waves radiate out in −x and +x directions with a speed of
c = gH. After these waves radiate out, the sea surface flatens and thus η = 0.

Figure 10: Schematic diagram showing the oceanic adjustment in the non-rotating case.

(b) Rotating case (f 6= 0). For one dimensional case, equation (14) becomes:
∂2η ∂2η ∂v
∂t2
− c2 ∂x 2 + f H ∂x = 0.
2
After gravity waves radiate out, that is, ∂∂t2η = 0 in the above equation, we obtain an
equilibrium state:
∂2η ∂v
−c2 ∂x 2 + f H ∂x = 0, and thus:
∂η
−g ∂x + f v = 0, or
∂η
f v = g ∂x .—Geostrophic balance.
Solutions to η [for derivation see Gill 7.2.1 P192] is:
x
η = η 0 [−1 + exp(− )], f orx > 0, (15a)
a
x
η = η 0 [1 − exp( )], f orx < 0, (15b)
a

where a = fc is called the Rossby radius of deformation. As we can see from the η solution,
sea level is deformed within the distance of “a”, and the geostrophic currents are confined
within the range of the Rossby radius. Outside this range, v is close to 0. Therefore, Rossby
radius is a natural scale for the large-scale ocean circulation. [For the first baroclinic mode
with c1 –3 m/s, a1 is about 30km at 45◦ N where f = 10−4 /s.]
Now we can see that, with effects of rotation, the equilibrium state of the ocean is in
“geostrophic balance”, rather than the “state of rest” as in the non-rotating case.

xiii
 Figure 11: Schematic diagram showing the equilibrium state in the rotating case.

We just saw that rotation can change the “equilibrium” state of the ocean, relative to
the non-rotating case. Next, we will consider how rotation modifies the gravity waves in the
system.
By assuming the wave form of:

η = η 0 expi(kx + ly − ωt), (16a)

u = u0 expi(kx + ly − ωt), (16b)
v = v0 expi(kx + ly − ωt), (16c)
substituting them into equations (11), and let the coefficient matrix equal “0”, we obtain
the dispersion relation

ω 2 = κ2 c2 + f 2 , (17)
where κ2 = k 2 + l2 .
∂ω
(i) First, the long waves become “dispersive”, since Cg = ∂κ
is a function of ω and thus
κ.
(ii) Second, the long gravity waves do not have “zero” frequencies anymore. Their “low-
est” frequency is at “f ”, which has a period (T = 2πω
) of a few days. That is, the longest
period for gravity waves is “days”. Their period range is “seconds–days”. They are con-
sidered “high” frequency in “oceanography”, since Rossby waves have periods of months to
years.

xiv
Figure 12: Schematic diagram showing the dispersion relation for gravity waves with the
influence of rotation.

4.7 Waves in an ocean with a varying f : Rossby waves

Up to now, we have considered the adjustment under gravity in a non-rotating ocean

with f = 0, and in a rotating ocean with f = constant. In these systems, gravity waves
exist and they are modified by rotation in the rotating case. Since the earth is rotating and
Coriolis parameter f = 2Ωsinφ, f in fact varies with latitude φ. Therefore, we can write
f = f0 + βy,
where f0 is the Coriolis parameter at latitude φ0 , say φ = 45◦ N, and y is meridional
distance from latitude φ to φ0 . It is referred to as mid-latitude “β-plane approximation”.
f = f0 is referred to as “f-plane approximation”.
Figure 13 Schematic diagram for β-plane.
A few relevant concepts. We often refer to:
f as planetary vorticity,
∂v
ζ = ∂x − ∂u
∂y
as relative vorticity,
f + ζ as absolute vorticity, and
ζ+f
H
as potential vorticity, where H is the depth of water column.
Again, we use the simple shallow water equation to demonstrate what β does to the
system. The equations are the same as those in the f-plane case discussed in the previous
class, except that f = f0 + βy here rather than f = f0 = constant.
∂u ∂η
− f v = −g , (18a)
∂t ∂x
∂v ∂η
+ f u = −g , (18b)
∂t ∂y
∂η ∂u ∂v
+ H( + ) = 0. (18c)
∂t ∂x ∂y

xv
 Figure 13: Schematic diagram for β-plane.

