26

5. LAMINAR FLOW: MORE APPLICA- rock is not very disturbed by the rock’s motion (that is,
TIONS the flow will vary with height, above the rock). Now
consider the same thing but with a rotating water tank.
We continue the previous discussion with more exam- The Taylor-Proudman result says that the flow at all
ples and applications from here and there. We start heights above the rock must be the same. This produces
with a simple discussion of flow in a (quasi) two- a Taylor column: the water above the rock forms a ver-
dimensional, rotating system – all sorts of fun things tical column which is fixed relative to the rock; water
happen when you’re rotating. We then look at two more not over the rock flows around this column, as though
mathematically involved examples of viscous flows. it were a solid object.
A. Geostrophic Flow ATMOSPHERIC CIRCULATION

What happens in a rotating system? Because our impor- One striking example of our first point is the at-
tant application is to terrestrial (geophysical) flows, we mospheric circulation around a pressure extremum:
can work in the incompressible limit. Referring back to the circulation is counterclockwise (seen from above)
chapter 1, the general equation of motion comes from around a low, and clockwise around a high – as shown
(1.20), but now has viscosity added, as per Chapter 2: in the figure.
Dv 1
= − ∇p + g + Ω2 R − 2Ω
Ω × v + ν∇2 v (5.1)
Dt ρ
pressure High pressure
We consider the limit which is interesting for terrestrial Low
coriolis coriolis

problems: steady (or nearly steady) flow, in which the

advective term (v·∇v) and the viscous term (ν∇2 v) are
small compared to the Coriolis force (2ΩΩ × v). In addi- Figure 5.1. Illustrating geostrophic flow around high and
tion we ignore the centrifugal force and gravity terms.1 low pressure centers. Following Kundu Figure 13.4.
We then have a situation in which the equation of mo-
tion simplifies to The jet stream is a related effect. To be specific, think
about the (north) polar jet stream. The atmosphere at
Ω × v + ∇p = 0
2ρΩ (5.2) the top of the world is cold, while that at lower latitidues
is warm; the hot and cold air masses meet in a region
This is known as geostrophic flow; such flows have in- called the polar front. The horizontal pressure gradient
teresting properites, as follows. across the polar front generates transverse winds, again
• The velocity is perpendicular to the pressure gra- by the Coriolis balance of (5.2); this large-scale circula-
dient. Thus, isobars are also streamlines; the flow is tion, from west to east, is called the polar jet stream. It
along lines of constant pressure. This is in marked typically lies at about 10 km altitude, has average speed
contrast to non-rotating systems, where we tend to ∼ 100 km/s, but can reach maximum speeds 300-400
think of pressure gradients driving the flows, or (from km/s. The northern hemisphere also has a sub-tropical
Bernoulli’s equation) think of pressure variations along jet, where the temperate mid-latitude air meets the hot,
streamlines. tropical air. These two jets are by no means steady –
• The flow must be two-dimensional. That is: take they wander about, and sometimes split, depending on
the curl of (5.2), to get ∇ × (ΩΩ × v) = 0. But if Ω k ẑ, the local thermodynamics of the atmosphere. Some of
say, this requires that dv/dz = 0. The flow cannot de- the wiggles and fluctuations in the jet streams are the
pend on z. This is known as the Taylor-Proudman theo- result of fluid instabilities – which we’ll discuss later in
rem, and will be important in our discussion of Rossby the course.
waves, below. It has one striking consequence: the Tay-
ROSSBY WAVES
lor column. Think about a rock on the bottom of a tank
of water. You tie a string to the rock and pull it through These waves exist only in particular geometries in a ro-
the water. If the tank is static, the water well above the tating system. For a simple version, consider a rotat-
ing layer of fluid, bounded by two planes which are not
1
Purists can note that both of these can be expressed as gradients
quite parallel. As in Figure 5.2, let the height of the
of a scalar: g = ∇Φg , and 2ΩΩ × (Ω Ω × r) = −∇(Ω2 R2 ) (in the fluid (parallel to the rotation axis ẑ) vary slightly. We
notation of §1.D.). Thus they can be subsumed into the pressure know from the Taylor-Proudman theorem that the fluid
term – where they will be only small corrections. flow must be independent of z. A “geostrophic” flow
 27

