You are on page 1of 33

EKMAN DYNAMICS

Geophysical Fluid Dynamics II midterm paper. Group 2: Emily Riley, Guy Cascella,
Pedro Di Nezio and Xiaofang Zhu.

March, 16th 2007.


Table of Contents

Introduction .............................................................................................................. 1

1. Boundary Layer Theory .................................................................................... 1

Interior flow dynamics........................................................................................... 4

Ekman layer.......................................................................................................... 4

Matching principle................................................................................................. 6

2. Quasigeostrophic dynamics in the presence of Ekman layers ........................ 10

Vorticity equation for an homogeneous fluid....................................................... 11

Spin down........................................................................................................... 12

Spin up ............................................................................................................... 13

3. Ekman dynamics in the presence of stratification ........................................... 14

The effect of turbulence and stratification........................................................... 14

The Ekman layer in a quasigeostrophic stratified fluid........................................ 15

4. The effects of stratification in the ocean Ekman layer..................................... 16

The effect of the diurnal cycle............................................................................. 18

5. The effects of stratification in the atmospheric Ekman Layer .......................... 20

Instability in the Ekman layer .............................................................................. 20

Critical parameters for instability in the Ekman layer .......................................... 22

6. Coupling of the atmospheric and ocean Ekman layers ................................... 23

Viscous coupling and classical double Ekman spiral.......................................... 23

Inertially coupled Ekman layers .......................................................................... 25

Conclusion ............................................................................................................. 29
Introduction
The small aspect ratio of geophysical fluids constrains their motions to be primarily
horizontal. However, friction between the main horizontal motion and the boundaries
acts to reduce the velocity creating a vertical shear. In such a circumstance the fluid
system exhibits two distinct behaviors:

• Interior flow. The flow at some distance from the boundaries where friction is
usually negligible.

• Boundary layer or Ekman layer. The flow near the boundaries where frictional
effects become important and the velocity of flow is brought down to zero at the
boundary.

To a first approximation geostrophic balance holds for interior fluid motion. As the fluid
approaches a rigid boundary; though, ageostrophic motions are induced by friction and
geostrophy breaks down. As a result, the fluid motion near the boundaries becomes a
balance between the viscous forces and the Coriolis force. The break down of
geostrophy leads to across-isobaric flow from high to low pressure and vertical motion
with the consequent energy dissipation.

This paper seeks to explore the effect of stratification on the dynamics of the Ekman
layer and its interaction with the interior flow in a quasigeostrophic framework. The
fundamental equations of motion are considered together with approximations and
scaling. Examples for both the atmosphere and ocean are included to highlight the
effects of stratification in specific conditions. To conclude, the interaction between the
atmospheric and oceanic boundary layers is covered. The basics of Ekman dynamics
are considered background knowledge in this paper; for more details the reader is
referred to Pedlosky (1987) section 4.1 – 4.4.

In section 1 the fundamentals of the boundary layer theory are presented as a


foundation for the rest of paper. Section 2 considers how motion in the Ekman layer is
communicated to the interior flow. Section 3 analyzes the importance of stratification.
Sections 4 and 5 explore the stratification effects in the Ekman layer for the ocean and
the atmosphere, while section 6 explains the coupling of the two.

1. Boundary Layer Theory


The content of this section is a summary from Pedlosky, 1987 (pp 200 – 212) where the
reader is referred for details. This section considers the motion in the Ekman layer
affected by friction in an idealized configuration. The results are afterwards easily
applied to either the ocean or atmosphere boundary layers. The motion of a
homogenous fluid bounded by two parallel planes perpendicular to the rotation axis is
considered. The upper plane imparts motion to the fluid and the lower plane rotates
fixed with the rotating frame of reference. figure 1 displays the details of the fluid
configuration.

1
r
The upper surface moves with a nondivergent velocity, uT and transfers momentum to
the bulk of the fluid through frictional coupling becoming the source for vorticity to the
system, ζ T . This idealized forcing is analogous to that of the wind stress over the upper
ocean Ekman layer. The matching principle and Ekman pumping/suction are the main
ideas developed to understand the relationship between the two regimes. The basics of
the classical Ekman spiral are considered background knowledge in this section; the
reader is referred to Pedlosky (1987) section 4.1 – 4.4 for more details.

We begin by introducing the non-dimensional momentum equations. The equations will


then be expanded keeping only the zero order terms for both the interior flow and the
Ekman layer flow. Finally, the matching principle will be explained in order to
understand how the fluid motion smoothly transitions from the Ekman layer flow to the
interior flow.

We start by introducing the following non-dimensional variables for time-space


coordinates, velocity and density:

( x ∗ , y ∗ ) = L ⋅ ( x , y ) , z∗ = Dz ,

D
( u ∗ ,v ∗ ) = U ⋅ ( u ,v ) , w ∗ = U w,
L

⎛L⎞
t ∗ = ⎜ ⎟t , p∗ = −ρgz + ρfULp ,
⎝U ⎠

where asterisks denote dimensional quantities and unprimed variables indicate non-
dimensional variables. D is half the distance between the upper and lower bounding
planes, L is a horizontal length scale and U is a characteristic velocity scale within the
fluid.

By applying these non-dimensional variables to the full momentum equations with


frictional effects included, we get the non-dimensional forms of the x, y, and z
momentum equations:

⎧ ∂u ∂u ∂u ∂u ⎫ ∂p E v ∂ 2 u E H ⎛ ∂ 2 u ∂ 2 u ⎞
ε⎨ + u +v +w ⎬−v = − + + ⎜ + ⎟, (1a)
⎩ ∂t ∂x ∂y ∂z ⎭ ∂x 2 ∂z 2 2 ⎜⎝ ∂x 2 ∂y 2 ⎟⎠

⎧ ∂v ∂v ∂v ∂v ⎫ ∂p E v ∂ 2v E H ⎛ ∂ 2 v ∂ 2v ⎞
ε⎨ + u +v +w ⎬ + u = − + + ⎜ + ⎟, (1b)
⎩ ∂t ∂x ∂y ∂z ⎭ ∂y 2 ∂z 2 2 ⎜⎝ ∂x 2 ∂y 2 ⎟

⎧ ∂w ∂w ∂w ∂w ⎫ ∂p ⎡ 2
2 Ev ∂ v E H ⎛ ∂ 2w ∂ 2w ⎞⎤
δ 2 ε⎨ +u +v +w ⎬ = − + δ ⎢ + ⎜ + ⎟⎥ , (1c)
⎩ ∂t ∂x ∂y ∂z ⎭ ∂z ⎢⎣ 2 ∂z
2
2 ⎜⎝ ∂x 2 ∂y 2 ⎟⎠⎥⎦

where

2
U
ε= , Rossby number,
fL

D
δ= , aspect ratio,
L

AV
EV = 2 , “vertical” Ekman number,
fD 2

AH
EH = 2 , “horizontal” Ekman number.
fL2

AV and AH are the vertical and horizontal turbulent viscosity coefficients respectively and
are considered constants. These turbulent or “eddy” viscosities come from the non-
linear terms in the momentum equations after applying Reynolds decomposition (i.e.
r r r r r r
replacing u with u = U + u ′ , where U is the non-turbulent mean flow and u ′ describes
the motion of the turbulent eddies). It is important to remind ourselves that the turbulent
viscosities are properties of the flow, unlike kinematic viscosities which are properties of
the fluid. The first term inside the curly brackets is the local velocity rate of change while
the remaining terms are the advection of velocity terms. The second term on the left
hand side (LHS) represents the Coriolis force. The first term on the right hand side
(RHS) is the pressure gradient force, while the last two terms represent vertical and
horizontal friction respectively. The scaling of the momentum equations (1) allows us to
determine the relative importance of the different small forces, which are important
when addressing geostrophic degeneracy.

