lecture - univ-brest.frmespages.univ-brest.fr/~scott/master_iuem/m2_physmar/... · 2018. 11....
TRANSCRIPT
Lecture 1: Introduction 1
Lecture
Lecture 1: Introduction 2
Plan for today
— Introduction : quick reminder of key concepts we will use.
You must interrupt me and tell me if you have not seen
some of this before.
— Sverdup theory (starts p.48 of these notes, material before
p.48 is background material I assumed you knew, and only
referred to as needed).
— Ekman currents and Ekman transport.
— The Wind-driven gyre (Stommel) theory.
— Observations will be referred to as needed, especially the
surface wind stress but also the zonally averaged potential
density to see the effects of the Ekman pumping on the
isopycnals.
Lecture 1: Introduction 3
Goal
— Our aim is to understand large-scale ocean circulation,
espeically the basin scale, midlatitudes circulation.
— The principle tools are the equations of ocean circulation and
the primary forcing of the circulation, especially the wind.
— We will focus on large-scale, equilibrium circulation. In
particular the balances known as geostrophy, hydrostatic
balance, thermal wind, and Ekman transport will be
introduced and used to understand the meridional
(north-south) variation in the thermocline.
Lecture 1: Introduction 4
Our Perpetual Ocean !
https://www.nasa.gov/topics/earth/features/
perpetual-ocean.html
background – stop me for more details ! 5
Notation and conventions
— We do not use inertial reference frames, but rather a
reference frame fixed to the rotating Earth. We will use
either spherical coordinates with origin at the centre of the
Earth or rectlinear (Cartesian) coordinates denoted (x, y, z)
or (x1, x2, x3) or simply xi with it understood that the index
i takes values 1, 2, 3 or equivalently values x, y, z. Generally z
(or r for spherical coordinates) will be the vertical direction.
— Vectors are all three dimensional Cartesian vectors (no
distinction between contravariant and covariant
components) and are indicated with an arrow, ~F .
— The velocity
~u =d
dt(x(t), y(t), z(t)) = (u, v, w) = (u1, u2, u3) (1)
background – stop me for more details ! 6
Or we can simply refer to an arbitrary component of the
vector, ui, with the index i taking the 3 possible values,
i ∈ 1, 2, 3.
background – stop me for more details ! 7
Physical oceanography in prespective
— The ocean (and most of the atmosphere) are governed by
the Navier-Stokes equations and the laws of classical
thermodynamics.
— But the full Navier-Stokes equations are too complex to
solve, and the resulting solutions too complex to describe
usefully.
— So we make gross simplifications, leading to key balances
that we give different names :
1. geostrophy
2. hydrostatic balance
3. thermal wind
4. Ekman transport
background : stops me for me info ! 8
Ocean bathymetry
— There are five principal ocean basins :
— Atlantic Ocean
— Pacific Ocean
— Indian Ocean
— Arctic Ocean
— Southern Ocean
— The ocean bathymetry is divided into several regions :
— Continental shelf of depth between 100 and 200 m.
— Abyssale plain, 3000 m to 6000 m.
— Continental slope, the region between the shelf and the
abyssal plain.
— Marine trenches, long thin valleys. For example the
Marianas Trench, trough in the Earth’s crust averaging
background : stops me for me info ! 9
about 2,550 km (1,580 mi) long and 69 km (43 mi) wide.
The maximum-known depth is Challenger Deep, at
10,994 metres
background : stops me for me info ! 10
background : stops me for me info ! 11
Large-scale circulation
quasi-two-dimensional
— The largest horizontal length scales in the ocean are the
basin scale, roughly the radius of the Earth or about
6400km, which is more than 1000 times the depth of the
ocean.
— As a result the ocean is circulation is mostly horizontal or
quasi two dimensional. This quasi two dimensional is, to a
very good approximation, in hydrostatic balance.
background : stops me for me info ! 12
Ocean forcing
Ocean currents are driven by 3 main mechanics
— Wind stress. The stress of the wind on the ocean surface
drives currents, called Ekman transport, that are the
principle mechanism setting the basin scale ocean thermal
structure. The resulting circulation is called the wind-driven
gyre circulation.
— Moon and Sun tidal force. The time varying
gravitational attraction of the Moon and Sun drives the
ocean tides on many time periods, the most prominant being
the principal lunar semi-diurnal tidal , called the M2 tide,
with period about 12 hrs and 25 minutes and principal solar
semi-dirurnal tide, called S2, with period exactly 12 hrs. The
direct motions resulting from this forcing are very large scale
background : stops me for me info ! 13
barotropic (essentially depth-independent) waves and
barotropic coastal Kelvin waves (waves that decrease in
amplitude exponentially from the coast) that we observe at
the shore. But these directly forced tidal motions also drive
turbulence and internal waves as they interact with the
rough sea floor, which in turn mix the ocean thereby
strongly affecting its thermal structure and resulting global
scale circulation.
— Heat fluxes, evaporation and freezing. These three
important mechanims take place at the ocean surface and
affect the sea water density. Evaporation increases the
salinity and therefore the seawater density. Cooling at high
latitudes increases the density. Freezing, in the formation of
sea ice, increases the salinity of the surrounding waters and
thereby increases the density. These mechanisms set the
density boundary conditions that influence the global scale
background : stops me for me info ! 14
ocean circulation. In the past some oceanographers believed
that these boundary conditions lead directly to a global
overturning circulation called the thermohaline circulation.
Now its is more widely appreciated that winds and tides are
essential in determining this global overturning circulation
and this term thermohaline circulation is less often used by
physical oceanographers.
— While all the above forcing mechanisms are important it is
sometimes difficult to separate a given current as due to the
given mechanism.
Ocean dynamics 15
Background material 2. The equations of
motion
Ocean dynamics 16
Fundamental equations
— The continuity equation
∂ρ
∂t+
∂
∂xj(ρuj) = 0, (2)
expresses the conservation of mass of the continuous fluid.
— The Mach number, M = Ucs
, where cs is the sound speed and
U is a typical ocean current velocity, is very small, which
implies that seawater is effectively incompressible ; the
volume of a fluid parcel undergoes negligible variations in
volume during its motion. The continuity equation Eq(2) in
this case simplifies to
∇ · ~u =∂
∂xiui = 0, (3)
meaning that the velocity field is non-divergent.