in this geometry would have to follow contours of con- The boundary conditions are: v = 0 at z = 0 (no-slip
stant depth h. Consider, then, a small radial displace- at the solid surface), and v = U ẑ as z → ∞. This
ment of a column of the fluid, to a region with h + δh. system solves to
This violates the T-P theorem; the column will try to h i
return to its initial position. But it has some inertia, so vx = U 1 − e−z/δ cos(z/δ)
it will overshoot. We thus have an oscillation: this is (5.4)
one version of a Rossby wave. vy = U e−z/δ sin(z/δ)
Rossby waves exist in the atmosphere, and in the
ocean; they are found to be low frequency ω < Ω with δ2 = ν/Ω. Thus, the velocity vector rotates
and long wavelength. The restoring force for terres- through a spiral, while it increases from zero at the
ground to ∼ U x̂ at z ∼> δ.
trial waves is due to variations in the depth of the atmo-
sphere parallel to Ω (here meaning the earth’s rotation). • Free boundary. Our driver here is the wind stress
In the atmosphere they are easily observed as the large- at the ocean surface, again in the x direction; call it τ x̂.
scale meanders of the mid-latitude jet stream. In the We can ignore pressure gradients in this case; thus our
ocean they are harder to find, being very small ampli- equations are, simply,
tude (several cm) and very long wavelength (hundreds
of km), but now detected in satellite data. d2 vx
−2Ωvy = ν
dz 2 (5.5)
d2 vy
Ω 2Ωvx = ν 2
dz
α Take z = 0 at the ocean surface, and negative below.
Our boundary conditions now are v → 0, z → −∞,
and dvy /dz = 0, dvx /dz = τ /ρν, z = 0 (refer back to
h
the definitions of stress in chapter 4). These solve to
 z π
vx = Vo e−z/δ cos − +
Figure 5.2. The geometry for a simple version of Rossby δ 4 (5.6)
waves. The curved arrow marks the deviation of the upper −z/δ
 z π
plate from parallel. Following Tritton Figure 15.13. vy = −Vo e sin − +
δ 4
√
THE EKMAN LAYER where Vo = τ /ρ 2Ων, and δ is the same as above.
Again, the velocity vector rotates through a spiral; note
Coriolis forces also do interesting things at boundaries. that it is at a 45◦ angle to ~τ at the ocean surface.
The two applications relevant to geophysical flows are Both cases have similar consequences. A prevail-
solid boundaries (such as land) and free boundaries ing wind or flow results in a transverse flow within the
(such as the ocean). In each case, the combination of Ekman boundary layer (this is Ekman transport). Mass
a driving wind (think of geostrophic flow)and viscosity conservation then forces a net vertical motion (why?)
within the boundary layer results in a transverse flow into or out of the layer (this is Ekman pumping). In
within that layer. In this section I present an overview; the atmosphere this is related to updrafts in low pres-
details will appear in the homework. sure regions (thus storm formation); in the ocean this
• Solid boundary. At some high altitude, let the is related to phenomena such as upwelling of cold sub-
wind velocity be U x̂. Thus, (5.2) reduces at high alti- surface water in regions with a steady prevailing wind
tudes to direction.
1 dp
2ΩU = − B. Viscous Flow: Time-Dependent Problems
ρ dy
Viscosity is dissipative: we expect a viscous flow to
(note dp/dx = 0, right?). Within the boundary layer
decay with time. In this section we consider solutions
we must keep the viscous terms. Here, (5.1) becomes
of the basic equations, (4.5) or (4.6), but here omitting
d2 vx gravity and the advective terms. Thus we want solu-
−2Ωvy = ν tions of
dz 2
(5.3)
d2 vy 1 dp ∂v
2Ωvx = ν 2 − ρ + ∇p = ρν∇2 v (5.7)
dz ρ dy ∂t
 28

and of boundary conditions vx → ±U, |y| → ∞):