The boundary conditions are

w =0 z = 0,1
(2a)
u =v =0 z=0

u = uT ( x, y ) z =1
(2b)
v = v T ( x, y ) z =1

Here, z = 0 (z = 1) is the bottom (top) boundary condition. By demanding w = 0 at the


boundaries means no flow will pass through the surfaces. Only an inviscid fluid can
satisfy the first boundary condition, (2a), (i.e. the interior flow). The existence of frictional
forces in the fluid allow the second boundary condition, (2b), to be satisfied, by imposing
a horizontally nondivergent velocity (uT, vT) that can be associated with the motion of
the rotating lid. The motion of the fluid will be driven by the momentum transferred by
this external forcing. For instance, this forcing can be associated with the wind stress
over the ocean surface when the idealized configuration is used to represent the effect
of frictional layers in the ocean.

3
Interior flow dynamics

Each dynamical variable in (1) is expanded as series of powers of a small parameter, ∆


which is in turn function of all nondimensional parameters:

u = u o ( x , y , z, t ) + ∆( ε, δ, E v , E H )u1( x , y , z , t ) + ... ., (3a)

v = v o ( x, y , z, t ) + ∆(ε, δ, E v , E H )v 1 ( x, y , z, t ) + ... ., (3b)

w = w o ( x, y , z, t ) + ∆(ε, δ, E v , E H )w 1 ( x, y , z, t ) + ... , (3c)

p = po ( x, y , z, t ) + ∆(ε, δ, E v , E H )p1 ( x, y , z, t ) + ... . (3d)

The small parameter, ∆ will become important in the next section when the higher order
terms are considered in a quasigeostrophic framework. If the expanded variables (3)
are then substituted into the non-dimensional momentum equations, (1a, b, c), and only
the O(1) terms are kept, the result for the horizontal velocity is geostrophic and
hydrostatic balance for the interior flow, i.e.

∂p o
uo = − , (4a)
∂y

∂p o
vo = , (4b)
∂x

∂p o
0= , (4c)
∂z

while the continuity equation becomes,

∂u o ∂v o ∂w o
+ + = 0. (5)
∂x ∂y ∂z

Ekman layer

Since we are interested not only in the motion of the fluid in the interior flow, but also on
the motion in the Ekman layer as well, we need a way to examine the fluid near the
boundaries z = 0,1 where the vertical derivative of the horizontal velocities are large. By
constraining ourselves to these regions near the boundaries, we force friction to enter
the equations. In order to explore the dynamics in the lower Ekman layer, region of the
flow near z = 0, we must rescale the vertical coordinate, z * = D ⋅ z so that the frictional
terms in the momentum equations are O(1) (the same procedure can be applied for the
upper Ekman layer, i.e. near z = 1). The new vertical coordinate is:

z
λ= , (6)
l

4
where l is the dimensionless boundary-layer thickness and is equal to:

2 Av
l = E v1/ 2 = D, (8)
f

E v ∂ 2 u E v ∂ 2v
constraining the friction terms , to be as large as the O(1) Coriolis
2l 2 ∂λ2 2l 2 ∂λ2
acceleration. The dimensional thickness associated l is represented by:

2 AV
l * = DE v1/ 2 = δ E = , (9)
f

becoming the appropriate scaling thickness when friction is important instead of D. This
depth scale is the Ekman layer thickness, δ E the parameter governing the e-folding
decay of classical Ekman spirals. The reader is referred to Ekman (1905) or Pedlosky
(1987) section 4.1 – 4.4 for more details.

The previous rescaling is key to understanding the flow in the Ekman layer, since it sets
a height scale where friction becomes comparable to the Coriolis term and thus with the
r
pressure terms. The rescaling takes place due to the rapidity in which u varies with z in
the friction layer. Put another way, the dynamical fields vary on a much smaller scale
than D in the frictional layer, thus making a new vertical coordinate based on the depth
of the frictional layer necessary.

The momentum equations, (1), are now rewritten in terms of λ

⎧ ∂u ∂u ∂u w ∂u ⎫ ∂p E v ∂ 2 u E H ⎛ ∂ 2 u ∂ 2 u ⎞
ε⎨ + u +v + ⎬−v = − + 2 2 + ⎜⎜ 2 + 2 ⎟⎟ , (7a)
⎩ ∂t ∂x ∂y l ∂λ ⎭ ∂x 2l ∂λ 2 ⎝ ∂x ∂y ⎠

⎧ ∂v ∂v ∂v w ∂v ⎫ ∂p E v ∂ 2v E H ⎛ ∂ 2v ∂ 2v ⎞
ε⎨ + u +v + ⎬ + u = − + + ⎜ + ⎟⎟ , (7b)
⎩ ∂t ∂x ∂y l ∂λ ⎭ ∂y 2l 2 ∂λ2 2 ⎜⎝ ∂x 2 ∂y 2 ⎠

1 ∂p δ ⎡ E v ∂ 2w ⎤ 2 EH ⎛ ∂ w ∂ 2w ⎞
2
⎧ ∂w ∂w ∂w w ∂w ⎫ 2
δ ε⎨
2
+u +v + ⎬=− + 2 ⎢ 2 ⎥
+δ ⎜ ∂x 2 + ∂y 2 ⎟⎟ ,
⎜ (7c)
⎩ ∂t ∂x ∂y l ∂λ ⎭ l ∂λ l ⎣ 2 ∂λ ⎦ 2 ⎝ ⎠

and the continuity equation becomes,

∂u ∂v 1 ∂w
+ + =0. (7d)
∂x ∂y l ∂λ

The variables in (7) are then expanded as before to give,

u = u~( x, y , λ, t , E v , E H , ε, δ) = u~o ( x, y , λ, t ) + ... , (10a)

5
v = v~( x, y , λ, t , E v , E H , ε, δ) = v~o ( x, y , λ, t ) + ... , (10b)

~ ~
W = W ( x, y , λ, t , E v , E H , ε, δ) = Wo ( x, y , λ, t ) + ... , (10c)

~ ( x, y , λ, t ) + ... ,
p = u~( x, y , λ, t , E v , E H , ε, δ) = p (10d)
o

where the tilde tells us the expansion is valid in the boundary layer near z = 0. Now, just
as the vertical coordinate had to be rescaled for the Ekman layer the aspect ratio and
D δ
vertical velocity must also be rescaled. The proper aspect ratio is no longer , but E ,
L L
as we are concerned now with the depth over which friction is important. The vertical
velocity then becomes,

⎛ δ ⎞ D δE UD
w * = O⎜ U E ⎟ = U = E v1 / 2 . (11)
⎝ L ⎠ L D L

If the rescaling of the vertical velocity, (11), is substituted into (7a, b, c) and each
boundary layer variable is expanded, as in (10), and only the O(1) terms are retained,
the momentum and continuity equations become,

∂p~ 1 ∂ 2 u~o
− v~o = − o
+ , (12a)
∂x 2 ∂λ2
~
∂p 1 ∂ 2v~o
~
uo = − o
+ , (12b)
∂y 2 ∂λ2

