Download as pdf or txt
Download as pdf or txt
You are on page 1of 57

3.

The Circulation of the Southern Ocean


In this chapter we discuss the circulation of the SO ... We start with a general description of
special features .. models ..

3.1. Some basic ingredients of Southern Ocean physics

Roughly 75% of the world ocean volume has temperatures below 4∘ C, connected with
only 2% of the ocean surface (at polar latitudes). Paleoceanographic data have revealed
that this was not always the case. Before Drake Passage opened due to continental drift
about 30 Myr ago, the climate of the ocean was considerably warmer. In the course of
the establishment of the Southern Ocean1 in its present shape the difference between sur-
face and bottom temperatures in equatorial regions changed from about 7∘ C to its present
value of about 26∘ C. The polar climate of the southern hemisphere got increasingly colder
by the growth of glacial ice on the Antarctic continent and the gradual development of
the sea ice cover around it, leading to the forming of deep cold water masses propa-
gating as Antarctic Bottom Water (AABW) to the adjacent northern ocean basins. The
opening of Drake Passage also established the strongest and longest current system in the
world ocean, the Antarctic Circumpolar Current (ACC), extending around the globe with
a length of roughly 24,000 km. The Antarctic water ring ranges from the Antarctic conti-
nent to about 50∘ S. As the most important link between the ocean basins of the Atlantic,
Pacific and Indian Oceans, the ACC serves as a conduit of all active and passive oceanic
tracers which affect Earth’s climate, notably heat and salt which strongly influence the
oceanic mass stratification, circulation and consequently the ocean heat transport, and
the greenhouse gas carbon dioxide and other chemical and biological components. But
in contrast to this strong zonal exchange brought about by the deep-reaching and strong
zonal current, these same characteristics of the ACC act to limit meridional exchange and
tend to isolate the ocean to the south from heat and substance sources in the rest of the
world ocean.

The Antarctic Circumpolar Current The ACC is the world ocean’s most intensive cur-
rent system, flowing round the Antarctic continent with a volume transport of roughly
130–140 Sv (that is about 1000 times the Amazon discharge; it would empty the Baltic
1
Many oceanographers refer to the region around the continent of Antarctica as the Southern Ocean. The
International Hydrographic Bureau, which is the authority responsible for the naming of oceanic fea-
tures, does not recognize a sub-region of the world ocean of that name but includes its various parts in
the other three oceans. From an oceanographic point of view, subdivisions of the world ocean should
reflect regional differences in its dynamics. The Southern Ocean certainly deserves its own name on that
ground (Tomczak and Godfrey 1994).

51
52 Some basic ingredients of Southern Ocean physics

Figure 3.1.: Upper left panel: Schematic map of major currents in the southern hemisphere oceans south
of 20∘ S. Depths shallower than 3500 m are shaded. The two major cores of the Antarctic Cir-
cumpolar Current (ACC) encircling Antarctica are shown: the Subantarctic Front and Polar
Front. F stands for front, C for Current and G for gyre. Upper right panel: Instantaneous
sea surface height [in meter] from the ACC model described in TURB . Lower panel: Mag-
nitude of the mean sea surface temperature gradient, from 44 months of observations by the
Along-Track Scanning Radiometer on the ERS-1 satellite. Superposed are positions of (from
north to south) the Subtropical Front, Subantarctic Front, Polar Front, South ACC Front, and
southern boundary of the ACC, taken from Orsi et al. (1995). From Hughes and Ash (2001).

Sea in three days). This is the largest transport rate in the ocean and can largely be ex-
plained by the mostly westerly, very intensive winds (see Figure 1.1) in the Southern
Ocean. Thermohaline processes in form of the stratification also play an important role.
From the picture of the global conveyor belt (see Figure ??) dies Bild fehlt jetzt it can be
seen that the deep structure of the ACC has a specialty: while in all other large-scale
current systems the deep current flows against the surface current, the ACC flows com-
pletely in one direction: the surface as well as the bottom currents are almost everywhere
eastward.

In the north, the Antarctic water ring is limited by the southern extent of the Sub-
tropical Convergence which winds itself around the globe as a nearly continuous but
diffuse water belt, a broad zone of transition between the tropical/temperate and the po-
3 The Circulation of the Southern Ocean 53

Fundamental Box 2 : The ACC transport

The meridional momentum balance of the ACC is basically geostrophic, i.e. the zonal current velocity (at
each geopotential level) is related to the meridional pressure gradient, resulting from a dip of about 1.5
m (from north to south) of the sea surface 𝑧 = 𝜁 across the current system, and the gradient of density in
the fronts as e.g. can be inferred from the WOCE-SR3 section in Figure 3.3. The surface pressure gradient
yields an overall eastward surface velocity 𝑢𝑔 (0) = −(𝑔/𝑓 )𝜁𝑦 , and the mass stratification yields a positive
shear 𝑢𝐺𝑧 = (𝑔/𝑓 )𝜌𝑦 of the geostrophic part of the current. The velocity thus diminishes with depth but
generally not as strongly as to imply a reversal of the flow.
The above ’thermal wind relation’ is utilized to infer from hydrographic section data the ’baroclinic’ trans-
port (normal to the section and referred to a common depth) or the DCL transport (referred to the bottom
depth or common deepest level (DCL) for station pairs) of the ACC. The average DCL transport through
Drake Passage (for six hydrographic sections of WOCE-SR1) is 136.7 ± 7.8 Sv, with about equal contribu-
tions from the Polar Front (57.5 ± 5.7 Sv) and the Subantarctic Front (53 ± 10 Sv) (Cunningham et al. 2003).
The mean transport south of Australia at SR3 is 147±10 Sv (Rintoul and Sokolov 2001). The transport south
of Australia must be larger than that at Drake Passage to balance the Indonesian throughflow, which is
believed to be of order 10 Sv. However, given the remaining uncertainty in the barotropic flow at both
locations, the agreement is likely to be fortuitous. Variability of transport at SR3 has been detected in a 6
years record of repeat hydrographic section (Rintoul et al. 2002). It is fairly small (1-3 Sv).
It should be borne in mind that the geostrophic transport is not the total transport. This is the integral
of absolute velocity across a section. The total volume transport through a section also contains the Ekman
transports (due to windstress 𝝉 0 and frictional bottom stress 𝝉 𝑏 ) and other contributions induced by non-
linearities and lateral friction F. Neglecting the nonlinear advective terms in the momentum balance, we
find the expression for the total transport U (vertical integral of absolute velocity)

𝑓 k × U = −ℎ∇𝑝𝑏 − ∇𝐸 + 𝝉 0 − 𝝉 𝑏 + F (3.1)

where 𝑝𝑏 is the bottom pressure and 𝐸 the baroclinic potential energy (see BARBI section 4.1). In large-
scale currents the bottom and lateral friction terms are usually small, and for north-south sections the
surface Ekman contribution can be neglected for a predominantly zonal windstress 𝝉 0 . The zonal trans-
port is then nearly geostrophic with respect to the two pressure gradient terms in (3.1).

lar oceans where the permanent thermocline reaches the surface. There are other (and
more pronounced) such fronts sketched in the upper left panel of Figure 3.1 and in more
detail in the lower panel. The most important are the Subantarctic Front and the Polar
Front at about 60∘ S. These fronts escort the ACC which is not a single current, but is
composed of several filaments. These are traced by the regionally and temporally highly
variable surface temperature gradient displayed in Figure 3.1, which shows that the ACC
is a fragmented system of more or less intense jet streams. The thermal fronts have a
close correspondence in density and extend to great depths, in most places to the bottom
(see below Figure 3.3), but can also be correlated with surface elevations as detected in
satellite altimetry data. At these fronts the ocean water masses are subject to an abrupt
change in their physical, chemical and biological features. We must not imagine them as
static lines; they are always shifting, extending, withdrawing, sometimes forming con-
vexities and strangulating intensive mesoscale eddies. This mesoscale turbulence, the
ocean weather, accompanies the current and exerts a strong influence on it.
54 Some basic ingredients of Southern Ocean physics

Eddies The zonal periodicity of the Southern Ocean creates a circumpolar pathway of
watermasses circling the globe and allowing the ACC to play a major part in the conveyor
belt circulation. But the zonallity also acts as a brake. In the basins which are zonally
blocked by continents we find a meridional exchange of heat accomplished by the time
mean gyre current systems. There is no significant mean transport of heat across the
latitudes of the ACC (DeSzoeke and Levine 1981). Instead, heat and other tracers must
be carried across the current by smaller-scale and time-varying features in the current
field, usually summarized as the meso-scale eddy field.

Transient eddies with scales of tens to a hundred km (much larger than the baroclinic
Rossby radius which is of order 10 km in the Southern Ocean, see Figure ??) are a very
prominent feature along the path of the ACC, as illustrated in a numerical experiment
of the Southern Ocean circulation of Figure ??. The mesoscale eddy activity leaves an
obvious imprint. The upper right panel of Figure 3.1 shows an instantaneous snapshot
of the sea surface height in a model of the the Southern Ocean circulation, and the lower
panel displays the sea surface temperature from satellite data. At large, these trajecto-
ries of the flow explicitly show the single branches of the Circumpolar Current; however,
each trajectory has considerable torsions with scales of several tens of kilometers, caused
by being carried along in the mesoscale eddies. This eddy activity reaches from the sur-
face to great depths and is characteristic for all large current systems. In the Circumpolar
Current, however, eddies are particularly pronounced. They are responsible for the pole-
ward heat transport across the current and the vertical transport of momentum into the
deeper layers of the ocean. Estimates of the meridional eddy heat flux from a number of

AUSSAF AUSSAF + Others


0
(a) (b)

1000
Depth (m)

2000

3000 shear, eddy


West NDP 90−day, eddy
North shear, all−freq
Central SENZ geog, all−freq
South NDP geog, all−freq

4000
−60 −20 20 −100 −60 −20 20
−2 −2
Heat Flux (kWm ) Heat Flux (kWm )

Figure 3.2.: Left: Profiles of cross-stream eddy heat flux observed in the AUSSAF experiment. (a) four
AUSSAF moorings, symbols indicate statistical significance (triangle 95 %, circle 90 %).
(b) representative profiles from all available SAF sites: AUSSAF (circles), southeast of New
Zealand (SENZ: Bryden and Heath 1985, triangles), northern Drake Passage (NDP: Nowlin
et al. 1985, squares). From Phillips and Rintoul (2000). [geog: northward velocity is used,
shear: velocity normal to mean current shear is used; the correlations are either determined
from all frequencies or from a bandpass covering the eddy frequencies]. Right: Gille 2003 ???.

moored instruments confirmed the southward transfer with sufficient magnitude to close
the overall heat budget (see left panel of Figure 3.2). Recently Gille (2003) confirmed the
enormous extrapolation from such point observations to the circumpolar area by deter-
3 The Circulation of the Southern Ocean 55

Fundamental Box 3 : Transient and standing eddies

In the description of the global atmospheric circulation it is custom to reduce the information contained
in observations by considering zonally averaged time mean fields and deviations from it (see e.g. Peixoto
and Oort 1992); e.g. the meridional velocity 𝑣 is split into its zonal-plus-time mean 𝑣, denoted by an
overbar, and deviation 𝑣 ′ so that 𝑣 = 𝑣 + 𝑣 ′ . The mean meridional heat flux (divided by 𝜌𝑐𝑝 ) is then
𝑣𝜃 = 𝑣𝜃 + 𝑣 ′ 𝜃′ which identifies a flux achieved by the zonal-plus-time mean fields and a flux induced
carried by the covariance of the deviation fields (the ’eddies’). Eddies in this definition have a nonlocal
character because they include the deviation from the zonal mean. Clearly, the combined zonal-plus-time
mean of the eddy field 𝑣 ′ vanishes, i.e. 𝑣 ′ = 0, but the time mean alone does not vanish. Denoting the
temporal average by cornered brackets, we obtain from 𝑣 ′ a separation into a time mean (’standing eddy’)
component and the time-varying (’transient eddy’) component, 𝑣 ′ = ⟨𝑣 ′ ⟩ + 𝑣 ★ ; and likewise the ’eddy’
heat flux, 𝑣 ′ 𝜃′ = ⟨𝑣 ′ ⟩ ⟨𝜃′ ⟩ + ⟨𝑣 ★ 𝜃★ ⟩, has the contributions from standing and transient eddies.
Motivated by the zonal unboundedness of the ACC, this separation of fields and covariances has been
applied to data for the belt of latitudes passing Drake Passage. But in a zonally oriented average most of
the ACC actually finds its representation in the averaged picture at latitudes north of the Drake Passage
belt. As a consequence we have standing eddy contributions which are strong compared to the transient
eddy contributions. The zonal average at Drake Passage latitudes does neither represent the local structure
of the current in Drake Passage nor the mean properties of the ACC. The zonal average does not separate
the time mean and the transient motion in a simple way. Zonal mean, standing eddy and transient eddy
components arise, and the standing eddy component is a major player. Dynamically it belongs to the time
mean flow.
More appropriate is an average oriented along mean streamlines or the contours of the time mean sea
surface height (SSH). This kind of averaging drastically reduces the standing eddy part (not entirely be-
cause the current may veer with depth) and reveals the true nature of the transient eddy field to transport
properties across mean streamlines.

zonal mean velocity mean tangential velocity


01
02

0 05 0 15
−200 −200
0 01

−400 −400
0 05 0 05

0
−600 01 −600
−1000 −1000

−2000 −2000
depth

depth

−3000 −3000

−4000 −4000

−5000 −5000
−75 −70 −65 −60 −55 −50 −45 −40 −35 −56 −54 −52 −50 −48 −46 −44
latitude latitude

The zonal average [right panel] and the path averaged tangential velocity (in ms−1 ) [left panel], following the time
mean SSH contours. The latitudinal coordinate is now oriented at the mean latitude of the SSH contours, the width
spans the entire ACC. From POP model data. POP???

mining the eddy heat flux from a combination of climatological hydrographic data and
the ALACE float trajectories tracing out the Southern Ocean in the last decade (see right
panel of Figure 3.2).

The Meridional Overturning From Figure 3.3, showing watermass properties on a sec-
tion between Australia and Antarctica (≈ 140∘ E), it becomes apparent that watermass
properties do penetrate across the ACC and, in fact, there is a prominent meridional cir-
56 Some basic ingredients of Southern Ocean physics

Figure 3.3.: Properties versus pressure along the WOCE SR3 repeat section between Australia and Antarc-
tica (≈140∘ E). Left: potential temperature. Middle: salinity. Right: neutral density. The
section was occupied in winter (September) 1996. From the Southern Ocean Hydrographic
Atlas (Orsi et al. ????). HASO?

culation associated with the predominantly zonal ACC. The figure implies this overturn-
ing circulation by the distribution of salinity, temperature and phosphate which reveal
the prominent watermasses: Antarctic Intermediate Water (AAIW) is seen in the blue core
of low salinity, Circumpolar Deep Water (CDW) is seen in the deeper brown core of high
salinity and AABW in the deep blue core of low potential temperature. The meridional
circulation was described as early as 1933 by Sverdrup (see also Sverdrup et al. 1942) and
has lately been interpreted as the Southern Ocean part of the ’global conveyor belt’ circu-
lation (Gordon 1986, Broecker 1991, Schmitz 1995). A sketch of the zonal and meridional
circulation is shown in Figure 3.4. The core of the eastward flowing ACC is associated
with the Polar Front and the Subantarctic Front. In the meridional-vertical plane an up-
per cell is formed primarily by northward Ekman transport beneath the strong westerly
winds and southward transport in the Upper Circumpolar Deep Water (UCDW) layer.
The lower cell is driven primarily by formation of dense AABW near the Antarctic con-
tinent and inflowing Lower Circumpolar Deep Water (UCDW or NADW). This deep,
relatively saline water spreads poleward and upwells towards the sea surface. Shallow-
ing of the isopycnals is evident as the deep water rises up towards the sea surface. There
it is partly transported southward, it cools and sinks, flooding the bottom layers with
waters of less than 0∘ C. This cold bottom water spreads well into the global oceans as
AABW. The inflow is balanced by a northward flow of lower salinity waters near 1000
m (AAIW) and by sinking of slightly lower salinity water along the continental slope of
Antarctica. This process (salty water in, fresher water out) removes the slight excess of
regional precipitation from the Southern Ocean.
3 The Circulation of the Southern Ocean 57

Figure 3.4.: A sketch of the ACC system showing the zonal flow and the meridional overturning circulation
and watermasses. Antarctica is at the left side. The east-west section displays the isopycnal
and sea surface tilts in relation to submarine ridges, which are necessary to sustain the bottom
formstress signatures (see Olbers et al. 2004). The curly arrows at the surface indicate the
buoyancy flux; the arrows attached to the isopycnals represent turbulent mixing. Redrawn
using a figure from Speer et al. (2000).

3.2. Homogeneous models

In this section we explore the momentum balance and the flow characteristics of a wind-
driven current in simple configurations of the Southern Ocean for different conditions, as
e.g. a flat-bottom ocean compared to an ocean with submarine topography, or different
scenarios of vertical compared to lateral momentum transport. In most of what follows
we will employ the Elementarstrom model (see Figure 3.5) which was introduced in sec-
tion 2, now, however, for southern hemisphere conditions where 𝑓 < 0. For simplicity
we consider an ocean with a southern solid boundary at 𝑦 = 𝑦𝑆 . It might be open or
closed (then we have a channel) at a northern latitude 𝑦 = 𝑦𝑁 . We assume that the wind-
stress is zonal and that it has limited meridional support. In some examples we will use
a sinusoidal form
( )
𝑦 − 𝑦𝑆
𝜏0𝑥 (𝑦) = 𝜏𝑠 sin 𝜋 (3.2)
𝑦𝑁 − 𝑦𝑆

and zero for 𝑦 > 𝑦𝑁 and start from zonally and temporally averaged equations

Δ𝑝
− 𝑓 𝑣¯ = − + 𝜏𝑧𝑥 + 𝐹𝑦𝑥 (3.3)
Δ𝑥
𝑓𝑢
¯ = 𝑝𝑦 + 𝜏𝑧𝑦 + 𝐹𝑦𝑦
−¯ (3.4)
𝑣¯𝑦 + 𝑤
¯𝑧 = 0 (3.5)

The term −Δ𝑝/Δ𝑥 only occurs in the presence of topography. It arises from the deep
pressure difference in the valleys between topographic peaks, as explained in Funda-
mental Box ??. The pressure is written as sum of the surface part and the baroclinic part,
58 Homogeneous models

Fundamental Box 4 : Bottom formstress

The pressure difference Δ𝑝 in the zonal momentum balance (3.3) is a shorthand notation of the sum over
all ridges blocking the zonal path at the depth 𝑧, of the pressure difference Δ𝑖 𝑝 = 𝑝𝑊 𝐸
𝑖 (𝑦, 𝑧) − 𝑝𝑖 (𝑦, 𝑧)
associated with the individual ridge numbered by 𝑖 (note that 𝐸/𝑊 refer to the eastern/western side of
the ridge, not of the intermediate valleys). Hence

Δ𝑝(𝑦, 𝑧) = Δ𝑖 𝑝(𝑦, 𝑧) (3.7)
𝑖

vanishes above the highest topography blocking the path at the depth 𝑧 = −𝐷(𝑦) and latitude 𝑦. The
vertical integral

0 0 ∫ 𝐸
1 ∑ 𝑥𝑖 (𝑧) ∂ℎ
∫ ∫
Δ𝑝 1 ∑
ℱ(𝑧) = − 𝑑𝑧 = − Δ𝑖 𝑝(𝑦, 𝑧)𝑑𝑧 = − 𝑝𝑏 𝑑𝑥 (3.8)
𝑧 Δ𝑥 Δ𝑥 𝑧 𝑖
Δ𝑥 𝑖 𝑥𝑊 (𝑧) ∂𝑥
𝑖

is zero for 𝑧 > −𝐷 and accumulates at the greatest depth (occurring in the deepest valley at the latitude
𝑦) the total bottom form stress

1 ∂ℎ
ℱ𝑏 = ℱ(−ℎ) = 𝑝𝑏 𝑑𝑥 = 𝑝𝑏 ℎ𝑥 = −ℎ𝑝𝑏𝑥 (3.9)
Δ𝑥 ∂𝑥

Note that the formstress becomes negative (taking eastward momentum out of the water and transferring
it to the solid earth) if on average 𝑝𝑊 > 𝑝𝐸 . In this case the geostrophic current within the valleys is
southward.

