Contents

3 Fundamental tools 3
3.1 Some basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
3.1.1 Shear stresses and Newton’s law of viscosity . . . . . 3
3.1.2 Pressure in a fluid . . . . . . . . . . . . . . . . . . . . 3
3.1.3 Tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
3.1.4 (Gauss’) Divergence theorem . . . . . . . . . . . . . . 3
3.1.5 Stokes’ theorem . . . . . . . . . . . . . . . . . . . . . . 4
3.2 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2.1 The material Derivative . . . . . . . . . . . . . . . . . 6
3.2.2 Streamlines and streamfunctions . . . . . . . . . . . . 7
3.2.3 Strain rates . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2.4 Vorticity and circulation . . . . . . . . . . . . . . . . . 8
3.2.5 Relative motion near a point . . . . . . . . . . . . . . 8
3.3 Equations of motion . . . . . . . . . . . . . . . . . . . . . . . 9
3.3.1 Motion in a rotating frame of reference . . . . . . . . 11
3.3.2 Thin shell approximation . . . . . . . . . . . . . . . . 13
3.3.3 The β-plane . . . . . . . . . . . . . . . . . . . . . . . . 14
3.4 Vorticity and Circulation . . . . . . . . . . . . . . . . . . . . . 17
3.5 Kinematical and dynamical approximations . . . . . . . . . . 18
3.5.1 Hydrostatic balance . . . . . . . . . . . . . . . . . . . 18
3.5.2 Hydrostatic approximation . . . . . . . . . . . . . . . 20
3.5.3 Shallow water approximation . . . . . . . . . . . . . . 20
3.5.4 Boussinesq approximation . . . . . . . . . . . . . . . 20
3.5.5 Rigid lid approximation . . . . . . . . . . . . . . . . . 20
3.6 Rossby number . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.7 Geostrophic and Thermal Wind Balance . . . . . . . . . . . . 20
3.8 The Rossby radius . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.9 The shallow-water equations . . . . . . . . . . . . . . . . . . 30

1
Page 2
Chapter 3
Fundamental tools

We have already discussed the Geostrophic equations, or geostrophic bal-

ance, in previous courses (Geophysical Fluid Dynamics, and maybe in Physics
of the Atmospheric too), but it is good to restate our assumptions and start-
ing point (repetita iuvant!).

3.1 Some basics

3.1.1 Shear stresses and Newton’s law of viscosity

3.1.2 Pressure in a fluid

3.1.3 Tensors

3.1.4 (Gauss’) Divergence theorem

The theorem relates a volume integral to a surface integral. Consider a vol-

ume V bounded by a closed surface A. Consider an infinitesimal surface
element dA whose outward unit normal is n. The vector ndA has magni-
tude dA and direction n.

3
Tbis very useful theorem relates a volume integral to a surface integral. Let V be a
volume bounded by a closed surface A . Consider an infinitesimal surface element
dA, whose outward unit normal is n (Figure 2. IO). The vector n d A has a magnitude
d A and direction n, and we sball write d A to mean the same thing. Let Q(x) be a
scalar, vector, or tensor field of any order. Gauss’ theorem states that

(2.30)

ndA- dA

Figum 2.10 Illustration of Gauss’ Ihcorcrn.

Gauss’ theorem states that the volume integral of the divergence of Q
is equal to the surface integral of the outflow of Q.

Z Z
∂Q
dV = dAi Q (3.1)
v ∂xi A

For a vector Q:

Z Z
∂Qi
dV = dAi Q , (3.2)
v ∂xi A

which is now called the Divergence Theorem. In vector notation

Z Z
∇ · Q dV = dA · Q (3.3)
v A

3.1.5 Stokes’ theorem

The theorem relates a surface over an open surface to a line integral. Con-
sider an open surface A, with bounding curve C. Let dr be an element of
the bounding curve whose direction is that of the tangent.

Page 4
Thcn the theorem stales that

(2.34)

which signifies that thc surface integral of the curl of a vector field u is equal to the
line integral of u along thc bounding curve.
The line integral of a vector u around a closed curve C (as in Figure 2.1 1) is called
the “circulation of u about C.” This can be used to define the curl o€ a vector through

I$

Figurc 2.11 lllustrdlion of SLokCs’ thcorcm.

Stokes’ theorem states that
Z Z
(∇×F) · dA = F · dr (3.4)
A C

the surface integral of the curl of a vector field F is equal to the line integral
of F along the bounding curve. The line integral of a vector around a
closed curve C is the circulation of the field about C.

Prove that div(curl u) = 0, for any vector u.

Page 5
3.2 Kinematics
• In the LAGRANGIAN description of motion, one essentially follows
the history of an individual particle. A flow variable F (r0 , t) and its
velocity is given by ui = d(ri )/dt

• In the EULERIAN description one focuses on what happens at a spa-

tial point r, so the flow variable is F (r, t).