~
∂p
0=− o
, (12c)
∂λ

and
~
∂W o ⎛ ∂u~o ∂v~o ⎞
= −⎜⎜ + ⎟⎟ . (12d)
∂λ ⎝ ∂x ∂y ⎠

Matching principle

We now have two representations for each variable, one for the interior (e.g. po) and
one for the boundary layer (e.g. ~p0 ). These two representations must merge smoothly
together at the boundary layer – interior flow interface, i.e.,
~ = lim p
lim p (13)
o o
λ →∞ z →0

6
This is considered the matching principle and represents the smooth transition from the
Ekman layer into the interior. It will be shown that this transition takes place by a small
vertical velocity being pumped out of the Ekman layer into the interior flow, via w. To
proceed, the equations for the boundary layer will be written in terms of the interior flow
so that w may be determined. To begin we note that p o and ~ p o are independent of z
and λ respectively (i.e. the fluid is homogeneous), so that for all λ in the boundary layer

∂~
po ∂po
= = v o ( x, y ) , (14a)
∂x ∂x

∂~
po ∂p
= o = −u o ( x , y ) . (14b)
∂y ∂y

This tells us that the horizontal pressure gradient in the Ekman layer is determined by
the interior horizontal pressure gradient. If the results in (14) are substituted into (12a,b)
the zero order horizontal momentum equations become:

1 ∂ 2u~o
v o ( x, y ) = v~o + , (15a)
2 ∂λ2

1 ∂ 2v~o
− u o ( x, y ) = −u~o + , (15b)
2 ∂λ2

The Ekman layer solution, satisfying the no-slip conditions on z = 0, is then derived, so
that u~o and v~o are known.

[ ]
u~o = u o ( x, y ) 1 − e − λ cos λ − v o ( x, y )e − λ sin λ
, (16)
[ ]
v~o = v o ( x, y ) 1 − e −λ cos λ − u o ( x, y )e −λ sin λ

~
Since u~o and v~o are known, we can solve for the O( E v1/ 2 ) term vertical velocity, W o , by
substituting the above equations into the continuity equation (12d) to get,

~ 1 ⎧ ∂v ∂u ⎫
[
Wo ( x, y , λ ) = ⎨ o − o ⎬ 1 − e −λ {cos λ + sin λ} ,
2 ⎩ ∂x ∂y ⎭
] (17)

As λ → ∞ , the upper edge of the boundary,

~ 1 ⎧ ∂v ∂u ⎫ 1
Wo ( x, y , ∞ ) = ⎨ o − o ⎬ = ζ o , (18)
2 ⎩ ∂x ∂y ⎭ 2

The following equation relates the O(1) interior vorticity, ζ o to the O( EV1/ 2 ) velocity in the
Ekman layer, thus showing that a small vertical velocity is pumped out of the boundary

7
layer into the interior flow. Physically, this is due to cross-isobar (non divergent) flow
from high to low pressure in the boundary layer.

The matching principle is then applied to recover an expression for the vertical velocity
but in variables corresponding to the interior flow as follows:
~
lim W ( x, y , λ ) = lim w ( x, y , z ) .
λ →∞ z→0

According the previous limit, the interior vertical velocity satisfies

E v1/ 2 ⎧ ∂v o ∂u o ⎫ E v1/ 2
w ( x, y ,0*) = ⎨ − ⎬= ζ0 , (19)
2 ⎩ ∂x ∂y ⎭ 2

to the lowest order, which establishes the lower boundary condition for the interior flow.
The 0 * notation is introduced to remind the reader that the previous expression (19) for
the vertical velocity is the result of limiting process in the approximate boundary
between the Ekman layer and the geostrophic interior.

The actual vertical distance from the boundary at which the vertical velocity is
represented by (19) is not z=0, since w ( x, y ,0) = 0 due to the boundary conditions (2a)
of the overall problem. This distance is of the order of a few length scales,
2 Av
l = E v1/ 2 = D so that the exponentials in (18) become sufficiently small. Figure 1
f
displays the setup considered in this analysis, including boundary conditions and the
different regimes within the fllow (i.e. Ekman layers and geostrophic interior).

If the same steps are applied to the upper surface of the Ekman layer the resulting
upper boundary condition for the interior flow is,

E v1/ 2
w ( x, y ,1* ) = {ζT ( x, y ) − ζ o ( x, y )}, (20)
2

where ζ T is the vorticity of the upper boundary and ζ o is the O(1) vorticity of the
interior. If ζ T > ζ o fluid moves outward in the upper Ekman layer, sucking fluid
vertically from the top of the interior. The same asterisk notation is applied here to
denote that the vertical velocity is evaluated in the boundary between the Ekman layer
and the geostrophic interior.

The importance of (19) and (20) is that any differences between the vorticity of the
interior flow and the vorticity of the boundary will produce an O( E v1/ 2 ) vertical velocity at
the edges of the interior flow. The small vortex tube stretching or compression of the
planetary vorticity that results from this is the primary mechanism by which the effects of
friction can be communicated into the interior.

8
r
ζT = ∇ × uT ⋅ kˆ
rotating lid ~
r w =W = 0
uT
z =1
upper Ekman layer few l
w z = 1*

Ev1/ 2
w= (ζT − ζ 0 )
2
r
geostrophic interior ζ 0 = ∇ × u0 ⋅ kˆ
Ev1/ 2
w= ζ0
2

w z =0*
lower Ekman layer few l
z =0

u=0
v =0
~
w =W =0

Figure 1 – Schematics showing the setup used to study the effect of frictional layers on a rotating fluid.
The whole configuration is rotating around a vertical axis with angular speed 2f, where f is the Coriolis
parameter. Different regimes are established in the fluid: an upper Ekman layer and a lower Ekman layer
bound a geostrophic interior. The system is driven by the vorticity input, ζ = ∇ × ur ⋅ kˆ imparted by the
T T

E 1/ 2
rotation of the upper lid. The vertical velocity, w = v (ζT − ζ 0 ) pumped out from the upper Ekman layer
2
r
into the geostrophic interior sets the geostrophic vorticity ζ 0 = ∇ × u0 ⋅ kˆ of the latter. A lower Ekman
layer is set by the interaction of the geostrophic interior with the lower boundary where fluid is pumped out
1/ 2
into the geostrophic interior at a rate w = Ev ζ 0 . Vertical coordinate, z is non dimensional. Both Ekman
2
layers have a thickness on the order of a few non dimensional Ekman depth, l.

In the next section will move beyond the O(1) terms, to explain quasigeostrophic
dynamics in the presence of friction.

9
2. Quasigeostrophic dynamics in the presence of Ekman layers
The content of this section is discussed primarily in Pedlosky, 1987 (pp 212 – 218)
where the reader is referred for details. The boundary layer equations were derived in
the previous section considering only O(1) terms for the interior flow, hence making the
flow geostrophic. Notwithstanding, in the presence of frictional layers, ageostrophic
velocities arise from three effects implicit in the scaled equations of motion for the
interior flow (1), namely:

1. Relative accelerations associated to the O(1) inertial terms produce O(ε)


horizontal velocities.

2. Vortex tube stretching by the Ekman suction/pumping velocity generates an


O(EV1/2) change in relative vorticity (19) and (20).

3. Horizontal diffusion of O(EH) momentum, represented by the terms with E H in


(1).