Figure 3.5.: The Elementarstrom system with topography. must redrawn, put in Ekman layers etc!!!

∫ 0
𝑝¯ = 𝑔 𝜁¯ + 𝑔 𝜌¯𝑑𝑧 = 𝑔 𝜁¯ + 𝑝𝑐𝑙𝑖𝑛 (3.6)
𝑧

We refer to the turbulence supporting the stress 𝝉 = (𝜏 𝑥 , 𝜏 𝑦 ) as ’billows’ transporting


horizontal momentum vertically, and the turbulence supporting the F = (𝐹 𝑥 , 𝐹 𝑦 ) terms
as ’eddies’, transporting horizontal momentum laterally. In some of the educational ex-
amples we will specify F as a downgradient diffusion of horizontal momentum though
it is generally agreed that this is not an appropriate representation of the lateral eddy
momentum transport.

It is commonly agreed and part of the Elementarstrom model that the stress 𝝉 in
the water column is very small beneath Ekman layer. We assume it to be zero in the
ocean interior, with the exception of a possible Ekman bottom boundary layer between
3 The Circulation of the Southern Ocean 59

𝑧 = −ℎ + 𝑑𝑏 and the bottom, located at 𝑧 = −ℎ. The surface Ekman layer has depth 𝑑𝑒 .
Here we put 𝜏𝑧𝑥 = 𝜏0𝑥 /𝑑𝑒 where 𝜏0𝑥 is zonal windstress, and assume that 𝜏 𝑦 ≡ 0 in the
Ekman layer and the interior (except possibly in the bottom layer).

The mass balance implies by vertical integration that 𝑈𝑥 + 𝑉𝑦 = 0 with (𝑈, 𝑉 ) =


∫0
¯ ¯
−ℎ (𝑢, 𝑣)𝑑𝑧. Zonal averaging yields 𝑉𝑦 = 0, hence 𝑉 = const. In an ocean with a southern
boundary where 𝑣¯ = 0 we thus have

∫ 0
𝑉¯ = 𝑣¯𝑑𝑧 = 0 (3.10)
−ℎ

Homogeneous flat bottom ocean First we consider a flat-bottom ocean with homo-
geneous density, i.e. ℎ = const, 𝜌 = const. The pressure difference term in (3.3) is then
absent, i.e. F = 0. Suppose also for the moment that eddies are absent. The zonal current
¯ is then geostrophic outside Ekman layers at the top and the bottom. Taking the vertical
𝑢
integral of (3.3), we notice the necessity of a frictional bottom stress (𝑥-component 𝜏𝑏𝑥 ) to
balance the zonal momentum,

0 = 𝜏0𝑥 − 𝜏𝑏𝑥 (3.11)

To sustain that stress, a current must be present rubbing on the floor, which can only
be a geostrophic current 𝑢 ¯𝐺 = −¯ 𝑝𝑦 /𝑓 = −(𝑔/𝑓 )𝜁¯𝑦 associated with meridional tilt of the
sea surface (see section 2.1.2 on the Elementarstrom system). In the barotropic state it
penetrates undiminished the whole water column and implies a stress (from (3.12))

1
𝝉𝑏 = 𝑔𝑑𝑏 𝜁¯𝑦 (1, −1) (3.13)
2

and hence from (3.11) we find

𝜁¯𝑦 = 2𝜏0𝑥 /(𝑔𝑑𝑏 ) or ¯𝐺 = −2𝜏0𝑥 /(𝑑𝑏 𝑓 )


𝑢 (3.14)

The sea surface is sloping upward to the north (for 𝜏0𝑥 > 0). The geostrophic flow occurs
at all depths and flows into the direction of the windstress. In addition, in terms of
transport, we have meridional transport 𝑉¯𝑒 in the surface Ekman layer and zonal and
meridional transports (𝑈¯𝑏 , 𝑉¯𝑏 ) in the bottom Ekman layer, given by

¯𝑒 , 𝑉¯𝑒 ) = (0, −𝜏 𝑥 /𝑓 )
(𝑈 and ¯𝑏 , 𝑉¯𝑏 ) = (𝜏 𝑥 /𝑓, 𝜏 𝑥 /𝑓 )
(𝑈 (3.15)
0 0 0
60 Homogeneous models

Fundamental Box 5 : Ekman spiral and transport in the southern hemisphere

The Ekman solutions for the southern hemisphere are given here for reference. When 𝑓 < 0 the real part
of the solutions for the surface Ekman layer is

[ ]
u𝐸 = 𝑑/(2𝐴𝑣 ) 𝑒𝑧/𝑑 (𝝉 0 + 𝝉 0 ) cos 𝑧/𝑑 + (𝝉 0 − 𝝉 0 ) sin 𝑧/𝑑
¬ ¬


where 𝑑 = 2𝐴𝑣 /∣𝑓 ∣ is the Ekman depth (which may be different for the surface and bottom Ekman
layers). For the bottom Ekman spiral we obtain

[ ]
u𝐸 = 𝑒−(𝑧+ℎ)/𝑑 −u𝐺 cos(𝑧 + ℎ)/𝑑 − u𝐺 sin(𝑧 + ℎ)/𝑑
¬

The bottom stress becomes


∂u𝐸
𝝉 𝑏 = 𝐴𝑣 = 𝑑∣𝑓 ∣/2(u𝐺 − u𝐺 ) (3.12)
∂𝑧 𝑧=−ℎ ¬

The top Ekman transport in the southern hemisphere is given by U𝑡𝑜𝑝


𝐸 = − 𝝉 0 /𝑓 , while the bottom Ekman
¬
transport becomes

U𝑏𝑜𝑡
𝐸 = 𝝉 𝑏 /𝑓 = 𝑑∣𝑓 ∣/(2𝑓 ) ( u𝐺 + u𝐺 ) = −𝑑/2 ( u𝐺 + u𝐺 )
¬ ¬ ¬

Hence we find 𝑉¯ = 𝑉¯𝑒 + 𝑉¯𝑏 = 0, as required by the mass balance. The implied meridional
circulation is closed by Ekman pumping arising from the windstress below the surface
layer and from frictional bottom stress, in the same manner as explained in section 2.1.2.
Note that the zonal geostrophic transport 𝑑𝑏 𝑢¯𝐺 = −2𝜏0𝑥 /𝑓 in the bottom layer is opposed
to and exceeds the frictional transport 𝑈¯𝑏 .

The standard parameters listed in Table 3.1 yield fairly reasonable values for the quan-
¯𝐺 ∼ 1.6 × 10−2 m/s, 𝐿𝑦 ℎ¯
tities of the above solution: 𝑢 𝑢𝐺 ∼ 64Sv, 𝐿𝑦 𝑈¯𝑏 ∼ −0.8Sv, 𝐿𝑥 𝑉¯𝑒 =
−𝐿 𝑉¯𝑏 = 16 Sv. An exception is the surface tilt. We find 𝜁¯𝑦 ∼ 2 × 10−7 which is an order
𝑥

of magnitude too small.

𝑓 −1.25 × 10−4 s−1 𝛽 = 𝑑𝑓 /𝑑𝑦 1.15 × 10−11 m−1 s−1 𝐴ℎ 104 m2 s−1
𝐿𝑦 1000 km 𝐿𝑥 20000 km ℎ 4000 m
𝑑𝑒 100 m 𝑑𝑏 100 m 𝜏0𝑥 10−4 m2 s−2
𝑁 1.2 × 10−3 s−1 𝜆 12.6 km 𝐾 1000 m2 s−1

Table 3.1.: List of parameters and standard values used for scale estimates in this section.

Next let us consider the presence of eddies so that the vertically integrated balances
3 The Circulation of the Southern Ocean 61

Figure 3.6.: Sketch of the solution of the homogeneous flat-bottom case. The zonal current is barotropic and
above the friction bottom layer in geostrophic balance with a sea surface tilt. The overturning
is due to the surface Ekman current, Ekman pumping in the interior and a frictional bottom
current. The fields are independent of the longitude 𝑥. Thiele

of zonal and meridional momentum become

¯𝑦𝑦
0 = 𝜏0𝑥 − 𝜏𝑏𝑥 + 𝐴ℎ 𝑈 ¯ = −𝑔ℎ𝜁¯𝑦 − 𝜏 𝑦
𝑓𝑈 (3.16)
𝑏

where the eddy momentum flux is expressed in the common diffusive form. Momen-
tum can now be exported in the meridional direction from the wind patch towards the
southern boundary where friction on the wall can work, or it can be exported towards
the north. The bottom friction is no longer needed to obtain a balance, but can we find a
reasonable solution for free-slip at the bottom, i.e. 𝝉 𝑏 ≡ 0? This assumption leads to

Hidaka’s dilemma The system (3.16) with zero bottom friction seems to be straightfor-
ward, but it turns out to be quite unphysical. From the zonal balance we infer the scaling
¯ ∼ (𝐿𝑦 )2 𝜏 𝑥 /𝐴ℎ so that the zonal transport scales as 𝐿𝑦 𝑈
𝑈 ¯ ∼ (𝐿𝑦 )3 𝜏 𝑥 /𝐴ℎ . With our
0 0
standard parameters this yields values exceeding 1000 Sv. More specifically, the above-
¯ 𝑑𝑦 = 𝜏𝑠 (𝑦𝑁 − 𝑦𝑆 )3 (4 +

mentioned sinusoidal windstress (3.2) yields a total transport 𝑈
2 3
𝜋 )/(2𝜋 𝐴ℎ ) (this is obtained for 𝑈¯ = 0 @ 𝑦 = 𝑦𝑆 ; 𝑈
¯𝑦 = 0 @ 𝑦 = 𝑦𝑁 ) amounting to 2240
62 Homogeneous models

Sv for our typical parameter values. The sea surface would increase by 6 m across the
ACC (a factor of 4 too large)! To arrive in a realistic size of the transport, we need to in-
crease 𝐴ℎ by more than a factor of 15 which becomes entirely unrealistic. Hidaka (1950??)
was the first to point out the dilemma of either choosing a reasonable value for the diffu-
sivity 𝐴ℎ , leading to a drastical oversize of the modeled ACC transport, or choosing an
unrealistic high 𝐴ℎ to arrive at the correct size of about 140 Sv.

The dilemma is mediated if we return to no-slip at the bottom. The geostrophic cur-
¯𝐺 = −(𝑔/𝑓 )𝜁¯𝑦 is, however, no longer the only driving of the frictional bottom layer,
rent 𝑢
but it is rather the interior velocity (¯
𝑢𝐼 , 𝑣¯𝐼 ) at the top of that layer. With a billow-less
interior this follows from

− 𝑓 𝑣¯𝐼 = 𝐴ℎ 𝑢
¯𝐼𝑦𝑦 ¯𝐼 = −𝑔 𝜁¯𝑦 + 𝐴ℎ 𝑣¯𝐼𝑦𝑦
𝑓𝑢 (3.17)

We may solve the entire problem for the sinusoidal wind, but note here that the lateral
Ekman number 𝐴ℎ /𝑓 (𝐿𝑦 )2 ∼ 10−4 is very small, and thus the interior current is well
approximated by the geostrophic part, i.e. u𝐼 ∼ = u𝐺 . This leads to 𝜏𝑏𝑥 = 𝑔𝑑𝑏 𝜁¯𝑦 /2 as before
in the eddyless case, and inserting now 𝜁¯𝑦 = −𝑓 𝑈 ¯ /(𝑔ℎ), derived from (3.16) with 𝜏 𝑦 =
𝑏
−𝑔𝑑𝑏 𝜁¯𝑦 /2 and 𝑑𝑏 ≪ ℎ, the balance (3.16) of the zonal flow becomes

1 𝑑𝑏 𝑓 ¯ ¯𝑦𝑦 = 0
𝜏0𝑥 + 𝑈 + 𝐴ℎ 𝑈 (3.18)
2 ℎ

The second term, resulting from the bottom stress, overcomes the lateral momentum
diffusion by one to two orders of magnitude so that the latter only plays a minor role
(except possibly in boundary layers). We find 𝑈 ¯ ∼ −2𝜏 𝑥 ℎ/(𝑑𝑏 𝑓 ) and 𝐿𝑦 𝑈
¯ ∼ 64 Sv and
0
𝜁¯𝑦 ∼ 10 , as in the eddyless case discussed above. Note that the bottom stress implies a
−7

frictional transport (𝑈¯𝑏 , 𝑉¯𝑏 ) = (𝑑𝑏 /2ℎ)𝑈


¯ (−1, 1). The same scaling argument used for the
interior flow is also valid for the upper layer, and we conclude that the flow in that layer is
almost contained in the Ekman transport 𝑉𝑒 = −𝜏0𝑥 /𝑓 . It balances the frictional transport
𝑉¯𝑏 in the bottom layer. In fact, the lateral momentum diffusion is of little significance in
this solution.

Homogenous ocean with topography Though formulated as a zonal average, the pre-
vious models of the ACC are valid in three dimensions: in the idealized condition with
no zonal perturbations the flow is strictly independent of the 𝑥-coordinate. Introduction
of a zonally varying topography ℎ = ℎ(𝑥, 𝑦) breaks the zonal symmetry. It leads to a
zonally undulating current and the presence of a deep pressure perturbations entering as
bottom formstress into the balance of the flow (see Box ...), i.e.

𝜏0𝑥 − 𝜏𝑏𝑥 + ℱ𝑏 + ℛ𝑥𝑏 = 0 (3.19)


3 The Circulation of the Southern Ocean 63

Example Box 12 : BARBI Southern Ocean model

The numerical experiments with the BARBI model (see section 4.1) for the circulation around Antarctica
(south of 20∘ S) have a horizontal resolution 2𝑜 × 1𝑜 , the horizontal viscosity is 𝐴ℎ = 4 × 104 m2 s−1 , and
the eddy diffusivity is 𝐾 = 4 × 103 m2 s−1 . The baroclinic cases have 𝑁 = 1.8 × 10−3 s−1 (discussed
later in section ??) which yields a baroclinic Rossby radius 𝜆 = 21 km. The model is forced with interpo-
lated annual mean windstress data from the NCEP reanalysis. The topography was interpolated from the
ETOPO20 dataset onto the model grid and smoothed with a two-dimensional symmetric filter. The Drake
Passage of the model is open between 62.5∘ S and 55.5∘ S. Figure ?? shows the the 𝑓 /ℎ-contours and the
zonal and meridional components of the windstress.

Model topography and forcing. Left: geostrophic contours 𝑓 /ℎ, CI = 5 × 10−9 m−1 s−1 . Middle and right: the
components of windstress [10−4 m2 s−2 ]. Bilder neu. CI??

where

∫ 0
𝑥
ℛ (𝑧) = 𝐹𝑦𝑥 𝑑𝑧 (3.20)
𝑧

denotes the integrated Reynolds stress divergence and ℛ𝑥𝑏 = ℛ𝑥 (−ℎ) is the bottom value.
The bottom formstress (BFS) ℱ𝑏 is defined in Fundamental Box ??. It depends on the zonal
pressure difference Δ𝑝(𝑦, 𝑧), appearing in (3.3). Neither of them can be determined from
zonal mean equations. In order to obtain a nonzero Δ𝑝 and ℱ𝑏 , there must be current
u at depth, and hence some functional dependence Δ𝑝 = Δ𝑝[ū, 𝑦, 𝑧] must exist. With
knowledge of this relation the model is closed and can be solved. We will work out
examples below (see the CDV section below). In general, however, the relation is too
complicated and not accessible.
We conclude this section on homogeneous ACC models with two barotropic simula-
tions of the BARBI model (see section 4.1), a flat bottom case FL and a topographic case
BT. Their outfit is briefly summarized in Example Box ??. The model has the realistic ge-
ometry, topography and wind-forcing of the Southern Ocean: the figures in the box show
the geostrophic contours 𝑓 /ℎ and the forcing. Streamfunctions and bottom pressure of
these experiments are displayed in Figure 3.7. The flat-bottom, homogeneous case (FL)
has an ACC transport of about 1100 Sv, The homogeneous ocean with topography (BT)
has very low ACC transport of 33 Sv. A stratified case NL with a transport of 140 Sv will
be discussed in section ?? below.
64 Homogeneous models

Figure 3.7.: Steady solutions obtained with the barotropic BARBI model (FL upper panels and BT lower
panels) forced by NCEP windstress. Left: streamfunction, CI = 10 Sv. Right: bottom pressure,
CI = 0.1 m2 s−2 . Positive contours are black, negative red.
Bilder neu. CI??

