Rotating, Homogeneous,

Incompressible Fluids 5

5.1 Equations of Motion for an Incompressible Homogeneous

Fluid

In Chap. 3, the governing equations applied to a homogeneous (ρ = constant),

incompressible (Dρ/Dt = 0) fluid in a rotating, non-inertial frame of reference were
derived. For an incompressible, homogeneous fluid these Eqs. (3.1b) and (3.56) are
simplified to become
∇ ·u =0 (5.1)[3.1b]
and
Du 1
+ 2Ω × u = − ∇ p + ν∇ 2 u. (5.2)[3.56]
Dt ρ
As a side product of these governing equations, an equation for the vorticity
balance was also derived. Simplifying again for homogeneous, incompressible fluids,
the vorticity Eq. (3.68) reads:

Dω A
= ω A · ∇u + ν∇ 2 ω A (5.3)[3.68]
Dt
where ω A = ω + 2Ω has been referred to as absolute vorticity. In addition, an equa-
tion governing the circulation around a closed curve was also derived, expressing
the balance of vorticity. For a homogeneous fluid, we have

dΓ d Sp
= ν ∇ 2 ω · n d S − 2Ω (5.4)[3.80]
dt S dt

where Γ is the circulation around the closed curve C enclosing the surface S, and S p
is the projection of the surface S on the plane perpendicular to the angular velocity
vector Ω. We recall the above result as Kelvin’s circulation theorem.

© Springer Nature Switzerland AG 2020 89

E. Özsoy, Geophysical Fluid Dynamics I, Springer Textbooks in Earth Sciences,
Geography and Environment, https://doi.org/10.1007/978-3-030-16973-2_5
90 5 Rotating, Homogeneous, Incompressible Fluids

Finally, Bernoulli’s theorem was stated for steady flow of a homogeneous incom-
pressible fluid, based on
∇ H = u × ω A + ν∇ 2 u (5.5a)[3.83a]
where

1 p 1
H= u·u+ − g · x − (Ω × x) · (Ω × x). (5.5b)[3.83b]
2 ρ 2

Note that p  represents the fluid pressure in (5.5b), while the notation p in Eq.
(5.2) represents the modified pressure, replacing the last three terms of (5.5b).
These basic equations will be considered in this volume, since our subject will only
cover the motion of homogeneous, incompressible fluids with respect to a rotating
frame of reference. We will elaborate the effects of rotation, leading to important new
types of behaviour of geophysical fluids (i.e. in fluid systems on a rotating earth).
The effects due to density inhomogeneity of the fluid (stratification) are deliberately
omitted in this volume, and will be studied under the title of Stratified Fluids, later
in this series.

5.2 Inviscid Rotating Flows

In this section, we first study the relatively simpler class of motions where the viscous
effects are to be ignored, i.e. by setting viscosity ν = 0 in Eqs. (5.1) and (5.2). Without
loss of generality, we assume the angular velocity vector is aligned with the z-axis
in a Cartesian coordinate system (x, y, z) with unit vectors (ı, j , k), i.e., Ω = Ωk.

∇ ·u =0 (5.6a)

Du 1
+ 2Ωk × u = − ∇ p (5.6b)
Dt ρ
We consider flows bounded by a container with solid boundary B with normal
vector n, where the normal velocity component should vanish:

u · n = 0 on B (5.6c)

The effects of rotation are expressed by the Coriolis term (second term in Eq. 1.2),
since the centrifugal force has already been included in the modified pressure p of
Eq. (5.2). We will now demonstrate novel effects in rotating fluid motion arising due
to this apparently minor modification of the governing equations. One of the most
important of these effects is the elasticity created in rotating fluid motion. This effect
is important, because the presence of a restoring mechanism allows particular types
of wave motions to be supported.
5.2 Inviscid Rotating Flows 91

In particular, if we seek oscillatory solutions in time, where we assume

u = Ueiωt
p = φeiωt ,
we obtain
∇ ·U=0 (5.7a)