A small parameter, identified as ∆ in (3), associated with these effects is needed to


expand the dynamical variables (u,v,w and p) in order to develop a quasigeostrophic
formulation. The ratio between vorticity changes due to vortex stretching and due to
advection, r is assumed to be O(1) in order to consider the situation where both effects
are comparable:
1/ 2
EV vorticity changes due to vortex streching
r = = = O(1) . (21)
ε vorticity changes due to advection

The ratio between horizontal diffusion of momentum and inertial accelerations (point 3,
above) is:

E H 2 AH 2
= = = O(Re −1 ) ,
ε UL Re

where Re is the Reynolds number of the interior flow:

UL
Re =
AH

This number is usually quite large for geophysical fluids, thus making Re-1 very small.
Regardless, the O(EH) terms are retained because the horizontal friction term become
important near vertical boundaries of an enclosed fluid.

If the interior flow variables u,v,w and p are expanded in series of ε and inserted in (1a,
b, c). The resulting O(1) terms are the geostrophic and hydrostatic equations as found
in the previous section (4a, b, c). The vertical velocity equations, (19) and (20), indicate

10
that w is O(EV1/2) at the boundaries, and since O(EV1/2) = O(ε) due to the assumption
implicit by (21), w0 must be zero. Consequently the O(ε) vertical velocities are:
1/ 2
E
w 1 ( x, y ,0*) = V ζ 0 ( x, y )

1/ 2
. (22)
E
w 1 ( x, y ,1*) = V (ζ T ( x, y ) − ζ 0 ( x, y ))

Vorticity equation for an homogeneous fluid

The fact that w0 must be zero for the interior flow makes also the O(1) horizontal velocity
field non-divergent and independent of z. If the O(ε) terms are kept when the x- and y-
momentum equations are expanded, and cross differentiation is applied to eliminate the
pressure gradient terms, the vorticity equation is obtained:

d0 ζ 0 ⎛ ∂u ∂v ⎞ 1 ⎛ ∂ 2 ∂2 ⎞
= −⎜⎜ 1 + 1 ⎟⎟ + ⎜ + ⎟ζ 0
dt ⎝ ∂x ∂x ⎠ Re ⎜⎝ ∂x 2 ∂y 2 ⎟⎠

∂w1 1 2
= + ∇ ζ0 .
∂z Re

The resulting equation indicates that the total change of O(1) relative vorticity is caused
by vortex tube stretching (first term, right hand side) and horizontal diffusion of vorticity
(second term, right hand side). Since only w1 depends on z, the previous equation can
be integrated in the z direction from z=0* to z=1*, the upper and lower boundaries
between the geostrophic interior and the frictional layers, yielding:

d0 ζ 0 1 2
= w1( x, y,1*) − w1( x, y,0*) + ∇ ζ0 .
dt Re

Once the results obtained in (22) for w 1 ( x, y ,1*) and w 1 ( x, y ,0*) are substituted in the
previous equation, the final form of the vorticity equation becomes:

d0ζ0 ⎧ ζ ⎫ 1 2
= −r ⎨ζ0 − T ⎬ + ∇ ζ0 . (23)
dt ⎩ 2 ⎭ Re

Since the O(1) variables are in geostrophic balance the vorticity equation can be written
in terms of the O(1) pressure field as follows:,

⎡ ∂ ∂p0 ∂ ∂p0 ∂ ⎤ 2 ⎡ 2 ζT ⎤
⎢ ∂t + ∂x ∂y + ∂y ∂x ⎥∇ p0 = −r ⎢∇ p0 − 2 ⎥ ,
⎣ ⎦ ⎣ ⎦

where frictional effects are neglected (terms proportional to Re-1).

11
The right hand side on the previous equation represents the sole direct effect of the
frictional Ekman layer on the interior flow (i.e. the damping of vorticity). The magnitude
of this effect in controlled by the value of r as it will shown in the next two subsections.

Spin down

Consider the vorticity equation (23) derived in the previous section, the term
proportional to Re-1 can be safely neglected since it vanishes away from the lateral
walls where horizontal gradients in the horizontal velocity field are small. Additionally in
the absence of forcing ( ζ T = 0 ) the resulting equation is:

d0 ζ 0
= −rζ 0 ,
dt

The advection of vorticity terms implicit in the total derivative can be neglected to
linearize the equation. Since d0 ζ 0 dt ~ ∂ ζ 0 ∂t now, the linearized equation can be
integrated in time to obtain:

ζ 0 ( t ) = ζ 0 ( 0 )e −rt . (24)

Equation (24) simply states that in the absence of forcing, Ekman layers act to dissipate
the vorticity of the geostrophic interior according to the time scales implicit by r. In other
words, r-1 is the time taken for the system to reduce its vorticity to e-1~0.36 of its initial
value. The time scales implicit in r are the ratio of the advective time scale L / U and the
spin down time τ = EV −1 / 2 / f . The corresponding dimensional value for r-1 is the spin
down time:

L −1 L ε 1
r = 1/ 2
= 1/ 2
= τ,
U U EV fEV

In the real ocean-atmosphere, this result is representative of the spin-down effect of the
atmospheric Ekman layer on the atmospheric interior flow. Typical values of τ are on
the order of 10 days for the atmosphere and 40 days for the ocean.

It is interesting to note that the same damping effect on the interior flow would result
from a simple linear drag (-ru0, -rv0) in the x- and y-momentum equations in the absence
of an Ekman layer. Typical values for the drag coefficient are on the order of 100 days,
indicating that a linear drag process is much less efficient in dissipating the vorticity of
the interior flow than the ageostrophic motions induced by frictional layers.

However, the real mechanism through which a given flow dissipates its vorticity by the
interaction with an Ekman layer is different than a linear drag process. For instance,
consider a cyclonic vortex over a surface (ocean/land) with vorticity ζ 0 ( 0 ) > 0 as the
schematic in figure 2 shows. The Ekman layer in the lower boundary pumps fluid into
the vortex’s low-pressure center establishing a gradient in vertical velocity, ∂w 1 / ∂z .

12
This vertical velocity shear is negative and by continuity considerations makes the
interior flow divergent according to:

∂w 1
− = ∇ ⋅ u1
∂z

This O(ε) divergent horizontal flow is across the O(1) geostrophic streamlines from low
pressure (vortex center) to high pressure (vortex outside), therefore work is done by
geostrophic interior to maintain the Ekman layer making the fluid spin down to the state
of rest.

Figure 2 – Schematics showing the different flow regimes in a low pressure center. The positive vorticity
ζ o in the interior flow pumps fluid out of the boundary layer. As a consequence an across isobar flow is
induced within the boundary layer.

Spin up

On the other hand, when the motion is driven by the steady vorticity input ζ T associated
with motion of the upper lid, and the effect of friction in lateral walls is neglected, the
vorticity equation is:

d0 ζ 0 ⎛ ζ ⎞
= −r ⎜⎜ ζ 0 − T ⎟⎟
dt ⎝ 2⎠

In the real ocean-atmosphere, this case is representative of the spin-up effect of the
surface wind stress on the upper ocean. The case where r >> 1 (short spin-up time
compared to the advective scale) simplifies the vorticity equation to:

13
ζT
ζ 0 =∇ 2 p0 = .
2

The previous equation indicates that the upper limit for the vorticity of the geostrophic
interior is half the vorticity input by the driving mechanism (either a rotating upper lid or
the wind stress).