The flat-bottom case FL has an almost zonal ACC. There is slight squeezing of stream-
lines in Drake Passage and a marked northward shift in the path of the current behind
that obstacle. Since the bottom formstress cannot operate, friction is the only momentum
sink: with the extremely high transport and the moderate viscosity we are indeed facing
Hidaka’s dilemma. The topographic case BT establishes a bottom formstress arising from
the surface pressure being out-of-phase with the submarine barriers of the flow (more
details follow below). It has a transport that is far too low to represent ACC conditions.
The current is mostly along the geostrophic contours 𝑓 /ℎ = const (compare Figures ??
and ??) which run mostly north of the Drake Passage belt (the range of latitudes passing
through Drake Passage). Embedded in the circumpolar flow are huge closed circulation
cells in areas with closed 𝑓 /ℎ contours, namely above the Midatlantic Ridge and around
the Kerguelen plateau, with transports exceeding 100 Sv.
The bottom formstress arises from a systematic phase shift of the pressure 𝑃 with
respect to the underlying topography ℎ. To obtain a net westward acceleration of the
eastward current, highs of 𝑃 must appear to the west and lows to the east of topographic
barriers in the path of the flow. In fact, the flow organizes such that this sink of eastward
momentum can become effective. Figure 3.8 displays the bottom pressure (blue curve)
at a central latitude through Drake Passage together with the corresponding ocean depth
(red curve), and in a similar way for a latitude north of Drake Passage. The shaping of
the formstress in the topographic case BT is apparent: in the Drake Passage belt we notice
3 The Circulation of the Southern Ocean 65

latitude −59.5 latitude −46.5


2
0 1
FL −0.5 0
−1 −1

1 1
BT

0 0

−1 −1
0 100 200 300
longitude longitude

Figure 3.8.: Bottom pressure – depth correlation as a function of longitude for two sets of latitudes and
the cases FL and BT. The left column is for a latitude 59.5∘ S in Drake Passage, the right
column is for a more northern latitude at 46.5∘ S. Bottom pressure: upper curves [blue], ocean
depth: lower curves [red]. Scaled (depth is shown as −ℎ/𝑚𝑎𝑥(ℎ), pressure is shown by
𝑃/𝑚𝑎𝑥(∣𝑃 ∣) + 1).

a westward shift of roughly 10∘ of 𝑃 with respect to ℎ and no such (or a less correlated)
pattern at latitudes to the north. In the flat-bottom case FL, high pressure appears as
well west of Drake Passage, but in contrast to BT the zonal gradients are much smaller.
There is a gradual increase all around Antarctica which is balanced by the drop in Drake
Passage. The bottom formstress in FL is, of course, zero as ℎ = const.
Here we mention the general finding that the formstress is generally counteracting the
wind-driving to a high degree (it is a ’formdrag’), eliminating Hidaka’s dilemma even for
a free-slip bottom. Furthermore, in realistic conditions (eddy resolution, see section ??)
the Reynolds stress term is found non-diffusive and small. The frictional bottom stress
can be neglected as well, so the momentum balance of the ACC reduces to the simple
form2

𝜏0𝑥 + ℱ𝑏 = 𝜏0𝑥 − ℎ𝑝𝑏𝑥 = 0 (3.21)

which is a balance between the applied windstress and transfer of zonal momentum into
the bottom, occurring independently at each latitude. Munk and Palmen (1951) were
the first discussing this balance of momentum for the ACC but surprisingly, much of
the research on the ACC after Munk and Palmen’s article forgot the importance of the
BFS and tried frictional balances, e.g. Hidaka and Tsuchiya (1953), Gill (1968) (see above
discussion). In a homogeneous ocean, as discussed at present, the bottom pressure is
entirely due to a tilt of the sea surface. To balance a windstress of 10−4 m2 s−2 , only a few
dynamic centimeters is required across a ridge of a width of 1000 km (higher surface on
the upstream side).
The balance of FL and BT is shown in Figure 3.9. In the belt of latitudes which pass
through Drake Passage we find the ’form-drag balance’ (3.21) between windstress and
2
Friction is still necessary to close the energy budget because formstress is not an energy sink.
66 Vertical momentum transport and the meridional overturning

FL BT
2 2

1 1

0 0

−1 −1

−2 −2
−70 −60 −50 −40 −30 −70 −60 −50 −40 −30
latitude latitude

Figure 3.9.: Zonally integrated momentum balance as function of latitude for the FL and BT cases. Drake
Passage is indicated by purple lines. Windstress: yellow, bottom formstress: black, friction:
green, pressure difference on continents: red. Units: 10−4 m2 s−2 .

bottom formstress in the case BT (for clear reasons not in FL). Outside this belt the balance
is mainly between windstress and the pressure difference ℎΔ𝑃 on the continents. This, in
fact, is the principle balance of momentum in the basin-wide gyre circulations embedded
between continents (see section 2).

Both simulations of the BARBI model are clearly quite unrealistic for the ACC. What
is missing in the homogeneous model? In the following discussion we demonstrate that
stratification puts the system into a very different flow regime. At first, the baroclinic
pressure contributes to the meridional pressure gradient 𝑝¯𝑦 such that a increasing com-
pensation of the surface pressure gradient occurs with increasing depth. The geostrophic
current thus diminishes with depth (see also Fundamental Box ??). The bottom form-
stress gets a baroclinic component as well, acting generally against the barotropic one,
and thus accelerates the eastward flow. Most important, however, is the influence of
baroclinicity on the overturning circulation. In a stratified medium the current cannot
easily cross isopycnals, and hence the zonal average, highlighted in the previous discus-
sion, cannot reflect the local overturning anymore. We investigate these processes in the
following sections.

3.3. Vertical momentum transport and the meridional


overturning

We return to the zonally averaged framework of section 3.1 and, for simplicity, focus
for the moment on the interior flow with Δ𝑝 = 0 and no eddies. Evidently, the flow is
driven by a frictional ’billow’ stress 𝜏 𝑥 which is supposedly weak, and thus the merid-
ional velocity in the interior must be small compared to the Ekman current in the sur-
face layer. In the idealized cases we assumed ’zero billows’ and obtain thus a vanishing
meridional current in the interior: the flow is strictly vertical, supported by the pumping
𝑤¯ = −∂(𝜏 𝑥 /𝑓 )/∂𝑦 out of the surface Ekman layer. Remember that the zonal average was
performed on 𝑧-levels such that the zonal average of the pressure gradient 𝑝𝑥 vanishes
exactly. The presence of deep topography, stratification and eddies modifies this state.
We refer to this framework as the Eulerian frame.
3 The Circulation of the Southern Ocean 67

The Eulerian circulation Λ(𝑦, 𝑧) may be defined by Λ𝑧 = −¯ ¯ and Λ(𝑦, 𝑧 = 0) =


𝑣 , Λ𝑦 = 𝑤
0, hence

∫ 0
Λ(𝑦, 𝑧) = 𝑣¯𝑑𝑧 (3.22)
𝑧

is vanishing also on the bottom 𝑧 = −ℎ by (3.10). We find the balance of the Eulerian
flow in terms of the streamfunction by vertical integration of (3.3),

− 𝑓 Λ = 𝜏0𝑥 − 𝜏 𝑥 + ℱ − ℛ𝑥 (3.23)

now written for a nonzero Δ𝑝. The Reynolds stress term ℛ𝑥 (𝑧) is induced by standing
and transient eddies. The bottom formstress ℱ(𝑧) is defined in Fundamental Box ??.

The Deacon cell With a small billow stress 𝜏 𝑥 below the mixed layer and small
Reynolds stresses, the Eulerian streamfunction Λ is generally dominated by the north-
ward Ekman transport −𝜏0𝑥 /𝑓 in the surface layer and a deep geostrophic return flow
(associated with ℱ) in the valleys between the topography peaks along the particular lat-
itude, or a frictional return flow in a bottom boundary layer if the model ocean is flat and
ℱ = 0. We exemplify this behavior with a simple eddy-resolving numerical model with
a flat bottom (Viebahn and Eden 2009), described in Example Box ?? and in the panels of
Figure 3.10. The Eulerian streamfunction is shown in the upper left panel. In the interior,
Λ is vertically constant and thus has a zero meridional velocity, but with a meridionally
varying windstress (and 𝑓 ) there is vertical pumping. This overturning cell is named af-
ter G EORGE D EACON3 . It does not reflect the more or less adiabatic motion in the ocean
interior with a local transport of active and passive tracers which is predominately along
isopycnals (the zonal mean buoyancy field is displayed in the upper right panel of Figure
3.10). As we shall demonstrate, the Deacon cell is a construct of the zonal averaging on
𝑧-levels.

The isopycnal circulation We may view the same system from a different perspective.
Realizing that the above zonally averaged state cannot reflect the local overturning in
a stratified ocean, it seems meaningful to perform the zonal average going along a lon-
gitude 𝑥 on isopycnals 𝜌 = const. To capture the transient eddy component, we have
to consider time-dependent quantities. The meridional velocity, expressed in isopycnal
coordinates (𝑥, 𝑦, 𝜌) (see Fundamental Box ??), is written as 𝑣(𝑥, 𝑦, 𝜌, 𝑡). The transport be-
∫𝜌
tween two isopycnals 𝜌 = 𝜌1 and 𝜌 = 𝜌2 becomes 𝜌21 𝑣(𝑥, 𝑦, 𝜌, 𝑡)𝜎𝑑𝜌 where 𝜎 = −∂𝑧/∂𝜌
measures the thickness between isopycnals. More accurately, the thickness is 𝜂 = 𝜎Δ𝜌
with a constant Δ𝜌 = 𝜌2 − 𝜌1 . The relevant isopycnal streamfunction for this configuration
is defined by
3
G EORGE D EACON, * †
68 Vertical momentum transport and the meridional overturning

Example Box 13 : Idealized ACC-Atlantic basin model

We use an eddy-permitting primitive equation model consisting of a zonally reentrant channel (the ACC
part) which is connected to a northern ocean basin enclosed by land (the Atlantic part). The equations
are formulated in Cartesian coordinates. The domain of the idealized ACC model extends over 𝐿 = 2500
km in zonal and meridional direction with 20 km horizontal resolution and 20 vertical levels with 50 m
thickness. There is only buoyancy in the model, which is proportional to temperature. The circulation
in the model is driven by a sinusoidal eastward windstress over the ACC part and a restoring boundary
condition at the surface for buoyancy 𝑏 with a restoring time scale of 30 days and target buoyancy 𝑏∗ , as
given in the left panel of the following figure.

4.5 2500 0.5

4 b*/∆b
T1/τ1 2000
3.5
T /τ
3 2 1

T3/τ1 1500
2.5 y [km]
T4/τ1 0
2
1000
1.5

1 500
0.5

0 0 −0.5
0 500 1000 1500 2000 2500 500 1000 1500 2000 2500
y [km] x [km]

Left: Target buoyancy 𝑏∗ /Δ𝑏 for the restoring boundary condition and different wind forcing scenarios imposed
in the idealized ACC model. Right: Time-mean zonal velocity at the surface from the idealized ACC model for a
windstress amplitude 0.25 × 10−4 m2 s−2 (contour interval is 0.1ms−1 ). White areas denote land mass.
The model uses a 𝛽-plane. The parameters are: buoyancy change Δ𝑏 = 30×10−3 m s−2 across ACC, restor-
ing time scale 𝑇𝑏 = 30 days, vertical viscosity 𝐴𝑣 = 10−3 m2 s−1 , vertical diffusivity 𝐾𝑣 = 10−4 m2 s−1 ,
biharmonic viscosity 𝐴ℎ𝑏𝑖 = 1012 m4 s−1 , bottom friction parameter 𝑟 = 10−5 s−1 . We will discuss simula-
tions with a flat bottom and with an idealized ridge system.
Note that the zonal symmetry of the flat-bottom version of the model leads to an absence of standing
eddies (there is a small amount due to the continents at the transition from the channel to the basin ge-
ometry). In the ACC part the zonal flow is predominantly eastward with a strong jet slightly northward
of the wind stress maximum (at about 𝑦 = 800 km) and then monotonically decreasing in all directions.
Nevertheless, the zonal flow extends through the whole water depth.

∮ ∫ 𝜌0 ∫ 𝜌0 ∮
𝜓(𝑦, 𝜌, 𝑡) = 𝑣(𝑥, 𝑦, 𝜌′ , 𝑡)𝜎 𝑑𝜌′ 𝑑𝑥 = 𝑣(𝑥, 𝑦, 𝜌′ , 𝑡)𝜎 𝑑𝑥 𝑑𝜌′ (3.24)
𝜌 𝜌

measuring the zonal mean transport between an arbitrary isopycnal 𝜌 and a reference
isopycnal 𝜌0 . Note that the isopycnal streamfunction is time-dependent and a correlation
(between 𝑣 and 𝜎). The time average of 𝜓 may thus be split into a part resulting from the
time quantities and an eddy part (see Fundamental Box ??). This property contrasts the
isopycnal streamfunction from the Eulerian one.

correct following description of isopyc circ The temporally averaged isopycnal stream-
function from the idealized ACC model is displayed in Figure 3.11. It clearly captures
different physics than the Eulerian projection: the deep zonal mean flow is now along
isopycnals and watermass formation in the near surface layers is visible. The Ekman
transport is masked in a shallow clockwise cell in the upper mixed layer which embeds
3 The Circulation of the Southern Ocean 69

0 0 0.03

−100 −100
0.025
−200 −200

−300 −300 0.02

−400
z [m] −400
0.015
−500 −500

−600 −600 0.01

−700 −700
0.005
−800 −800

−900 −900 0
0 500 1000 1500 2000 2500 0 500 1000 1500 2000 2500
y [km] y [km]

0 8 0

−100 6 −100

−200 4 −200

−300 −300
2
−400 −400
z [m]
0
−500 −500
−2
−600 −600
−4
−700 −700

−800 −6 −800

−900 −8 −900
0 500 1000 1500 2000 2500 0 500 1000 1500 2000 2500
y [km] y [km]

Figure 3.10.: Time-zonal mean (upper left) Eulerian (Λ), (lower left) eddy (−𝐵) and (lower right)
residual streamfunction (𝜙) from the idealized ACC model for a windstress amplitude
0.25 × 10−4 m2 s−2 . Contour interval is 0.2 Sv in each case, and zero lines are thick. Also
shown is the time-zonal mean buoyancy distribution (upper right) with a contour interval of
0.001m/s2 and a thick 0.007m/s2 line.

0.03

0.025
4

0.02 2
Buoyancy [m/s2]

0.015 0

−2
0.01

−4

0.005

−6

0
−8
500 1000 1500 2000 2500
y[km]

Figure 3.11.: Isopycnal streamfunction of the experiment with the idealized ACC model shown in Figure
3.10. more text

besides the Ekman flow a more horizontal recirculation in a boundary layer flow occur-
ring at the western sides of the continents and islands blocking the zonal averaging path,
and having different density classes. These currents actually contribute by a standing
70 Vertical momentum transport and the meridional overturning

eddy part contained in the time mean of 𝜓: this becomes obvious by splitting the covari-
ance of 𝜎 and 𝑣 following the procedure explained in Fundamental Box ??. Any recircu-
lation on 𝑧-levels averages out in the Eulerian mean but not in the isopycnal framework
where isopycnal recirculations cancel. Note especially that the Deacon cell is absent in
the isopycnal circulation.

Figure 3.12.: Sketch showing idealized trajectories of a water particle in the ACC moving on density sur-
faces. The trajectories are shown in three dimensions and projected onto a constant longitude
plane [left]. Redrawn from ????

How does the Deacon cell come about in the zonal average on 𝑧-levels? A schematic
view of the isopycnal circulation is presented in Figure 3.12. On each single isopycnal
the motion is slightly descending and southward, balanced by upward and northward
movement upstream of Drake Passage. The resulting circulation, integrated at constant
latitude and depth for a particular density surface, is an overturning cell with a vertical
extent of only a few hundred meters. Deeper density surfaces show similar overturning
cells, with northward branches at the same depth as the southward branch of the cell
related to lighter water, so the zonally integrated cell including all density classes repre-
sents a meridional overturning penetrating to great depth, without a need for any water
particles to traverse the full depth range.

Interfacial formstress Evidently, 𝜎 ˆ𝑣 defines a zonal mean isopycnal velocity variable


(see Fundamental Box ??), where the hat denotes the zonal mean taken on an isopycnal.
The appropriately averaged zonal momentum balance becomes (the temporal tendency
is omitted for simplicity; the tilde notation has been omitted: all quantities in the follow-
ing paragraph are meant to be isopycnal)

( )
− 𝑓𝜎 ˆ𝑥 + 𝜏 𝑥 + ℱ̂ − ℛ̂𝑥
ˆ𝑣 = −𝜎𝑀 (3.27)
𝜌

Here, 𝑀 (𝑥, 𝑦, 𝜌, 𝑡) is the Montgomery potential (pressure in isopycnal coordinates), and


the quantities ℱ̂ and ℛ̂𝑥 are the isopycnal counterparts of the bottom formstress and the
appropriate Reynolds stress divergence (to be in congruence to the Eulerian discussion,
these quantities are integrated over the density variable). Imagine now a stack of well-
mixed isopycnal layers (an isopycnal model) and pick out one bounded by the interfaces
𝜌 = 𝜌1 at 𝑧 = −𝜂1 and 𝜌 = 𝜌2 at 𝑧 = −𝜂2 . We integrate (3.27) over this layer
3 The Circulation of the Southern Ocean 71

Fundamental Box 6 : Equations of motion in isopycnal coordinates

The transformation from Cartesian coordinates (x, 𝑧) to the isopycnal coordinates (x, 𝜌) is defined by the
depth 𝑧 = 𝑧(x, 𝜌) of an isopycnal 𝜌(x, 𝑧) = const (𝜌 is the potential density). The relations between the
gradients (and other derivatives) of a field 𝜒(x, 𝑧) = 𝜒(x,
˜ 𝜌) = 𝜒(x, 𝑧(x, 𝜌), taken either on a depth level
or on an isopycnals, are

(∇𝜌 )𝜒 = (∇𝑧 )𝜒 + 𝜒𝑧 (∇𝜌 )𝑧 and 𝜒𝑧 = 𝜒


˜ 𝜌 𝜌𝑧 (3.25)

where (∇𝜌 ) is the gradient taken on isopycnals etc. For the pressure one finds (∇𝑧 )𝑝 = (∇𝜌 )(˜
𝑝 + 𝑔𝜌𝑧) =
(∇𝜌 )𝑀 , using 𝑝𝑧 = −𝑔𝜌. The field 𝑀 = 𝑝˜ + 𝑔𝜌𝑧 is called Montgomery potential. The equation of motion
−𝑓 𝑣 = −𝑝𝑥 + 𝜏𝑧𝑥 (used in this form in this section) become in isopycnal form

𝑣 = −𝜎𝑀𝑥 − 𝜏˜𝜌𝑥
− 𝑓 𝜎˜ (3.26)

where 𝑀𝑥 = 𝑀 (𝑥, 𝑦, 𝜌)𝑥 is the gradient on isopycnals and 𝜎 = −∂𝑧/∂𝜌 = −𝑧𝜌 . Note that 𝑀𝜌 = 𝑔𝑧.