We combine the above three equations to write a single equation in v alone, and assume
periodic wave solution v − −e−iωt , we obtain:
2 2
3 2∂
v 2 2 ∂v 2∂ v
ω v + ωc − f0 vω + c β i + ωc = 0, (19)
∂x2 ∂x ∂y 2
√
where c = gH is phase speed for long surface gravity waves. Assuming wave form:
v − −ei(kx+ly−ωt) in equation (19) and re-arrange the equation, we obtain the dispersion
relation:

β 2 ω 2 − f02 β2
(k + ) + l2 = + . (20)
2ω c2 4ω 2
Let’s look at the simple case with l = 0, and we shall see what waves are available in the
system. The dispersion relation becomes:

β 2 ω 2 − f02 β2
(k + ) = + . (21)
2ω c2 4ω 2
xvi
 (i) For high frequency waves with large ω, the terms with ω in the denominator can be
ignored. Thus,
ω 2 −f 2
k 2 = c2 0 , or

ω 2 = k 2 c2 + f02 . (22)
These are long surface gravity waves under the influence of “f”–they are also called
Inertial gravity waves.

(ii) For low frequency waves with small ω, we can drop ω 2 term. Then we have:
β 2 β2 f02
(k + 2ω ) = 4ω 2 − c2 . Rewrite it,

s s
β 1 β2 f02 β f02 ω 2
k=− ± − 4 = − (1 ± 1 − 4 ). (23)
2ω 2 ω 2 c2 2ω c2 β 2

Figure 14: Schematic diagram showing the dispersion relation for inertial gravity waves
and Rossby waves.

To further understand the wave properties, we need to obtain simplified results. Since
f02 ω 2
c2 β 2
is a small number given that ω is small, the above equations can be approximated by:
β f 2ω2
k = − 2ω [1 ± (1 − 2 β02 c2 )].

xvii
 Choosing the “-” sign we have,
β f 2 ω2 ωf 2
k = − 2ω (1 − 1 + 2 β02 c2 ) = − βc02 , or

βc2
ω=− k = −cr k. (24)
f02

• Since cg = ∂ω∂k
= −cr is independent of ω, these waves are non-dispersive and both
ω
phase (cp = k = −cr ) and group velocity (energy cg ) propagate westward. From the
dispersion relation we can see these waves are: long Rossby waves.

• No matter f is positive (Northern Hemisphere) or negative (Southern Hemisphere),

long Rossby waves propagate westward, both for energy and phase.
2
• Long Rossby wave speed decreases with the increase of latitude: cg = cp = −cr = − βc
f02
.
As the absolute value of f0 increases poleward, Rossby waves speed decreases.

Choose the “+” sign and ignore the second order ω term:
β
ω=− . (25)
k
Dispersive since cg = ∂ω
∂k
= kβ2 is a function of k and thus ω. cp = ωk = − kβ2 . These
waves are short Rossby waves: Their group velocity propagates eastward but phase propagates
westward. These can be seen from the dispersion curve. Short Rossby waves are very short
so that mixing acts strongly on these waves. They are easily killed by mixing and thus very
hard to propagate far away from the forcing region. Also, in real open ocean they are not
strongly excited by the “large-scale” winds because their wavelengths are too small. Near
the western boundary, these waves are killed by the strong mixing there and thus contribute
to the WBC.

• Existence of Rossby waves are due to β: the variation of planetary vorticity

(f ) with latitude (φ), and β = ∂f
∂y
. If we set β = 0 then the dispersive relation
will be just the inertial gravity waves.

• Rossby waves are low-frequency waves. Their periods are T > 30 days. That
is, if the forcings are at periods shorter than a month, Rossby waves will
not be excited.

• Their motion satisfies “Quasi-geostrophy”. Quasi-geostrophy is a complicated

concept and has precise definition. But we will not cover it here. However, we will
show the quasigeostrophic motion of Rossby waves in Figure 15.

4.8 Effects of side boundaries: coastally trapped waves

xviii
Figure 15: Schematic diagram showing Rossby wave propagation.

xix
Figure 16: Longitude-time plot of Sea Surface Height (SSH) anomaly: Observed mid-
latitude Rossby waves in the Pacific.
xx
 The presence of lateral boundary due to the continental margins distinguishes the ocean
circulation from the atmospheric circulation. As we will see, the coastal ocean acts as a
waveguide. Information can be transported rapidly from low latitudes to high latitudes. In
Figure 17, we can see SSH anomalies hug the eastern Pacific coasts in the off equatorial
latitudes.