• In the Eulerian case, d/dt gives the local rate of change of F at each
point r and is not the total rate of change seen by a fluid particle ...

3.2.1 The material Derivative

We seek to calculate the rate of change of F at each point following a par-
ticle of fixed identity.

DF ∂F ∂F
= + ui (3.5)
Dt ∂t ∂xi

The material Derivative DF Dt is made of (1) the local rate of change at a

given point (zero for steady flows...) and (2) the advective derivative.
∂F
∂t is the local rate of change of F at a given point.
∂F
ui ∂x is the advective derivative, it is the change in F as a result of ad-
i
vection of the particle from one location to another where F is different.

Page 6
3.2.2 Streamlines and streamfunctions
The streamline
• At t = t0 , streamlines are curves that are tangent to direction of flow.
• For unsteady flows, streamlines change with time.
Let ds = (dx, dy, dz) be an element of arc length along a streamline,
and let u = (u, v, w) be the local velocity vector along that streamline,
then dx/u = dy/v = dz/w.

• Close to a solid boundary, streamlines are parallel to that boundary.

• The direction of the streamline is the direction of the fluid velocity.
• Fluid can not cross a streamline.
• Streamlines can not cross each other.
• Any particle starting on one streamline will stay on that same stream-
line.
• In unsteady flow, streamlines can change position with time.
• Streamlines are a family of curves that are instantaneously tangent
to the velocity vector of the flow. These show the direction a fluid
element will travel in at any point in time.
• Pathlines are the trajectories that individual fluid particles follow.
These can be thought of as a ”recording” of the path a fluid element
in the flow takes over a certain period. The direction the path takes
will be determined by the streamlines of the fluid at each moment in
time.
• For a steady flow, the two are the same.

Page 7
The streamfunction

3.2.3 Strain rates

3.2.4 Vorticity and circulation
3.2.5 Relative motion near a point

Page 8
3.3 Equations of motion
There can be two kind of forces acting on fluids. Body forces, and we will
restrict our attention, for now, to gravitational force per unit mass
∂( gz)
g = −∇( gz) = −k̂ = −k̂g (3.6)
∂z
and surface forces, which can be normal or tangential to the fluid. Normal
forces will be relate to pressure, whereas tangential forces will be related
to shear stresses.
In order to derive a principle of conservation of momentum we will
start by applying Newton’s law of motion to an infinitesimal element of
fluid. The continuity equation, for an element of fluid of constant density
is
∂ρ
+ ∇ · (ρu) = 0 (3.7)
∂t
and we multiply this by u:

∂(ρu) ∂(ρu2 ) ∂(ρuv) ∂(ρuw)

+ + + = fx (3.8)
∂t ∂x ∂y ∂z
∂(ρv) ∂(ρvu) ∂(ρv2 ) ∂(ρvw)
+ + + = fy (3.9)
∂t ∂x ∂y ∂z
∂(ρw) ∂(ρwu) ∂(ρwv) ∂(ρw2 )
+ + + = fz (3.10)
∂t ∂x ∂y ∂z
(3.11)
which for a constant density reduces to
 
∂
ρ +u·∇ u = f (3.12)
∂t
If we express our body force per unit volume ρg, we arrive to the
Cauchy equation of motion
D ui ∂τij
ρ = ρgi + (3.13)
Dt ∂x j
where the stress tensor τij includes all surface forces. Using the constitu-
tive equation for a Newtonian fluid, we now arrive to the Navier-Stokes
equation
D ui
ρ = ρg − ∇ p + µ∇2 u (3.14)
Dt

Page 9
which reduces to the Euler equation under the assumption of frictionless
flow
D ui
ρ = ρg − ∇ p (3.15)
Dt

Page 10
3.3.1 Motion in a rotating frame of reference
Eq.3.28 is valid for an inertial or fixed frame of reference. But in GFD we
measure positions and velocities relative to a frame of reference fixed on
the surface of the Earth, which rotates w.r.t. to a frame inertial.
Let’s have a frame of reference (x1 , x2 , x3 ) rotating at a uniform angular
velocity Ω w.r.t. a fixed frame (X1 , X2 , X3 ). Any vector P is represented by
P = P1 i1 + P2 i2 + P3 i3 (3.16)
For a fixed observer, the directions of the rotating unit vectors (i1 , i2 , i3 )
change with time. The time derivatives of P is thus
 
dP d
= ( P1 i1 + P2 i2 + P3 i3 ) =
dt I dt
(3.17)
dP1 dP2 dP3 di1 di2 di3
i1 + i2 + i3 + P1 + P2 + P3
dt dt dt dt dt dt
For an observer rotating with (x1 , x2 , x3 ) the rate of change of P is equal
to the first three terms in Eq.3.17, and so
   
dP dP di di di
= + P1 1 + P2 2 + P3 3 (3.18)
dt I dt R dt dt dt

X1

x1 Ω

x2
i1

i2
X2

i3

x3
X3

Figure 3.1: Coordinate frame (x1 , x2 , x3 ) rotating at angular velocity Ω with

respect to a fixed frame (X1 , X2 , X3 ).

