pace.

At one extreme, we find

whereas differences in atmospheric pressure play no signifi

viscosity, even after viscosity magnification by turbulenc

and the combination of the two wind systems drives significan
basins: North and South Atlantic, North and South Pacific, an

, a layer in which the vertical stratification is erased every

seasonal thermocline retreats; it is permanently stratifie

pressure fluctuations in the upper layers, it is believed that, in first approximation, the study of

, ocean circulation theory was long in coming, chiefly for lac

beginning with the first correct theory for the Gulf Strean (S
circulation theory is a significant body of knowledge (Warre

Large-scale flows in the main thermocline and abyss are slow a

coefficients (see Equation

ssible fluid)

This set of five equations for five unknowns (

only way for a parcel of fluid to change its vorticity is to adju

of fluid parcels induced by vertical

d and its flow is nearly geostrophic.

 Abyssal flow is extremely slow, and we may very well take

If we define the meridional transport as the vertically integ

d flow component in the north-south

ch we define as the vertical

 Ideally, we would wish to impose a boundary condition on the fl

zonal flow and its vertical integral (U

In the ocean interior, the flow is approximately geostrophic

ities are difficult to measure

hich the pressure field

oriously difficult to obtain directly from the deep

ocean. First, they are almost always fluctuating at the mesos
a constraint upon, rather than the forcing of, the flow.

In Sverdrup dynamics, the Montgomery potential defined as

flow contains only the spatial derivatives:

is conserved along the flow

e pressure field from pressure

the density field, is of greater use with data.

Specifically, the form is

 is conserved along the flow. Thus the potential
vorticity defined above in (

wherever the flow is on a large scale and dissipation is weak, t

value along the way, a moving parcel is confined to stay on

: The same parcel is also confined to stay on a surface
 Along its journey towards the equator, a fluid column experie
low the flow in the vertical
finition of the trajectory, no transverse flow

zero transverse flow below, the horizontal velocity vector t

), we concluded that the simplified Sverdrup dynamics do not

 quires that water flow toward
the equator must be compensated elsewhere by poleward flow, but poleward flow violates
Sverdrup dynamics. Thus, this return flow must exist on a scal

ward Sverdrup flow in the basin’s interior creates a velocity

Sverdrup flow Sverdrup flow

The two possible configurations for a northward boundary cur

Sverdrup flow that exists across most of an ocean basin in the m

ct configuration.

a, the northward flow has positive

Thus, if flow is returning on the eastern side, it has positive

is still negligible, the return flow has negative

go from about zero in the Sverdrup flow to a negative value in th

ward flow (the Sverdrup transport) occupying most of the doma

is naturally identified with the Gulf Stream of the North Atla

and Kuroshio in the North Pacific, for example). The circulat

which flow eastward to the north and westward to the south. The

westward intensification by Henry Stommel, who provided the first correct

exerted by a surface stress on a viscous fluid. The chief reaso

flow transverse to the winds, which converges, resulting in a downward flow into the ocean

is about 30 m per year), this vertical flow squeezes water parc

fluid parcels flatten and widen, and to conserve their circula

waters run into a region of slower flow, veering westward and gathering into a zonal flow
that intensifies downstream. Upon arriving at the western boundary, this flow turns into a
swift poleward flow, so swift that relative vorticity become

n boundary layer where the flow

) are likely. Stratification

is another aspect that requires ample consideration. Briefly, the effect of stratification is to
decouple the flow in the vertical and thus to make it respond le
of stratification causes baroclinic

an air–sea heat flux is created, resulting in the cooling of th

culation pattern. The interested reader will find additiona

turn our attention more specifically to the upper of these two

th and South Pacific, and Indian

Oceans. They reveal similar patterns in all five oceans. The p
 ) in the five
dle panel: North and South Pacific.
gence of the flow inside the mixed layer (Cushman-Roisin, 198
, which can be defined as the deposition of fluid formely belong

flowrate per unit horizontal area (

Consider a fluid column of infinitesimal width and extending f

thermocline upward to the base fluid of the Ekman layer. The density is this fluid column is
state of mixing (by definition).

represents subduction proper (flow into the

Ekman pumping, and convergence of geostrophic flow in the sea

terior systematically reflect surface water properties of l

descent into the stratified thermocline.