Ocean dynamics 17
— The momentum equation for a continuous fluid can be
written in the reference frame fixed to the solid Earth :
ai =DuiDt
+ Coriolis + Centripetal (4)
= −1
ρ
∂p
∂xi− gδi3 + Viscose forces. (5)
The RHS of Eq(5) is the net force per unit mass, the viscose
terms result from the divergence of the deviatoric stress
tensor. The RHS of Eq(4) includes inertial terms that do not
appear in an inertial reference frame. Because the reference
frame rotates with the Earth we have the additional
acceleration terms arising from the acceleration of the fixed
coordinates, the Coriolis acceleration and Centripetal
acceleration. We discuss these more in detail next.
— In Eq(4) we have the acceleration ai of a fluid parcel relative
to an inertial reference frame. We have decomposed this into
Ocean dynamics 18
3 contributions :
1. The acceleration of the fluid parcel relative to the
reference frame of the Earth,
DuiDt
(6)
2. The Coriolis acceleration of a point moving with velocity
~u in the rotating reference frame of the Earth,
Coriolis = 2~Ω× ~u (7)
where ~Ω is the rotation rate of the Earth expressed as a
vector pointing in the direction of the axis of rotation
and magnitude equal to the rate of rotation. Note that
this term vanishes if the fluid is stationary ! The Coriolis
acceleration is perpendicular to the current and the
Ocean dynamics 19
Coriolis force per unit volume is
~Fcor = −ρ2~Ω× ~u. (8)
You can feel the effect of the Coriolis force if you try to
rotate a spinning object, like a hair dryer or power drill.
3. The centripetal acceleration of a fixed point of the
rotating reference frame (accelerating because the Earth
is spinning on its axis).
Centripetal = ~Ω× ~Ω× ~r, (9)
where ~r is the position vector of the fluid parcel relative
to the centre of the Earth. Because this last term
depends only upon position we can consider this as a
modification of the acceleration of gravidty ~g. We define
the effective gravity
~g∗ = ~g − ~Ω× ~Ω× ~r, (10)
Ocean dynamics 20
which is no longer directed toward the centre of the
Earth. In fact this term has deformed the solid Earth
such that the Earth is no longer precisely a sphere but
has radius is about 22 km greater at the equator than at
the poles. To simplify the notation, we will never include
the star, and always just write ~g it being understood that
this term includes the gravity and a small correction
from the centripetale acceleration.
Ocean dynamics 21
Oceanic reference frame
— We naturally use a reference frame fixed to the rotating
Earth. Because of the Earth’s rotation, this frame is not
inertial, and the additional acceleration terms discuss above
arise.
— The most natural coordinate system in this reference frame
is a spherical coordinate system (r, θ, φ), with origin at the
centre of the Earth, with r measuring the distance from the
centre of the Earth, the polar angle θ measuring the angle
from the axis of rotation (so θ = 90 − lat, where lat is the
latitude in degrees, and the azimuthal angle φ measure the
angle from Greenwich meridian (where longitude is zero so
that φ is the longitude). We idealize the Earth as being
exactly spherical with radius R corresponding to the
Ocean dynamics 22
sealevel. Then relation then to the more familiar latitude
(lat) and longitude (lon) and altitude z are
r = R+ z, θ = 90 − lat φ = lon (11)
This is the best coordinate system to use for dealing with
large (basin size) calculation.
— In practice we often want to analyze local processes. Then
the more familiar Cartesian coordinate system is adequate
and simpler. We choose an origin at some convenient
location (latc, lonc) and apply a simple map projection to
local Cartesian coordinates
x = R cos(lat)(lon− lonc),
y = R(lat− latc),
z = z (12)
We will use the conventional unit vectors ~i,~j,~k along the
Ocean dynamics 23
x, y and z axes.
Ocean dynamics 24
Equations of motion in the oceanic
reference frame
— First we must find how the Coriolis acceleration can be
written for the three components of velocity (u, v, w). The
vector ~Ω points along the axis of the Earth, so geometry
gives us that at latitude lat this vector has components
2~Ω = 2Ω cos(lat)~j + 2Ω sin(lat)~k,
= f∗~j + f~k. (13)
What is Ω = ‖~Ω‖ ? It is the rate of rotation of the Earth in
an inertial reference frame. The stars provide an (extremely
good) approximation to an inertial reference frame so that
we can measure the rotation rate relative to the stars. The
Ocean dynamics 25
Earth takes by definition 24 hours to find the Sun in the
same position in the sky. But that means it turns slightly
more than 2π radians in a day, in fact it turns
2π +2π
365.2425rad day−1 =⇒ Tsidereal = 23 hr 56 min 4.09 sec
(14)
And so
2Ω =2π radian
(23 · 60 · 60 + 23 · 60 + 4.09)second= 7.292× 10−5
rad
s(15)
Ocean dynamics 26
— Recall our convention for velocity in Eq(1). We have
Coriolis = 2~Ω× ~u = det
~i ~j ~k
0 f∗ f
u v w
,
= (−fv + f∗w)~i+ fu~j − f∗u~k. (16)
— Denoting the viscose force as ~F we can write the equations
of motion Eq(5) as
Du
Dt+ f∗w − fv = −1
ρ
∂p
∂x+ Fx,
Dv
Dt+ fu = −1
ρ
∂p
∂y+ Fy,
Dw
Dt− f∗u = −1
ρ
∂p
∂z− g + Fz,
∇ · ~u = 0. (17)
Ocean dynamics 27
— The first approximation we make is called the Boussinesq
approximation. As noted earlier, the density in the ocean
does not changes much, at most a few percent, from the
value 1028 kg m−3. So we will make very little error in the
first two equations by setting ρ = ρ0, a constant reference
density in the first two equations.