Page 11
 Each unit vector i traces a cone with radius sin α, where α is a constant
angle. i changes in time dt as di = sin αdθ which is the length travelled by
the top of i. The rate fo change is thus
 
di dθ
= sin α = sin αΩ (3.19)
dt dt

The direction of the rate of change is thus perpendicular to the plane

(Ω, i), hence
di
= Ω×i (3.20)
dt
for any rotating vector i, giving us
   
dP dP
= +Ω×P (3.21)
dt I dt R

Applying this rule to the position vector r

   
dr dr
= +Ω×r (3.22)
dt I dt R
or
u I = uR + Ω × r (3.23)
Applying this rule to the velocities
   
duI duI
= + Ω × uI (3.24)
dt I dt R

 
duI d
= (u R + Ω × r) R + Ω × (u R + Ω × r)
dt I dt
    (3.25)
duR dr
= +Ω× + Ω × u R + Ω × ( Ω × r)
dt R dt R

Hence, accelerations in the two frames are relates as

   
duR duI
= − 2Ω × uR − Ω2 r (3.26)
dt R dt I

The second term of the r.h.s is the Coriolis acceleration and the last term
the centripetal acceleration. This last term is added to the Newtonian grav-
ity as an effective gravity
g = gn + Ω2 r (3.27)

Page 12
The apparent force Ω2 r will be zero at the poles.
The momentum equations are now
Du 1
= g − ∇ p + ν∇2 u − (2Ω × u) (3.28)
Dt ρ
It is clear that the Coriolis force (−2Ω × u) will deflect a particle to the
right of its direction in the northern hemisphere (right-hand rule). As the
Coriolis force constantly acts normal to the fluid path, it will not accelerate
the particle (in fact, Coriolis does not play any role in the energy equation).

3.3.2 Thin shell approximation

A scale analysis of the continuity equation reveals that, for typical length
scales much larger than typical vertical scales, L ≫ H, horizontal veloci-
ties must be much larger than the vertical ones, U ≫ W.
Now, decomposing the angular velocity vector into its three compo-
nents (Fig.3.2), we have
Ωx = 0
Ωy = Ω cos θ
Ωz = Ω sin θ

The Coriolis term, assuming U ≫ W, has the following components

h i
2Ω × u = 2Ω (−v sin θ )î + (u sin θ ) ĵ − (u cos θ )k̂ (3.29)

and defining the Coriolis parameter as f = 2Ω sin θ, which is now clearly

twice the angular velocity and hence a (planetary) vorticity
2Ω × u = (− f v)î + ( f u) ĵ − (2Ωu cos θ )k̂ (3.30)
But the vertical component of the Coriolis force, 2Ωu cos θ, is negligible
compared to the dominant terms in the vertical equation of motion, namely
the pressure gradient and the gravitational acceleration. Our final set of
momentum equations reduces to
Du 1 ∂p
− fv = − + ν ∇2 u (3.31)
Dt ρ ∂x
Dv 1 ∂p
+ fu = − + ν ∇2 v (3.32)
Dt ρ ∂y
Dw 1 ∂p
= − − g + ν ∇2 w (3.33)
Dt ρ ∂z

Page 13
 Ω
z

θ
Ωy Ωz

Figure 3.2: Components of the angular velocity vector for a point on the sphere.

Or
Du 1
+ 2Ω × u = − ∇ p − g + ν∇2 u (3.34)
Dt ρ
where ν = µ/ρ is the kinematic viscosity.

3.3.3 The β-plane

A first approximation is to set the Coriolis parameter, f , to a constant
value. This approximation, denoted the f -plane, is useful in some very
idealized studies when the westward propagation of disturbances is not
of interest or it is purposely neglected. But for large-scale dynamics it
is not appropriate, when flows occurring over large horizontal scales are
of interest. Rossby waves depend on variations of f , it is their restoring
mechanism, for example, and the large-scale dynamics of the ocean will
this be affected by latitudinal variations in the Coriolis parameter. An ap-
proximation can be done, however, to make equations more tractable, and
it consists of considering a cartesian plane over which f does vary, so ne-
glecting spherical coordinates.
The β-plane approximation is useful to avoid the sphericity and stay-
ing in a cartesian plane, yet retaining the dynamical effects of sphericity
itself.