A general result for any value of r is obtained if one requires the vertical velocity in the
lower boundary layer being pumped out is balanced by the vertical velocity being
suctioned into the surface layer:
1/ 2 1/ 2
EV EV
w ( x, y ,0*) = ζ 0 ( x, y ) = (ζT ( x, y ) − ζ 0 ( x, y )) = w ( x, y ,1*) ,
2 2

or

1
ζ 0 ( x, y ) = ζT ( x , y ) .
2

3. Ekman dynamics in the presence of stratification


The effect of turbulence and stratification

Geophysical flows have large Reynolds numbers associated with their inherently
turbulent character. The equations of motion used in previous sections rely on the
Reynolds decomposition for turbulent fluxes assuming constant enhanced eddy
viscosities as a closure scheme for turbulence. With the insight gained on Ekman
dynamics, it is clear that in a shear flow such as in an Ekman layer, the turbulence is not
homogeneous. Rather, the turbulence is more vigorous where the shear is greater and
also partially suppressed in the proximity of the boundary where the size of turbulent
eddies is restricted.

In the absence of an exact theory of turbulence, several schemes have been proposed.
For instance, a vertical eddy viscosity (AV) linearly increasing with depth has been
proposed to account for the inhibitory effect of the boundary upon turbulence (Madsen,
1977). Additionally eddy viscosities dependent on the bottom stress have been
proposed to obtain vertical velocities no longer directly proportional to the relative
vorticity of the geostrophic flow, but on a more complicated function of the geostrophic
flow (Cushman-Roisin and Malăcĭc, 1997).

The intensity of the surface heat flux plays an important role in the dynamics of the
surface atmospheric boundary layer during daytime over land and above warm currents
at sea. In such situations, the Ekman dynamics give way to convective motions which
are inextricably linked with the convection controlled by the intensity of the surface heat
flux.

14
Evidently, other processes besides geostrophic wind, as in the simplified model of
section 1, affect the dynamics of Ekman layers in the real ocean/atmosphere.

The Ekman layer in a quasigeostrophic stratified fluid

In this section the effect of stratification on frictional layers will be analyzed by focusing
on its effect on the O(ε) vertical velocity, w1. The content of this section is discussed
primarily in Pedlosky, 1987 (pp 360 –362) were the reader is referred for details.

For a homogeneous fluid the vertical velocity pumped out of the Ekman sets an
important boundary condition for the geostrophic interior dynamics, consequently we
seek to determine under what conditions stratification alters this condition.

Contrary to the homogenous z-momentum equation (4c), the presence of vertical


pressure gradients in the stratified Ekman layer is given by a hydrostatic balance, since:

~ ∂λ = ~
∂p θ0 ,
0

where λ is the non dimensional vertical coordinate in the frictional layer defined in (6),
~
and θ0 is the potential temperature within the frictional layer.

Meanwhile, the thermodynamic equation,


~
d 0 θ0
+ S ⋅ w1 = 0 ,
dt

allows us to examine if the presence of stratification,

N 2D 2
S= 2 2 ,
f L

( N 2 is the stratification frequency) may give rise to an O(1) pressure vertical gradient.
An order of magnitude for the pressure vertical gradient within the frictional layer can be
−1/ 2
obtained from the fact that λ = O(EV ) :

~ ∂λ = O(E1/2 ⋅ S ⋅ w ) .
∂p0 V 1

w1 is scaled as w 1 = O(EV1 / 2 ε) following the result obtained for a homogeneous fluid


(21) to obtain:

~ ∂λ = O(S ⋅ EV ) .
∂p0
ε

Thus, if

15
S ~ ∂λ = 0 ,
EV < O( ) ⇒ ∂p 0
ε
the equations of motion remain the same as in the homogeneous model. Consequently
the effect of stratification becomes negligible on the Ekman pumping/suction velocities
obtained for the homogeneous case.

Representative orders of magnitude of the variables for synoptic/mesoscale motions in


the atmosphere/ocean are included in table 1. The grayed rows indicate that the
condition EV < O(S ε) is generally fulfilled in the geophysical flows of interest.

Atmosphere Ocean
N2 [s-2] 10-5 10-8
Ε 10-1 5 10-3
S 10-1 0.2 10-1
S/ε 1 2 10-1
EV 10-4 10-5

Table 1 – Orders of magnitude of dynamical variables involved in the discussion of the effect of
stratification on boundary layers. The values correspond to synoptic/mesoscale motions in the
atmosphere/ocean.

Furthermore, studies of Barcilon and Pedlosky (1967a,b), indicate that a stratification of


order E1/2 represented an important transition in the dynamics of the fluid. This occurs
because the order of the vertical velocity is E1/2. For stratifications less than this value
Ekman layers play a dominant role in the dynamics. For stratifications greater than this,
the stratification inhibits the interior vertical velocity to be less than O(E1/2).

At this point the reader should have a firm understanding of the fundamentals of the
Ekman layer. In the following sections we will explore the ramifications of stratification
in the Ekman layer for the ocean (section 4) and the atmosphere (section 5), while
section 6 will explain the coupling of the two.

4. The effects of stratification in the ocean Ekman layer


The upper ocean Ekman layer problem was defined and solved more than a century
ago by Ekman (1905) for homogeneous conditions. Ekman also anticipated that the
eddy diffusivity can not be considered constant when the density of the upper ocean is
not uniform. Additionally, there are fundamental aspects of the oceanic Ekman layer
problem that remain unsettled, including even important external variables, such as heat
flux. The great and enduring difficulty is that turbulent stress is important at lowest order
in the Ekman layer arising from the Reynolds decomposition. Measurement of Ekman
layers presents similar challenges, for instance the observed upper ocean current

16
includes significant contributions from tides, internal waves and quasi-geostrophic
motions that make complex the isolation of the wind driven velocity.

Different regimes are identified for the upper ocean Ekman layer:

1. The winter Ekman layer (Krauss, 1993; Schundlich and Price, 1998). In this case
the upper ocean is well mixed and the dynamics are well described by the linear
homogeneous Ekman model.

2. The Ekman layer under ice (McPhee, 1990).

3. The Ekman layer under fair weather conditions (Price and Sundermayer, 1999).
Under fair weather the temporal variability of stratification due to the diurnal cycle
is found to play a major role in the structure of the Ekman layer.

Since the concern of this review is on the effect of stratification on Ekman dynamics, we
will focus on the work by Price and Sundermayer (1999). This study takes advantage of
observations obtained in the western Sargasso Sea during the LOTUS3 program, the
California current and a transect along 10°N in the Pacific Ocean. The dataset are
useful to study the structure of the Ekman layer under fair weather when stratification
effects are thought to be important.

In the three datasets the observed structure of the observed Ekman layer velocity has a
shape resembling that of a classical Ekman spiral. The observed current speed
decreases with depth following an e-folding spiral, however it decreases with depth
more rapidly than the current vector rotates to the right. The spirals also appear to be
flattened or compressed in the downwind direction. The total effect is that the direction
and speed change following different depth scales, Lθ and LS respectively. The ratio
between these depth scales is found to be Lθ LS ~ 2 − 3 while its value is 1 for a
classical Ekman layer.
r
The following equation for the Ekman velocity U E as a complex variable, U E makes
explicit how a flattened Ekman spiral can be related to different depth scales for the
speed and the direction,

⎡ ⎛ z ⎞ ⎛ z ⎞⎤
U E = U 0 ⎢exp⎜⎜ ⎟⎟(1 + i ) exp⎜⎜ i ⎟⎟⎥ .
⎣ ⎝ LS ⎠ ⎝ Lθ ⎠⎦

Moreover, the study carries out a preliminary evaluation of the observations by linear-
fitting them to a constant turbulent coefficient (AZ) model akin to that of Ekman. It differs
from the classical Ekman model in that the lower boundary condition is:
r
∂v
= 0, (24)
∂z z =−H

17
where H is taken to be a constant reference depth identified with the depth of the
seasonal o semi permanent stratification instead of an infinite depth as in Ekman’s
study. The obtained turbulent coefficients are found to explain more than 80% of the
variability for the three datasets.