1
iωU + 2Ωk × U = − ∇φ. (5.7b)
ρ

U · n = 0 on B (5.7c)
We can then transpose these equations in better manageable form by first taking
the divergence and later the curl of (5.7b). Firstly, the divergence results in

1
iω∇ · U + 2Ω∇ · (k × U) = − ∇ 2 φ
ρ

and making use of (5.7a) and vector identity (1.27c) yields

∇ 2 φ − 2Ωρk · ∇ × U = 0 (5.8a)

Secondly, the curl gives

1
iω∇ × U + 2Ω∇ × (k × U) = − ∇ × ∇φ ≡ 0
ρ

by vector identity (1.27i). Then, by using vector identity (1.27d), one obtains

iω∇ × U − 2Ω(k · ∇)U = 0. (5.8b)

Multiplying (5.8a) by iω and (5.8b) by 2Ωρk and adding together, with some
cancellations, gives

iω∇ 2 φ − (2Ω)2 ρk · (k · ∇)U = iω∇ 2 φ − (2Ω)2 ρk · (k · ∇)k · U = 0. (5.9)

Then dot product of k with (5.7b) yields

1
iωk · U = (k · ∇)φ
ρ

which can be substituted in (5.9) to obtain

 2
2Ω
∇ φ−
2
(k · ∇)2 φ = 0. (5.10)
ω
92 5 Rotating, Homogeneous, Incompressible Fluids

∂
Noting that k · ∇ = ∂z , the open form of (5.10) can now be written as

∂2φ ∂2φ 2 ∂ φ
2
+ + (1 − λ ) = 0, (5.11a)
∂x 2 ∂ y2 ∂z 2

where
2Ω
λ=.
ω
The boundary condition (5.7c) can be transposed in terms of φ by first dot multi-
plying (5.7b) by k × n and making use of vector identities (1.11a) and (1.11c), then
combining with the boundary condition (5.7c), to yield

− λ2 n · ∇φ + 4(n · k)(k · ∇φ) + 2iλ(k × n) · ∇φ = 0 on B. (5.11b)

(5.11a) is the governing equation with prescribed boundary condition (5.11b) applied
to the present case of inviscid, homogeneous, unsteady motions.
The striking feature that emerges from Eqs. (5.11a,b) is the possible change of
regime that we may expect from the studied flows in respect to the parameter λ. In
particular, the nature of the solution will depend on the value of λ = 2Ω ω , which is
in the form of an inverse ratio of the frequency to planetary (inertial) frequency. It is
immediately observed that the equations are of parabolic (Laplacian) form if λ < 1
i.e. for the super-inertial frequencies ω > 2Ω. On the other hand, the equations are
of hyperbolic form if λ > 1 i.e. for the sub-inertial frequencies ω < 2Ω.
Based on these results, it is expected that the solutions would be smooth in
parabolic regime, ω > 2Ω. On the other hand, for values of λ > 1, the flow will
be in hyperbolic regime which allows wave-like solutions. For instance, an oscilla-
tory source placed at (0, 0, 0) with small displacements aligned with the z-coordinate
in the hyperbolic regime would create wave-like solutions spreading along charac-
teristics in the form of cones

z = ±λ(4 − λ)1/2 (x 2 + y 2 )1/2 .

5.2.1 Inertial Motion—Unsteady (Periodic) Uniform Flow

To see the restoring mechanism of the Coriolis term in more detail, consider an
inviscid fluid (ν = 0) with vanishing pressure gradients (∇ p = 0). In the absence
of a pressure gradient, the flow is uniform in space, subject to the fictitious Coriolis
force. Equation (5.2) becomes

Du
= −2Ω × u, (5.12)
Dt
i.e., the fluid acceleration is balanced only by the restoring Coriolis force (per unit
mass) −2Ω × u. Without loss of generality, we assume that the angular velocity
5.2 Inviscid Rotating Flows 93