Page 14
Figure 3.3: A cartesian reference system (x,y,z) and its associated spherical sys-
tem (r, θ, ϕ) around the point (a, θ0 , ϕ0 ). The plane z = 0 (or β-plane) is tangent
to the sphere around the point (a, θ0 , ϕ0 ). The approximation tan(θ − θ0 ) ≈
(θ − θ0 ) is well justified for small variations in latitude. On the β-plane, the ro-
tation vector is kΩ sinθ, where sinθ ≈ sinθ0 + (y/a)cosθ0 .

The plane z = 0, what will be called the β-plane is tangent to the sphere
in (a, θ0 , ϕ0 ). For small variations in latitude we can approximate tan(θ −
θ0 ) ≈ θ − θ0 . hence, our meridional cartesian coordinate is
y = a ( θ − θ0 )
z = r−a
x = (ϕ − ϕ0 ) a cosθ0 ,
where r is the distance of the fluid from the center of the sphere, θ is the
latitude, ϕ the longitude, and a is the radius of the Earth.
Hence, latitude θ is a linear function of y
y
θ = θ0 + . (3.35)
a
Now, for small variations in latitude we have:
y
sinθ ≈ sinθ0 + cosθ0 , (3.36)
a
as a truncated series around θ0 . And we can express f as the following:
2Ω
f = 2Ωsinθ = 2Ωsinθ0 + cosθ0 y = f 0 + β y. (3.37)
a

Page 15
 1.0 3.5
3.0
0.5 2.5

β [*10 −11 ]
f/(2 ∗ Ω)

2.0
0.0
1.5

−0.5 1.0
0.5
−1.0 0.0
90S 60S 30S 0 30N 60N 90N 90S 60S 30S 0 30N 60N 90N
Latitude Latitude

Figure 3.4: The Coriolis parameter f and its meridional gradient β as a function
of latitude.

Where we have introduced β = 2Ω a cosθ0 .

But what is β? We have gone from f = 2Ωsinθ to f = f 0 + βy. The
dependence of f on latitude is conserved because of the linear relation-
ship between f and y. This is an important result: we are not working
on spherical coordinates but the dynamical effects of sphericity are re-
tained.
β is called the gradient of planetary vorticity given that:

∂f 1 ∂f 2Ω
( θ = θ0 ) = ( θ = θ0 ) = cosθ0 = β. (3.38)
∂y a ∂θ a

Typical mid-latitude values for f and β are 10−4 s−1 and 10−11 m−1 s−1
(Fig. 3.4).
In conclusion, we have the β-plane approximation as

f = f 0 + βy (3.39)

For relatively large areas, with θ varying over a few tens of degrees, be-
tween mid-latitudes and the equator, the tangent plane approximation is
called β-plane. This approximation is only valid if

βy
βy ≪ f 0 or ≪ 1. (3.40)
f0

For even smaller variations in θ the f -plane is used, where

f = f 0 = 2Ω sinθ0 . (3.41)

Page 16
3.4 Vorticity and Circulation

Page 17
3.5 Kinematical and dynamical approximations
3.5.1 Hydrostatic balance
The vertical component (the component parallel to the gravitational force,
g) of the momentum equation is

Dw 1 ∂p
=− − g, (3.42)
Dt ρ ∂z

where w is the vertical component of the velocity and g = − gk. If the

fluid is static the gravitational term is balanced by the pressure term and
we have
∂p
= −ρg, (3.43)
∂z
which is called the hydrostatic balance, or hydrostasy. Scaling analysis shows
that the hydrostatic balance is the dominant balance within the vertical
momentum equation , so long as the vertical length scales of motion are
much smaller than the horizontal length scales. Such scales are relevant for
large-scale ocean climate modeling, and global ocean models typically as-
sume a hydrostatic balance, and this constitutes a basic assumption of the
primitive equations. Integrating the hydrostatic balance vertically from
the ocean surface η determines the pressure at a point in the ocean column
Z η
p(z) = p a + g dz′ ρ(z′ ), (3.44)
z

where p a is the sea surface pressure resulting from external forcing (e.g.,
atmospheric loading, sea ice, ...).

Scaling and aspect ratio

For a Boussinesq fluid, the momentum equations are

Du
+ f × u = −∇ϕ (3.45)
Dt
Dw ∂ϕ
= − + b, (3.46)
Dt ∂z
where ϕ = p/ρ0 and buoyancy b = − gρ/ρ0 . In the case of f = 0 the
horizontal momentum equation reduces to

Du
= −∇ϕ (3.47)
Dt

Page 18
and a scaling for the horizontal equation is

U Φ LU
∼ , or ∼ Φ, or U 2 ∼ Φ. (3.48)
T L T
Using mass conservation to scale vertical velocities we obtain