Results in the same study obtained using a depth dependent coefficient AV(z) are also
unable to explain the flatness of the observed spiral, indicating that the classical
diffusion model is unable to account for the observed flattened spirals.

The effect of the diurnal cycle

Since the diurnal cycle is an important mode of variability of the upper 10 to 30 m of the
water column, the study develops a model to include this effect on AV in attempt to
explain the mentioned flatness of the Ekman layers. Observations obtained in the
subtropical North Pacific, show the effect of the covariation of the turbulent diffusivity
with stratification during the diurnal cycle on the wind driven current (figure 3).
(a) (b)

(c) (d)

18
Figure 3 – Observed diurnal cycle of temperature and current from Price et al, 1986. The uppermost
dashed vector shows the wind stress direction.

During the evening and early morning the ocean is neutrally stratified, i.e. a density
mixed layer was observed to a depth of 20 – 30m. Comparatively little vertical shear is
observed consistent with a homogeneous Ekman layer (figure 3a). As the surface layer
is warmed approaching noon the current becomes trapped in the warmed surface layer
(figure 3b). The current turns to the right of the wind stress due to the Earth’s rotation
(i.e. Coriolis force) as the day progresses (figure 3c) until sundown, by then the jet has
turned ~90° to the right of the wind stress (figure 3d). During the evening and early
morning the current speed is reduced as a result of vertical mixing due to cooling.

The effect of the diurnal cycle on the Ekman layer is modeled as follows. The
assumptions are derived from the observations of velocity and temperature:

1. A very large diffusivity AV0 = 1000 10-4 m2 s-1 is assumed within a mixed layer of
thickness h(t) and AV0 = 0 below. This condition aims to model that most of the
vertical shear occurs with stable stratification conditions, and that it is comparably
small within the mixed layer.

2. The depth of the mixed layer h(t) undergoes a diurnal cycle with a saw tooth
shape. Its minimum is at the sea surface during daytime and its maximum is 50m
during nighttime. This condition accounts for the observed steep reduction of
turbulence and diffusivity below the mixed layer.

3. Under fair weather conditions the depth of the mixed layer has a large amplitude
diurnal cycle.

4. Same boundary conditions as in the classical mode (24).

These rules can be synthesized as:

AV(z,t) = AV0 If 0 > z > -h(t)

AV(z,t) = 0 if -h(t) > z

A Reynolds decomposition of the turbulent stresses allows makes the behavior of the
model due to the time dependent turbulent coefficient explicit:

∂v ∂v ′ ∂V
τ = AV = AV′ + AV ,
∂z ∂z ∂z

The second term on the right hand side is analogous to that of the linear model. The
eddy term (first on the right hand side) can be as large as the mean term and is not
parallel to the mean stress.

A critical part of the model is to define the stratification represented by the depth of the
semipermanent stratification H and the mixed layer depth h(t). H is assumed to be the

19
top of the seasonal main thermocline. Since we are concerned with short time scale
variability, H can be taken as constant. Under fair weather conditions the diurnal
variability of h(t) is modeled as a saw tooth function with maximum during the night and
minimum during the day. The night-time mixed layer thickness is equal to H and the day
time is a function of the wind stress and daily maximum heat flux. For more details on
the model the reader is referred to Price and Sundermeyer (1999). To conclude, the
model is evaluated with observations of wind stress, surface heat fluxes and the depth
of the semipermanent stratification from climatology and compared with current meter
observations.

The results of the study confirm that the time varying stratification due to the diurnal
cycle is able explain the observed flat Ekman spirals trapped in the mixed layer. While
the diurnal cycles dominates the stratification variability in the upper 10-20m, other
effects such as internal waves and seasonality are important, especially in tropical
regions where the Ekman layer is deeper than the diurnal warm layer.

5. The effects of stratification in the atmospheric Ekman Layer


Instability in the Ekman layer

One of the most prominent visual effects of Ekman layer instability in the atmosphere is
the creation of cloud streets (figure 4). They are a direct result of longitudinal roll
vortices (figure 5) which are derived from underlying instability mechanisms. These
longitudinal vortices are characterized by counterrotating roll vortices that fill the depth
of the boundary layer. These processes only take place when certain conditions are
met. Observations have shown that the optimal conditions for the formation of these
features are (1) moderate background winds/vertical shear (~5 m/s) and (2) a weakly
unstably stratified boundary layer (LeMone 1973). Moreover, the roll axis of the
longitudinal vortices is generally found to be approximately parallel to the geostrophic
flow (Lilly 1966). That having been said, the two distinct mechanisms via which
instability is manifested in the Ekman Layer, shear and buoyancy instabilities, will now
be discussed.

20
Figure 4 – Cloud streets off the Georgia coast (from Brown, 1974).

21
Figure 5 – Schematic of longitudinal roll vortices.

Buoyancy instability (or convective instability) is a direct result of the thermal


stratification of the boundary layer. Since temperatures in this layer decrease with
height, the mechanism for buoyancy processes is present most of the time. In contrast
to our assumed background conditions, it should be noted that increased surface
heating will increase vertical velocities such that the longitudinal rolls will be unable to
form, thus ending the possibly for the formation of cloud streets. Convective instability,
on its own, produces quasi-2 dimensional cells which are not organized linearly.
However, when a background mean flow (shear) is present the cells are allowed to form
into longitudinal roll vortices.

Stratification also plays a role in dynamic instability. In the atmosphere, horizontal layers
may be distinguished by differences in velocity magnitude, direction, or both (Brown
1980). Under the restricted conditions noted earlier, it turns out that dynamic instability
is only created via the presence of an inflection point in the velocity field. This inflection
point typically occurs at the interface between two layers, and inevitably leads to a
vorticity maximum at the inflection point. The magnitude of the instability created is
determined by the steepness of the velocity gradient across the interface, which is
determined by the ratio of the directional and speed sheer across the Ekman Layer.
Frictional forces are sometimes not strong enough to counteract this shear forcing, and
thus any perturbation imposed on the new field will generate waves. These waves will
amplify until they change the overall flow to a more stable pattern. This change in flow
may result in the formation of longitudinal rolls.

Critical parameters for instability in the Ekman layer

Both experimentation and theoretical analysis have confirmed that flow in the Ekman
Layer becomes unstable for Re > 55 (e.g. Brown 1972, Lilly 1966, Kaylor and Faller
1972), but it is not unusual, and actually fairly typical, to find Re values on an order of
magnitude greater. However, another parameter exists, the Richardson Number (Ri),
which may either amplify or damp the effects of an unstable Reynolds Number. Here, in
the presence of stratification, the Richardson Number is defined as:

N2 buoyant suppression of turbulence


Ri ≡ 2
= .
⎛ du ⎞ shear generation of turbulence
⎜ ⎟
⎝ dz ⎠

When Ri > .25 at the inflection point, it is found that instability is completely suppressed,
regardless of the aforementioned conditions. It should be noted, though, that the
majority of studies which make this assumption of the “critical Richardson Number”
have not included the effects of rotation.