— We can write the equations of motion
Du
Dt+ f∗w − fv = − 1
ρ0
∂p
∂x+ Fx,
Dv
Dt+ fu = − 1
ρ0
∂p
∂y+ Fy, (18)
— In the third equation we have to be careful because the
gravity is such an important term. Instead we multiply
Ocean dynamics 28
through by ρρ0
to obtain
ρ
ρ0
(Dw
Dt− f∗u
)= − 1
ρ0
∂p
∂z− ρ
ρ0g +
ρ
ρ0Fz. (19)
We have seen previously that the primary balance in this
equation is the hydrostatic balance, the underlined terms on
the RHS. The other terms provide at most a small
correction to this hydrostatic balance. So we argue thatρρ0≈ 1 and this is a good enough approximation for all the
small correction terms,
Dw
Dt− f∗u = − 1
ρ0
∂p
∂z− ρ
ρ0g + Fz. (20)
— In summary we can write the incompressible, Boussinesq
Ocean dynamics 29
equations of motion
Du
Dt+ f∗w − fv = − 1
ρ0
∂p
∂x+ Fx, (21)
Dv
Dt+ fu = − 1
ρ0
∂p
∂y+ Fy, (22)
Dw
Dt− f∗u = − 1
ρ0
∂p
∂z− ρ
ρ0g + Fz, (23)
∇ · ~u = 0. (24)
Ocean dynamics 30
The large scale motions
— We now scale the terms in the Boussinesq equations by
associating a typical value for horizontal velocity U , vertical
velocity W , horizontal length scale L, vertical length scale
H. As we are interested in the large scale motions, we
assume aspect ratio α = H/L is small :
α = H/L 1. (25)
— First consider the incompressibility condition.
∂u
∂x+∂v
∂y= −∂w
∂z
O
(U
L
)= O
(W
H
), =⇒ W = αU. (26)
Ocean dynamics 31
So the ratio of the two Coriolis terms in the Eq(21)
f∗w
fv=
cos(lat)
sin(lat)α (27)
which is small everywhere accept very near the Equator,
permitting us to simplify the first equation to
Du
Dt− fv = − 1
ρ0
∂p
∂x+ Fx, (28)
for large scale motion away from the Equator.
— An important observation in fluid mechanics is that the
pressure gradient term is almost always important ; no term
dominates the pressure gradient term.
— Now consider the vertical momentum equation Eq(23). The
pressure gradient in this equation is 1/α times stronger than
in the horizontal equations, and yet the terms on the LHS of
Eq(23) are similar to or smaller than those on the LHS of
Ocean dynamics 32
the horizontal momentum equations :
O(f∗u) = 2Ω cos(lat)U ≈ O(fu) = O(fv) = 2Ω sin(lat)U
(29)
accept near the Equator. Furthermore,
O
(Dw
Dt
)= αO
(Du
Dt
)= αO
(Dv
Dt
)(30)
The viscose terms involve the velocities and only in special
circumstantce could be much larger than the acceleration
terms. So outside the equator region, there is nothing to
balance this strong vertical pressure gradient force for the
large scale circulation except the gravity term. The primary
balance must be the hydrostatic balace we found for a static
fluid :∂p
∂z= −ρg (31)
Ocean dynamics 33
where p is the fluid pressure, z the vertical coordinate in our
inertial reference frame with Cartesian coordinates, ρ is the
fluid density, and g = 9.81 m s−2 the acceleration due to
gravity near the surface of the Earth. This is the very
important hydrostatic balance.
— In summary, for the large scale, non Equatorial ocean we
have found the simpler set of equations :
Du
Dt− fv = − 1
ρ0
∂p
∂x+ Fx, (32)
Dv
Dt+ fu = − 1
ρ0
∂p
∂y+ Fy, (33)
1
ρ0
∂p
∂z= − ρ
ρ0g, (34)
the hydrostatic Boussinesq equations.
Ocean dynamics 34
Geostrophic currents
— For time scales T long compared to the sidereal day, the
relative acceleration terms DuDt and Dv
Dt will be small relative
to the Coriolis terms, except possibily very near the
Equator. Notice we are considering the time scale following
the fluid parcel.
— We can anticipate when the time scales will be long using
the advection time scale T = L/U . Then we find the ratio of
the accerlation and Coriolis terms is
O(DuDt
)O (vf)
=U
L/U
1
Uf=
U
fL≡ Ro. (35)
This ratio is of fundamental importance in geophysical fluid
dynamics, and defined as the Rossby number.
Ocean dynamics 35
Table 1 – Rossby number of typical ocean (and atmospheric ) phe-
nomena
Feature L U Ro
Gulf Stream ring 100 km 1 m s−1 0.1
Gulf Stream 50 km 1 m s−1 0.2
Mid ocean eddy 50 km 0.1 m s−1 0.02
Rossby wave 1000 km 0.1 m s−1 0.001
Rossby wave (atmosphere) 1000 km 10 m s−1 0.1
Anticyclone (midlatitude “high”) 2000 km 10 m s−1 0.05
Cyclone (midlatitude “low”) 1000 km 20 m s−1 0.2
Category 3 hurricane 500 km 50 m s−1 1
Tornado 100 m 50 m s−1 5000
— So for small Ro we can ignore the relative acceleration term
Ocean dynamics 36
in the equations of motion, giving
−fv = − 1
ρ0
∂p
∂x+ Fx, (36)
+fu = − 1
ρ0
∂p
∂y+ Fy, (37)
1
ρ0
∂p
∂z= − ρ
ρ0g, (38)
— The viscose terms are generally of less secondary importance
accept in turbulent boundary layers.
— The conclusion is that the large scale, long time scale
motions outside turbulent boundary layers and not near the
equator are in geostrophic balance,
−fvg = − 1
ρ0
∂p
∂x, fug = − 1
ρ0
∂p
∂y(39)
where we have used a subscript g to emphasize that these
Ocean dynamics 37
currents are the geostrophic currents. One can take this as a
definition of the geostrophic currents, regardless of the
setting, and say that the currents are well approximated by
the geostrophic currents for the large scale, long time scale
motions outside turbulent boundary layers and not near the
equator.
In vector form
f~k × ~ug = − 1
ρ0∇p,
or ~ug =1
fρ0~k ×∇p. (40)
Notice the geostrophic currents are, surprisingly, orthogonal
to the pressure gradient. This relation works even in the
Southern Hemisphere where f is negative. In the NH
(northern hemisphere) where f > 0 Eq(40) implies that, if
you stand with your back to the wind (or current) the higher
Ocean dynamics 38
pressure is on your right. Furthemore, around a low pressure
system (or sea surface depression) the winds (or currents)
rotate in the same sens as the Earth’s rotation ; that is why
low pressure atmospheric systems are called cyclones and
high pressure systems are called anticyclones. In the NH, the
Earth appears to spin anticlockwise when you look down at
the North Pole.
— Note that Eq(40) implies that if we knew the pressure field
we could calculate the winds and currents. In the
atmosphere, there is a global network of ballons that
measure the temperature of the air to determine the
pressure, and thereby determine the large scale wins. For the
ocean, since the early 1990s, the sea surface height is
monitored using several satellites equiped with radar
altimeters that measure the sea surface height η to with a
cm or so precision.