∂w
∇z · u + = 0. (3.49)
∂z
A scaling of this equation is

U W
+ =0 (3.50)
L H
H
W = U = αU (3.51)
L
where α ≡ HL is the aspect ratio between the typical horizontal and vertical
scales. The advective terms in the vertical momentum equation scale as

Dw W U U H U2 H
∼ = W = ( U) = 2 . (3.52)
Dt T L L L L
Now we can use the scaling for the horizontal and vertical motions, to-
gether with the aspect ratio of their typical scales, to reveal the condition
for hydrostasy.
For hydrostatic balance to hold, the ratio of advective terms to the pres-
sure gradient term in (3.46) must be

| DDwt |
≪1 (3.53)
| ∂ϕ
∂z |

This implies that

| DDwt | U 2 H/L2  H 2
∼ ∼ ≪ 1. (3.54)
| ∂ϕ U 2 /H L
∂z |

In other words, the aspect ratio should be

 H 2
2
α ≡ ≪1 (3.55)
L
for the advective terms in the vertical momentum to be neglected. The
hydrostatic balance is then a small aspect ratio approximation.

Page 19
3.5.2 Hydrostatic approximation
3.5.3 Shallow water approximation
3.5.4 Boussinesq approximation
3.5.5 Rigid lid approximation

3.6 Rossby number

We consider the dynamical balance in the horizontal components of the
momentum equation. In the horizontal plane (along geopotential sur-
faces) we find that the Coriolis term is much larger than the advective
terms and the dominant balance is between Coriolis and the horizontal
pressure force. The balance is called geostrophic balance, and it occurs when
the Rossby number is small.
The horizontal momentum equation is

∂u 1
+ (v · ∇)u + f × u = − ∇z p, (3.56)
∂t ρ

where v = (u, v, w) and u = (u, v, 0). A scaling analysis of the second

(U 2 /L) and third ( f U) terms, where U is the approximate magnitude of
the horizontal velocities and L is a typical length scale over which that
velocity varies, reveals the importance of rotation. The ratio of the sizes of
the advective and Coriolis terms defines the Rossby number:

U
Ro ≡ (3.57)
fL

The Rossby number characterizes the importance of rotation in a fluid.

It is the ratio of the magnitude of the relative acceleration to the Corio-
lis acceleration, and it is of fundamental importance in geophysical fluid
dynamics.

3.7 Geostrophic and Thermal Wind Balance

If the Rossby number is sufficiently small, then the rotation term domi-
nates the nonlinear advection term, and if the time period of the motion
scales advectively (or there are no accelerations) then the rotation term also

Page 20
Figure 3.5: Schematic of a geostrophically balanced flow with a positive
value of the Coriolis parameter f . Flow is parallel to the lines of constant
pressure. Cyclonic flow is anticlockwise around a low pressure region.
[from Vallis (2006)]

dominates the local time derivative. The only term that can then balance
the rotation term is the pressure term, leaving us with

1 ∂p
fv ≈ (3.58)
ρ ∂x
1 ∂p
fu ≈ − . (3.59)
ρ ∂y

This balance is known as geostrophic balance, and is one of the pillars of

geophysical fluid dynamics. We can now define geostrophic velocities as

1 ∂p 1 ∂p
f ug = − f vg = (3.60)
ρ ∂y ρ ∂x

and for flows with a low Rossby numbers, u ≈ u g and v ≈ v g .

A geostrophic flow is parallel to lines of constant pressure (isobars). If
f > 0, after a pressure gradient is initiated somehow, the fluid starts to
move down the gradient. Then, the fluid experiences the Coriolis force to
the right and therefore swings to the right. The fluid eventually moves
along isobars (along the slope, not down it), with the pressure force down

Page 21
the slope balanced by the Coriolis force up the slope. In the northern hemi-
sphere, the flow is anticlockwise round a region of low pressure and clock-
wise around a region of high pressure.
Consider now a plane horizontal flow in which density does not vary
along the fluid path (the Boussinesq approximation). In this case the con-
tinuity equation reduces to

∂u ∂v
+ = 0. (3.61)
∂x ∂y

We can now define a function ψ( x, y, t) such that

∂ψ
u ≡ − , (3.62)
∂y
∂ψ
v ≡ , (3.63)
∂x
and Eq.3.61 is thus satisfied, and this is called a streamfunction.
Returning to our geostrophic balance, if the Coriolis force is constant
and if density does not vary in the horizontal, the geostrophic flow is hor-
izontally non-divergent

∂u g ∂v g
∇ z · ug = + = 0, (3.64)
∂x ∂y
p
and we may define a geostrophic streamfunction, ψg , by ψg ≡ fρ, and

∂ψ ∂ψ
ug ≡ − , vg ≡ . (3.65)
∂y ∂x

Thermal wind
Thermal wind balance arises when combining the geostrophic and hydro-
static approximations. They are useful in elucidating how temperature
differences in the horizontal can lead to vertical variations in geostrophic
velocities, hence the term thermal wind equations.
Taking the vertical derivative of the geostrophic equations for a Boussi-
nesq fluid

∂p
ρ0 f ∂ z u = − ∂ z (3.66)
∂y
∂p
ρ0 f ∂ z v = ∂ z . (3.67)
∂x

Page 22
 y
f>0
ρ1
ρ2
ρ3

Figure 3.6: Schematic of thermal wind balance in the northern hemisphere.