Z y
ω
∂ω U 2
= ν∇2ω (5.8) v(y, t) = 1/2
e−u /4νt du
∂t (πνt) 0
  (5.12)
y
The ∇2 term makes these “diffusive” DE’s, and we can = U erf √
use standard methods for their solution. Let’s do this 4νt
by example. The last expression simply notes that this defines the
error function, erf(x). Thus, the |y| → 0 region obeys
1.SIMILARITY METHODS IN A DIFFUSION
vx ∝ y, and the outer has √ vx → ±U , as required; the
EQUATION
transition occurs at y ∼ 4νt. THus, viscosity spreads
Here’s one good approach to solving diffusion equa- the transition region out with time.
tions. Take a Cartesian system, ignore the pressure gra-
3. FLOW ABOVE AN OSCILLATING PLATE
dient, and take v = v(y, t)x̂ as a simple geometry. The
basic equation is then Another variant is an infinite, flat plate which moves
back and forth parallel to itself. The basic equation
∂v ∂2v for fluid above the plate is again (5.9), and the bound-
=ν 2 (5.9)
∂t ∂y ary conditions are v(0, t) = U cos ωt, and that v stays
bounded as y → ∞. To solve this, we try a separable
To set up a similarity transform, we note that there is solution, v(y, t) = eiωt f (y). Note, we are looking for
no physical scale applied to this problem (no edges of the “steady” solution, established after initial transients
finite length, no rocks
√ in the flow). We also notice that have gone away. Plugging this test solution in and do-
the quantity η = y/ νt is dimensionless. We therefore ing algebra, we find
guess that the (normalized) solution, v(y, t)/U can be
√  r 
ω
expressed as a function of η alone, call it f (η). Now do v(y, t) = U e −y ω/2ν
cos ωt − y (5.13)
some chain rules (you should check this!) 2ν

∂v Uη ′ ∂v U ∂2v U The cosine term here represents a signal propagating

=− f ; = ηf ′ ; 2
= η 2 f ′′ away in the y direction, while the exponential is a decay
∂t 2 t ∂y 2 ∂y 4
(5.10) in the y direction. Thus, the flow resembles
p a damped
(here, f ′ = df /dη, and etc). Putting these into (5.9), transverse “wave”, with wavelength 2π 2ν/ω in the
our DE becomes simpler, and can be integrated by sim- y direction, propagating
√ in the y direction, with ampli-
−y ω/2ν
pler methods: tude falling off as e .
Z
′2
4. IRROTATIONAL VORTEX DECAY
f + 2ηf = 0 ; f (η) = A e−η dη ′ + B (5.11)
′′ ′
A third application is the time-dependent evolution of
the irrotational vortex. Switching to cylindrical coordi-
and the integration constants A, B (and/or limits of the
nates, we have
integral) are chosen for the boundary conditions of the  
problem. ∂v ∂ 1 ∂
=ν (rv) (5.14)
∂t ∂r r ∂r
2. SMOOTHING OUT A VELOCITY JUMP
The interesting solution of this equation, for a system
Consider planar flow with a sharp jump in the velocity. which initially obeys v = K/2πr everywhere, is
At t = 0, take
K  2

v(r, t) = 1 − e−r /4νt (5.15)
vx = +U , y>0 2πr
Thus: we find v ∝ r for r → 0 (that is the core of
= −U , y<0 the vortex); v ∝ 1/r for r → ∞ (the outer part of
the vortex); and the transition
√ occurs at a radius which
Viscosity will clearly try to smooth out this jump. We grows with time, as r ∝ 4νt. Thus, viscous dissi-
again have (5.9) as the basic equation; and (from the pation causes the core of the vortex to spread. Some
last section), we can find its solution, with reasonable details of this may appear in the homework.
 29