If we apply a scaling to the definition of the Richardson Number, it can be seen that the
underlying dynamical mechanisms necessary for the generation of instability in the

22
atmospheric Ekman layer typically exist. Assuming the longitudinal roll vortices fill the
depth of the Ekman layer (e.g. Brown, LeMone, Kraus), we have:

N2 N2 10 -5 s -1
Ri ≡ ~ = = 0.1
⎛ du ⎞
2
⎛U ⎞
2
⎛ 10 m/s ⎞
⎜ ⎟ ⎜⎜ ⎟⎟ ⎜ ⎟
⎝ dz ⎠ ⎝ δE ⎠ ⎝ 1000 m ⎠

The derived value is clearly less than the theoretical critical Richardson Number, which
would indicate that instability is generally possible. However, we must recall that
significant vertical motions associated with moderate to strong convection will inhibit the
formation of longitudinal vortices (i.e. the Brunt-Vaisala frequency increases in value for
convective processes). Thus, again, ideal circumstances must be present for the
formation of these instabilities.

Until this point we have assumed a weakly unstably stratified system, inclusive of
variations of temperature and velocity with height. In a study by Kaylor and Faller
(1972), it was shown that similar instabilities (here, internal waves) could be created in a
stably stratified environment. In this case, internal waves take the place of the
longitudinal rolls created in the unstable scenario. The internal waves are generated
when the frequency of the shear flow is close to the frequency of a possible internal
gravity wave. In such a case, resonance occurs and the internal wave can grow. Similar
values for the critical parameters are still assumed (see Kaylor and Faller 1972)
throughout the model simulation.

Overall, the main take away point from this section is that instabilities may result from
both stably and unstably stratified atmospheric Ekman Layers. Thus, the effect of
stratification on the Ekman Layer is critical to understanding the totality of its dynamics.

6. Coupling of the atmospheric and ocean Ekman layers


The coupling of the atmospheric and ocean Ekman layers is best understood by
considering the matching boundary conditions at the interface of the two. We start by
introducing the classical form of the double Ekman Spiral, as well as the viscous
coupling conditions. Then we introduce the concept of inertial coupling.

Viscous coupling and classical double Ekman spiral


r
The velocity field, u in both the atmosphere and ocean is assumed to have two
components,
r r r
u = ug + uE .

This decomposition is done to represent the response of the fluid to the two driving
r r
mechanisms. u g is pressure driven component of the velocity and u E is the component
associated with the vertical stress gradient.

23
The steady state equations of motion for both the ocean and atmosphere are solved.
r
The horizontal velocity, u = (u,v ) is represented from now on as a complex variable,
U = u + iv , where i is the complex unit. Since we have two second order equations in U,
four boundary conditions are needed.

Two of them exist at the interface and become the conditions for viscous coupling:

Viscous coupling

Continuous velocity at the interface U 0 (0 ) = U Eair (0 ) + U gair (0 ) = U Eocean (0 ) + U gocean (0 )

Continuous vertical stress at the ∂U air ∂U ocean


τ s = ρ air AZair = ρ ocean AZocean
interface ∂n 0
∂n 0

And the remaining two boundary conditions are in the non-slip conditions for the velocity
U at the atmosphere upper boundary and at the ocean bottom:

U air = U gair z→∞


U ocean = 0 z→∞

The solutions are the classical double Ekman spiral (see figure 6):
z z
( i − 1)z − i
U air = U gair − ( −U Eair ) = U gair [ 1 − ( 1 − µ ) exp ] = U gair − U gair ( 1 − µ )e DEair e DEair , (25a)
DEair

z z
( 1 − i )z − i
U ocean = U gair µ exp = U gair µe DEocean e DEocean , (25b)
DEocean

where the parameters are:

ρ air A
µ= ( air ) , the diffusion capacity with a typical of µ = 0.005 ,
ρ water Aocean

0. 5
⎡ 2A ⎤
DEair = ⎢ air ⎥ , the Ekman depth for the atmosphere,
⎣ f ⎦

0. 5
⎡ 2A ⎤
DEocean = ⎢ ocean ⎥ , the Ekman depth for the ocean.
⎣ f ⎦

24
The velocity profile of the atmosphere takes the form of the classic Ekman spiral given
by equation 25a. The atmosphere velocity U decreases to zero approaching the
interface, remaining purely geostrophic away from the Ekman layer. The corresponding
Ekman depth, DEair gives a measure of the depth over which the effects of the interface
are important.

The velocity profile of the ocean is given by equation 25b. The ocean total velocity
decays with depth to zero at the bottom. The corresponding Ekman depth, DEocean gives
a measure of the depth over which the effects of the interface are important.

Figure 6 – Double Ekman Spiral. The frame of reference moves with the drift velocity at the ocean
surface, U 0 . The surface drift and Ekman spiral in the ocean are magnified about 30 times to make them
comparable to the velocities in the atmosphere. (Kraus et al, 1994). The ocean Ekman spiral is
represented relative to the drift velocity U 0 . In other words, when the drift velocity U 0 is vectorially added
to the displayed ocean profile it takes the form of classical Ekman spiral.

Inertially coupled Ekman layers

The interaction between fluids with large density contrasts such as the ocean and the
atmosphere is dominated by waves. To account for the effect of waves on coupled
Ekman layers, let us suppose now that the interfacial Lagrangian velocity can be
→ →
partitioned into two parts: a non-wave induced part u 0 and wave induced part uL in fluid

25

1 (atmosphere) and a wave induced component ε u L in fluid two (ocean), where
ε = (ρ1 ρ 2 )1 2 is the squared root of the ratio between the densities of the two fluids.
Unlike the viscous case, the addition of a wave component, makes the velocity
discontinuous. Simultaneously the shear stress continuity condition across the interface
remains the same as in the viscous case. Figure 7 shows the conceptual distribution of
velocities in both boundary layers.

Figure 7 – Schematics showing two inertially coupled planetary boundary layers. z B is the wave
r r
boundary. The height at which u = uG1 defines the planetary boundary layer thickness.

A bulk relationship for the surface shear stress is applied both for the atmosphere and
ocean with transfer coefficient for momentum, K L :
r r r r r r r
τ s = ρ1K L u1 − u 0 − u L (u1 − u 0 − u L ) (air) (26a)

r r r r r r r
τ s = ρ 2 K L u 0 − εu L − u 2 (u 0 − Au L − u 2 ) (water) (26b)
r
Eliminating u L , which is an arbitrary wave velocity, from equation 26a and 26b and
comparing with 26a, the set of inertial coupling conditions is simplified. As long as there
is wave present in the interface (which is most time the case), the follow inertial coupling
relationships hold:

26
Inertial coupling boundary condition

r 1 r r 1 r r ⎡r r 1 r r ⎤
τS = ρ1K L u1 − u 0 − (u1 − u 0 ) ⋅ ⎢u1 − u 0 − (u1 − u 0 )⎥ (27a)
4 ε ⎣ ε ⎦
r r r r
εu1 + u 2 = 2εu L + (1 + ε )u 0 (27b)

In the previous subsection we found the integrating the shear stress for viscous
r r r
coupling, results in τS ∝ (u1 − u 2 ) . In the inertial coupling case will have same relation
only if ε = 1 . But in the case of fluids with big density contrast, such as air and ocean
ε << 1 , thus the difference between formulations becomes significant.

Here we will not go through step by step how the solution was attained, rather,
important steps will be presented and emphasis will be placed on analysis of results as
well as how it is related to the classical viscous coupling results.