Shown are surfaces of constant density, or isopycnals. Density increases
with depth and latitude, ρ3 > ρ2 > ρ1 . The termal wind associated
with this density field is eastward, or out of the page, and decreases with
depth. The same eastward thermal wind velocity would have resulted in
the southern hemisphere, with ρy < 0 and f < 0.

Combining these with the hydrostatic balance, ∂z p = −ρg, and changing

the order of differentiation for p, gives

ρ0 f ∂ z u = g ∂ y ρ (3.68)
ρ0 f ∂z v = − g ∂ x ρ. (3.69)

These equations represent the thermal wind balance, and the vertical deriva-
tive of the geostrophic wind is the ‘thermal wind’. Thermal wind balance
says that the geostrophic velocity has a vertical thermal wind shear in
case where density has a horizontal gradient.
In general, zonally averaged ocean temperature decrease poleward due
to the differential heating received from solar radiation. Neglecting salin-
ity effects on density, this poleward reduction in temperature corresponds
to a poleward increase in density. Also, for a stably stratified fluid, density
increases with depth. In a zonally-averaged flow, ∂ x ρ = 0, and so thermal
wind reduces to
g
∂z u = ∂y ρ (3.70)
ρ0 f
This equation is telling us that, if temperature falls in the poleward direc-

Page 23
tion, ∂y ρ > 0, then the zonally-averaged thermal wind is eastward. Wind
shear also increases as we move upward in the ocean, ∂z u > 0, which
yields a surface intensified zonal velocity field. Thermal wind, although
diagnostic, represents a valid steady state balance of a frictionless rotat-
ing fluid. That is, in the presence of rotation, a flow can exist in steady
state with nonflat isopycnals. Vertical integration of the thermal wind re-
lation, along with knowledge of the geostrophic velocity at a point along
the integration path, allows for determination of the full geostrophic ve-
locity in terms of density. However, the baroclinic density field (with a
horizontal gradient) is related to the baroclinic component of the velocity
field through thermal wind balance. The barotropic flow component has
zero vertical shear.

3.8 The Rossby radius

The Rossby radius of deformation is a length scale of fundamental impor-
tance in atmosphere-ocean dynamics. It describes the horizontal scale at
which rotation effects become as important as buoyancy effects. For exam-
ple, in the first stage of an adjustment problem, first the disturbance has a
small structure and gravity dominates with a very large pressure gradient.
Later, as the perturbation spreads over a larger horizontal scale, Coriolis
becomes more important and of similar magnitude as the pressure gra-
dient, and thus rotation causes a response that is much different from a
non-rotating case.
Using a geostrophic flow, it is easy to show that the Rossby radius of
deformation, Ld , is
Ld = c/| f | = ( gH )1/2 /| f | (3.71)
where c is the phase speed of the gravity wave. For the deep ocean, where
H= 4 km and c= 200 m/s, the Rossby radius is about 2000 km. Which is
much larger than depth, so the hydrostatic approximation is valid. How-
ever, the ocean is not only in rotation but also stratified, and so what is
more important is not the barotropic radius of deformation but rather the
baroclinic ones
Ld = cn /| f | (3.72)
where cn are baroclinic gravity wave phase speeds. So the Rossy radius is
directly related to the phase speed of long, baroclinic gravity waves, which
is also a very useful parameter in the study of ocean wave dynamics. A
global atlas of the first baroclinic gravity-wave phase speed, c, has been

Page 24
computed on a 1-degree global grid from observations (Fig. 3.7) as follows
Z 0
1
cn ∼ N dz (3.73)
nπ −H

where N is the buoyancy frequency.

Now, the first baroclinic Rossby radius, given that c1 =1-3 m/s, is Ld ∼10-
30 km with values increasing towards low latitudes (Fig. 3.8). Mesoscale
eddies have the size of the first baroclinic Rossby radius, therefore in order
to resolve mesoscale eddies and associated fluxes, ideally an ocean model
should have at least two grid points within Ld . It is clear from Fig. 3.9
that standard global ocean model can resolve mesoscale fluxes up to ∼25◦ ,
poleward of that latitude fluxes need to be parameterized. Benefits of hav-
ing fine-resolution ocean models is illustrated in Fig. 3.10, where eddies
and filaments are ubiquitous in the fine-resolution version of the model
whereas a laminar ocean is simulated in the (standard!) 1◦ version.