Ocean dynamics 39
— The oceanic pressure field determined from the η using the
Eq(34)
−∫ η
z
∂p
∂zdz =
∫ η
z
ρgdz, (41)
p(x, y, z)− patm = gρ(η − z), (42)
where we have introduced the mean density,
ρ =1
(η − z)
∫ η
z
ρdz. (43)
For shallow depths ∇ρ is negligible. Furthermore typically
we ignore the pressure gradients from the atmosphere, which
are small because of the much larger length scales in the
atmosphere ; i.e. we generally assume that ∇patm gρ0∇η.
— Using these assumptions and p from Eq(42) for the
Ocean dynamics 40
geostrophic current relation Eq(40) we find
~ug =gρ0f~k ×∇η, (44)
because ∇z = 0 because z is a constant height level.
— The equation Eq(44) is valid near the surface (but below the
turbulent boundary layer where shear stresses are
important ; i.e. the so-called Ekman layer discussed later)
but at shallow enough depths that the horizontal variations
in density are not important. Observations reveal that the
horizontal mean density variations compensate the pressure
gradient induced by the sea surface height variations. In the
mid latitudes the geostrophic current becomes negligible
below about 1000 to 1500 m depth. In the tropics and
equatorial ocean, the geostrophic currents are even more
strongly surface trapped.
— Since 1992, accurate global sea surface height η observations
Ocean dynamics 41
have been available, permitting the global calculation of near
surface geostrophic currents. The orbital period of one
satellite is about 10 days, so the time resolution was initially
about 20 day (Nyquist period = shortest period
unambiguously resolvable at 10 sampling). When several
satellites observe the sea surface height simultaneously this
sampling is of course improved. I was just starting my career
at this time and I recall many researchers were initally
skeptical of the validity of this data. But it gradually because
trusted and has since revolutionized physical oceanography.
Ocean dynamics 42
Background Material 3. Thermal wind
Ocean dynamics 43
Thermal wind
— The geostrophic balance holds throughout the ocean for
large-scale flows, i.e.
Ro =U
fL 1, (45)
away from the equator and away from turbulent boundary
layers.
— We exploited this balance in the previous section for the
determination of near-surface geostrophic currents from sea
surface height observations, one of its important
applications.
Ocean dynamics 44
— The geostrophic balance
f~k × ~ug = − 1
ρ0∇p,
or ~ug =1
fρ0~k ×∇p. (46)
leads to another important relation for the vertical
derivative of the geostrophic currents. Taking the vertical
derivative of Eq(40) we find
∂
∂z~ug =
∂
∂z
1
fρ0~k ×∇p,
=1
fρ0~k ×∇ ∂
∂zp,
(47)
Ocean dynamics 45
Now we use the hydrostatic relation Eq(31)
∂
∂z~ug = − g
fρ0~k ×∇ρ. (48)
This vertical gradient in geostrophic current, Eq(48), is valid
in many conditions : for atmospheric winds as well as ocean
currents. For historical reasons, it is called the thermal wind,
even when applied to ocean currents.
— Returning to the discussion of the surface trapped current
geostrophic currents, we can say that it is the thermal wind
that tends to compensate the near-surface geostrophic
currents reducing the geostrophic current with depth so that
it eventually vanishes around by 1500 m depth in the mid
latitudes, shallower in the tropics.
— Eq(48) is useful in many applications. Before the
development of satellite altimeter, the currents in the ocean
were either measured directly with current meters or
Ocean dynamics 46
inferred from the Eq(48). In the latter case, we need a level
of reference zref for which the geostrophic currents ~ug(zref)
are known. Then Eq(48) can be integrated from this level∫ z
zref
∂
∂z′~ugdz
′ =
∫ z
zref
− g
fρ0~k ×∇ρdz′,
~ug(z) = ~ug(zref)−∫ z
zref
g
fρ0~k ×∇ρdz′ (49)
Historically it was assumed that there was a so-called level
of no motion around 2000 m depth, so this was choosen as
zref = −2000m. That is, the geostrophic currents were
considered so weak at this level that they could be
neglected, ~ug(zref) ≈ 0. Then temperature and salinity
measurements throughout depths z > zref where enought to
determine ρ, the horizontal gradient of which could be
integrated in Eq(49) to make these current estimates.
— With the advent of the Argo programme (global distribution
Ocean dynamics 47
of floats that divide up and down in the upper 1500m depth
measuring T and S and park at 1000 m depth for 10 days),
we can do better than the level of no motion assumption.
The Argo floats effectively measure the ocean current
velocity at 1000 m. (The Argo float parks at 1000 m depth
for 10 days before returning to the surface to communicate
its position and data to satellites. From these 10-day float
displacements one can infer the ocean current at the parking
depth. ) Furthermore, we have regular global measurements
of upper ocean density from the T and S measurements
taken by the Argo floats as they dive up and down.
Ocean dynamics 48
IUEM Lecture. Sverdup theory, Ekman
currents and Ocean General Circulation
Sverdrup Theory 49
Sverdrup Theory
Reference : See Chapter 7 of the course notes “Elements de
dynamique des fluides planetaires,” by Xavier Carton.
— Our immediate goal is to derive the Sverdrup relation
βvg = f∂w
∂z(50)
where vg is the merdional geostrophic velocity, w is the
vertical velocity, f is the local vertical component of the
Coriolis parameter f = 2Ω sin(θ), θ is the latitude and Ω is
the rate of rotation of the Earth with respect to the fixed
stars.
— This surprisingly simple relation
1. applies to a good approximation for the incompressible
Sverdrup Theory 50
geostrophic circulation,
2. is useful, when combined with Ekman theory to explain
the basin-scale circulation driven by the wind-stress curl.
Sverdrup Theory 51
Derivation of the Sverdrup relation
— Our starting point is the following two approximations.
— The geostrophic balance
~ug =1
fρ0~k ×∇p,
(51)
where p is the pressure, ρ0 is the fluid density, and ~k is a
unit vector in the local vertical direction. We will use these
relations on a basin scale so we use a local Cartesian
coordinate system (constructed from a map projection from
Sverdrup Theory 52
the spherical coordinates, see Eq(12) ) :
ug = − 1
fρ0
∂p
∂y,
vg =1
fρ0
∂p
∂x(52)
where x is zonal coordinate, y is the meridional coordinate,
z is the local vertical coordinate.