Figure 17: Observed SSHA.

a) Coastal Kelvin wave.

To simplify things, we look at the shallow water equation with f = f0 = const, so Rossby
waves do not exist in the system.
Boundary conditions are y = 0, v = 0. We find solutions with v = 0 everywhere. The
equations become:
∂u ∂η
= −g , (26a)
∂t ∂x
∂η
f u = −g , (26b)
∂y
∂η ∂u
+H = 0. (26c)
∂t ∂x
By assuming a waveform solution you can easily derive the dispersion relation as:

xxi
 Figure 18: Schematic diagram showing the coastal vertical wall.

√
ω 2 = k 2 c2 , where c = gH, or
ω = ±kc.
This is the same dispersion relation for “gravity waves in a non-rotating” system.

I will not derive the solution in detail here, but I’ll write it down.
The solution for η is composed of two pieces F and G:
fy
F = Ae c F ′ (x + ct),
−f y
G = Be c G′ (x − ct),
where both F ′ and G′ are arbitrary functions. That is, one wave propagates westward
(F ) and the other one propagates eastward (G). [Determine wave propagation direction:
Find constant phase line. Φ = x + ct = const = 0, and Φ = x − ct = const = 0. As time
goes by (t increases), x increases in +x direction for G but in −x direction for F .]
As y increases to infinity, F becomes infinity. For a perturbation that is initiated near
the coast, amplitudes of the waves increase to infinity away from the coast. This is against
the energy conservation and thus this piece of solution is impossible in real ocean.
As y increases to infinity, G = 0 and thus is the reasonable solution.
Therefore, the solution is:
−f y
G = Be c G′ (x − ct). (27)
That is, only the gravity waves that propagate eastward (or with the coast to its right
in the northern hemisphere) are the reasonable solutions for the coastal waves. They are
referred to as “coastal Kelvin waves”.
The dispersion relation for the coastal Kelvin waves are:

ω = kc, (28)

xxii
 since only the one propagates with the coast to its right exists in NH in the system.

• Coastal Kelvin waves dispersion character is the same as those long gravity
waves in a non-rotating fluid. They are non-dispersive and they propagate
with the coast to their right (left) in the NH (SH). Their structures, how-
ever, are trapped to the coast with an e-folding scale of fc , which is the
Rossby radius of deformation. This is the reason for them to be called “coastal
Kelvin wave” because they are trapped to the coast.

• Coastal Kelvin waves can be generated by surface wind forcing close to the shore, tidal
forces, reflection of other waves incident on a coast.

In Figure 17, the SSH anomalies that hug the costs in subtropical-mid latitudes are
coastal Kelvin waves generated by the incident equatorial Kelvin wave signals.

b). Continental shelf waves.

When the depth of the ocean at the coast varies, the Kelvin waves are modified, and a
new class of waves arise. These waves are called: Continental shelf waves, or topographic
Rossby waves.

Figure 19a:

In this case, the shallow water equations on a f-plane are:

xxiii
 Figure 19b: Schematic diagram for shelf topography.

∂u ∂η
− f v = −g , (29a)
∂t ∂x
∂v ∂η
+ f u = −g , (29b)
∂t ∂y
∂η ∂(Hu) ∂(Hv)
+ + = 0. (29c)
∂t ∂x ∂y
In this case H = Hy and solution is more complicated to solve analytically. So we just
provide some basic discussion regarding their energy dispersion.
Continental shelf waves propagate in the direction of coastal Kelvin waves,
with the coast to its right (left) in NH (SH). A typical off-shore length scale is
30–60 km with c−2m/s. However, unlike Kelvin waves continental shelf waves are
dispersive. The mechanism for its propagation can be demonstrated by potential
vorticity conservation, which is Q = ζ+f H
= const and illustrated in Figure 20. [In
an inviscid, unforced ocean, the potential vorticity is conserved for a moving
water column.]

4.9 Equatorially trapped waves

xxiv
 Figure 20: Shelf waves.