Fig. 5.1 Coriolis y

acceleration applied to a
particle moving with velocity
uh in a rotating frame with
angular velocity Ωk about
the perpendicular axis

uh
−2Ωk × uh

Ωk
x
z

vector is aligned with the z-axis in a Cartesian coordinate system (x, y, z) with unit
vectors (ı, j , k), i.e., Ω = Ωk. Then, the only component of velocity u = (u, v, w)
contributing to the right hand side of (5.6) is uh = (u, v, 0) in the plane perpendicular
to Ω = Ωk), so that u can be replaced by uh . The direction of the restoring force
is at right angles to the lateral component of fluid velocity uh , and its sense is to the
right of this vector:
For a uniform flow in infinite domain, the nonlinear advection terms can be
neglected, so that (5.12) becomes

∂uh
+ 2Ωk × uh = 0 (5.13)
∂t
Cross multiplying with −2Ωk and adding with the time derivative of the above
equation yields
∂ 2 uh
+ (2Ω)2 uh = 0 (5.14)
∂t 2
This equation has sinusoidal solutions (harmonic motion), analogous to a spring-
mass system. For example, we can use the initial conditions (Fig. 5.1):

uh (0) = U0 (5.15a)
∂u h
(0) = −2Ωk × U0 (5.15b)
∂t
94 5 Rotating, Homogeneous, Incompressible Fluids

Fig. 5.2 Inertial motion y − yc

y
U0
(x0 , y0 ) α

yc x − xc
(xc , yc )

Ωk
x
z xc

The solution follows as

uh = U0 cos 2Ωt − k × U0 sin 2Ωt. (5.16)

The components (u, v), of the velocity uh = (u, v, 0) are

u = U0 cos(2Ωt − α) (5.17a)
v = −U0 sin(2Ωt − α) (5.17b)
where α is the angle that the initial velocity vector U0 makes with the x-axis, and
U0 = U0 .
For small amplitude motions, the displacements (x, y) of a material point (fluid
particle) with respect to its initial position (x0 , y0 ) can be obtained by integrating
(5.17a,b) with respect to time (Fig. 5.2):

U0
(x − x0 ) = [sin(2Ωt − α) + sin α] (5.18a)
2Ω
U0
(y − y0 ) = [cos(2Ωt − α) − cos α] (5.18b)
2Ω

These can be combined by eliminating t between (5.18a,b), to yield

 2
U0
(x − xc )2 + (y − yc )2 = (5.19)
2Ω

where xc , yc are appropriate constants determined from (5.18).

5.2 Inviscid Rotating Flows 95

π π
t2 = 4Ω t2 = 4Ω

u1 π
t1 = 0 t3 = 2Ω
y y

u2 u3 u2

Ωk Ωk
x x
z z
Fig. 5.3 Demonstration of elasticity in a rotating fluid by the action of Coriolis force as restoring
agent

In the above solutions, both the sense of rotation of the velocity vector uh and the
trajectory (x(t), (y(t)) in (5.18) are in the clockwise direction. Each particle rotates
π
clockwise, and comes to its initial position after one inertial period TI = 2Ω 2π
=Ω .
This inertial motion demonstrates the restoring effects of the Coriolis force. Because
it arises due to the inertia, without any external (surface or body) forces, this motion
is considered to be a free oscillation in a rotating fluid, corresponding to a natural
frequency of 2Ω.

5.2.2 “Elasticity” in a Rotating Fluid—Restoring Effects of Coriolis

The previous example on inertial motion illustrates the restoring mechanism rotating
fluids. Since particles displayed return to their initial positions after one characteristic
(inertial) period, the fluid acts as if it has some special form of elasticity, whereby
particles are forced into closed circular trajectories.
To further demonstrate the elastic behavior, consider a closed material curve C
whose projection in the lateral plane (⊥ to Ω) is C p (Fig. 5.3).
Suppose that a motion is generated in the fluid such that it will cause a positive
rate of expansion in the lateral plane, i.e. with