— The second starting assumption is the incompressibility
condition :
∇ · ~u = 0, (53)
which in the local Cartesian coordinate system becomes
∂u
∂x+∂v
∂y+∂w
∂z= 0 (54)
— Applying the incompressibility Eq(54) to the geostrophic
Sverdrup Theory 53
currents Eq(52) gives
−∂w∂z
=∂ug∂x
+∂vg∂y
,
=∂
∂x
(−1
ρ0f
∂p
∂y
)+
∂
∂y
(1
ρ0f
∂p
∂x
),
=
−1
ρ0f
(∂2p
∂x∂y
)+
1
ρ0f
(∂2p
∂x∂y
)+
(1
ρ0
∂p
∂x
)∂
∂y
(1
f
),
= −(
1
ρ0f2∂p
∂x
)∂f
∂y,
= −vgβ
f. (55)
βvg = f∂w
∂z(56)
— When does this apply ? When the circulation is geostrophic
Sverdrup Theory 54
and incompressible. These are very general conditions that
apply for the large scale, interior circulation (away from
turbulence boundary layers near the surface and bottom of
the ocean) away from the Equator.
— We can see this in more detail by considering which terms in
the complex momentum balance were ignored in the
geostrophic relation. These were the relative acceleration and
the turbulence stresses, the justification of which we now
recall.
— Recall geostrophy applies for the large scale L circulation
Ro =U
fL 1, (57)
with time scale T long compared to the “pendulum day”
(1/f),
1
T f. (58)
Sverdrup Theory 55
so that we can ignore the relative acceleration
∂ui∂t
+ uj∂ui∂xj
(59)
in the ith momentum equation, and away from the Equator
and the boundary layers where turbulence stresses∂σji
∂xj
become important (especially at the air-sea interface at the
ocean surface and the sea-solid-Earth interface at the ocean
bottom).
— Recall incompressibility applies when the current speed U is
much less than the speed of sound cs. The continuity
equation Eq(2) in this case simplifies to
∇ · ~u =∂
∂xiui = 0, (60)
meaning that the velocity field is non-divergent.
— To go beyond the Sverdrup relation we must consider these
Sverdrup Theory 56
turbulence boundary layers in more detail, especially the
surface layer and its relation to the surface wind stress.
wind-driven circulation 57
The wind-driven circulation
Reference : See Chapter 6 of the course notes “Elements de
dynamique des fluides planetaires,” by Xavier Carton.
— The wind is a principle source of the ocean circulation. The
mechanism is via the Ekman transport, which we now
describe.
— The theory of Ekman currents was first discovered by the
Swedish scientist Vagn Walfrid Ekman in the early 1900s in
attempting to explain the observations by Arctic explorer
Fridtjof Nansen that his ship drifted about 30 to the right
of the wind direction.
— The wind applies a stress ~τs at the surface of the ocean that
depends upon the near surface wind ~U10, air density
ρair ≈ 1.2 kg m−3, and sea surface roughness expressed as a
wind-driven circulation 58
so-called drag coefficient CD which in turn is a function of
‖~U10‖ = U10 and density stratification. An order of
magnitude estimate of CD is O(10−3) except in extreme
conditions such as hurricanes. The wind stress is generally
estimated by the quadratic drag formula
~τs = ρairCDU10~U10 (61)
where ~U10 is the wind at the reference height of 10 m above
the sea surface (a height convenient for observation from a
large ship).
— The wind applies a stress ~τs enters the equations of motion,
the horizontal components of the hydrostatic Boussinesq
Eqs(32,33), as a boundary condition for the frictional forces
wind-driven circulation 59
per unit mass, denoted Fx and Fy in
Du
Dt− fv = − 1
ρ0
∂p
∂x+ Fx, (62)
Dv
Dt+ fu = − 1
ρ0
∂p
∂y+ Fy. (63)
These are written for a general (not necessarily Newtonian)
continuous fluid as
ρ0Fx =∂
∂xiσxi =
∂
∂xσxx +
∂
∂yσxy +
∂
∂zσxz,
ρ0Fy =∂
∂xiσyi =
∂
∂xσyx +
∂
∂yσyy +
∂
∂zσyz. (64)
We seek equations valid in the upper boundary layer next to
the air-sea interface where the horizontal scales are set by
the length scales of the winds (generally hundreds to
thousands of kilometers for the most energetic scales) which
are much much greater than the vertical scales (observed to
wind-driven circulation 60
be tens of meters). This boundary layer is now called the
Ekman layer, and generally lies within the upper part of the
mixed layer.
wind-driven circulation 61
zx
-hEs
0
-H+hEb
-H
hEs
hEb
(H-hEb-hE
s)
surface Ekman layer
bottom Ekman layer
interior layer
Figure 1 – Three ocean layers : (i) surface Ekman layer, (ii) ocean
interior layer , (iii) bottom Ekman layer.
wind-driven circulation 62
To a very good approximation we retain only the
components with vertical derivatives :
ρ0Fx =∂
∂zσxz, ρ0Fy =
∂
∂zσyz, (65)
with boundary conditions σxz = τs,x and σyz = τs,y.
— As argued above, when the time scale following the fluid
parcel is long compared to the f−1, so time scale longer than
a few days outside the Equatorial region, the relative
acceleration Dui/Dt is small and the Eqs(36) and (37) apply
−fv = − 1
ρ0
∂p
∂x+
1
ρ0Fx, (66)
+fu = − 1
ρ0
∂p
∂y+
1
ρ0Fy, (67)
(68)
wind-driven circulation 63
— Replacing the frictional terms by Eq(65) we arrive at
−fv = − 1
ρ0
∂p
∂x+
1
ρ0
∂
∂zσxz, (69)
+fu = − 1
ρ0
∂p
∂y+
1
ρ0
∂
∂zσyz, (70)
— We pause to interpret this physically : The surface Ekman
layer also involves a 3-way momentum balance between
Coriolis, horizontal pressure gradient, and vertical derivative
of the horizontal turbulent stress.
— We now write the total current in Eqs(69) and (70) as the
sum of the geostrophic currents ~ug and the Ekman currents
~uE
~u = ~ug + ~uE . (71)
Substituting Eq(71) into Eqs(69) and (70) gives and using
wind-driven circulation 64
Eq(40) we find
−fvE =1
ρ0
∂
∂zσxz, (72)
+fuE =1
ρ0
∂
∂zσyz, (73)
That is, the Ekman currents result from a balance between
the wind stress and the Coriolis force.