∫ 𝜌1 ∫ 𝜌1 ( ) 𝜌 1
−𝑓 ˆ𝑣𝑑𝜌 = −
𝜎 ˆ𝑥 𝑑𝜌 + 𝜏 𝑥 + ℱ̂ − ℛ̂𝑥
𝜎𝑀 (3.28)
𝜌2 𝜌2 𝜌2

The pressure-thickness correlation can be cast in a simpler form: 𝜎𝑀𝑥 = −(𝑧𝑀𝑥 )𝜌 +


𝑧(𝑀𝑥 )𝜌 = −(𝑧𝑀𝑥 )𝜌 + 𝑔𝑧𝑧𝑥 , using 𝜎 = −𝑧𝜌 and 𝑀𝜌 = 𝑔𝑧. The zonal average eliminates
the last contribution 𝑔𝑧𝑧𝑥 = 𝑔(𝑧 2 /2)𝑥 , and finally the integration over the density range
yields

∫ 𝜌1
𝜎𝑀 2 𝑝2𝑥 − 𝜂ˆ (3.29)
ˆ𝑥 𝑑𝜌 = 𝜂ˆ 1 𝑝1𝑥
𝜌2

which may be interpreted as a vertical divergence of a stress 𝜂ˆ 𝑝𝑥 fluxing zonal momen-


tum through the layer interfaces: the layer of ocean gains 𝑥-momentum by the amount
1 𝑝1𝑥 from the fluid above 𝑧 = −𝜂1 (𝑥) and loses 𝜂ˆ
𝜂ˆ 2 𝑝2𝑥 to the fluid below 𝑧 = −𝜂2 (𝑥). Thus
for infinitesimally distant isopycnals the vertical divergence of the interfacial formstress
𝜂ˆ𝑝𝑥 enters the momentum balance. The similarity of the construction of 𝜂ˆ 𝑝𝑥 to the bottom
formstress is evident (see Fundamental Box ??), but here the isopycnal layer depth 𝜂 and
the pressure 𝑝 are still time-dependent. Evidently, only deviations of 𝜂 and 𝑝 from their
zonal mean 𝜂ˆ and 𝑝ˆ contribute to the correlation 𝜂ˆ𝑝𝑥 = −ˆ 𝜂𝑥 𝑝, revealing that the isopycnal
must be inclined for the stress to be nonzero. Though frequently viewed as a vertical
transport – because isopycnal layers are generally stacked vertically – the process is ac-
tually a zonal (horizontal) transport of zonal momentum across the inclined isopycnals.
To become a vertical transport, the imparted momentum must be redistributed within
the layer by small-scale billow turbulence: the layer must be effectively mixed which is
facilitated by the density being approximately constant with the layer.
The above described stress is called interfacial formstress (IFS), more precisely after
time averaging by which a standing and a transient component result. Let us denote the
72 Vertical momentum transport and the meridional overturning

z
Pressure on isopycnal

High Low

z= -d(x)
Isopycnal surface
x

Figure 3.13.: Schematic sketch demonstrating the interfacial formstress for an isopycnal interface in the
water (shown is the zonal depth section). There is higher pressure at the depth of the density
surface where it is rising to the east compared with where it is deepening to the east. This
results in an eastward pressure force (interfacial form stress) on the water below. This is
related to the fact that the northward flow occurs where the vertical thickness of water above
the density surface is small, and southward flow occurs where the thickness is large; so there
is a net southward mass flux at lighter densities due to the geostrophic flow. The same kind
of pressure force acting on the sloping bottom topography leads to the bottom form stress.
Redrawn from Rintoul et al. (2001).

latter by ⟨𝜂 ∗ 𝑝∗𝑥 ⟩ where the star names the transient eddy part and the bracket time plus
zonal mean on isopycnals. Equating the zonal pressure gradient with the northward
geostrophic velocity, 𝑓 𝑣𝑔∗ = 𝑝∗𝑥 , and the layer depth fluctuation with (potential) density
anomaly, 𝜂 ∗ = 𝜌∗ /¯ 𝜌𝑧 , we find that the interfacial formstress relates to the meridional
eddy density flux, − ⟨𝜂 ∗ 𝑝∗𝑥 ⟩ = −(𝑓 /¯ 𝜌𝑧 ) 𝑣𝑔∗ 𝜌∗ . Ignoring salinity contributions in these
〈 〉

relations, the density can be replaced by temperature, and thus a poleward eddy flux of
heat is just a downward transport of zonal momentum by IFS in the water column. These
processes are strictly coupled. The transient eddies which carry the poleward heat flux
shown in Figure 3.2 thus establish a downward transport of momentum.

In summary, we note that though horizontal pressure gradients can only establish a
transfer of horizontal momentum in horizontal direction they transport horizontal mo-
mentum across tilted surfaces from one piece of ocean to another. A layer bounded by
tilted isopycnals is thus forced by stresses (IFS) at the bounding top and bottom surfaces,
in the same way as the Ekman layer is driven by frictional stresses at top and bottom. Ob-
viously, to obtain a non-zero IFS, the pressure must vary at the isopycnal depth in a way
that an out-of-phase part with respect to the depth variations is present, as elucidated in
the sketch of Figure 3.13. Deriving the relation (3.29), it was assumed that the isopycnal
strip does not run into the bottom nor touches the sea surface. If this situation occurs,
additional pressure terms arise from the bounding outcrops. These terms present a flux
of horizontal momentum through these boundaries into the strip. For the bottom contact
the corresponding flux is part of the bottom formstress.

The residual circulation The Eulerian view of the overturning can be extended to a
closer correspondence with the isopycnal framework, using the TEM approach (Trans-
3 The Circulation of the Southern Ocean 73

−100 −100 −100 4

−200 −200 −200 3

2
−300 −300 −300
1
−400 −400 −400
z [m]
0
−500 −500 −500
−1
−600 −600 −600
−2

−700 −700 −700 −3

−800 −800 −800 −4


0 200 400 600 800 1000 0 200 400 600 800 1000 0 200 400 600 800 1000
y [km] y [km] y [km]

−100 −100 −100 0.02

−200 −200 −200

0.015
−300 −300 −300

−400 −400 −400


z [m]
0.01
−500 −500 −500

−600 −600 −600


0.005

−700 −700 −700

−800 −800 −800 0


0 200 400 600 800 1000 0 200 400 600 800 1000 0 200 400 600 800 1000
y [km] y [km] y [km]

Figure 3.14.: Time-zonal mean (upper row) residual streamfunctions (𝜙) and (lower row) mean buoyancy
(¯𝑏 = −𝑔 𝜌¯) distributions from the idealized ACC model for amplitudes (from left to right)
(0.5, 0.75, 1.0) × 10−4 m2 s−2 of the sinusoidal wind stress. The contour intervals are (up-
per) 0.2Sv and (lower) 0.001 m/s2 . Thick lines are (upper row) zero lines and (lower row)
0.007 m/s2 lines.

formed Eulerian Mean, Andrews and McIntyre, 1976; see section ??). It acknowledges
that the transport (advection) of zonally averaged tracers is not only performed by the
time-zonal mean Eulerian streamfunction Λ. Eddies advect density and tracers as well.
This fact can be inferred from the projection of the eddy density flux u′ 𝜌′ into the compo-
nent normal and along the mean isopycnal,

u′ 𝜌′ = 𝐵 ∇𝜌¯ − 𝜅∇¯
𝜌 (3.30)
¬

with u = (𝑣, 𝑤) and ∇ = (∂/∂𝑦, ∂/∂𝑧). As before, the bar is the time and zonal mean
on 𝑧-levels. Forming the divergence of this expression reveals that 𝜅 is an eddy-induced
diapycnal diffusivity (if positive), and −𝐵 is an eddy streamfunction advecting the density
𝜌¯(𝑦, 𝑧) in addition to the Eulerian flow. The advection of 𝜌¯ is thus achieved by the residual
streamfunction 𝜙 = Λ − 𝐵, and the diapycnal transport is represented by the diffusivity
𝜅. This decomposition is shown in Figure 3.10 (Eulerian streamfunction upper left panel,
eddy and residual streamfunctions lower panels). Figure 3.14 shows the residual stream-
function and the buoyancy field of the same model, now restricted to the ACC part, for
three amplitudes of the windstress.
confusion of notation 𝜅, 𝐾 in TURB
The residual streamfunction is governed by

− 𝑓 𝜙 = 𝑓 𝐵 + 𝜏0𝑥 − 𝜏 𝑥 + ℱ − ℛ𝑥 = −𝑓 𝑣 ′ 𝜌′ /¯
𝜌𝑧 + 𝜏0𝑥 − 𝜏 𝑥 + ℱ − ℛ𝑥 (3.31)

𝜌∣2 . In
found by use of (3.23). The eddy term 𝑓 𝐵, computed from (3.30), is 𝑓 u′ 𝜌′ ⋅ ∇𝜌¯/∣∇¯
¬
the second form of the balance the 𝐵-term follows from the approximation (??) assuming
74 The dynamical balance of the zonal flow

𝜌¯𝑧 ≫ 𝜌¯𝑦 , discussed in section ?? on TEM. It shows that the eddies accomplish a vertical
transport of zonal momentum by the flux 𝑓 𝑣 ′ 𝜌′ /¯ 𝜌𝑧 . We find here resemblance to the
interfacial formstress terms (3.29). We shall refer to the Eulerian form of this flux also
as IFS. Comparing (3.31) with isopycnal counterpart (3.28), we note that the residual
streamfunction is constructed to resemble the isopycnal one.

3.4. The dynamical balance of the zonal flow

The time-zonal average of the zonal momentum balance is central in the dynamics of the
ACC because the main external forcing – the windstress – is zonal and the flow is more
or less zonal as well. The balance has been written in various forms in the preceding
sections, namely in the Eulerian form (3.23), in the isopycnal form (3.27) (not yet time-
averaged) and in the residual form (3.31). The IFS and BFS contributions to the physics of
zonal currents can be embraced by a simple conceptual model. Consider a strip of ocean
from Antarctica to the northern rim of the ACC and split the water column into three
layers (which may be stratified), separated by isopycnals (see Figure 3.15 and the sketch
in Figure 3.4). The upper layer reaches from the sea surface 𝑧 = 𝜁 to some isopycnal at
depth 𝑧 = −𝜂1 and includes the Ekman layer. The intermediate layer lies below with
base at 𝑧 = −𝜂2 which is above the highest topography in the Drake Passage belt (the
range of latitudes which run through Drake Passage). These two layers are ’unblocked’
by topography. The lower layer reaches from 𝑧 = −𝜂2 to the ocean bottom at 𝑧 = −ℎ and
is thus ’blocked’ beneath the depth of the highest topography. We apply a time-isopycnal
average to the balance equation (3.27) of zonal momentum and integrate over the three
layers to obtain

− 𝑓 𝑉1 = − ⟨𝜂1∗ 𝑝∗1𝑥 ⟩ + 𝜏0𝑥 − 𝜏1𝑥 − ℛ𝑥1


−𝑓 𝑉2 = ⟨𝜂1∗ 𝑝∗1𝑥 ⟩ − ⟨𝜂2∗ 𝑝∗2𝑥 ⟩ + 𝜏1𝑥 − 𝜏2𝑥 − ℛ𝑥2 (3.32)
−𝑓 𝑉3 = ⟨𝜂2∗ 𝑝∗2𝑥 ⟩ − ℎ𝑝𝑏𝑥 + 𝜏2𝑥 − 𝜏𝑏𝑥 − ℛ𝑥3

The depth and zonally integrated northward volume flux in each layer is denoted by
𝑉𝑖 , 𝑖 = 1, 2, 3 (layer 3 = 𝑏 is the bottom layer). Furthermore, 𝑝𝑖 is the pressure at the
respective layer depths, 𝑝𝑏 the bottom pressure, and the brackets denote time and zonal
mean. An ACC path-following mean should better be considered to abandon standing
eddies. In this case the coordinate 𝑥 is along the specific path. As before, the star denotes
the deviation from this average, 𝜏0𝑥 is the windstress, 𝜏𝑖𝑥 the frictional stresses at inter-
faces, 𝜏𝑏𝑥 the frictional bottom stress, and ℛ𝑥𝑖 the integrated divergence of the appropriate
lateral Reynolds stress. Note that the surface term 𝜁𝑝0𝑥 drops out in the first equation
because the surface pressure is 𝑝0 = 𝑔𝜁.

Dividing the above momentum balances by 𝑓 , we recognize the components of the


meridional overturning circulation: eddy-driven and windstress-driven parts as well as
the geostrophic contribution associated with the bottom formstress. The contributions
3 The Circulation of the Southern Ocean 75

Figure 3.15.: Zonal section of the observed potential density at 60∘ S, taken from the gridded data of the
Hydrographic Atlas of the Southern Ocean (Olbers et al. 1992). The section is viewed from
the south.

from interfacial friction and Reynolds stresses are small and will thus be mostly aban-
doned in the following. The wind-driven component −𝜏0 /𝑓 (the Ekman transport) in the
top layer and the geostrophic component in the bottom layer, ℎ𝑝𝑏𝑥 /𝑓 , also appear if the
flow is averaged between geopotential levels (constant depth). As discussed above, these
Eulerian quantities form the Deacon cell of the Southern Ocean meridional overturning
(see section 3.3).


Balance of total momentum Since 𝑖 𝑉𝑖 = 0 at any latitude by mass balance (ne-
glecting the very small effect of precipitation and evaporation on the mass balance), the
overall balance of zonal momentum is between the applied windstress, the bottom form-
stress (BFS), the frictional stress on the bottom, and the vertical integral of the Reynolds
stress contributions,


𝜏0𝑥 − 𝜏𝑏𝑥 − ℎ𝑝𝑏𝑥 − ℛ𝑥𝑖 = 0 (3.33)
𝑖

The frictional bottom stress is generally small in the ocean. Lateral eddy momentum
fluxes in the ACC turned out to be rather small (compared to the windstress) as well and
indifferent in sign (e.g. Morrow et al. 1994, Phillips and Rintoul 2000, Hughes and Ash
2001). Hence the momentum put into the ACC by windstress is transferred at the same
latitude to the solid Earth by bottom formstress. This balance has been confirmed in most
numerical models which include submarine topographic barriers in the zonal flow and
have a realistic (small) magnitude of the Reynolds stress divergence. Eddy effects are
thus rather unimportant in the vertically integrated balance. It is worth mentioning that
some coarse OGCMs do not confirm (3.76), see e.g. Cai and Baines (1996). The reason
is that such models use very large lateral viscosities so that the parameterized Reynolds
stresses become large, even though the simulated current is broad and smooth. In models
with a flat ocean bottom either bottom friction may get importance and/or the neglected
Reynolds terms could come into play.
If the BARBI model of the Southern Ocean, introduced in Example Box ?? and more
generally in section 4.1, is considered with topography and stratification in a simulation
76 The dynamical balance of the zonal flow

Figure 3.16.: Steady solution obtained with the baroclinic BARBI model (NL) forced by NCEP windstress.
Left: streamfunction, CI = 10 Sv. Middle: baroclinic potential energy, CI = 1000 m3 s−2 .
Right: bottom pressure, units: CI = 0.1 m2 s−2 .

NL latitude −59.5
2
1
NL
1 0

−1
0 latitude −46.5
2
−1 1
0
−2
−70 −60 −50 −40 −30 −1
0 100 200 300
latitude longitude

Figure 3.17.: Left: Zonally integrated momentum balance as function of latitude for the baroclinic case
NL. Drake Passage is indicated by purple lines. Windstress: yellow, bottom formstress:
black, friction: green, pressure difference on continents: red. Units: 10−4 m2 s−2 . Right:
Bottom pressure – depth correlation as a function of longitude for two sets of latitudes. The
upper panel is for a latitude 59.5∘ S in Drake Passage, the lower one is for a more northern
latitude at 46.5∘ S. Bottom pressure: upper curves [blue], ocean depth: lower curves [red].
Scaled (depth is shown as −ℎ/𝑚𝑎𝑥(ℎ), pressure is shown by 𝑃/𝑚𝑎𝑥(∣𝑃 ∣) + 1).

NL, displayed in Figure 3.16. It has a relative small effect of friction (strictly speaking:
lateral momentum and vorticity diffusion; this property is discussed further below). The
streamfunction, baroclinic potential energy and bottom pressure fields of the case NL
contrast the previous homogeneous simulation with a considerable amount of realism.
The zonal transport is now 140 Sv, and a structured sized circumpolar flow appears with
little apparent influence of the underlying topography. Note that the potential energy
follows quite closely the streamlines while the bottom pressure has a tendency to follow
the 𝑓 /ℎ contours, as was found for the barotropic case BT (see Figure 3.7). Nevertheless,
the baroclinicity introduced by the baroclinic potential energy 𝐸 — in the momentum
balance as a pressure gradient and in the vorticity as the JEBAR term (see section 4.1)
— breaks the constraint of free flow along 𝑓 /ℎ contours which shaped the barotropic
simulation in the case BT. The zonally averaged balance of zonal momentum shown in
3 The Circulation of the Southern Ocean 77

Figure 3.17 (left panel) reveals the windstress-formstress balance in the latitude band of
Drake Passage with very little contribution from friction. The figure presents the total
meridional extent of the model, and the pressure force acting on the continents north
and south of Drake Passage latitudes comes into play. This term is disregarded in the
conceptual model (3.32) and therefore not present in (3.76) which must be replaced by


𝜏0𝑥 − 𝜏𝑏𝑥 − ℎ𝑝𝑏𝑥 − Δ𝐸 𝑐𝑜𝑛𝑡𝑖𝑛𝑒𝑛𝑡𝑠 − ℛ𝑥𝑖 = 0

(3.34)
𝑖

if continents are blocking the averaging path in addition to submarine topography. Wind-
stress and the two pressure terms are the principle contributors to the balance of momen-
tum in the basin-wide gyre circulations embedded between continents.

Finally, the right panel of Figure 3.17 shows the pressure-depth correlation of case NL
in the same manner as Figure 3.8 for FL and BT. As in FL we find high pressure upstream
of ridges and low pressure downstream in the Drake Passage band, leading to a sink of
eastward momentum.

−4 −3
x 10 x 10
2
1

0 0

−1

−1
−2
0 500 1000 1500 2000 2500 0 500 1000 1500 2000 2500

Figure 3.18.: Upper panels: Sea surface height (in m) and streamfunction of the vertically integrated trans-
port (in Sv) of the the idealized ACC model (see Example Box ??). Lower panels: zonal mean
balance of total momentum, indicating in the lower left panel the dominant balance between
windstress [yellow] and negative bottom formstress [black] at each latitude. The other terms
are: bottom friction [green], pressure difference on continents [red], vertically integrated
Reynolds stress [blue], mean advection [magenta]. Lower right: baroclinic formstress [red],
barotropic formstress [green], total formstress [black] (units Nm−2 ). units correct?
thick lines!
78 The dynamical balance of the zonal flow

The role of barotropic and baroclinic bottom formstress Figure 3.18 exemplifies the to-
tal zonal momentum balance with results from the eddy-resolving idealized ACC model,
here integrated with a submarine topography: a meridional ridge in the ’Drake Passage’
and another one stretching through the middle of the attached basin and the southern
channel. It is instructive to write the pressure 𝑝 as sum of the baroclinic (density-related)
∫0
part 𝑝𝑐𝑙𝑖𝑛 = 𝑔 𝑧 𝜌𝑑𝑧 and the barotropic (surface-related) part 𝑔𝜁. The right-hand panel
of Figure 3.18 displays the balance with the pressure terms ℎ(𝑝𝑐𝑙𝑖𝑛
𝑏 )𝑥 and 𝑔ℎ𝜁𝑥 separated:
individually they are much larger than the zonal windstress by about an order of mag-
nitude but of opposite sign, and thus they nearly cancel. While the total bottom form-
stress ℎ𝑝𝑏𝑥 clearly takes out the momentum put in the ocean by windstress, we note that
the baroclinic part ℎ(𝑝𝑐𝑙𝑖𝑛
𝑏 )𝑥 does not have the corresponding sign: the baroclinic bottom
formstress accelerates the eastward current (the center is roughly 800 km from the south-
ern boundary). A summary of the balance of zonal momentum in the ACC is displayed
in the sketch of Figure 3.19.

Figure 3.19.: Sketch of the zonal balance of momentum for the ACC. The system is viewed from Antarc-
tica. The flow establishes a high of barotropic (surface) pressure and a low of baroclinic
(hydrostatic) pressure upstream of a zonal ridge and a corresponding low/high downstream.
The associated barotropic and baroclinic bottom formstresses almost balance; their residual
counteracts the windstress. The wind drives a northward Ekman transport in the surface

layer ( ), and corresponding to the bottom formstress there is southward geostrophic return

flow in the valleys between the ridges which partly block the zonal path ( ).

A depth-pressure correlation can in fact be seen in circumpolar hydrographic sections


passing through Drake Passage around Antarctica, e.g. as shown in Figure 3.15. From
∫0
the density 𝜌 we can infer the baroclinic bottom pressure 𝑝𝑐𝑙𝑖𝑛 = 𝑔 𝑧 𝜌𝑑𝑧 contained in the
mass stratification. It is obvious in the above section that there is more lighter water to
the west of the submarine ridges than to the east. Surprisingly, the BFS derived from such
a pattern accelerates the eastward current, acting thus in cooperation with the eastward
windstress — a feature of the ACC dynamics which will be reconsidered below.