We noted earlier that a lateral boundary in the ocean can act as a waveguide,
which guides the waves to propagate along it. These waves (coastal Kelvin waves
and shelf waves) are “trapped” to the coast and thus perturbations associated
with these waves decay away from the coast. For coast Kelvin waves, their
amplitudes decay with an e-folding decay scale of Rossby radius a = fc . In fact,
it is f that makes the coastal Kelvin waves trapped to the coast”.
As we shall see next, the equator acts as a waveguide. Waves are trapped to
the equator because of “β, the variation of f with latitude”, rather than “f ”.
Because at the equator, f = 2Ωsinφ = 0. but β 6= 0.
Here in this class, we will just focus on discussing the physics without provid-
ing the mathematical details. For each type of wave, we will discuss the following
aspects associated with the equatorially trapped waves:

• energy dispersion, phase speed (cp ) and group velocity (cg );

• Solution character;

• Symmetric property;

• Forcing.

4.9.1. The equatorial Kelvin wave

xxv
 a. Dispersion relation. The dispersion relation for the equatorial Kelvin wave
is:

ω = kc, (30)
where√c is a barotropic or baroclinic mode speed. If the ocean is barotropic,
c = c0 = gH = 200m/s for H = 4000m. Very fast. For a continuously stratified
ocean, we have one barotropic mode and infinite number of baroclinic modes.
Usually, the first baroclinic mode has a speed of c1 = 2 − 3m/s.
As we can see, the equatorial Kelvin wave has the same dispersion relation as
the coastal Kelvin wave and as the gravity waves in a non-rotating fluid (f = 0).
They are non-dispersive since cg = ∂ω ∂k
= c is independent of frequency. cg = cp .
Both phase and group velocity (energy) propagate eastward.
Equatorial Kelvin waves propagate eastward.

Figure 21: Dispersion relation curve for Kelvin wave.

b. Solution. The solutions to Kelvin wave are:

βy 2
η = e− 2c G′ (x − ct),
2
− βy
u = gc e G′ (x − ct),
2c

v = 0.
As we can see from the above solutions, Kelvin waves propagate eastward
with a speed of c. Their amplitudes are the largest at the equator q
and decay
poleward to each side of the equator with an e-folding scale of y = 2c
β
. They
are like two coastal Kelvin waves propagating back-to-back along the equator

xxvi
(where f = 0), one is to the north of the equator and the other to the south. In
other words, the equator behaves like an artificial coastline for each hemisphere.

Figure 22: Schematic for Kelvin waves.

Often,
q a naturally occurring meridional length scale in the tropics is:
c
ae = 2β (see Gill 11).
It is called the “equatorial radius of deformation”, and is analogous to the
Rossby radius of deformation at off-equatorial latitudes.
Again, the equatorially trapped nature of waves results from the presence of β. If we set
β = 0 in the above solutions then the waves are not trapped anymore! Recall that the trapping
of coastal Kelvin waves is due to the presence of “f ”.
For the first baroclinic mode with c1 − 3m/s, a1 − 250km.
(visit http://www.pmel.noaa.gov/tao/vis/gif/t-dyn-med.gif, which shows real
time temperature and dynamic height associated with Kelvin waves pauses).

c. Symmetric property. As we see from the solutions, Kelvin waves are symmet-
ric about the equator. They are symmetric waves. Both u and η are symmetric
about the equator. v = 0.

d. Forcing. At the equator, the Kelvin waves are mostly generated by the
changing winds. That is, variations of winds with time.
Assume the ocean is calm initially. Then all of a sudden westerly winds
(winds blowing from west to east) start to blow along the equator. You will see
Kelvin wave pause associated with a high sea level and a deepened thermocline
in a simple 2-layer system. This is because westerly winds will cause eastward
currents initially, and Coriolis force off the equator tend to pull the motion
toward right (left) in the NH (SH). As a result, there is water converges on
the equator and thus pushes the thermocline downward and raise the sea level.
Amplitudes of Kelvin waves are proportional to the “zonal integral” of the westerly winds
(derivation is not provided).

xxvii
Figure 23: Schematic diagram showing equatorial Kelvin waves forced by westerly winds.

Forced Kelvin and Rossby waves from a linear model. Gill 11.11 (cited from
McCreary 1978). Coastal Kelvin waves are clearly seen along the eastern ocean
boundary after the equatorial Kelvin waves arrive at the boundary. At first, the
coastal Kelvin waves are trapped to the coast because they propagate fast (at
speed of c). Then as time goes by, they become entrapped due to the westward
radiation of the Rossby waves. We can see from the figure that speeds of Rossby
waves decrease with latitude increase.