— A turbulent closure is required to be more quantitative. But
independent of the turbulent closure we can find the very
wind-driven circulation 65
important Ekman transport
Mx =
∫ η
−Hρ0uE(z)dz =
∫ η
−H
1
f
∂
∂zσyzdz,
=1
fτs,y, (74)
My =
∫ η
−Hρ0vE(z)dz = −
∫ η
−H
1
f
∂
∂zσxzdz,
= − 1
fτs,x, (75)
where H is a depth sufficiently below the turbulent
boundary layer such that the stress has dropped to a
negligible amount. Mx and My and the eastward and
northward components of the depth-integrated mass
transport associated with the Ekman currents. Because they
depend only upon the stress boundary condition at the top
of the ocean, which is set by the winds, this incredible
wind-driven circulation 66
theory lets us infer a complex, difficult to observe oceanic
mass transport with an atmospheric variable.
— The Ekman transport equations Eq(74) and Eq(75) can be
written in vector form
~M = − 1
f~k × ~τs. (76)
Check your understanding : This implies to the Ekman
transport in the Northern Hemisphere is directed (a) parallel
to the wind, (b) 90 to the right of the wind, or (c) 90 to
the left of the wind ?
— Historically is was not feasible to observe the wind stress
throughout the world ocean on a daily basis.
Oceanographers collected observations throughout the world
over many years and created approximate atlases of the
monthly means values taken (so averages of a given month
with observations from many different years). This common
wind-driven circulation 67
averageing practice leads to climateological variables. Much
of the data came from ships of opportunity, these are
commercial ships that take climate data. As a result the
shipping lanes between North America and Europe and
Asian were well sampled, but much of the Southern Ocean
and Arctic were very poorly sampled.
— The 1980s saw a revolution in atmospheric observations with
the development of the technology to estimate surface wind
stress from satellite observations. The satellite-based radar
infer the sea surface roughness from echo return of radar
pulses. The surface wind stress is then calibrated to give
surface wind stress from detailed analysis of given sites.
Because winds decorrelated quickly in time, higher temporal
resolution is needed ; several daily wind stress products are
available, and some with multiple times per day.
— The divergence of the Ekman transport, ∇ · ~M , leads to
wind-driven circulation 68
vertical velocity wE called the Ekman pumping upwelling,
wE =1
ρ0∇ · ~M. (77)
This relation can be derived using the continuity equation
and the definition of ~M in Eq(74) and Eq(75).
— From Eq(77) and Eq(76) we find that the Ekman pumping
is given by the wind-stress curl
wE =1
ρ0~k ·(∇× ~τs
f
). (78)
We will use this later !
— The most dominant wind patterns are the prevailing easterly
trade winds in the tropics and westerly winds between about
30 and 60 in both hemispheres, see Fig. 2.
— There is a positive wind-stress curl centred are 60N,
resulting in upwelling of the subpolar gyres in the North
wind-driven circulation 69
Atlantic and Pacific oceans. This brings up nutrient rich
water to upper ocean that makes these waters very
productive marine life.
— We can understand the narrow thermocline in the equatorial
region from upwelling in this region due to the Ekman
pumping associated with the prevailing easterly trade winds.
wind-driven circulation 70
Figure 2 – Annual mean surface wind stress (vector) and zonal
component (colour), (Talley et al., 2011, Fig. 5.6). Red τx > 0 ; blue
τx < 0.
— The convergence of the Ekman transport, −∇ · ~M , leads to a
negative vertical velocity wE called the Ekman pumping
wind-driven circulation 71
downwelling. This is most prevalent in the subtropical gyre
regions of the mid latitudes, especially in the North
Hemisphere. The result is a much deeper thermocline. The
cold water necessary for the efficient OTEC is much deeper
in the subtropical gyre.
— In summary, we can understand the zonally averaged density
structure in the upper ocean, Fig. 3, based upon the global
wind patterns and Ekman theory.
wind-driven circulation 72
Figure 3 – Zonally averaged annual mean potential density in
the global ocean. Units are kg m−3, expressed as a departure from
1000 kg m−3.
Sverdrup Transport 73
Derivation of the Sverdrup transport
Reference : See §7.1 of the course notes “Elements de dynamique
des fluides planetaires,” by Xavier Carton.
— We can integrate the geostrophic meridional velocity vg in
the vertical over its range of validity from the base z = −hsEof the surface turbulence boundary layer (Ekman layer) to
the top of the bottom turbulence boundary layer
z = −H + hbE .
— First note that
β =∂2Ω sin θ
∂y=
2Ω
a
∂ sin θ
∂θ,
=2Ω cos θ
a, (79)
Sverdrup Transport 74
where again a is the radius of the Earth. Thus
f
β= 2Ω sin θ
(2Ω cos θ
a
)−1= a tan θ. (80)
which in the Sverdrup relation Eq(56) we obtain
vg =f
β
∂w
∂z= a tan θ
∂w
∂z,∫ −hs
E
−H+hbE
vgdz = a tan θ
∫ −hsE
−H+hbE
∂w
∂zdz,
VS = a tan θ [w(z)]−hs
E
−H+hbE
(81)
— Now we can use Eq(78) for the vertical velocity at the base
of the surface Ekman layer,
w(−hsE) = wE =1
ρ0~k ·(∇× ~τs
f
). (82)
— At the base of the inertior layer, z = −H + hbE , we
Sverdrup Transport 75
traditionally set the vertical velocity to zero,
w(−H + hbE) = 0, commonly used assumption (83)
This assumption can be violated when the bottom
topography is strongly sloping, or when there are strong
boundary layers. The latter possibility will be discussed in
more detail later.
— The meridional volume transport, VS , then becomes
VS = a tan θwE = a tan θ1
ρ0~k ·(∇× ~τs
f
),
=f
β
1
ρ0~k ·(∇× ~τs
f
). (84)
This volume transport is called the Sverdrup transport.
— If we write the depth-integrated volume transport VS as the
product of a depth-averaged velocity v times the depth of
the ocean interior H − hsE − hbE ≈ H, since the Ekman layers
Sverdrup Transport 76
are 10s of meters thick while the ocean is thousands of
meters deep, then
VS = vH. (85)
In fact, we note that for a homogeneous interior,
ρ(z) = ρ0,−H + hbE < z < −hsE , we have constant interior
geostrophic velocity, vg = v.