C. Creeping Flow The first is vr = 0 at the surface (recall vr ∝ dψ/dθ,

and we want vr = 0 for all θ); the second is vθ = 0
Now, consider very viscous flow; flow at very low at the surface; and the third is uniform flow at infinity.
Reynolds number. This is called creeping flow.2 In Choosing a separable solution, ψ(r, θ) = f (r) sin2 θ
creeping flow we assume the inertial term v ·∇v can be (to match the distant boundary), we find the solution is
dropped compared to the viscous and pressure gradient
terms: this means the flow speed is very slow. We con- a3
 
2 2 1 3a
tinue to assume steady state and incompressible flow. ψ(r, θ) = U r sin θ − + 3 (5.22)
2 4r 4r
The basic equaiton is, then,
From the stream function, we can find the velocity com-
∇p = −ρν∇ × ω = ρν∇2 v (5.16) ponents:
Taking the curl of this, recalling ω = ∇ × v is the a3
 
3a
vorticity, we have vr = U cos θ 1 − +
2r 2r 3
(5.23)
a3
 
∇2ω = 0 (5.17) 3a
vθ = −U sin θ 1 − − 3
4r 2r
Now, we consider flow around a sphere of radius a,
placed in a stream flow with asymptotic velocity U . We We’ll need these in the next step.
want to find the drag force between the sphere and the Next, we want to find the drag force on the sphere
rock. The solution is long but straightforward ... here – which is the integral of the stress over the surface of
goes. the sphere. Separating out Σij as the non-pressure com-
First, we need to find the velocity field. Pick polar ponent of the stress tensor, the component of the drag
coordinates so that θ = 0 corresponds to the direction force in the U direction3 is
of downstream flow. The only nonzero component of ω
is then FD = [−p cos θ + Σrr cos θ − Σrθ sin θ]r=a (5.24)

For the pressure, we go back to (5.16) and solve for p

 
1 ∂(rvθ ) ∂vr
ωφ = − (5.18) (taking zero pressure at infinity, or solving for the over-
r ∂r ∂θ
pressure):
But this is axisymmetric flow, so we can use a stream
function (as in Chapter 2); in terms of ψ, the vorticity 3aρνU cos θ
p(r, θ) = − (5.25)
is 2r 2

1

1 ∂2ψ 1 ∂

1 ∂ψ
 (Note that we couldn’t use Bernoulli’s law here; why?)
ωφ = + 2 (5.19) Thus, the pressure is highest at the forward stagnation
r sin θ ∂r 2 r ∂θ sin θ ∂θ point, and lowest (negative in fact) at the rear stagna-
Putting this into (5.32), we get a fourth order equation tion point. For the rest of the stress tensor, we use the
for ψ: velocity solution, from (5.23), to get

3a3
 
2 ∂vr 3a
∂2
 
sin θ ∂ 1 ∂ Σrr = 2ν = 2νU cos θ − 4
+ 2 ψ=0 (5.20) ∂r 2r 2 2r
∂r 2 r ∂θ sin θ ∂θ
3νU a3
 
∂ vθ
  1 ∂vr
Σrθ = ν r + =− sin θ
This is now something can be solved. Our boundary ∂r r r ∂θ 2r 4
conditions are
(5.26)
ψ(a, θ) = 0 Putting this into the expression for fD , multiplying by
4πa2 and integrating over θ, gives the drag
dψ/dr(a, θ) = 0 (5.21)
1 2 2 FD = 6πρνaU (5.27)
ψ(∞, θ) = U r sin θ
2
3
Compare our earlier discussion in Chapter 3, in and around
2
For once the name makes sense! (3.24).
 30

Thus, we get a finite drag due to the viscosity. This is

Stokes’ law.

References
I’ve mostly followed Kundu, who does the more
mathematical examples. He also has a good chapter
on geophysical fluid dynamics. Tritton is also good for
geostrophic flows – as always his discussions are help-
ful (not just pure math).