∂u ∂v
∇h · u h = + > 0.
∂x ∂y

This outward motion along the material curve C is going to give rise to Coriolis
forces in the clockwise direction along the curve since the induced force is to the
right hand side of the motion (in the northern hemisphere). This is also seen exactly
by Eq. (5.4) (Kelvin’s theorem) since an increase in the projected area S p enclosed
by curve C leads to a negative contribution to the circulation. On the other hand,
clockwise motion along the material curve will give rise to Coriolis forces in the
96 5 Rotating, Homogeneous, Incompressible Fluids

inward direction (i.e. with ∇h · uh < 0) then the material line C will then tend to
contract.
Thus the fluid is seen to resist elastically to any motion that would cause displace-
ment of fluid elements leading to a change in the projection of an are enclosed by a
curve of such elements.
The relative importance of Coriolis effects is determined by the inverse of the
Rossby number R0 = U0 /L 0 Ω0 measuring the ratio of Coriolis terms to other inertial
terms (cf. Eq. 4.4). When Ro  1, the elasticity effect of rotation is expected to be
dominant.

5.2.3 Geostrophic Motion: Steady Flow with a Pressure Gradient

 
When the flow is steady ∂u ∂t = 0 , inviscid (ν = 0), and if the Rossby number
Ro  1, then the nonlinear term u · ∇u is negligibly small compared to the Coriolis
term 2Ω × u). In this limit the momentum Eq. (5.2) becomes

1
2Ω × u = − ∇ p (5.20)
ρ

Without loss of generality we can let Ω = Ωk be aligned with the z-axis of a

Cartesian coordinate system (x, y, z) with unit vectors (ı, j , k), such that

1
2Ωk × u = − ∇ p. (5.21)
ρ

The continuity equation

∇ ·u =0 (5.22)
complements the momentum equation. In principle (5.21) and (5.22) should be suf-
ficient to solve for the unknowns u and p. However, it turns out that the so-called
geostrophic motion by these equations have some very special characteristics.
First, by taking the curl of (5.21) we can show that

1
∇ ×k×u=− ∇ ×∇p ≡ 0 (5.23)
2Ωρ

by virtue of (1.27i). Then, by making use of (1.27d) the l.h.s. is

∇ × k × u ≡ k∇ · u − k · ∇u = 0, (5.24)

of which, the first term on the r.h.s. vanishes by (5.22). Then (5.24) states that

∂u
k · ∇u = =0 (5.25)
∂z
5.2 Inviscid Rotating Flows 97

Fig. 5.4 Flow bounded by

two rigid surfaces

Ωk

z h

expressing the fact that the velocity field has to be two-dimensional; u = u(x, y)
only. On the other hand, (5.21) dictates that

∂p
k·∇p = = 2Ωρk · (k × u) = 0, (5.26)
∂z

so that we find the pressure to be also two-dimensional, p = p(x, y) only.

The above results, namely that none of the flow variables depend on the vertical
coordinate z, indicates that the flow is essentially two-dimensional and occurs in the
(x, y) plane (Fig. 5.4).
By virtue of the above results, (5.21) and (5.22) can equivalently be written as

1
2Ωk × uh = − ∇h p (5.27)
ρ
∇h · uh = 0, (5.28)
where uh (x, y) = (u, v, 0) is the horizontal velocity vector, p = 
p(x, y) is the
 pres-
∂
sure, such that both variables are independent of z, and ∇h = ∂x , ∂∂y , 0 is the
horizontal gradient operator in the (x, y) plane.
The result, that the rotating flow at the limit Ro → 1 must be two dimensional,
is known as the Taylor–Proudman theorem. Consider flow bounded by two rigid
surfaces
Φ1 = z − f 1 (x, y) = 0, Φ2 = z − f 2 (x, y) = 0. (5.29a,b)
If a mound was placed in an otherwise constant depth motion, would bypass the
mound (Fig. 5.5).
Since the flow is two-dimensional, any fluid column that is initially vertical will
remain vertical. However, while moving, the net height of the column h would have
98 5 Rotating, Homogeneous, Incompressible Fluids