4.9.2. The equatorial Rossby wave.

a. Dispersion relation.
ω = − k2 + ββk
(2n+1)
,
c
where n = 1, 2... is the order number for Hermite function (or Parabolic
cylinder function).
For long Rossby waves, k 2 is small and thus the dispersion can be written as:
ck
ω = − 2n+1 .
c
Cg = − 2n+1 is independent of ω, and thus long Rossby waves are non-dispersive,
and cg = cp = const. Therefore, both phase and group velocity propagate west-
ward, as the long Rossby waves in mid-latitude.
2π
For c1 = 3m/s, n = 1, and k = 1000km , T = 2π
ω
= 35days, which is also consistent
with the highest frequency (shortest period) for baroclinic Rossby waves.
For short Rossby waves, k 2 is larger and thus,

xxviii
 ω = − βk . The same as dispersion relation as the short Rossby waves at mid-
latitude Rossby waves. They are dispersive waves.
cg = kβ2 and cp = − kβ2 .
Short Rossby waves group velocity is eastward, but phase speed is westward. They are
two short to be efficiently excited in the ocean interior. Mixing also acts strongly
on these short waves. As discussed in the mid-latitude Rossby waves section,
these waves exist only near the western boundary. Their energy is killed by
mixing and thus forming the WBC.

b. Solution. The solutions for equatorial Rossby wave involve Hermite function.
That is, theqsolutions are expanded in qy direction by Hermite function.
βy 2
vn = Dn ( 2β c
y)cos(kx − ωt) = 2−n/2 Hn ( βc y)e− 2c cos(kx − ωt),
√ √
c
√ Dn+1 ( 2β y) nDn−1 ( 2β y)
η = 2g 2βc[ kc−ω c
+ kc+ω
c
]sin(kx − ωt),
q √ 2β √ 2β
D ( y) nD ( y)
un = βc 2
[ n+1kc−ω c − n−1 kc+ω
c
]sin(kx − ωt),
where Dn is parabolic cylinder function of order n, and Hn is a Hermite
polynomial of order n.
For even numbers of “n”, Dn is symmetric; For odd numbers of “n”, Dn is
antisymmetric.
From these solutions, we can see that amplitudes of Rossby waves decay poleward
and oscillatory
q in y. They are equatorially trapped: with an e-folding scale of
c
ae = 2β .
c. Symmetric property. We use symmetric properties of u or p (or η) to
define symmetric properties of waves. That is, if u or p (or η) are symmetric
(antisymmetric) about the equator, then we referred to these waves as symmetric
(antisymmetric) waves.
From the above solutions, if vn is antisymmetric, un and η n are symmetric.
Rossby waves can be both symmetric or antisymmetric, depending on the forc-
ing structures. If the forcing is symmetric, symmetric Rossby waves will be
excited. If the forcing is antisymmetric, then antisymmetric Rossby waves will
be generated.
d. Forcing. Long equatorial Rossby waves can be efficiently excited by the
change of large spatial scale winds. See Figure 24. Since westerly winds cause
equatorial convergence due to Coriolis effects, in the off equatorial region for
both NH and SH, divergence is generated due to water flows toward the equa-
tor. Therefore, an equatorial deepened thermocline (which represents high sea
level for baroclinic modes) occurs, and off equatorial shallowed thermocline is
produced. Since the forcing winds are symmetric about the equator, only sym-
metric Rossby waves are generated.

BOUNDARY REFLECTION FOR ROSSBY AND KELVIN WAVES

xxix
 Physics: Kelvin waves propagate eastward, both for phase and group velocity.
After Kelvin waves impinge onto the eastern ocean boundary, part of the energy
is reflected back into the ocean interior as “long Rossby waves”, since the long
Rossby waves’ energy propagate westward and thus can take the energy away
from the eastern boundary; part of the energy propagates poleward via coastal
Kelvin waves. Therefore, variability that initiates at the equator can propagate
to the mid-latitude quickly through coastal Kelvin waves.
Long Rossby waves propagate westward. As long Rossby waves impinge onto
the western ocean boundary, part of the energy is reflected back into the ocean
interior as Kelvin waves, and part of it as short Rossby waves, both of which
have eastward group velocity and thus can take the energy away from the western
ocean boundary. Since short Rossby waves are short and slow, their energy is
primarily killed by mixing near the western boundary.
Now we have a question: why don’t we have coastal Kelvin waves that take
the energy poleward near the western boundary? [Prompt the class]
4.9.3. Equatorially trapped inertial gravity waves