lated area thus reflects the den-

a theory using an inviscid and non-diffusive fluid seems appr

The theory has been extended to continuous stratification by

In these zones, the flow is circulating without surface conta

effectively filling the abyss of the world ocean. Known areas

Surface flow

Abyssal flow

A highly simplified model of the abyssal circulation, with “l

first two equations yields

 ) from a flat bottom yields the

confined to small marginal regions), the abyssal flow must be n

latitudes should create flow away from, not toward, the poles
is that the flow coming from the high-latitude regions is confi
al flow consists of the return
flow toward higher latitudes.

demands that the flow carried by the interior (boundary integ

flow ( ) be compensated by the exiting western boundary layer flow (
boundary-layer flow is as strong as the source, while the northward Sverdrup flow is
equal to the source implying that the half the flow is pure reci

, 1958) to verify their findings.

. The source is insufficient to sustain the required upwellin

northward boundary-layer flow across the equator is necessa
a situation is prevailing in the North Pacific.

, requiring zero normal flow at the eastern boundary, consist

ry layer feeding this flow must

The velocity field (

f geophysical fluid dynamics

assumptions was that of a flat bottom. The oceanic bathymetry

les and passages guiding the flow.
Among the special features that ridges and passages inflict on the flow are concentrated zonal

A milestone in numerical ocean modeling was the first

 A few trajectories of the abyssal flow in a basin sector (left p

LODYC, ORCA configuration; Madec

). The release to the scientific

code was based on straightforward second-order centered fin

This model enabled the first general circulation studies wit

 ) was also ill-fitted to simulate weak bottom slopes and the as

resentation of overflows typical

(Dietrich, 1998), which uses a modified Arakawa “A” with four

weak slope is awkwardly approximated by a flat bottom stretch

oped. Perhaps the most significant change in terms of numeric

providing great flexibility in terms of geographical covera

 schematic overflow where dense bot-

Real flow

An unstructured finite-volume approach is a generalization of the finite-volume approach

, in which integration is performed over each finite volume, w
ng between the finite volumes
arises naturally through the fluxes across the shared interf

in finite-element methods are of a

Beside the widespread finite-volume and finite-elements met

of the flow (ratio of vertical scale to horizonatal scale – see

r formation does influence broader-

scale flow and needs to be included in these models. The brutal
 Example of a finite-

eventually found to be flawed (Cessi, 1996), for it generates

finement with flexible horizontal grids, and today’s hydrost

of free-surface models because of their added flexibility an

) remains a difficult problem because

specification of the eddy viscosity and diffusivities, part

vertical gridding is again considered crucial, and a signifi

finite-volume integration

cell, or we can first perform a change of coordinates after whi

llowed by finite differencing runs

in fixed coordinates

tion on a fixed grid. The discrete equations to be solved by the

along the flow. An important expression in the coordinate tra

substitutes for the vertical velocity. It is defined as

is the vertical velocity of the flow relative to the moving

is density and if density is conserved with the flow

boundaries, because the flow must follow that material bound

of change of variables and the definition of the vertical velo

particularly well suited for integration over a finite volum

is specified. The choice is at the modeler’s discretion.

very popular in coastal modeling. This coordinate is defined

 e efficiently used because they all

Beckers, 1992). Although the problem was initially identifi

 is defined as

e density and pressure fields

same depth as the horizontal gradient. For simple finite diff

Alternatively, for a fixed vertical grid and given slopes, th

grid that must be sufficiently fine to resolve the slopes accur
 l grids. Since stratification on

re large, leading to a significant

order finite differencing (using

of average density profiles

specialized finite differencing (

Once the grid is defined, and the shortest resolved scale know
bottom, which the flow is forced to follow, a terrain-followi

is the presence of additional terms and non-constant coeffic

Baroclinic instability releases potential energy by flatte

ount for such a flattening since by
orcing them to flatten out.
 Vertical section across a density field with frontal structu

nsity field in time, so that some time later

existing slope, the flattening of isopycnals slows down over

the isopycnals, with the coefficient

dimension of a diffusion coefficient (length squared per tim
pressed in Cartesian coordinates, it is easily verified that

tively advects the density field and, with the chosen signs, l

amical equation, reflecting the

, and Griffies (1998) shows how one may combine bolus advectio
s both “diffusion” coefficients

reader is referred to Griffies