The dominating equations are still the same, with assumption that total velocity could be
partitioned to geostrophic velocity and Ekman frictional velocity. The only difference
here is we use inertial coupling boundary condition rather than viscous boundary
condition. The results are presented using a set of new variable

The following notation is introduced to discuss the results obtained from equations 27,
1/ 2 r
where the Uˆ i = ρ1 u i are the corresponding inertially weighted velocities. Additionally,
the inertially weighted shear across planetary boundary layer, i.e. planetary shear is
defined as:
r
G = ρ1 [(u g 1 − u 0 ) − (u g 2 − u 0 )].
1/ 2 r r r r

1/ 2
α = ⎛⎜ εκ ⎞
⎟ , κ here is Von Carmon’s constant , and K I = 1 / 4K L and
⎝ 2K I ⎠

r
τS
FˆS = is the normalized wave boundary shear.
(K I rτ S ) 1/ 2

The vector form results are:

Uˆ E 1 = −Uˆ E 2 , (28a)

Uˆ E 2 =
1
((2 + α )Gx + αGy , ((2 + α )Gy − αGx )) , (28b)
((2 + α ) 2 + α 2 )

27
α
FˆS = ((α + 1)G x − G y ,G x + (α + 1)G y ) . (28c)
(1 + α ) 2 + 1)

Interpretation of the results:

From equation 28a, we get that the inertially weighted Ekman velocities have the same
modulus, but opposite direction. If we write results in polar coordinates, as shown by
figure 9, we found that there is always π / 4 angle difference between the wave
r
boundary shear stress τ s and the inertially weighted Ekman velocity Uˆ E1 , i.e.
δ + γ = π / 4 . This relationship is also valid for the classical Ekman spiral.

Figure 9 – Orientation of inertially weighted shear and velocities.

From equation 28b, we obtain a solution that only depends on the parameter α :
r r r
εu L − u 2 + u 0
α = . (29)
wE

Where w E is the Ekman scale velocity for the ocean. Equation 29 indicates that α is
the ratio of the velocity shear in the wave boundary layer to the velocity shear in the
Ekman layer. Different values of α yields different set of results associated with
different physical phenomena:

1. The Ekman Limit (α → 0) . For α → 0 , then δ → 0 , γ → π / 4 (see figure.9), this


is the Ekman limit in which the surface shear stress lies at a 45o angle to the left
of the inertially weighted geostrophic shear (northern hemisphere).

2. The Stokes limit ( α → ∞ ). For α → ∞ , then δ → π / 4 , γ → 0 , thus in the stokes


limit, the surface shear stress lies in the direction of the planetary shear.

28
3. Naturally occurred sea state ( δ ~ 36 o , γ ~!9 o ). In reality, the Ekman spiral is more
close to the Stokes limit.

For more details on the content of this subsection the reader is referred Bye (2002).

Conclusion
In the presence of boundaries the velocity field of a geostrophic flow must go to zero,
producing vertical stresses which make frictional effects important. A distinct dynamical
regime is set near the boundaries. The vertical scale over which vertical stresses are
important is set by the Ekman depth and defines the thickness of the Ekman layer.

The interaction of the Ekman layers with the geostrophic interior gives rise to vertical
motions upsetting the non-divergent character of geostrophic flows. Particularly, the
motions of the atmosphere Ekman layer are driven by the vorticity input from the
geostrophic wind. On the contrary, the motions of the ocean Ekman layer are set by the
vorticity input from frictional velocity at the interface.

Atmosphere/ocean flows, are characterized by very small Ekman numbers indicative of


the minor role friction plays in the momentum balance. However, in the absence of
forcing, Ekman layers act to dissipate the vorticity of geostrophic flows through cross
isobar flow induced by vertical motions. This dissipation takes place on a time scale
controlled by the Ekman number and the Coriolis parameter.

Turbulence and stratification are inextricably linked since stratification acts to inhibit
vertical motions. For synoptic/mesoscale motions in the atmosphere/ocean, the effect of
stratification becomes negligible on the Ekman pumping/suction velocities.
Consequently the results obtained for the homogeneous case are also valid in the
presence of stratification. Nevertheless, in some particular conditions such as the
Ekman layer under fair weather, stratification effects become important.
Parameterization schemes are usually implemented to model the Ekman dynamics in
these cases.

To conclude, the interfacial boundary condition between two coupled fluids such as the
ocean and the atmosphere is the fundamental aspect to represent the dynamics of the
Ekman layers in each fluid. While viscous coupling gave us classical double Ekman
spiral, inertial coupling seems to be more close to reality, given the ubiquity of waves at
the ocean/atmosphere interface.

References

Barcilon, V., and J. Pedlosky, 1967a: Linear theory of rotating stratified fluid motions. J.
Fluid Mech., 29, 1–26.

Barcilon, V., and J. Pedlosky, 1967b: A unified Linear theory of homogenous and
stratified rotating fluids. J. Fluid Mech., 29, 609–621.

29
Brown, R. A., 1974: Analytic Methods in Planetary Boundary Layer Modeling, 150 pp.,
John Wiley, New York, 1974.

Brown, R. A., 1972: The Inflection Point Instability Problem for Stratified Rotating
Boundary Layers, J. Atmos. Sci., 29, 850–859.

Brown, R. A., 1980: Longitudinal Instabilities and Secondary Flows in the Planetary
Boundary Layer: A Review. Rev. Geophys. Space Phys., 18, 683–697.

Bye, J. A. T., 2002: Inertially Coupled Ekman Layers. Dyn. Atmos. Oceans, 35, 27–39.

Cushman-Rosin, B., Malăcĭc, V., 1997: Bottom Ekman Pumping with Stress-Dependent
Eddy Viscosity. J. Phys. Oceanogr., 27, 1967–1975.

Ekman, V. W. , 1905: On the influence of the earth’s rotation on ocean-currents. Ark.


Math. Astron. Fys., 2, 1–53.

Kaylor, R., and A. J., Faller, 1972: Instability of the Stratified Ekman Boundary Layer
and the Generation of Internal Waves. J. Atmos. Sci., 29, 497–509.

Kraus, E. B., and J. A., Businger, 1994: Atmosphere-Ocean Interaction, 2nd Edition.
Oxford University Press, Oxford.

Krauss, W., 1993: Ekman drift in homogeneous waters, J. Geophys. Res., 98, 20, 187–
20,209.

LeMone, M. A., 1973: The Structure and Dynamics of Horizontal Roll Vortices in the
Planetary Boundary Layer. J. Atmos. Sci., 30, 1077–1091.

Lilly, D. K., 1966: On the Instability of Ekman Boundary Flow. J. Atmos. Sci., 23, 481–
494.

Madsen, O. S., 1977: A realistic model of the wind-induced Ekman boundary layer. J.
Phys. Oceanogr., 7, 248–25.

McPhee, M. G., 1990: Small-scale processes, in Polar Oceanographic Part A, Physical


Science, chap 6, pp. 287-344, Academic, San Diego, California.

Pedlosky, J., 1987: Geophysical Fluid Dynamics, 710 pp., Springer, New York, NY.

Price, J. F., and M. A. Sundermeyer, 1999: Stratified Ekman Layers. J. Geosphys. Res.,
104, 467–494.

Price, J. F., Weller, R. A., and Schudlich, R. R., 1986: Wind-driven ocean currents and
Ekman transport, Science, 238, 1534–1538.

Schudlich, R. R., and J. F. Price, 1998: Observations of the seasonal variation in the
Ekman layer, J. Phys. Oceanogr., 28, 1187–1024.

30
31

You might also like