Adiabatic regime If the flow conserves potential density, i.e. if there is no diapycnal
mixing or surface forcing anywhere, then there cannot be transport across isopycnals
and, by mass conservation, the meridional transport in each layer must vanish, 𝑉𝑖 = 0.
Neglecting all friction terms, we find that the interfacial formstress 𝜂𝑖∗ 𝑝∗𝑖𝑥 is vertically
3 The Circulation of the Southern Ocean 79

constant and equal to the windstress 𝜏0𝑥 and to the bottom formstress (in models with a
flat bottom we must replace the bottom formstress by the frictional bottom stress),

⟨𝜂𝑖∗ 𝑝∗𝑖𝑥 ⟩ ≃ 𝜏0𝑥 ≃ ℎ𝑝𝑏𝑥 (3.35)

Then, in each layer, the meridional mass fluxes induced by windstress and pressure gra-
dients are compensated. We will proceed with this scenario of ’constant vertical momen-
tum flux’ in section ??. In passing we note that many layer models, even in an eddy-
resolving state, approximately satisfy (3.35) if the Reynolds stress divergence is small.
This applies particularly to the quasi-geostrophic models where, with eddy resolution,
Reynolds stresses are found small and even upgradient.

Diabatic regime The real ocean is diabatic, i.e. there is mixing across isopycnals by
small-scale turbulence and air-sea fluxes, but it is still in debate if it occurs predomi-
nantly between the outcropping isopycnals in the surface layer or in the interior as well
(see section ??). The meridional overturning transports 𝑉𝑖 at a certain latitude circle can
be non-zero only if there is exchange of mass between the layers south of the respective
latitude — implying conversion of watermasses south of the ACC. In fact, by mass con-
servation, the 𝑉𝑖 equal the net exchange with the neighboring layers over the area south
of the respective latitude. At the same time, referring to (3.32), the overturning trans-
ports imply a Coriolis force in the individual isopycnal layers which is in balance with
the vertical divergence of the interfacial formstress. The divergence of the heat flux due
to transient eddies can clearly be deduced from Figure 3.2 (we have shown that roughly
− ⟨𝜂 ∗ 𝑝∗𝑥 ⟩ ≈ −𝑓 𝑣𝑔∗ 𝜃∗ /𝜃¯𝑧 ). Eddy effects at the respective latitude and diabatic interior ef-
〈 〉

fects of small-scale turbulence occurring to the south must thus adjust according to mass
and momentum requirements of the zonal current and the meridional overturning. The
isopycnal analysis of the zonal momentum balance in eddy-resolving models, as e.g. the
isopycnal analysis of the idealized ACC model, exemplify the importance of diabatic pro-
cesses and the inapplicability of the adiabatic regime: there is a net meridional circulation
at all depths in balance with a divergent IFS and the windstress.

3.5. The adiabatic regime

The ACC is the outstanding example in the ocean circulation for diapycnal transport
of momentum by eddies in the form of IFS. At intermediate depths where the bottom
formstress does not yet work and billow turbulence is marginal the residual circulation
– the meridional overturning – is driven by windstress and the divergence of IFS, which
is immediately apparent from (3.31) or (3.32). In the adiabatic regime these processes
are in exact balance, and the residual circulation vanishes, i.e. 𝑣¯ = 0, 𝜙 ≡ 0. Of course,
this is a highly unrealistic situation, and we weaken the condition to the assumption that
−𝑓 𝑣¯ and −𝑓 𝜙 are small in the respective balances (3.31) or (3.32). We proceed with the
level-averaged form here and neglect the Reynolds stress term so that
80 The adiabatic regime

𝜌𝑧 + 𝜏0𝑥 − 𝜏 𝑥 + ℱ = 0
− 𝑓 𝑣 ′ 𝜌′ /¯ (3.36)

The zonal current is geostrophic, i.e. 𝑓 𝑢 ¯𝐺 = −¯


¯ = 𝑓𝑢 𝑝𝑦 . Some simple models of the ACC
and the overturning follow from these equations. Note that the vanishing of the interior
meridional overturning circulation in such models may rightly be questioned (see next
section).

Johnson-Bryden model For the intermediate layer, defined in the previous section,
neglecting the billow-term 𝜏 𝑥 and omitting the bottom formstress term ℱ, we get the
Johnson-Bryden relation (Johnson and and Bryden 1989),

𝑓 𝑣 ′ 𝜌′
= 𝜏0𝑥 (3.37)
𝜌¯𝑧

The standing eddy component will be neglected (which is a severe assumption because
it exceeds the transient component in realistic conditions) or, if the mean is interpreted
as average along the current path, 𝜏0𝑥 is not the zonal windstress but rather the path fol-
lowing component (we use dashes for the transient eddy component in the following
arguments). According to the formula the ’northward’ eddy density flux 𝑣 ′ 𝜌′ in the cir-
cumpolar belt of the ACC, suitably normalized, equals the ’zonal’ windstress 𝜏0𝑥 . In a first
step Johnson and Bryden parameterize the transient lateral eddy flux by a down-gradient
form,

𝑣 ′ 𝜌′ = −𝐾 𝜌¯𝑦 (3.38)

and find that the windstress and the eddy diffusivity constrain the slope of the isopy-
cnals, 𝑠 = −¯ 𝜌𝑧 = 𝜏0𝑥 /(𝑓 𝐾). Such a relation is roughly consistent with the ob-
𝜌𝑦 /¯
served slopes in the ACC belt if the eddy diffusivity is of order 𝐾 ∼ 103 m2 s−1 (take
𝑠 = 10−3 , 𝜏0𝑥 = 10−4 m2 s−2 ). Replacing the lateral density gradient using the thermal
wind relation, 𝑓 𝑢
¯𝐺𝑧 = 𝑔 𝜌¯𝑦 , we obtain

𝑓2
𝐾 ¯𝐺𝑧 = 𝜏0𝑥
𝑢 (3.39)
𝑁2

where the vertical density gradient is replaced by the squared Brunt-Väisälä frequency
𝑁 2 = −𝑔 𝜌¯𝑧 . Apparently, 𝐾(𝑓 /𝑁 )2 defines an equivalent diffusivity for the vertical
momentum transfer which is achieved by lateral density diffusion (see e.g. Rhines and
Young 1982, Olbers et al. 1985). We notice here the same equivalence between vertical
momentum transfer and horizontal heat transfer by eddies as in section 3.3.
3 The Circulation of the Southern Ocean 81

In a second step Johnson and Bryden used Green’s form (Green 1970, Stone 1972)
of the diffusivity 𝐾 = 𝛼ℓ2 /𝑇 . It is obtained for a baroclinically unstable zonal current,
where ℓ is a measure of the eddy transfer scale, and 𝑇 = 𝑁/∣𝑓 𝑢 ¯𝐺𝑧 ∣ is the Eady time of
growth of the unstable eddies (Eady 1949, see also section ??). The constant 𝛼 measures
the level of correlation between 𝑣 ′ and 𝜌′ in the density flux (𝛼 = 0.015 ± 0.005 according
to Visbeck et al. 1997). Johnson and Bryden’s result for the ACC transport is obtained by
relating the turbulence scale ℓ to the baroclinic Rossby radius 𝜆 = 𝑁 ℎ/(∣𝑓 ∣𝜋). For ℓ = 𝜋 2 𝜆
we get their estimate of the shear

)1/2
𝜏0𝑥
(
𝑁 (𝑧)
𝑢
¯𝐺𝑧 = 2 2
⋅ (3.40)
𝜋 𝛼ℎ ∣𝑓 ∣

The shear and thus also the transport is proportional to the square root of the wind stress.
Note that this result is due to the particular parameterization of the diffusivity 𝐾. With
some reasonable values of parameters a transport (relative to the bottom) of about hun-
dred Sv relative to the bottom can be evaluated. A vertically constant 𝐾 yields a linear
relation between wind and shear, and with other assumptions for ℓ or 𝑇 other power
laws can be obtained.
The Johnson-Bryden model has been a much discussed as a theory of the ACC trans-
port. Attempts to verify the square-root relation with numerical models are plentiful
(e.g. Gnanadesikan and Hallberg 2000 with coarse-resolution models with simple geom-
etry, Gent et al. 2001 for coarse-resolution global models, Tansley and Marshall 2001 for
two-layer channel models) but generally without success. This is not surprising in view
of the many assumptions put together in this model. First there is the assumption of the
adiabatic state of the flow (zero diapycnal mixing) which is violated in the real ocean but
also in numerical models operating on 𝑧-levels (isopycnal models may be adjusted close
to an adiabatic state). Secondly, it is not clear whether the Eady model and other details
of baroclinic instability theory are appropriate in the ACC eddy field. Certainly some of
these features are generated by baroclinic instability, but they are not in the initial growth
state but rather in some state of equilibration. Furthermore, the above parameterizations
are not always implemented in coarse OGCMs.

Extensions The Johnson-Bryden model (3.39) is dynamically incomplete as it does not


satisfy a closed momentum balance. Integrating (3.36) from top to bottom (with ℱ = 0),
we note that the balance 𝜏0𝑥 = 𝜏𝑏𝑥 between wind stress and frictional bottom stress must
be valid, as for the homogeneous model in section 3.1. The model can straightforwardly
be extended to include a frictional bottom boundary layer, along the way outlined in
section 3.1. Integrating (3.39), the current profile

0
𝑁 2 𝜏0𝑥 ′ 𝑁 2 𝜏0𝑥

𝑔 𝑔¯
¯𝐺 (𝑧) = − 𝜁¯𝑦 −
𝑢 𝑑𝑧 = − 𝜁𝑦 + 𝑧 (3.41)
𝑓 𝑧 𝑓 2𝐾 𝑓 𝑓 2𝐾

The second relation is found, taking for simplicity all coefficients constant: we obtain a
linear profile (the model is easily extended to any 𝑁 (𝑧) profile). The velocity at the top of
82 The diabatic regime

the bottom boundary layer at 𝑧 = −ℎ + 𝑑𝑏 is used to evaluate the frictional bottom stress
as

𝑁 2𝜏 𝑥
( )
1 1 𝑔
𝜏𝑏𝑥 = 𝑑𝑏 ∣𝑓 ∣¯
𝑢𝐺 (−ℎ + 𝑑𝑏) = − 𝑑𝑏 𝑓 − 𝜁¯𝑦 − (ℎ − 𝑑𝑏 ) 2 0 (3.42)
2 2 𝑓 𝑓 𝐾

and equating this result with 𝜏0𝑥 , the associated gradient of the surface displacement fol-
lows as

2𝜏 𝑥 𝑁2
( )
1
𝜁¯𝑦 = 0 1 − (ℎ − 𝑑𝑏 )𝑑𝑏 (3.43)
𝑔𝑑𝑏 2 𝑓𝐾

The first term is equal to the slope resulting from the homogeneous model in section 3.2.
A reasonable size of 𝜁¯𝑦 ∼ 0.7 × 10−6 is found for our standard parameters (see Table 3.1).
Note that the surface slope relates to the isopycnal slope 𝑠 = 𝜏0𝑥 /(𝑓 𝐾) of the adiabatic
regime roughly by 𝜁¯𝑦 /𝑠 ∼ −ℎ𝑁 2 /𝑔 ∼ −6 × 10−4 . The transport in this balanced model
becomes

0
1 (ℎ𝑁/𝑓 )2
∫ ( )
𝑦 2ℎ 𝑦
𝐿 ¯𝐺 𝑑𝑧 = 𝐿 −
𝑢 + 𝜏0𝑥 (3.44)
−ℎ 𝑓 𝑑𝑏 2 𝐾

evaluated as 145 Sv. The first term is identical to transport in the homogeneous model
(see (3.14) and yields 65 Sv; the second contribution results from eddy processes and
yields 80 Sv. In addition to the geostrophic transport there is also a frictional transport
arising from the bottom stress which is identical to the one in (3.15) of the homogeneous
model and thus negligible. Note that the diffusivity 𝐾 appears as 𝐾/𝜆2 in the above
expression where 𝜆 = ℎ𝑁/𝑓 is the internal Rossby radius. We comment on this property
below in section 3.8.
A further extension of the Johnson-Bryden model is to include bottom topography
and the associated bottom formstress ℱ. We have previously argued that in this case, the
frictional bottom stress may be abandoned, and the momentum balance takes the form
(3.21). In lack of a physical parameterization of the stress ℱ we write ℱ(𝑧) = ℱ[¯𝑢𝐺 , 𝑧]
for the stress at the depth 𝑧, accounting for a functional dependence of the formstress
on the current field. Using the geostrophic relation (3.41) the momentum balance (3.21)
results in a condition 𝜏0𝑥 + ℱ[−(𝑔/𝑓 )𝜁¯𝑦 + 𝑢
¯𝑐𝑙𝑖𝑛
𝐺 , 𝑧 = −ℎ] = 0, determining the gradient
of the surface displacement, as before in the frictional model. In general, however, the
functional dependence of the bottom formstress is unknown. Simple examples will be
discussed in section 3.9.

3.6. The diabatic regime

Much of the recent perception of the ACC circulation is centered on the meridional over-
turning and ventilation of water masses in the Southern Ocean. The classical view (Sver-
drup et al. 1942) of water mass storage and spreading is updated in Figure 3.4 where the
3 The Circulation of the Southern Ocean 83

role of eddies in the unblocked part of the water column is highlighted. We have pointed
out at various places in the previous sections that eddies and turbulent mixing might
accomplish a major task in shaping and balancing the overturning circulation, and it re-
mains to put up some simple models to demonstrate how it might work. We refer to the
model of Marshall and Radko (2003) and of Olbers and Visbeck (2005). Both models are
zonally averaged and neglect the standing eddy component.
We assume that all mixing and watermass formation processes take place in an upper
layer of the ocean – basically a turbulent layer where the Ekman transport and pumping
is established by the wind, and buoyancy is imprinted on the surface waters by heat and
freshwater exchange with the overlying atmosphere, while the ocean interior is void of
billow turbulence, but eddies are present that transport and mix substances along isopy-
cnals. We see this scenario as an extreme. The real ocean might have substantial mixing
by turbulence in the interior as well (see e.g. Heywood et al. 2002, Naveira Garabato et
al. 2004). Eddies might contribute by a diapycnal flux to watermass formation, too. How-
ever, the above simplified view allows for an analytical treatment.
The dynamical part is given by the zonal average of the zonal momentum balance,
written as (3.31) for the residual streamfunction, and is augmented by the zonally aver-
aged balance of density. The zonal current is geostrophic. The model is thus governed
by

− 𝑓𝜙 = 𝑓 𝐵 + 𝜏0𝑥 − 𝜏 𝑥 + ℱ (3.45)
𝜌¯𝑡 + 𝒥 (𝜙, 𝜌¯) = ∇ ⋅ 𝜅∇¯
𝜌 + 𝐽𝑧 (3.46)
𝑓𝑢
¯ = −¯
𝑝𝑦 (3.47)

The Reynolds stress is neglected. The density equation, derived in section ?? on the Trans-
formed Eulerian Mean (TEM) theory, will be used in the steady form. Concerning the
eddy streamfunction 𝐵 and the diapycnal diffusivity we step back to the exact relations
in terms of the eddy fluxes 𝑣 ′ 𝜌′ and 𝑤′ 𝜌′ (the prime denotes transient eddy components
and standing eddies are ignored), given by

𝑣 ′ 𝜌′ 𝜌¯𝑧 − 𝑤′ 𝜌′ 𝜌¯𝑦 𝑣 ′ 𝜌′ 𝜌¯𝑦 + 𝑤′ 𝜌′ 𝜌¯𝑧


𝐵 = − 𝜅 = − (3.48)
∣∇¯𝜌∣2 ∣∇¯𝜌∣2

because the approximation used in (3.31) becomes invalid in the mixed layer. The merid-
ional density flux 𝑣 ′ 𝜌′ = −𝐾 𝜌¯𝑦 is still parameterized as in (3.38) with a thickness diffu-
sivity 𝐾. Using the downgradient form of the lateral eddy flux, one finds

𝑤 ′ 𝜌′
( )
𝐾 1 𝜅
𝐵 = (𝜅 − 𝐾)𝑠 and = + 1 + 2 𝐵 = −𝐾𝑠 + (1 + 𝑠2 ) (3.49)
𝜌¯𝑦 𝑠 𝑠 𝑠

in terms of the slope 𝑠 = −¯ 𝜌𝑧 of the isopycnals. Note that 𝐾 generally differs from the
𝜌𝑦 /¯
diapycnal diffusivity 𝜅. Both these diffusivities must in turn be parameterized in terms
84 The diabatic regime

of resolved fields, but in the following we simply assume their spatial structure as given
functions of (𝑦, 𝑧) (for simplicity either taken constant or taken from a numerical model).
Finally, 𝐽 is a vertical flux of density carried by small-scale turbulence. It is only relevant
in the upper near-surface layer where, at the air-sea interface, it equals the surface density
flux, 𝐽(𝑦, 𝑧 = 0) = 𝐽0 (𝑦).

Figure 3.20.: Set-up of the zonal mean overturning model. There is an Ekman layer at the top where the
stress is modeled by a body force, spreading the windstress into a layer of constant depth
𝑑. The stress in the interior is zero. A bottom layer is included in which either a frictional
stress or a bottom formstress becomes active. The Ekman layer coincides with the surface
mixed layer in which the density is vertically constant. The interior density field shows
sloping isopycnals. The ocean system interacts with the overlying atmosphere via the zonal
windstress 𝜏0𝑥 and the density flux 𝐽0 . Thiele

We follow the most simple concepts which have partly been used before. The interior
ocean is assumed adiabatic with respect to small-scale mixing, i.e. 𝐽 ≡ 0. We also assume
here a vanishing diapycnal eddy flux, hence 𝜅 ≡ 0, so that the density balance becomes
adiabatic, 𝒥 (𝜙, 𝜌¯) = 0. The consequence is that isolines of the residual streamfunction
and the density coincide. Furthermore, from 𝜅 = 0 it follows that the eddy density flux
is strictly along isopycnals, i.e. 𝑣 ′ 𝜌′ 𝜌¯𝑦 + 𝑤′ 𝜌′ 𝜌¯𝑧 = 0. Then 𝐵 = −𝑣 ′ 𝜌′ /¯
𝜌𝑧 = −𝐾𝑠 as in
𝑥
(3.31) and the previous section. The stress 𝜏 is assumed nonzero only in the surface
Ekman layer which is embedded in the upper ocean layer. If the bottom formstress ℱ is
absent, there is also a frictional bottom layer. Hence in the interior part of the ocean, the
streamfunction is given by

𝜙 = 𝐾𝑠 + 𝑀 (interior for ℱ = 0) (3.50)

with 𝑀 = −𝜏0𝑥 /𝑓 denoting henceforth the Ekman transport.


3 The Circulation of the Southern Ocean 85

−5
x 10
28 28 1.5 15

surface density [kg/m3]


500 m density [kg/m3]
27.5 27.5 1 10

net heatflux [W/m2]


E − P [kg/m2/s]
0.5 5
27 27
0 0
26.5 26.5
−0.5 −5
26 26 −1 −10

25.5 25.5 −1.5 −15


−75 −70 −65 −60 −55 −50 −45 −40 −35
latitude
−6 −6 −
x 10 x 10 x 10
1 3 2 2.5

Ekman transport [m2/s]


Ekman pumping [m/s]
2 2
density flux [kg/m2/s]

wind stress [m2/s2]


0.5 1
1 1.5

0 0 0 1

−1 0.5
−0.5 −1
−2 0

−1 −3 −2 −0.5
−75 −70 −65 −60 −55 −50 −45 −40 −35
latitude

Figure 3.21.: The forcing data from the NCAR-NCEP reanalysis. Upper: net heat flux (black), unit
Wm−2 , evaporation minus precipitation (red), unit kg m−2 s−1 ). The densities (surface in
blue, units kg m−3 , 500 m depth in green) are taken from the WOCE climatology. Lower:
windstress (green), unit m2 s−2 , Ekman transport (black), unit ms−1 , Ekman pumping
(blue), unit ms−1 , density flux (red), unit kg m−2 s−1 . The data are obtained by an ACC-path
following average. From Olbers and Visbeck (2005).