Fig. 5.5 Geostrophic motion

in a basin with variable
bottom topography has to
follow closed contours of
constant depth

h0

to adjust itself to the distance of separation between the two surfaces, requiring that
(Fig. 5.6)
DΦ1 DΦ2
= 0, =0 (5.30)
Dt Dt
Since Φ1 , Φ2 are material surfaces according the (1.41). Substituting (5.29a, b)
(Fig. 5.7):
DΦ1
= u · ∇Φ1 = uh · ∇h f 1 − w = 0, on z = f 1 (5.31a)
Dt

DΦ2
= u · ∇Φ2 = uh · ∇h f 2 − w = 0, on z = f 2 (5.31b)
Dt
Subtracting (5.31b) from (5.31a) and since uh = uh (x, y) only, we have

w |z= f1 −w |z= f2 = uh · ∇( f 1 − f 2 )
= uh · ∇h
(5.32)
Dh
=
Dt

On the other hand, since w = w(x, y) only ( ∂w ∂z = 0), the vertical velocity w at
the upper surface can not be different from that at the lower surface, i.e. the l.h.s. of
(5.32) must vanish, so that
Dh
=0 (5.33)
Dt
5.2 Inviscid Rotating Flows 99

h0

Fig. 5.6 Geostrophic motion in a container with sliced cylinder bottom topography is not possible,
since the flow has to follow constant depth contours

h0

Fig. 5.7 Geostrophic flow passing a mound has to follow constant depth

This result indicates that any moving fluid column must preserve its height in
geostrophic motion, i.e. the fluid column moves along a very special trajectory that
would make h = constant. In a closed container, this would mean that fluid columns
could only move along closed contours having h = const, if such closed contours
exist. If there are no such closed contours, geostrophic motion would not be possible.
Such columns which are identified with their constant thicknesses in geostrophic
motion are called Taylor columns, since G. I. Taylor was the first to discover them.
The flow modelled by Eqs. (5.21) and (5.22) [equivalently (5.27) and (5.28)] is called
100 5 Rotating, Homogeneous, Incompressible Fluids

Fig. 5.8 Isobars in weather

charts serve as streamlines
which are parallel to the flow

geostrophic flow. It is a steady approximation to the governing equations for inviscid,

homogeneous, incompressible rotating fluids in the limit Ro → 0.
On the other hand, it can immediately be seen that geostrophic flow is in fact
degenerate or indeterminate, i.e. while such a flow would exist, it is impossible to
obtain a “solution” to equations. This is seen if pressure is eliminated from (5.27),
by first rearranging such that
1
2Ωuh = k × ∇h p
ρ
then taking divergence of both sides
1
∇ · uh = ∇ · (k × ∇h p).
2Ωρ
Utilizing (1.27c) and (1.27i), the above equation is equivalent to
1
∇ · uh = − k · (∇h × ∇h p) ≡ 0,
2Ωρ
i.e. the same thing as Eq. (5.28). This result shows that both of the statements (5.27)
and (5.28) are equivalent to each other, i.e. one of the two equations is redundant.
Because there is only one independent equation with insufficient information to solve
for the two variables u and p, a simultaneous solution can not be obtained; which
shows that geostrophic flow is indeterminate or degenerate.
Since there are two unknowns uh and p in (5.27) and (5.28) it is only possible to
infer one of these fields from given values of the other field. For example if pressure
is given we can infer the velocity distribution, or vice versa. This diagnostic method
is often used for interpreting weather charts or interpreting hydrographic fields in
oceanography (Fig. 5.8).
Equation (5.27) can be put into the form
1
uh = k × ∇h p
2Ωρ
  (5.34)
p
= k × ∇h ,
2Ωρ
5.2 Inviscid Rotating Flows 101