a. Dispersion relation.
ω 2 = k 2 c2 + βc(2n + 1),
where n = 1, 2... is the order number for Hermite function (or Parabolic cylin-
der function).
These waves are called inertial gravity waves, which is similar to the inertial
gravity waves at mid-latitude. They are gravity waves under the influence of
β (f ) at the equator (mid-latitude). They are dispersive waves, and their energy and
phase can propagate both eastward and westward.
p
Their minimum frequency is: ω min = βc(2n + 1). For n = 1, c1 = 3m/s, and
2π
k = 1000km ,
ω min = 2π × 3.8day −1,
and thus the longest period is 3.8 day. So gravity waves periods are shorter
than a few days.

b. Solution. The solutions for equatorially trapped inertial gravity waves have
the same form as the solutions for Rossby waves. They are oscillatory in y direction
and at the same time q
decay poleward. They are equatorially trapped: with an e-
c
folding scale of ae = 2β .
c. Symmetric property. We use symmetric properties of u or p (or η) to
define symmetric properties of waves. That is, if u or p (or η) are symmetric
(antisymmetric) about the equator, then we referred to these waves as sym-
metric (antisymmetric) waves. Inertial gravity waves can be both symmetric or
antisymmetric, depending on the forcing structures. If the forcing is symmetric,
symmetric gravity waves will be excited. If the forcing is antisymmetric, then
antisymmetric gravity waves will be generated.

xxx
 d. Forcing. Winds associated with synoptic weather phenomena, or gusty
winds with periods of a few hours.

4.9.4. The mixed Rossby gravity wave: Yanai wave

a. Dispersion relation.
k = − ωβ + ωc .
This wave is interesting because (i) as frequency is high (ω is large), the
dispersion becomes k = ωc , which is Kelvin wave’s dispersion relation; (ii) as
frequency is low, the dispersion becomes k = − ωβ , which is short Rossby wave’s
dispersion relation. Therefore, energy dispersion of this type of wave behaves
like short Rossby wave at low frequency, and like Kelvin wave (whose restoring
force is gravity) at high frequency. So they are called “mixed Rossby gravity
waves”, or Yanai waves since Yanai first discovered it.
So, at low frequency, Phase speed of Yanai waves is westward but group
velocity is eastward. At high frequency, phase speed of Yanai waves is eastward
and group velocity is also eastward. Group velocity of Yanai waves is eastward. Phase
speed can be eastward or westward, depending on the frequency. They are dispersive
waves. But at high frequency they are close to be non-dispersive.

See Figure 30 for Yanai wave dispersion curve on dispersion relation diagram.
b. Solution. The solutions for Yanai wave are the n = 0 case for the solutions
of Rossby waves.
q Therefore,
v0 = D0 ( 2βc
y)cos(kx − ωt).
c. Symmetric property. As we can see, v is symmetric for Yanai waves and
thus antisymmetric for u and p. Yanai waves are antisymmetric waves; their zonal
velocity u and pressure p are antisymmetric about the equator, and meridional
velocity v is symmetric about the equator.
d. Forcing. Oscillatory meridional winds or anti-symmetric part of zonal
winds.

xxxi
Figure 24: Wind-driven equaotiral Kelvin and Rossby waves from a 1 21 -layer model.

xxxii
Figure 25: Schematic diagram for the first a few Dn .

xxxiii
Figure 26: Observed equtorial Rossby waves from satellite SSH anomaly.

xxxiv
Figure 27: Schematic diagrams for equatorial Kelvin and Rossby waves boundary reflection.

Figure 28: Schematic diagram for coastal Kelvin waves.

xxxv
Figure 29: Dispersion relations for equatorial Kelvin and Rossby waves.

Figure 30: Yanai wave dispersion curve on dispersion relation diagram.

xxxvi
Figure 31: Structures for Yanai waves.

xxxvii