In the upper layer the diapycnal diffusivity 𝜅 is generally nonzero. In fact, if the layer
𝜌𝑧 = 0, 𝑠 = ∞), the diapycnal direction is horizontal and then 𝜅 ≡ 𝐾.
is vertically mixed (¯
Furthermore, 𝐵 = 𝑤′ 𝜌′ /¯ 𝜌𝑦 . We will treat the billow stress as a body force 𝜏 𝑥 = 𝜏0𝑥 𝐸(𝑧)
where 𝐸(𝑧 = 0) = 1 and zero at the mixed layer base (see Figure 3.20). Then

𝑤 ′ 𝜌′
𝜙 = + 𝑀 𝑇 (𝑧) (mixed layer) (3.51)
𝜌¯𝑦

where 𝑇 (𝑧) = 1 − 𝐸(𝑧) which equals 1 in the interior. Note that 𝜙 is basically unknown
unless the vertical density flux is parameterized. As discussed in the previous sections,
the balance of zonal momentum (see (3.45)) requires a stress layer above the bottom. The
stress may appear either in the form of a frictional bottom stress or a bottom formstress
(or both). We treat this problem at the end of this section.
To develop our model of the mean density structure, we need the size and pattern of
the surface momentum and density fluxes. The estimates, shown in Figure 3.21, are based
on the adjusted NCAR-NCEP reanalysis data, discussed in section 1. Windstress and
density flux data were obtained by averaging along mean ACC streamlines by finding the
latitude of 3∘ C isotherm at a depth of 200 m. All meridional sections were then shifted by
their departure from the zonally averaged ACC position, sorted along their new ’latitude
circles’ and subsequently averaged. The maximum Ekman transport in these data is
about 30 Sv, occurring at 51∘ S with upwelling to the south and downwelling to the north
of this latitude. The surface density loss depends on the combined air-sea heat and fresh
water flux. Consistent with other estimates we find oceanic heat gain of about 10 Wm−2
86 The diabatic regime

−7 −4
x 10 x 10
27.5 2 1.5

surface density [kg/m3]

density flux [kg/m2/s]

windstress [m2/s2]
1
27 1

26.5 0.5
−1

26 −2 0
0 500 1000 1500 2000 2500 3000
y [km]
3 −6
density [kg/m ] residual circ [Sv] pumping [10 m/s]
0 27.4 2
27.2
5
27 1
−1000
26.8
0 0
26.6
−2000 26.4
−5 −1
26.2

−3000 26 −2
0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000
y [km] y [km] y [km]

Figure 3.22.: The Marshall-Radko model with constant thickness diffusivity 𝐾 = 800 m2 s−1 and idealized
forcing field as indicated in the upper panel. The conversion of the streamfunction unit to Sv
is performed using the zonal path length at 60∘ .The vertical velocities (residual 𝜙𝑦 [black],
Ekman 𝑀𝑦 [red] and eddy-induced (𝐾𝑠)𝑦 [blue]) of the solution are shown in the lower right
panel.

over the core of the ACC and cooling of a similar magnitude north and south of the
stream. The fresh water flux at the surface shows a net freshening south of 50∘ S of about
10−5 kgm−2 s−1 and about a similar amount of net evaporation around 30∘ S, causing a
significant modification of the surface density. Only in the most northern part is the
freshwater contribution to the density flux smaller than the air-sea heat flux. Our final
estimate of air-sea density flux yields a reduction of the surface density south of 48∘ S and
an increase to the north of 48∘ S. However, the expected error on the surface density flux
as a difference of two not well-known quantities could be rather large.

A diagnostic model Marshall and Radko (2003) put forward a partly diagnostic model
of the overturning. The upper layer between 𝑧 = 0 and 𝑧 = −𝑑 is assumed vertically
mixed, and its meridional density profile 𝜌¯𝑑 (𝑦) is given. The fields in the interior, how-
ever, are predicted.
In a diagnostic mode, by vertical integration of the density balance (3.46) in the mixed
layer, the streamfunction 𝜙𝑑 (𝑦) = 𝜙(𝑦, 𝑧 = −𝑑) at each latitude along the mixed layer
base

(∫ 0 )
𝜙𝑑 𝜌¯𝑑𝑦 = 𝐽0 + 𝐾𝑑𝑧 𝜌¯𝑑𝑦 = 𝐽𝑒𝑓 𝑓 (3.52)
−𝑑 𝑦

may be inferred from the local values of the air-sea flux 𝐽0 , the mixed layer density gradi-
ent 𝜌¯𝑑𝑦 and the diffusivity 𝐾. By matching to the interior representation (3.50) at 𝑧 = −𝑑,
3 The Circulation of the Southern Ocean 87

Fundamental Box 7 : Characteristics of the interior problem

The density field in the interior is governed by

𝒥 (𝜙, 𝜌¯) = 𝜙𝑦 𝜌¯𝑧 − 𝜙𝑧 𝜌¯𝑦 = 0 (3.53)

The characteristic curves (𝑦(𝑡), 𝑧(𝑡)) of this differential equation are determined by
𝑑𝑦 𝑑𝑧
= −𝜙𝑧 = 𝜙𝑦 (3.54)
𝑑𝑡 𝑑𝑡

representing a system of ordinary differential equation of Hamiltonian form (see Appendix ??). The
’Hamiltonian’ 𝜙 is conserved along the characteristics, 𝑑𝜙/𝑑𝑡 = 𝑦𝜙 ˙ 𝑧 = 0. Likewise, the density
˙ 𝑦 + 𝑧𝜙
is constant on these curves: from (3.53) we find 𝑑𝜌/𝑑𝑡 = 𝑦𝜌 ˙ 𝑧 = 0. Hence knowledge of the charac-
˙ 𝑦 + 𝑧𝜌
teristics is equivalent to solving the interior problem.
To solve (3.54), 𝜙 = 𝐾𝑠 + 𝑀 must inserted, leading to a complicated problem. Finding the characteristics,
however, is simpler if (3.53) is rewritten as

𝜙𝑦 + 𝑠𝜙𝑧 = 0 (3.55)

and the characteristic curves (𝑦 ′ (𝑡), 𝑧 ′ (𝑡)) of this differential equation are determined by

𝑑𝑦 ′ 𝑑𝑧 ′
=1 =𝑠 (3.56)
𝑑𝑡 𝑑𝑡

As before, from 𝑑𝑦 ′ − 𝑠𝑑𝑧 ′ = 0, we find that density is constant along these curves, and the derivative
𝑑𝜙/𝑑𝑡 of 𝜙(𝑦 ′ (𝑡), 𝑧 ′ (𝑡)) along such a characteristic is zero, so 𝜙 = 𝐾𝑠 + 𝑀 = const on each characteristic.
Integration of (3.56) requires initial conditions 𝑦 ′ (𝑡 = 0) = 𝑦 ★ , 𝑠(𝑡 = 0) = 𝑠𝑑 . Then 𝐾𝑠 + 𝑀 (𝑦) =
𝐾(𝑦 ★ , −𝑑)𝑠𝑑 (𝑦 ★ ) + 𝑀 (𝑦 ★ ). Initial data for 𝑠 are required on a non-isopycnal curve, in our case the depth
level 𝑧 = −𝑑. Elimination of 𝑡 from this problem yields (3.57).

the slope 𝑠𝑑 (𝑦) = (𝜙𝑑 (𝑦) − 𝑀 (𝑦))/𝐾𝑑 (𝑦) of isopycnals along the mixed base can be com-
puted.
In the diagnostic mode the direction of the residual circulation is found to be governed
by the signs of the (modified) surface flux 𝐽𝑒𝑓 𝑓 and the gradient 𝜌¯𝑑𝑦 of the surface density.
Following the observations, shown in Figure 3.21, the gradient is negative. We expect a
negative 𝐽𝑒𝑓 𝑓 because 𝐽0 < 0 in the main area of interest, and because the diffusive
contribution is likely small if 𝜌¯𝑑𝑦 and 𝐾 are close to being constant. Then we find 𝜙𝑑 > 0, a
northward residual circulation. Furthermore, with a stable stratification the slope 𝑠𝑑 must
be negative, and hence the eddy-driven circulation is southward, 𝐾𝑠 < 0, compensating
the Ekman circulation 𝑀 > 0 in parts. We return to this compensation effect later in the
prognostic model.
The depth-latitude dependence 𝑧𝜌 = 𝑧𝜌 (𝑦 ★ , 𝑦) of an isopycnal starting at a point 𝑦 =
𝑦★, 𝑧= −𝑑 on the mixed layer base, with the streamfunction value 𝜙𝑑 (𝑦 ★ ) attached to it,
can be found by integrating the ordinary differential equation

𝜙𝑑 (𝑦 ★ ) − 𝑀 (𝑦)
( )
𝜌¯𝑦 𝑑𝑧
𝑠 = − = = (3.57)
𝜌¯𝑧 𝑑𝑦 𝜌 𝐾
88 The diabatic regime

which can be performed analytically if 𝐾 is constant (the problem is easily solved for a
factorized diffusivity 𝐾(𝑦, 𝑧) = 𝐾1 (𝑦)𝐾2 (𝑧)). With the isopycnal structure, the density
field and the residual circulation are determined. We exemplify this diagnostic model in
Figure 3.22 where the driving fluxes 𝐽0 and 𝜏0𝑥 are taken as sinusoidal and the surface
density gradient as constant, with amplitudes and scales resembling the observed fields
shown in Figure 3.21 in the latitude range from 70∘ S to 40∘ S. The thickness diffusivity is
assumed constant. The figure shows a reasonable density field (compare with Figure 3.3)
and upwelling of deep water (NADW), coming from below 1500 m at the northern rim
of the modeled domain of about 10 Sv. In the mixed layer the water is made lighter and
descends, leaving the domain above 1500 m depth in the north as AAIW. The vertical
velocities at the mixed layer base (lower right panel) reflect the approximate compen-
sation of the Eulerian and the eddy-induced circulation. It should be mentioned that
this solution is sensitive to changes in any of the prescribed fields and particularly to the
thickness diffusivity 𝐾. Lowering (increasing) 𝐾 makes the isopycnals flatter (steeper).
It may occur that isopycnals cross, in which case the model with zero diapycnal diffusion,
of course, becomes unrealistic.
−100 −100 −100 3500

−200 −200 −200 3000

−300 −300 −300 2500

−400 −400 −400 2000


z [m]
−500 −500 −500 1500

−600 −600 −600 1000

−700 −700 −700 500

−800 −800 −800 0


0 200 400 600 800 1000 0 200 400 600 800 1000 0 200 400 600 800 1000
y [km] y [km] y [km]
−100 −100 −100 4

−200 −200 −200 3

2
−300 −300 −300
1
−400 −400 −400
z [m]
0
−500 −500 −500
−1
−600 −600 −600
−2

−700 −700 −700 −3

−800 −800 −800 −4


0 200 400 600 800 1000 0 200 400 600 800 1000 0 200 400 600 800 1000
−100 −100 −100 4

−200 −200 −200 3

2
−300 −300 −300
1
−400 −400 −400
z [m]
0
−500 −500 −500
−1
−600 −600 −600
−2

−700 −700 −700 −3

−800 −800 −800 −4


0 200 400 600 800 1000 0 200 400 600 800 1000 0 200 400 600 800 1000

Figure 3.23.: Upper row: thickness diffusivity 𝐾 (unit [m2 /s]) diagnosed from the idealized eddy-resolving
ACC model described in Example Box ??. The experiments are those from Figure 3.14 for
amplitudes (from left to right) (0.5, 0.75, 1.0) × 10−4 m2 s−2 of the sinusoidal wind stress.
The contour intervals for the diffusivities are 200m2 /s and the 2000 m2 /s line is thick.
Middle and lower row: time-zonal mean residual streamfunction (𝜙) and mean buoyancy
(¯𝑏 = −𝑔𝜌) obtained for the diagnostic Marshall-Radko model with 𝐾 and surface fluxes from
the eddy-resolving model. The contour intervals are 0.2 Sv and 0.001 m/s2 , respectively.
Thick lines are zero lines for the streamfunction and 0.007 m/s2 lines for the buoyancy.

Finally, we demonstrate that the conceptual framework of the interior residual over-
turning, given by the diagnostic model, is appropriate as long as the correct eddy dif-
fusivity variability is taken into account. We compare the application of equation (3.57)
with the numerical results of the idealized eddy-resolving model. Figure 3.23 shows
3 The Circulation of the Southern Ocean 89

Example Box 14 : Approximate solution of the equation (3.63)

To gain some insight into the behavior of (3.63) we assume all coefficients 𝑀, 𝒦𝑚 , 𝐾𝑑 , 𝜅 and 𝑄 =
𝑔𝐽0 /(𝜌0 𝑁 2 ) to have an identical 𝑦-dependence, however, with different amplitudes 𝑀0 , 𝒦0 , 𝐾0 , 𝜅0 and
𝑄0 . Allowing for the slight inconsistency to take the diffusion coefficient outside the derivative, an equa-
tion with constant coefficients is obtained,

𝒦0 𝑠′ = 𝑀0 𝑠 + 𝐾0 𝑠2 − 𝜅0 (1 + 𝑠2 ) − 𝑄0 (3.58)

Depending on Δ = 𝑀02 + 4(𝐾0 − 𝜅0 )(𝑄0 + 𝜅0 ) the solution is


⎧ √ (√ )
⎨ − Δ tanh Δ
(𝑦 − 𝑦 ) for Δ > 0
1 𝑀0 𝐾0 −𝜅0
( √2𝒦0
0
𝑠=− √ ) (3.59)
2 𝐾0 − 𝜅0 ⎩ + −Δ tan −Δ
(𝑦 − 𝑦0 ) for Δ < 0
𝐾0 −𝜅0 2𝒦0

For a realistic situation 𝑄0 must be negative. If Δ < 0 a ’blow up’ develops in the tan-solution, the slope
tends to infinity, and obviously a realistic model demands that the diapycnal mixing increases, i.e. 𝜅0 must
depend on 𝑠.
For Δ > 0 the solution is exponential. As 𝒦0 is attached to the meridional diffusion of density, we notice
from (3.59) that diffusion influences the lateral scale of the solution but not the overall magnitude. The
overall magnitude of the slope increases (becomes more negative) with the increase of the Ekman transport
and the diapycnal diffusion and decreases with a larger (more negative) surface density flux.

the residual streamfunction 𝜙 in the Southern Ocean part of the model, for three differ-
ent wind stress amplitudes, prescribing 𝜙(𝑧 = −150m) and 𝐾(𝑦, 𝑧) by the results of the
eddy-resolving numerical model (see Figure 3.14). The diagnostic model (??) in conjunc-
tion with the diagnosed eddy diffusivity 𝐾 reproduces the extent of the overturning cell
and the slopes of the streamlines very well for each wind stress amplitude. In particular,
the slight deepening of the zero line (thick) is captured almost perfectly. Small deviations
from the eddy-resolving numerical model results represent diabatic effects not included
in the adiabatic diagnostic model.

A prognostic model for the slope In a prognostic mode the surface density is not pre-
scribed but must be calculated from the equations (3.45) and (3.46). We must solve these
balances for the upper layer. Then, (3.52) is still valid but cannot be used to compute the
streamfunction 𝜙𝑑 in the mixed layer because it is the equation determining 𝜌¯𝑑 (𝑦). As the
slope is infinite in the mixed layer, we must load more physical features into the model
in order to provide a slope profile at some depth to feed into the characteristic equation
for the interior, either as described in Fundamental Box ?? or via 𝜙𝑑 (𝑦 ★ ) in (3.57).
We derive an equation determining the slope 𝑠𝑑 just below the mixed base. The mixed
layer is vertically completely mixed (hence 𝜅 = 𝐾) and absorbs the entire billow stress
𝜏 𝑥 as before. Immediately below the mixed layer, before entering the adiabatic interior,
the eddy flux can still be diapycnal in part so 𝜅 is nonzero and generally not equal to 𝐾.
Expressing 𝜙𝑑 in the vertically integrated density balance (3.52) of the mixed layer by the
general form (3.51) this balance is now written as


𝑀 𝜌¯𝑑𝑦 − 𝑤′ 𝜌′ −𝑑 = 𝐽0 + (𝒦𝑚 𝜌¯𝑑𝑦 )𝑦 (3.60)
90 The diabatic regime

Here, the integral 𝒦𝑚 of 𝐾 over the mixed layer is introduced, appearing also in (3.52).
We divide by the density gradient 𝜌¯𝑧 , appropriate just below the mixed layer base, and
evaluate the vertical density flux 𝑤′ 𝜌′ just below the base using (3.49). Then we obtain

𝐽0 (𝒦𝑚 𝜌¯𝑑𝑦 )𝑦
(𝑀 + 𝐾𝑑 𝑠𝑑 )𝑠𝑑 − 𝜅(1 + 𝑠2𝑑 ) = − − (3.61)
𝜌¯𝑧 𝜌¯𝑧

where 𝑠𝑑 = −¯ 𝜌𝑧 . To convert (3.61) to an equation for the slope, a fairly ’mild’ as-
𝜌𝑑𝑦 /¯
sumptions must be made: in the diffusion term

(𝒦𝑚 𝜌¯𝑑𝑦 )𝑦 (𝒦𝑚 𝜌¯𝑧 𝑠𝑑 )𝑦


= − ≈ − (𝒦𝑚 𝑠𝑑 )𝑦 (3.62)
𝜌¯𝑧 𝜌¯𝑧

the 𝑦-dependence of 𝜌¯𝑧 is ignored. In the density forcing term, 𝜌¯𝑧 is replaced by a the
squared Brunt-Väisälä frequency 𝑁 2 = −𝑔 𝜌¯𝑧 /𝜌0 below the mixed layer which later will
be regarded as prescribed. We arrive at the equation

𝑔𝐽0 /𝜌0
(𝒦𝑚 𝑠𝑑 )𝑦 = (𝑀 + 𝐾𝑑 𝑠𝑑 ) 𝑠𝑑 − 𝜅(1 + 𝑠2𝑑 ) − (3.63)
𝑁2

determining the slope at (or just below) the mixed base. The eddy-induced advection
(second term on the left-hand side) and the surface flux (if 𝐽0 is negative; last term) tend
to increase the slope towards positive numbers, and only the Ekman advection (if 𝑀 is
positive; first term) and diapycnal mixing (third term) can counteract this tendency. In
a realistic situation where 𝑠𝑑 = −𝑂(10−3 ) and ∣𝑀 𝑠𝑑 ∣ ∼ 𝐾𝑑 𝑠2𝑑 = 𝑂(10−3 ) the diapycnal
diffusivity must be fairly small, i.e. 𝜅 ∼ 𝑂(10−3 ) or smaller. This requirement hints at
a dependence of 𝜅 on the slope: for the infinite slope in the mixed layer we have 𝜅 =
𝐾, reducing to much smaller values just below the mixed layer base where the slope
becomes finite, and then approaching zero in the adiabatic interior.