Page 25
Figure 3.7: A global map of the first baroclinic gravity wave phase speed
and its zonal mean. [data from Chelton et al., 1998]

Page 26
Figure 3.8: A global map of the first baroclinic Rossby radius of deforma-
tion and its zonal mean. [data from Chelton et al., 1998]

Page 27
 R. Hallberg / Ocean Modelling 72 (2013) 92–103 93

Fig. 1. The horizontal resolution needed to resolve the first baroclinic deformation radius with two grid points, based on a 1/8! model on a Mercator grid (Adcroft et al., 2010)
Figure 3.9: The oceanic resolution needed to resolve the Rossby Radius of
on Jan. 1 after one year of spinup from climatology. (In the deep ocean the seasonal cycle of the deformation radius is weak, but it can be strong on continental shelves.) This
model uses a bipolar Arctic cap north of 65!N. The solid line shows the contour where the deformation radius is resolved with two grid points at 1! and 1/8! resolutions.
deformation in an ocean model [from Hallberg et al., 2013].
of spin-up from climatology. At the coarse resolution that is typical 2. The test configuration and model
of the ocean components of CMIP5 coupled climate models (nom-
inally 1! resolution), an ocean model only resolves the deformation Phillips (1954) analyzed the baroclinic instability that arises in
radius in deep water in a narrow band within a few degrees of the a simple two-layered quasigeostrophic model of a geostrophically
equator; any important extratropical eddy effects will need to be sheared flow in a reentrant channel. This problem has the advan-
parameterized. At a much higher resolution, such as a 1/8! Merca- tage that many of the properties of the eddies, including necessary
tor grid, the deformation radius is resolved in the deep ocean in the conditions for the growth of instabilities, the growth rate, energet-
tropics and mid-latitudes, but even in this case eddies are not re- ics and vertical structure of the exponentially growing linear
solved on the continental shelves or in weakly stratified polar lat- modes can be calculated analytically, as has been documented in
itudes. An unstructured and adaptive grid ocean model could help many textbooks on geophysical fluid dynamics (e.g. Pedlosky,
to address this issue, but such models are not yet in widespread 1987; Vallis, 2006).
use for global ocean climate modeling, and even then computa- This study examines instabilities of a stacked shallow water
tional speed may dictate the use of models that do not resolve variant of the Phillips problem, which is described by the momen-
mesoscale eddies everywhere. tum and continuity equations:
In this paper, a series of numerical simulations of a variant of $ %
@un " #
^ % r & un & un ¼ 'r M n þ 1 kun k2
the Phillips (1954) model of baroclinic instability are used to þ f þk
@t 2
examine the effects of resolution on a numerical model’s ability
' r % T ' dn2 cD ku2 ku2 ; ð1Þ
to exhibit the net overturning circulation driven by mesoscale ed-
dies. The effects of a commonly used parameterization of eddy ef- h " # " #i
fect, both on the models’ explicitly resolved eddies and on the net @hn
þ r % ðhn un Þ ¼ ð3 ' 2nÞ c g3=2 x ' g3=2;Ref ' r % K h rg3=2 :
overturning, are examined. Based on these results, a simple pre- @t
scription is offered for the typical situation in global ocean mod- ð2Þ
els, where eddies are resolved in only part of the domain and in Here un is the horizontal velocity in layer n, where n = 1 for the
that portion it is desired that the model be allowed to explicitly top layer and n = 2 for the bottom layer. hn ¼ gn'1=2 ' gnþ1=2 is the
simulate their effects, but in the remainder of the domain that thickness of layer n, which is bounded above and below by inter-
eddies be entirely parameterized. Specifically, the eddy diffusivi- faces at heights gn'1=2 and gnþ1=2 . These equations are solved in a
ties should be multiplied by a ‘‘resolution function’’, ranging from 2000 m deep channel that is 1200 km long and reentrant in the
Figure 3.10: The same ocean model at different horizontal resolutions, in-
0 to 1, of the ratio of the baroclinic deformation radius to the
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi x-direction, and 1600 km wide in the y-direction with vertical
model’s effective grid spacing, D e ¼ ðDx2 þ Dy2 Þ=2. The resolu- walls at the northern and southern boundaries. The Coriolis param-
creasing from right to left.
tion function that works best for the cases presented here rapidly eter, f, varies linearly in the y-direction between 6.49 & 10'5 s'1
makes a transition from 1 when this ratio is greater than a value and 9.69 & 10'5 s'1, following the common b-plane approxima-
of about 2 (the exact value is not very important and can be cho- tion. The horizontal stress tensor, T, is parameterized with a shear
sen to be higher) to 0 for larger values. In the idealized case pre- and resolution dependent Smagorinsky biharmonic viscosity (Grif-
sented here, this prescription is found to give a reasonable fies and Hallberg, 2000). The Montgomery potentials, Page 28
representation of the net eddy-driven overturning over a wide M n ¼ p=q0 þ gz, in the two layers are given by a vertical integration
range of resolutions. of the hydrostatic equation, so that
Let’s now go a little ahead of ourselves. Consider that the Corio-
lis parameter is not constant and is actually a function of latitude
f (y). The nondivergent condition ∇ · ( f u) = 0 is satisfied by the
geostrophically balanced flow. Cross-differentiating Eq.3.60 gives