— Substituting Eq(85) into Eq(84)
Hvβ =f
ρ0~k ·(∇× ~τs
f
). (86)
we can interpret it in terms of a vorticity balance. The large
scale wind stress curl is positive for example in the subpolar
gyres in the North Atlantic and North Pacific (see Fig. 2),
imparting positive vorticity (cyclonic, or in the same sense
as the rotation of the Earth) on the ocean. The relative
vorticity ζ f of the ocean interior being weaking relative
Sverdrup Transport 77
to the planetary voriticty f , the dominant term balancing
this voriticity input is the merdional advection of planetary
voriticity Hvβ ; i.e. v > 0 as the water drifts towards the
pole, gaining planetary vorticity. In the subtropical gyres in
the North Atlantic and North Pacific (see again Fig. 2), the
wind stress curl is negative and as a result v < 0 as the
water drifts towards the equator.
— Similarly we can interpret the Sverdrup relation Eq(56)
βvg = f∂w
∂z(87)
in terms of vorticity balances. When a water parcel, which is
spinning at f about the local vertical due to the Earths
rotation is subjected to vortex stretching (∂w∂z > 0), this
tends to increase the vorticity. This positive forcing term is
balance by the merdional advection of planetary voriticity
βvg > 0 as the water parcel moves poleward. This occurs in
Sverdrup Transport 78
the interior of the Northern Hemisphere subpolar gyres
where the Ekman pumping is upward ; the positive wind
stress curl drives upwelling. In the subtropical gyres the
negative wind stress curl drives downwelling, tending to
reduce vorticity which is achieved by drifting equatorward.
— Finally we note a limitation of the Sverdrup theory. Writing
the geostrophic velocity vg in terms of the pressure in the
Sverdrup relation Eq(94), we have
β1
ρ0f
∂p
∂x= f
∂w
∂z. (88)
On the other hand the RHS Eq(100) is determined by the
Ekman pumping and the local wind stress curl. Consider a
constant latitude, y =constant, where ∂w∂z > 0, for example
in the NA subpolar gyre. Integrating Eq(100) from the
Sverdrup Transport 79
western boundary, the pressure will monitonically increase
p(x, y) =ρ0f
2
β
∫ x
0
∂w
∂z(x′, y)dx′. (89)
On the other hand, we must meet the boundary condition
p(x, y) =constant along the boundary so that geostrophic
flux into the wall vanishes. Clearly this is impossible.
— This suggests that the friction term that we neglected above,
and which depends upon second derivatives of the pressure
field, becomes important in a boundary layer and physically
dissipates the vorticity provided by the wind and
mathematically allows two BCs to be meet on pressure on
either side of the basin.
bottom Ekman layer 80
Bottom Ekman Layer
bottom Ekman layer 81
Bottom Ekman Layer
— Like the surface Ekman layer, the bottom Ekman layer also
involves a 3-way momentum balance between horizontal
pressure gradient, Coriolis, and vertical derivative of the
horizontal turbulent stress.
— But in the bottom Ekman layer source of the turbulence
stresses is the interaction with the interior circulation and
the sea floor.
— A more complete description of this friction would involve
using a quadratic drag law as we used for the surface wind
stress, Eq(61). But the resulting equations become difficult
to treat analytically (see (Arbic and Scott , 2008) for a
numerical treatment).
— We follow the treatment of Stommel (Stommel , 1948) who
bottom Ekman layer 82
simplified the interaction by parameterizing the bottom
stress as linearly dependent upon the interior geostrophic
velocity above the bottom Ekman layer, ~ub
~τb = ρhbE |f |
2~ub. (90)
Comparing Eq(90) with Eq(61)
~τs = ρairCDU10~U10 (91)
we see that the velocity-dependent factor CDU10 has been
replaced by a constant hbE |f |/2.
— Using the linear bottom stress Eq(90) in the relation we
bottom Ekman layer 83
derived earlier between Ekman pumping velocity and stress
wE =1
ρ0~k ·(∇× ~τb
f
),
=1
ρ0~k ·(∇× ρhbE |f |~ub
2f
),
= sign(f)hbE2~k · (∇× ~ub) ,
= sign(f)hbE2ζb, (92)
where ζb is the vorticity of the interior geostrophic flow
above the Ekman layer. As anticipated by the argument
above, this term will turn out to be important only in a
narrow western boundary layer, where ζb is much larger than
in the deep-ocean interior.
Stommel Gyre 84
Stommel Gyre
Stommel Gyre 85
Stommel Gyre
— Stommel (1948) realised that the Sverdrup circulation could
be closed via a narrow boundary current in which the effects
of turbulent stresses become important and locally break the
Sverdrup meridional transport Eq(84).
— Correspondingly, we reconsider Sverdrup transport Eq(93),
replacing the bottom BC of vanishing vertical velocity,
Eq(83), with the Ekman pumping Eq(92) from the linear
drag, giving
VS = a tan θ [w(z)]−hs
E
−H+hbE
,
=f
β
(1
ρ0~k ·(∇× ~τs
f
)− sign(f)
hbE2ζb
)(93)
— A priori this boundary current could be either on the left or
Stommel Gyre 86
the right of the basin.
— But there is a simple argument that says that it should be
on the western boundary. Consider the NH subtropical gyres
where the vorticity input from the wind stress curl in the
central basin is negative, and the Sverdrup transport is
southward. The narrow boundary current must provide a
compensating, intense northward transport wherein the term
−hbE
2 ζb must be large and positive. This requires a negative
relative vorticity ζb < 0. An intense, north-flowing boundary
current on the eastern boundary would create a region of
positive vorticity, so −hbE
2 ζb < 0 would have the wrong sign.
On the other hand, a north-flowing boundary current on the
western boundary would create a region of negative
vorticity, so −hbE
2 ζb > 0 would have the correct sign.
— The argument still holds in the NH subpolar gyres with all
the signs reversed by the boundary current still on the
Stommel Gyre 87
western boundary.
— Consider the SH subtropical gyre. The easterly trade winds
to the north and the westerlies to the south (see Fig. 2)
create positive wind stress curl. Choosing our y-axis of the
local Cartesian coordinate system still pointing in the
northward direction even in the SH, we have f < 0 but
β > 0. There are two f in Eq(93) so their sign cancels, so
positive wind stress curl drives a northward (i.e.
equatorward) VS . The narrow boundary current must
provide a southward (poleward) flow in a region such that
f(−sign(f)hbE
2 ζb) < 0. This must be a narrow and positive
vorticity region, which we have seen must be adjacent the
western boundary.