In the following we regard 𝑀 and 𝑄 = 𝑔𝐽0 /(𝜌0 𝑁 2 ) as given functions of latitude 𝑦


and, for simplicity, the mixing coefficients 𝒦𝑚 = 𝑑𝐾𝑑 , 𝐾𝑑 , 𝜅 as constant. The differential
equation (3.63) for 𝑠𝑑 is of the Riccati type. Though some analytical techniques exist to
reduce the nonlinear Riccati differential equation to simpler linear equations of second
order or to find the general solution if a particular solution is known, exact analytical
solutions for a reasonable meridional dependence of the above coefficient functions can
not be given. An approximate analytical solution is discussed in the Example Box ??.

We discuss some numerical solutions of (3.63), shown in Figure 3.24. First consider
the case 𝐽0 = 𝜅 = 0 so that the right-hand side of (3.63) is given by 𝜙𝑑 𝑠𝑑 , written in
terms of the residual streamfunction 𝜙𝑑 = 𝑀 + 𝐾𝑑 𝑠𝑑 . The zero curve of 𝜙𝑑 is shown
in the black dashed line of the left figure in each panel of Figure 3.24. If 𝑠𝑑 is above
this curve (but negative) and 𝜙𝑑 thus northward, 𝑠𝑑 will becomes more negative until it
3 The Circulation of the Southern Ocean 91

slope x 103 streamfcts [Sv] pumping [10−6 m/s] density slope x 103 streamfcts [Sv] pumping [10−6 m/s] density
0 30 2 27.5 0 30 2 27.5

−0.2 20 −0.2 20
1 27 1 27
10 10
−0.4 −0.4
0 0 26.5 0 0 26.5
−0.6 −0.6
−10 −10
−1 26 −1 26
−0.8 −20 −0.8 −20

−1 −30 −2 25.5 −1 −30 −2 25.5


0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000
y [km] y [km] y [km] y [km] y [km] y [km] y [km] y [km]

slope x 103 streamfcts [Sv] pumping [10−6 m/s] density slope x 103 streamfcts [Sv] pumping [10−6 m/s] density
0 30 2 27.5 0 30 2 27.5

−0.2 20 −0.2 20
1 27 1 27
10 10
−0.4 −0.4
0 0 26.5 0 0 26.5
−0.6 −0.6
−10 −10
−1 26 −1 26
−0.8 −20 −0.8 −20

−1 −30 −2 25.5 −1 −30 −2 25.5


0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000
y [km] y [km] y [km] y [km] y [km] y [km] y [km] y [km]

Figure 3.24.: Some solutions of the prognostic model (3.63) for different 𝒦𝑚 = [1, 3] × 𝑑𝐾𝑑 where 𝑑 =
100 m and 𝐾𝑑 = 1500 m2 s−1 , displayed by the blue and red curves, respectively. The
Brunt-Väisälä frequency below the mixed layer, entering the density forcing, is chosen 𝑁 =
3 × 10−3 s−1 . The dashed curves refer to the eddy components, the full curves to the residual
components. The Ekman transport and pumping are displayed in the black full curves. In
the slope figures the black dashed curve is 𝑀 (𝑦) + 𝐾𝑑 𝑠𝑑 = 0. The last figure in each panel
shows the density 𝜌𝑑 inferred from the slope, as explained in the text. The dashed black line
is the approximate observed surface density. Upper left: 𝐽0 = 𝜅 = 0. Upper right: 𝐽0 = 0
and 𝜅 = 10−4 m2 s−1 . Lower left: 𝐽0 as shown in Figure 3.22 and 𝜅 = 10−4 m2 s−1 . Lower
right: same but with a different initial condition.

hits 𝜙𝑑 = 0 and turns to increase. The residual circulation reverses here to a southward
direction. Depending on 𝐾𝑑 and 𝒦𝑚 this reversion occurs at different latitudes (see upper
left panel of Figure 3.24). The residual streamfunction may start from the initial value
being northward or southward and may reverse as described above (see second panel
of Figure 3.24). In a southward case the Eulerian circulation is overcompensated by the
eddy-induced component. Introduction of a small diapycnal diffusivity yields a minor
modification of this behavior. The model becomes meaningless if the slope proceeds
towards positive numbers. For the first two cases in Figure 3.24 this would occur north of
the integration interval, for the third case we find this behavior inside because the surface
forcing is turned on and overcomes the negative tendency of the 𝜙𝑑 𝑠𝑑 -term for small slope
values (see the red curves applying to the largest 𝒦𝑚 ). Here the blow-up singularity,
explained in Example Box ??, comes into play. With a smaller 𝒦𝑚 or a larger (negative)
initial value (see fourth panel) this behavior can be avoided in the present integration
interval, but what is really required is a physically motivated parameterization in the
model to counteract such a behavior. With a positive slope the surface density would
increase towards the north and the air-sea exchange of density would change (see below).

Knowing the slope 𝑠𝑑 = −𝜌𝑑𝑦 /𝜌𝑧 we can determine density in the mixed layer by
approximating the vertical density gradient 𝜌¯𝑧 by the given 𝑁 2 , so far only done for the
forcing term in (3.63). The result is displayed in last figure of each case in Figure 3.24,
92 The diabatic regime

slope x 103 streamfcts [Sv] pumping [10−6 m/s] density density flux [10−7 kg/m2/s]
0 30 2 27.5 5

20 4
−0.2
1 27
3
10
−0.4 2
0 0 26.5
−0.6 1
−10
0
−1 26
−0.8 −20 −1

−1 −30 −2 25.5 −2
0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000
y [km] y [km] y [km] y [km] y [km]

Figure 3.25.: Some solutions of the prognostic model (3.63), using the restoring flux (3.64) with 𝐽˜0 =
0 and two different 𝒦𝑚 = [1, 3] × 𝑑𝐾𝑑 [in blue and red]. The right figure displays the
diagnosed density flux of the corresponding solutions. The restoring parameter has the value
𝛼𝑠 = 1.5 × 10−6 ms−1 . The other parameters and the colors are as in the lower right case of
Figure 3.24.

showing a reasonable success: for comparison the approximate profile of the observed
surface density (see Figure 3.21 or Figure 3.22) is included by the black dashed line.
The above model is unsatisfactory for two reasons. First, the model depends on the
specified Brunt-Väisälä frequency below the mixed layer, and second, as already indi-
cated above, prescription of the density flux is not acceptable. It should not be the mod-
eler but rather the ’atmosphere’ to choose the flux and react on the ocean’s surface con-
dition. To implement a feed-back in a simple way a restoring-type flux

𝐽0 (𝑦) = 𝐽˜0 (𝑦) + 𝛼𝑠 (¯


𝜌𝑠 (𝑦) − 𝜌¯𝑑 (𝑦)) (3.64)

should be used. A tunable parameter 𝛼𝑠 (with dimension ms−1 ) and an ’atmospheric’


restoring density 𝜌¯𝑠 (𝑦) then enter the model. A prescribed climatological flux 𝐽˜0 may be

density [kg/m3] residual circ [Sv] zonal vel [m/s] pumping [10−6 m/s]
0 0.15 2
5
0.1
−1000 27
0 1
0.05
−2000 −5
26.5 0 0
−10
−3000 −0.05
26 −1
−15 −0.1
−4000
−20 −2
0 1 2 3 0 1 2 3 0 1 2 3 0 1 2 3
y [103 km] y [103 km] y [103 km] y [103 km]

Figure 3.26.: Solution for the ocean interior with the same coloring and parameters as the Marshall-Radko
model of Figure 3.22.
3 The Circulation of the Southern Ocean 93

taken as zero or the one as in the previous examples. The performance of this model is
demonstrated in Figure 3.25. For the ’blue’ solution the interior density, residual stream-
function, zonal velocity (see next paragraph) and pumping velocities are displayed in
Figure 3.26.

Giving up to prescribe 𝑁 2 and proceeding to a fully prognostic model, i.e. resolving


the transition layer between the mixed layer and the adiabatic interior with appropri-
ate parameterizations, is a fairly complicated task. The reader is referred to Olbers and
Visbeck (2005) where prognostic equations are discussed for slope in this transition layer.

Balance of zonal momentum and the zonal transport The model is completed by the
determination of the zonal current which has a geostrophic balance as given by (3.47).
The baroclinic pressure 𝑝¯𝑐𝑙𝑖𝑛 is determined by interior density field (see Figure 3.26). The
gradient 𝑔 𝜁¯𝑦 of the surface pressure follows for the vertically integrated balance of zonal
momentum, as below in the respective model discussed in the chapter. In case of a flat
bottom we have to work with the balance between the applied windstress and the fric-
tional bottom stress, 𝜏0𝑥 = 𝜏𝑏𝑥 , and use the equations of the bottom Ekman layer. In the
presence of topography we have the balance between the windstress and the bottom
formstress, 𝜏0𝑥 = −ℱ𝑏 but can proceed only if a parameterization of ℱ𝑏 is given. Both
routes result in a condition on 𝜁¯𝑦 . Given this gradient, the velocity profile 𝑢
¯(𝑧) is deter-
mined and the total transport can be calculated. A ’cheap’ way around the problem of
the surface pressure is to assume a level of no motion at the bottom, i.e. calculate the DCL
transport (see Fundamental Box ??). This solution is depicted in Figure 3.26.

3.7. The local balances of momentum and vorticity

The previous discussion in this chapter mainly dealt with zonally or ACC-path averaged
properties and balances of the Southern Ocean circulation. We continue with a particular
view on the local balances of momentum and vorticity in the light of the BARBI model,
specifically the most realistic case NL discussed in section 3.4, which includes topography
and stratification. For a coarse model of the circulation the vertically integrated balance
of momentum is given by (3.1). The flow is, of course, predominantly in a geostrophic
balance. Thus, deviations from this state are of interest. These are revealed by writing
the equations in natural coordinates (𝑠, 𝑛) oriented at the flow direction with 𝑠 along 𝜓-
contours and 𝑛 normal; here 𝜓 is the streamfunction of the horizontal transport, U =
∇𝜓. Note that 𝑈 ⊥ ≡ 0 so we obtain
¬

∂𝑝𝑏 ∂𝐸 ∂𝑝𝑏 ∂𝐸 ∥
𝑓 𝑈 ∥ = −ℎ − + 𝜏0⊥ + 𝐹 ⊥ −ℎ − + 𝜏0 + 𝐹 ∥ = 0 (3.65)
∂𝑛 ∂𝑛 ∂𝑠 ∂𝑠

Two pressure terms appear, one associated with the bottom pressure 𝑝𝑏 and one aris-
ing from the baroclinic potential energy 𝐸. The geostrophic terms are now only in the
balance of momentum normal to streamlines and are, of course, highly dominating that
94 The local balances of momentum and vorticity

a: −ℎ∂𝑝𝑏 /∂𝑛 − ∂𝐸/∂𝑛 b: 𝜏0⊥ c: 𝐹 ⊥


d: −ℎ∂𝑝𝑏 /∂𝑠 − ∂𝐸/∂𝑠 e: 𝜏0 f: 𝐹 ∥

Figure 3.27.: Upper row: local momentum balance normal to streamlines for case NL. a) normal gradient
of bottom pressure and baroclinic potential energy (the corresponding Coriolis force 𝑓 𝑈 ∥ is
indistinguishable from this figure). b) normal component of windstress. c) normal compo-
nent of friction term. Lower row: same for the components parallel to streamlines. Units:
10−2 m2 s−2 . The contours show the streamfunction, CI = 25 Sv. Note the difference in the
scales.

component, at least by two orders of magnitude (see upper left panel of Figure 3.27).
Windstress and the lateral friction are very much smaller, the latter occurs only very lo-
calized. The balance of the parallel component of momentum is between the pressure
gradients and the windstress. Overall, there is surprisingly little effect from the friction,
though the lateral viscosity is not small (see Example Box ??). The partition between the
two pressure gradients is of interest. The normal gradients are of similar size, as demon-
strated in Figure 3.28. However, the bottom pressure gradient is very localized, occurring
in worm-like features oriented at the streamlines, whereas the 𝐸 gradient term is a broad
scale structure. It has an imbedded filamented pattern which just compensates the bot-
tom pressure worms, so that the total pressure force becomes smoother (shown in Figure
3.27). The along-streamline pressure gradients are much smaller than the normal com-
ponents (lower roe of Figure 3.28). They have a similar pattern and size but of mostly an
opposing sign. Note that −ℎ∂𝑝𝑏 /∂𝑠 generates the bottom formstress in the along-stream
mean, balancing the along-stream mean of the parallel windstress.

The pressure forces enter the balance of vorticity as a torque-like term. As explained
in section ?? there are two forms of vorticity. The vorticity balance of the depth integrated
momentum U is
3 The Circulation of the Southern Ocean 95

a: −ℎ∂𝑝𝑏 /∂𝑛 b: −∂𝐸/∂𝑛

c: −ℎ∂𝑝𝑏 /∂𝑠 d: −∂𝐸/∂𝑠

Figure 3.28.: The normal and parallel pressure gradient terms for case NL. Units: 10−2 m2 s−2 . The
contours show the streamfunction, CI = 25 Sv.

∂𝜓
𝛽 = 𝒥 (𝑝𝑏 , ℎ) + curl 𝝉 0 + curl F (3.66)
∂𝑥

The bottom torque 𝒥 (𝑝𝑏 , ℎ) appears here as source of vorticity. All terms in this balance
contribute in the BARBI experiment with a similar overall size but different patterns (not
shown). The 𝛽-term in (3.66) is most important after the current leaves Drake Passage,
heading northward, and some other areas where the current must deviate from being
almost zonal due to islands or other massive changes in the topography. The bottom
torque, arising from depth gradients, has a spotty structure but has locally the highest
amplitudes. Friction is again of minor importance. The ACC flow is definitely not in a
Sverdrup balance anywhere.

The other form of vorticity equation is built from (3.1) by elimination of the pressure
𝑝𝑏 , resulting in

( )
𝑓 1 𝝉0 F
𝒥 𝜓, = 𝒥 (𝐸, ) + curl + curl (3.67)
ℎ ℎ ℎ ℎ

It describes the balance of vorticity of the depth-averaged velocity U/ℎ and is forced by
another torque-like term, the JEBAR term 𝒥 (𝐸, 1/ℎ). The dominant terms in the balance
96 The local balances of momentum and vorticity

a: 𝒥 (𝜓, 𝑓 /ℎ) b: −𝒥 (𝐸, 1/ℎ)

c: −curl 𝝉 0 /ℎ d: −curl F/ℎ

Figure 3.29.: Vorticity balance of depth-averaged current for case NL. a) topographic-planetary 𝛽-term. b)
JEBAR. c) curl of windstress. d) friction term. Units: 10−12 s−2 . The contours show the
streamfunction, CI = 25 Sv.

(4.46) are – by at least an order of magnitude – the topographic-planetary Jacobian and


the JEBAR term. They largely oppose and cancel each other, as obvious from Figure 3.29.
Windstress and friction are an order of magnitude smaller. The approximate compensa-
tion of the two terms in (4.46) may be traced back to the predominance of the geostrophic
terms in the balance of momentum, or likewise to an approximate compensation of the
barotropic component (associated with the planetary Jacobian) and the baroclinic com-
ponent (associated with JEBAR) of the vertical velocity.
Remember that (4.46), with the time tendency term retained, is one of the prognostic
equations of the BARBI model. The other dynamical equation is the balance of baroclinic
potential energy, written here in the reduced form

∂𝐸 𝑁 2 ℎ2 𝑁2 ℎ2 𝝉 0
[ ]
+ 𝒥 (𝐸, ) − 𝒥 (𝜓, ℎ2 ) = − curl + 𝐾∇2 𝐸 (3.68)
∂𝑡 6 𝑓 6 𝑓

To derive this form from the complete BARBI model, the balance of baroclinic momentum
is taken for simplicity as geostrophic, and advection of the perturbation density, buoy-
ancy forcing and dissipation are ignored. The terms of the balance (4.47) are displayed
in Figure 3.30, revealing again the dominance of two terms, namely the two Jacobians on
the left-hand side of (4.47). The potential energy balance thus suffers from the dominance
of the geostrophic terms in a similar way as the vorticity balance. While the geostrophic
3 The Circulation of the Southern Ocean 97

a: −(𝑁 2 /6)𝒥 (𝐸, ℎ2 /𝑓 ) b: (𝑁 2 /6)𝒥 (𝜓, ℎ2 )

c: 𝐾∇2 𝐸 d: (𝑁 2 /6)curl 𝝉 0 ℎ2 /𝑓

Figure 3.30.: Terms of the potential energy balance (4.47) for case NL. a) barotropic divergence term
b) baroclinic divergence term. c) eddy diffusion term. d) Ekman pumping term. Units:
10−4 m3 s−3 . The contours show the streamfunction, CI = 25 Sv.

terms entered there as almost compensating torques, they come into the balance of poten-
tial energy as integrals of vertical velocities which appear in (4.47) as pumping up/down
the background stratification. We may rephrase the statement about the dominance of
the geostrophic terms as one about a partial compensation of the bottom-induced verti-
cal barotropic velocity and the geostrophic vertical baroclinic velocity. Both act on the
background stratification and thus generate potential energy.

3.8. The ACC transport in BARBI

The above discussion repeats in parts the BARBI concepts of the wind-driven circulation
in the presence of stratification worked out in section ??. There, we were lead to the
baroclinic Stommel equation, using the similarity of the barotropic streamfunction 𝜓 and
the baroclinic streamfunction 𝜙 = 𝐸/𝑓 and the occurrence of the bottom torque term in
both BARBI balances, if properly written. Here we gain the same result: 𝜓 ≈ 𝜙 derives
from the dominance of the Jacobians in (4.46) and (4.47). Expressing these balances in the
equivalent form
98 Wave-induced bottom formstress

( ) ( )
𝛽 ∂𝜓 1 1 𝝉0 F
+ 𝑓𝒥 𝜓, = 𝒥 𝐸, + curl + curl
ℎ ∂𝑥 ℎ ℎ ℎ ℎ
(3.69)
ℎ2 𝝉
( ) ( ) ( )
1 ∂𝐸 1 1 𝑓 0
𝛽 + 𝜅∇2 𝐸 + 𝒥 𝐸, = 𝑓𝒥 𝜓, + curl
2ℎ𝑓 ∂𝑥 ℎ ℎ 2ℎ3 𝑓

to have identical Jacobians, eliminating them and using 𝜓 ≈ 𝜙, we find

3 ∂𝜓 1 𝑓 ℎ2 𝝉 0 𝝉0 F
𝛽 + 𝜅∇2 𝜓 = 2
curl + ℎ curl + curl (3.70)
2 ∂𝑥 2 2ℎ 𝑓 ℎ ℎ

In these relation we have introduced 𝜅 = 𝐾/𝜆2 where 𝜆 = 𝑁 ℎ/( 6∣𝑓 ∣) is the local inter-
nal Rossby radius in the BARBI framework. The viscous term in (3.70) is small and may
be abandoned, likewise the 𝛽-term arising in the wind forcing (the topographic terms
cancel exactly). Then we arrive at the baroclinic Stommel equation4

∂𝜓 1
𝛽 + 𝜅∇2 𝜓 = curl 𝝉 0 (3.71)
∂𝑥 3

which immediately allows to derive an equation for the ACC transport. Taking the zonal
mean of (3.71) we obtain

¯
1 ∂𝑈 ¯
𝐾𝑓 2 ∂ 𝑈 ∂𝜏 𝑥
𝜅 = 2 2 2 ≈ 0 (3.72)
3 ∂𝑦 ℎ 𝑁 ∂𝑦 ∂𝑦

which may be integrated twice to yield the total transport, roughly 𝐿𝑦 𝑈 ¯ ∼


𝑦 2 𝑥
(1/2)𝐿 (ℎ𝑁/𝑓 ) 𝜏0 /𝐾. Surprisingly this is identical to the estimate from the adiabatic
model in section 3.5. However, the agreement is accidental because (3.72) yields the total
transport in a model which regards the presence of standing waves and surface displace-
ment, both of which are absent in the baroclinic part of (3.44).