Fig. 5.9 A ball of mass m

m
left to roll along a slope y
x
mg
α

so that the velocity vector is perpendicular to the pressure gradient, and its sense is
p
such that it takes high pressure to its right hand side. In fact, Ψ = ( 2Ωρ ) acts as the
stream function for the two dimensional flow; comparing (5.34) with (2.34). Around
low-pressure centers L, the flow is cyclonic, i.e. it rotates in the anti-clockwise sense;
and around high pressure centers H it is anti-cyclonic (i.e., rotation in clockwise
sense). We must finally note that, to remove the geostrophic indeterminacy, we must
include other effects in the dynamics, such as friction, unsteady variations, etc. The
inclusion of these effects can be in the form of small corrections if Ro  1, but
nevertheless they would render the equations determinate.

Exercises

Exercise 1
Consider a ball of mass m released from rest at the origin (x = 0, y = 0) on an
inclined plane as shown. The plane makes a small inclination angle α with the
horizontal. The coordinates (x, y) are aligned with the inclined plane, where x is
downward of the slope.
At the latitude φ where the experiment is performed, the Coriolis parameter is
calculated from f = 2Ω sin φ, reflecting the earth’s rotational effects. Linear friction
opposite to the direction of motion and proportional to the velocity of motion u =
(u, v) with respect to the coordinates (x, y) is represented by the friction factor k
(Fig. 5.9).
By simple mechanical arguments, we can show that the motion is governed by

du
− f v = g sin α − ku
dt
dv
+ f u = −kv
dt
Find the position of the ball as a function of time by solving the above equations
with initial condition u(t = 0) = 0.
For realistic values of parameters, assume the experiment is done at latitude φ =
30 ◦ N, on a slope of α = 10◦ and k = 0.01. What would be the space and time scales
of the motion?
Sketch the motion for different values of f and k and discuss the motion for the
following cases:
102 5 Rotating, Homogeneous, Incompressible Fluids

Fig. 5.10 Flow past a w(x)

mound between two
horizontal plates
y
H(x)
H0 U0 U (x)

 

(i) f = 0, k = 0;
(ii) f = 0, k = 0;
(iii) f = 0, k = 0;
(iv) f = 0, k = 0.
(v) What would change if the initial conditions would be changed? For instance,
what would be the trajectory if an initial force was applied to give initial acceleration
a0 to the small ball?
Exercise 2
Consider an incompressible fluid of constant density confined between two plates of
infinite horizontal extent. A mound of length 2 in the x-direction and uniform in
the y-direction is placed on the bottom, perpendicular to a flow with uniform speed
U0 in the x-direction. The total thickness of fluid between the solid boundaries is
H0 , constant outside the mound region, and H (x) at the mound (Fig. 5.10).
By making use of the governing equations with inviscid solid boundary conditions
and assuming small Rossby number and negligible friction, investigate if it would
be possible for the flow to pass over the mound.
Show that a flow with uniform profile U (x) adjusted to the mound would only be
able to pass over the mound if fluid was either sucked out or injected in at the upper
boundary, with a vertical velocity

d
w(x) = −U0 H0 (ln[H (x)]) in the region −  < x < , at y = H0 .
dx
Where is it, along x, that the fluid is sucked out and where is it injected in?
Exercise 3
The linearized momentum equation for an incompressible fluid rotating with angular
velocity Ω = Ωk about the z-axis, subject to linearized viscous friction can be
written as
∂uh
+ 2Ωk × uh + μuh = 0.
∂t
This equation would represent inertial motions modified by frictional effects, due
to the additional term.
5.2 Inviscid Rotating Flows 103

(i) Solve the above equation with the initial condition for horizontal velocity,
uh (0) = U0 .
(ii) Based on an oscillatory solution, what kind of motion is expected? Which
processes do terms in the governing equation represent, and how do they affect the
motion?
(ii) Determine the components u, v of the complex vector variable u = (u, v) =
u + iv as a function of time, and obtain linear displacements of a particle initially
placed at (x0 , y0 ) where the initial velocity is given as u0 .
Plot the velocity components and particle trajectory as a function of time. Compare
the cases μ = 0 and μ = 0.