Stommel Gyre 88
Stommel’s idealized subtropical gyre
— To obtain a simple idealized model to illustrate these ideas,
Stommel (1948) considered the case of a homogeneous
interior water column, ρ = ρ0 =constant. As we mentioned
briefly earlier, this implies vg = v, so we can replace
VS = Hvg = (H/fρ0)∂p/∂x in the Sverdrup transport,
VS =f
β
(1
ρ0~k ·(∇× ~τs
f
)− hbE
2ζb
),
H
fρ0
∂p
∂x=f
β
(1
ρ0~k ·(∇× ~τs
f
)− hbE
2
1
ρ0f∇2p
),
fhbE2H∇2p+ β
∂p
∂x=f2
H~k ·(∇× ~τs
f
)(94)
— This is a second-order linear inhomogeneous PDE for the
Stommel Gyre 89
pressure p. The wind stress curl represents a prescribed
forcing term.
— Stommel considered a rectangle basin on the β-plane,
f = f0 + βy. We consider the example of zonal extent
Lx = 5000km and meridional extent Ly = 3300km (as in
Carton §7.2).
— Following Stommel we prescribe a zonal wind stress
τx = −τ0 cos(π
Lyy), τy = 0,
~k · ∇ × ~τsf0
= τ0π
Lysin(
π
Lyy) (95)
noting that the β-plane approximation ignores the variation
in 1/f for simplicity.
— In the central basin (away from the western boundary) the
Stommel Gyre 90
Sverdrup transport applies and gives
∂p
∂x=
f0βH
τ0π
Lysin(
π
Lyy). (96)
This is easily integrated to give the interior solution,
pI(x, y) = p(x0, y) + (x− x0)f0βH
τ0π
Lysin(
π
Lyy). (97)
Imposing the boundary condition pI(Lx, y) = 0 (implying a
choice of p(x0, y) ) gives
pI(x, y) = (x− Lx)f0βH
τ0π
Lysin(
π
Lyy). (98)
— We now add on a narrow boundary layer solution pS(x, y),
with p(x, y) = pS(x, y) + pI(x, y). In the narrow boundary
layer,
∂
∂x ∂
∂y=⇒ ∇2 =
∂2
∂x2(99)
Stommel Gyre 91
and the dissipation is much larger than the forcing so the
full PDE Eq(94) reduces to
f0hbE2H∇2p+ β
∂p
∂x= 0,
f0hbE2H
∂2p
∂x2+ β
∂p
∂x= 0. (100)
This can be integrated to give
∂pS∂x
(x, y) = A(y)e− (x−x1)
LS , (101)
where LS is the Stommel boundary layer width,
LS =f0h
bE
2βH. (102)
and
A(y) =∂pS(x1, y)
∂x(103)
Stommel Gyre 92
and x1 = 0 is the western boundary.
— Another integration gives
pS(x, y) = pS(0, y)−A(y)LSe−x/LS , (104)
— This boundary solution pS must be matched to the interior
solution pI(x, y). This means when well outside the
boundary layer ( x/LS 1 ) but still near the western
boundary (x/Lx 1) (which is possible because LS Lx),
the boundary solution approaches the interior solution
pS → pS(0, y)→ −Lxf0βH
πτ0Ly
sin(πy
Ly). (105)
This gives our function pS(0, y) (it comes from the interior
solution), so that
p = pS + pI = −A(y)LSe−x/LS + (x− Lx)
f0βH
τ0π
Lysin(
π
Lyy)
(106)
Stommel Gyre 93
— We now impose p(0, y) = 0.
p(0, y) = pS + pI = −A(y)LS +−Lxf0βH
τ0π
Lysin(
π
Lyy),
=⇒ A(y) = −LxLS
f0βH
τ0π
Lysin(
π
Lyy). (107)
— Finally the total soluton is
p(x, y) =f0βH
τ0π
Lysin(
π
Lyy)[(x− Lx) + Lxe
−x/LS ]. (108)
Terminology in climate science 94
Terminology in climate science
— Monthly mean climatology : mean of each month, taken from
data over many years. For example, March SST is obtained
by averaging data only from the month of March but over
many years.
— Season climatology : like the monthly mean climatology, but
grouping months together to form seasons. Typically we
devide the year into 4 seasons of 3 months each. There are
different conventions, but a typical choice is January,
February and March (JFM) for winter, April, May, June
AMJ for spring, JAS for summer, OND for autumn. But
another commonly found convention is DJF for winter etc.
Again the JFM climatology will be an average over many
years of historical data.
Terminology in climate science 95
— Meridional average. Is an average in the meridional
direction. Meridians run North-South so this is an average
over latitude at fixed longitude.
— Zonal average. The zonal direction is East-West so this is an
average over longitude at fixed latitude.
Summary 96
Thesaurus
— Diapycnale, is a surface of constant density.
— downwelling (see also upwelling)
— Ekman transport
— Ekman pumping
— geostrophy
— Potential temperature, θ, is the temperature that a water
parcel would have if moved adiabatically to a reference
pressure. The surface is the most common reference level.
— S, denotes salinity in PSU (practical salinity units). The
salinity is the concentration of dissolved salt.
— T , denotes temperature.
— θ, denotes potential temperature.
— thermal wind
Summary 97
— thermocline
— upwelling (see also downwelling)
Summary 98
Practical resources for later use
— The MatLab routines that implement the EOS for seawater,
based upon the PSU for salinity, are available here http://
www.cmar.csiro.au/datacentre/ext_docs/seawater.htm
— Notice that they encourage the user to update to the latest
version of these routines, which are based upon absolute
salinity rather than PSU. The problem here is that you will
find most available historical data in PSU, as we discussed in
class. If you are working in a coastal area (as opposed to a
floating deep-sea platform) then the key quantity you need
to verify to know if your EOS calculations are accurate are
the proportions of the various salts.
Summary 99
References
Arbic, B. K., and R. B. Scott (2008), On quadratic bottom drag,
geostrophic turbulence, and oceanic mesoscale eddies, J. Phys.
Oceanogr., 38, 84–103.
Stommel, H. (1948), The westward intensification of wind-driven
ocean currents, Trans. Am. Geophs. Union, 29, 202–206.
Talley, L., et al. (2011), Descriptive Physical Oceanography : An
Introduction, Elsevier Science.
Vallis, G. K. (2006), Atmospheric and Oceanic Fluid Dynamics :
Fundamentals and Large-scale circulation, 744 pp., Cambridge
University Press.