∂f
v g + f ∇ z · ug = 0 (3.74)
∂y

Using mass continuity leads to

∂w
βv g = f , (3.75)
∂z
∂f
where β ≡ ∂y . This is a geostrophic vorticity balance, also called
Sverdrup balance. In a Sverdrup balance, the vertical velocity re-
sults from an external agent, most notably wind stress. It states that
the vertical shear in the vertical velocity balances a meridional cur-
rent, with the Coriolis parameter f and the planetary vorticity gra-
dient β determining the sense and strength of the meridional flow.
A vertical velocity shear arises when there is a nonzero curl in the
wind stress acting on the ocean surface. Vorticity is then transferred
to the ocean via frictional effects causing Ekman pumping or suction.
These effects alter the vertical structure of the vertical velocity and,
through Sverdrup balance, induce a meridional flow.

Page 29
3.9 The shallow-water equations
To describe large-scale oceanic, and atmospheric, motions, where the hor-
izontal scale is much larger than the vertical scale, we can use a set of
simplified equations that retain the necessary ingredients of the fluid mo-
tion but use some useful approximations. We will thus consider a fluid
in hydrostatic balance of constant density and, for simplicity, we will also
consider a flat bottom. The necessary condition of the shallow-water equa-
tions is that the horizontal length scale must be much larger than the ver-
tical scale over which the fluid develops so that L >> H.
If the fluid is in hydrostatic balance
∂p
= −ρg. (3.76)
∂z
Then the total pressure will be

p( x, y, z, t) = −ρgz + p′ . (3.77)

Pressure must vanish at the surface, so that p = 0 at z = η

p = p0 + p ′ = 0 (3.78)

and at z = η we have
p′ = ρgη (3.79)
Our total pressure will then be

p( x, y, z, t) = ρg(η ( x, y) − z) (3.80)

This means that the horizontal gradient of pressure, and the flow, is inde-
pendent of depth
∇ p = ρg∇η (3.81)
and the horizontal momentum equations reduce to
Du 1
= − ∇ p = − g∇η (3.82)
Dt ρ
We can now easily add rotation to our shallow-water momentum equa-
tions
Du 1
+ f × u = − ∇ p = − g∇η (3.83)
Dt ρ
The continuity equation is obtained by the mass balance within an in-
finitesimal column of fluid. The mass flux passing through a section of the

Page 30
 η
Fm

h H

Z=0 δy
δx

Figure 3.11: Schematic of a flat-bottomed shallow-water system and mass

balance within a column of fluid.

column is Fm = ρu( H + η )δy and the difference between the fluxes into
and out of the section is given by

∂
δxδy [ρu( H + η )] (3.84)
∂x
Considering the total volume, the net rate of change is

∂h ∂ ∂
+ [u( H + η )] + [v( H + η )] = 0 (3.85)
∂t ∂x ∂y

which is the new continuity equation for the shallow-water system

∂h ∂ ∂
+ (uh) + (vh) = 0 (3.86)
∂t ∂x ∂y

∂h
+ ∇ · (uh) = 0 (3.87)
∂t
and if the perturbation is small and H is constant, mass continuity reduces
to the linear equation
∂η
+ H∇ · u = 0 (3.88)
∂t
If there is flux by advection this is balanced by a net increase in mass and
an increase in height, giving rise to a vertical velocity, so that the mass
convergence is balanced by the increase in height allowing for a dynami-
cal surface elevation. This will be the basis for the propagation of waves
within the rotating shallow-water system.

Page 31
Exercices
1. Use ϕ = p/ρ0 and the definition of buoyancy b = − gρ/ρ0 to rewrite
the hydrostatic balance and thermal wind equations.

2. Where is thermal wind velocity directed in the southern hemisphere,

considering a poleward increasing (decreasing) density (tempera-
ture)? (see Fig. 3.12)

3. How is thermal wind shear changed as we approach the poles?

Figure 3.12: Zonal-mean potential density in the latitudes of the Drake

Passage.

Page 32
Bibliography

Vallis, G. K., Atmospheric and Oceanic Fluid Dynamics, Cambridge Univ.

Press, 745 pp, 2006.

33