3.9. Wave-induced bottom formstress

The above considerations miss so far an analytical treatment of the influence of topogra-
phy (loosely speaking, the formula relating ℱ𝑏 to 𝑈¯ ). The formstress in equations (3.19)
and (3.21) contains the part of the bottom pressure which is out of phase with variations
of the topography along the zonal path of integration. Some insight into the mechanism
which the flow uses to generate the bottom formstress has been gained from heavily trun-
cated images of the full dynamics, so called low-order models, where the flow fields are
represented by very few spectral components (see section ??), a few ’waves’. The simplest
of such low-order models is the one by Charney and DeVore (1979; CDV in the following)
which resolves only one Rossby wave.
4
It worth mentioning that 2/3 of the windstress forcing in this equation results from the vorticity balance
and 1/3 from the potential energy balance, the latter arises from Ekman pumping acting on the back-
ground stratification.
3 The Circulation of the Southern Ocean 99

Fundamental Box 8 : Derivation of the Charney-DeVore model

The barotropic Charney-DeVore low-order model follows from the vorticity equation, supplemented by
the zonal momentum balance,
[ ]
∂ 2 𝑓0 𝑏
∇ 𝜓 + u ⋅ ∇ ∇2 𝜓 + 𝛽𝑦 + = 𝜖∇2 (𝜓 ∗ − 𝜓) (3.73)
∂𝑡 𝐻
∂𝑈 𝑓0 ∂𝜓
= 𝜖 (𝑈 ∗ − 𝑈 ) + <𝑏 > (3.74)
∂𝑡 𝐻 ∂𝑥

governing a barotropic zonally unbounded flow over a sinusoidal topography in a zonal channel. Here,
𝜓 is the streamfunction, u = (𝑢, 𝑣), 𝑢 = −𝜓𝑦 , 𝑣 = 𝜓𝑥 the velocity, 𝑈 is the zonally and meridionally
averaged zonal velocity and 𝜖∇2 𝜓 ∗ = −𝜖∂𝑈 ∗ /∂𝑦 is the vorticity and 𝜖𝑈 ∗ the zonal momentum imparted
(per unit of time) into the system, e.g. by thermal forcing or, in an oceanic application, by wind stress.
Furthermore, 𝜖 is a coefficient of linear bottom friction and 𝑓0 a constant Coriolis parameter. The last
term in the latter equation is the force exerted by the pressure on the bottom relief, the bottom formstress
(the cornered brackets denoted the average over the channel domain). The momentum input 𝜖𝑈 ∗ is thus
balanced by bottom friction and bottom formstress. 𝑈 ∗ is usually taken constant, i.e. there is no vorticity
forcing.
The depth of the fluid is = 𝐻 − 𝑏 and the topography height 𝑏 is taken sinusoidal, 𝑏 = 𝑏0 cos 𝑘𝑥 sin 𝑘𝑦
with 𝑘 = 2𝜋/𝐿 where 𝐿 is the length and 𝐿/2 the width of the channel. To derive the CDV system (3.76) a
heavily truncated expansion
1
𝜓 = −𝑈 𝑦 + [𝐴 cos 𝑘𝑥 + 𝐵 sin 𝑘𝑥] sin 𝑘𝑦 (3.75)
𝑘

is used, representing the flow in terms of the zonal mean 𝑈 and a wave component with cosine and
sine amplitudes 𝐴 and 𝐵. To derive the equation for 𝑈 , the representation is implement into the zonal
momentum balance and integrated over the whole channel domain. To derive the equation for 𝐴 [𝐵],
insert the representation into the vorticity equation, multiply by cos 𝑘𝑥 [sin 𝑘𝑥] and integrating over the
whole channel domain.

3.9.1. The barotropic Charney-DeVore low-order system

The barotropic Charney-DeVore (CDV) low-order system describes some aspects of zonal
flow (with current 𝑈 ) over sinusoidal topography of height 𝑏 = 𝑏0 cos 𝑘𝑥 sin 𝑘𝑦 with 𝑘 =
2𝜋/𝐿 where 𝐿 is the length and 𝐿/2 the width of the channel. It operates on a 𝛽-plane
and thus, a Rossby wave is generated with amplitudes 𝐴 (in phase with the topography,
i.e. a sine component) and 𝐵 (out of phase, i.e. a cosine component). The CDV system,
derived in Fundamental Box ??, is governed by

1
𝑈˙ = 𝜖 (𝑈 ∗ − 𝑈 ) + 𝛿𝐵
4
𝐴˙ = −𝑘𝐵 (𝑈 − 𝑐𝑅 ) − 𝜖𝐴 (3.76)
𝐵˙ = 𝑘𝐴 (𝑈 − 𝑐𝑅 ) − 𝛿𝑈 − 𝜖𝐵.

where 𝑐𝑅 = 𝛽/𝑘 2 is the speed of the planetary barotropic Rossby wave and 𝛿 = 𝑓0 𝑏0 /𝐻
measures the topography height5 . Each balance has a frictional term ∼ 𝜖 deriving from
5
The√solution of the linearized form of the CDV model (3.76) is an oscillation with the frequency
± (𝑘𝑐𝑅 )2 + 𝛿 2 /4, representing a mixed planetary-topographic wave.
100 Wave-induced bottom formstress

bottom friction. The zonal current is barotropic and forced by a stress, written for sim-
plicity as 𝜖𝑈 ∗ . The zonal balance in (3.76) is the time-dependent version of (3.21).

CDV model rel vort = 0 β=0


formstress and friction / ep [m]

200 2 200

150 1.5 150

100 1 100

50 0.5 50

0 0 0
0 20 40 60 80 0 20 40 60 80 0 20 40 60 80
U [m/s] U [m/s] U [m/s]

Figure 3.31.: Graphical display of the steady solutions of the CDV model showing ℱ𝑏 /𝜖 and 𝑈 ∗ −𝑈 versus
𝑈 . The parameter values are appropriate for a latitude of 55∘ S (𝑓0 = −1.1 × 10−4 s−1 ) and
𝜖 = 10−6 𝑠−1 , 𝐻 = 5, 000𝑚, 𝐿 = 10, 000𝑘𝑚. Various values for 𝑈 ∗ are used: the leftmost
line represents the momentum balance for typical ACC conditions (𝑈 ∗ = 20𝑚𝑠−1 , solid
black line), for the other lines 𝑈 ∗ is increased by a factor of 2, 4, 6 ⋅ ⋅ ⋅ (dashed blue lines),
respectively. The left panel applies to the complete CDV model. The middle panel is for a
version neglecting of the relative vorticity in (3.73), leading to a cancelation of the 𝐵𝑈 and
𝐴𝑈 terms in (3.76), the right panel for the 𝑓 -plane (𝛽 = 0). The formstress is plotted for
𝑏0 = 100𝑚 [red curve] and 𝑏0 = 200𝑚 [green curve] for the first two case and 𝑏0 = 500𝑚
[red curve] and 𝑏0 = 1500𝑚 [green curve] for the case 𝛽 = 0. Notice the difference in the
ℱ𝑏 /𝜖-axes.

CDV model rel vort 0 β 0


100 100 100

90 90 90

80 80 80

70 70 70

60 60 60
U U U
50 50 50

40 40 40

30 30 30

20 20 20

10 10 10

0 0 0
0 500 1000 1500 2000 0 1000 2000 3000 4000 5000 0 500 1000 1500 2000
b0 b0 b0

Figure 3.32.: The bifurcation of 𝑈 (𝑏0 ) for the cases shown in Figure 3.31, using 𝑈 ∗ =
20, 40, 60, 80, 100 ms−1 . Left: complete CDV model. Middle: model with cancelation of
the relative vorticity. Right: model for 𝛽 = 0. The black branches are stable, the red branches
are unstable. Notice the difference in the 𝑏0 -axes.

The wave induces a pressure field which acts against the zonal acceleration. At the
upstream side of the hills the fluid must be lifted, thus making high pressure, at the
downstream side a pressure low follows. The naturally westward propagating wave
gets stationary by eastward advection in the zonal current and friction: it is locked in
resonance with the mean flow and produces a formstress which becomes a nonlinear
functional of the zonal velocity,
3 The Circulation of the Southern Ocean 101

1 1 𝛿 2 𝜖𝑈
ℱ𝑏 [𝑈 ] = 𝛿𝐵 = − 2 (3.77)
4 4 𝜖 + 𝑘 2 (𝑈 − 𝑐𝑅 )2

derived from the two wave equations in (3.76). The total momentum balance in the set
(3.76), written in steady state now as

𝜖(𝑈 ∗ − 𝑈 ) + ℱ𝑏 [𝑈 ] = 0 (3.78)

then determines the zonal transport 𝑈 . Three equilibria are found if 𝑈 ∗ is well above
𝑐𝑅 but not too large, two are stable circulation regimes. For solutions in the resonant
range (𝑈 close to 𝑐𝑅 ) the friction in the momentum balance is negligible, these solutions
are balanced by formstress. The off-resonant solutions are controlled by friction. It is
remarkable that the formstress amplitude 𝐵 is proportional to 𝜖 (as shown in (3.77)): fric-
tion is essential to shift the pressure field out-of-phase with respect to the topography. Of
course, for zero friction the flow would not become steady. But is important to note that
friction plays a twofold role in the balance of the mean zonal flow: there is a direct fric-
tional effect on 𝑈 , manifested here by the bottom friction, and an indirect effect through
the feedback by the topographically induced waves where friction enables to build up
the phase shift and generate bottom formstress.
Significant sizes of the formstress can only arise if the topography is sufficiently high,
if Rossby waves propagation is present (i.e. 𝛽 ∕= 0), and if the forcing is sufficiently strong
(see left panels of Figure 3.31 and Figure 3.32). The three possible steady states which
then exist can be classified according to the size of the mean flow 𝑈 compared to the
wave amplitudes. The high zonal index regime is frictionally controlled, the flow is intense
and and the wave amplitude is low. The low zonal index regime is controlled by formstress,
the mean flow is weak and the wave is intense. The intermediate state is transitional, it is
unstable to perturbations. This ’formdrag instability’ works obviously when the slope of
the resonance curve is below the one associated with friction (here 𝜖 + ∂ℱ𝑏 /∂𝑈 > 0, see
Figure 3.31) so that a perturbation must run away from the steady state. Apparently, the
criterion of instability is always satisfied for the intermediate state whereas the other two
states are always stable.
Two ingredients are important to generate a significant resonance. It weakens when
the Rossby waves are filtered (which yields the 𝑓 -plane approximation used for the pre-
vious simple ACC models; in this case only topographic waves are present: right panels
of Figure 3.31 and Figure 3.32). If the relative vorticity is neglected, the resonance disap-
pears, and the momentum balance of this barotropic model is linear in 𝑈 (middle panels
of Figure 3.31 and Figure 3.32). Then multiple steady states do not exist. In these cases
the formstress is very small compared to the forcing and the flow is directly controlled
by the bottom friction: we find 𝑈 ∼ = 𝑈 ∗.
Charney and DeVore (1979) have developed this model for atmospheric flow regimes.
In the ocean the resonant solutions cannot be realized because flow speeds are much less
102 Wave-induced bottom formstress

than speeds of barotropic Rossby waves. The forcing 𝑈 ∗ relates to the windstress by
𝜖𝑈 ∗ = 𝜏0𝑥 /𝐻, and reasonable values for the windstress and the bottom friction allow only
for a frictionally controlled solution in which nonlinearities are irrelevant. The solution
then becomes

1 𝜏0𝑥 /𝜖
ℱ𝑏 [𝑈 ] = − 𝜖𝑈 (𝑏0 /𝐻)2 (𝑎𝑘)2 𝐻𝑈 = (3.79)
4 1 + 14 (𝑏0 /𝐻)2 (𝑎𝑘)2

where 𝑎 = ∣𝑓0 /𝛽∣ is the Earth radius times tangent of latitude and 𝜖2 ≪ (𝛽/𝑘)2 was
assumed for simplicity. The transport in this barotropic model decays away from the
frictional solution 𝐻𝑈 = 𝜏0𝑥 /𝜖𝐻 (with hundreds of Sv transport) with increasing height 𝛿
of the topography (see middle panel of Figure 3.32). The drag of the formstress increases
quadratically with the height of the topography and, because (𝑎𝑘)2 ≫ 1, attains higher
values than friction for moderately sized submarine ridges.
It is worth mentioning that much of its interesting dynamics of CDV – as the oc-
currence of multiple steady states – is lost by incorporation of more and more spectral
components arising for more complex topographies.

3.9.2. A baroclinic CDV low-order model

In a baroclinic extension the Charney-DeVore resonance can operate in realistic ACC con-
ditions. This will be discussed in the following.
A two-layer zonal channel with quasigeostrophic dynamics (see section 4.2) is con-
sidered, as before with a sine-shaped topography in the zonal direction. The stream-
functions of the flow in the two layers are expressed by suitably chosen sinusoidal struc-
ture functions. The resulting system is written in terms of 6 amplitudes representing the
barotropic transport 𝑇 (upper layer plus lower layer transport), the baroclinic transport
𝑆 (upper layer minus lower layer transport), and respective barotropic and baroclinic
sine and cosine components, 𝑇𝑠 , 𝑆𝑠 and 𝑇𝑐 , 𝑆𝑐 (more accurately, the 𝑇 and 𝑆 quantities
are based on barotropic and baroclinic velocities). The cosine variables generate the
barotropic and baroclinic formstress parts. The effect of transient eddies is parameter-
ized by the downward transfer of momentum (the interfacial formstress IFS, see section
??) via friction acting on the interface of the two layers with a coefficient 𝜅. In addition
we have lateral diffusion of momentum with a viscosity 𝜖.
We have scaled these variables6 and the dimensionless form of the governing differ-
ential equations becomes
6
The scaling and coefficients are as follows: All transport variables are scaled by 𝜋 2 /(∣𝑓0 ∣𝑌 2 ) to yield di-
mensionless 𝑇, 𝑆, 𝑇𝑐 , ⋅ ⋅ ⋅ . Parameters are 𝑏 = 𝑌 /(𝜋𝐿), 𝜖 = 𝜋 2 𝐴ℎ /(𝑏∣𝑓0 ∣𝑌 2 ), 𝜅 = 𝜋 2 𝐾/(𝑏∣𝑓0 ∣𝑌 2 ), 𝜆 =
√ ′ 𝛽2 = 2𝑌 𝑑𝑓 /𝑑𝑦/∣𝑓0 ∣ where 𝐿 is the zonal length and 𝑌 the width of the channel. 𝑅 =
𝜋𝑅/𝑌,
(𝑔 /𝑓0 )𝐻1 𝐻2 /𝐻 is the internal Rossby radius of a 2-layer fluid with densities 𝜌1 , 𝜌2 , mean layer
thicknesses 𝐻1 , 𝐻2 , 𝐻 = 𝐻1 + 𝐻2 and reduced gravity 𝑔 ′ = 𝑔(1 − 𝜌1 /𝜌2 ). Furthermore ℎ = 𝐻1 /𝐻
is a thickness ratio. Coupling coefficients are ℓ = 3𝜋 2 𝑏2 /8, 𝑚 = 64𝜋 2 /3, 𝑛 = 16/3. The scaled
depth is 1 + (𝛿/𝜋) sin(2𝜋𝑥/𝐿), thus 𝛿/𝜋 is the relative height of the topography. The forcing ampli-
tudes are 𝑊0 = (3/16)𝜋 2 𝜏0 /(𝑏𝐻𝑌 𝑓02 ), 𝑄0 = (3/32)𝜋 3 𝐵0 /(𝑏∣𝑓0 ∣3 𝑌 2 ) in terms of the zonal windstress
𝜏 = 𝜏0 sin2 (𝜋𝑦/𝑌 ) and the surface buoyancy flux 𝐵 = 𝐵0 cos(𝜋𝑦/𝑌 ) sin(𝜋𝑦/𝑌 ). Here, 𝜏 and 𝐵 are
dimensioned m2 s−2 and m2 s−3 , respectively.
3 The Circulation of the Southern Ocean 103

𝑇˙ = 𝑊0 + 𝛿 (𝑇𝑐 − ℎ𝑆𝑐 ) − 𝜖𝑇 (3.80)


𝜆 2 𝑊0 𝑄0 𝜆2
𝑆˙ = 𝛿 (𝑇𝑐 − ℎ𝑆𝑐 ) + ℓ [𝑇𝑐 𝑆𝑠 − 𝑇𝑠 𝑆𝑐 ] − 𝜅 + 𝜆2 𝜖 𝑆
( )
+ − (3.81)
ℎ 1−ℎ 1−ℎ
˙
𝑇𝑐 = 𝛽𝑇𝑠 − 2𝛿 (𝑇 − ℎ𝑆) − 𝑚 [𝑇 𝑇𝑠 + ℎ (1 − ℎ) 𝑆𝑆𝑠 ] − 𝜖𝑇𝑐 (3.82)
2𝜆2
𝑆˙𝑐 = 𝛽𝜆2 𝑆𝑠 + 𝛿 (𝑇 − ℎ𝑆) − 𝑛 [𝑇 𝑆𝑠 − 𝑆𝑇𝑠 ] − 𝜅 + 𝜆2 𝜖 𝑆𝑐
( )
(3.83)
1−ℎ
𝑇˙𝑠 = −𝛽𝑇𝑐 + 𝑚 [𝑇 𝑇𝑐 + ℎ (1 − ℎ) 𝑆𝑆𝑐 ] − 𝜖𝑇𝑠 (3.84)
𝑆˙𝑠 = −𝛽𝜆2 𝑆𝑐 + 𝑛 [𝑇 𝑆𝑐 − 𝑆𝑇𝑐 ] − 𝜅 + 𝜆2 𝜖 𝑆𝑠
( )
(3.85)

The forcing appears in 𝑊0 (windstress) and 𝑄0 (external buoyancy flux). Terms derived
from nonlinear advection are found in the cornered brackets (ℓ, 𝑚, 𝑛 are numerical cou-
pling coefficients depending only on the channel dimensions, see footnote). The respec-
tive term in the zonal baroclinic balance (3.81), the ℓ-term, is the interfacial formstress
induced by the standing eddies. The 𝜅-term in that equation is the corresponding tran-
sient eddy interfacial formstress. The baroclinicity enters via the internal Rossby radius
𝜆, and the scaled topography height is 𝛿. Lateral friction operates in the barotropic equa-
tions (for 𝑇, 𝑇𝑐 and 𝑇𝑠 ) and interfacial friction in addition in the baroclinic equations (for
𝑆, 𝑆𝑐 and 𝑆𝑠 ). The 𝑊0 -term in the baroclinic equation (3.81) arises from the Ekman pump-
ing acting on the background stratification. The sine and cosine equation (3.82) to (3.85)
describe planetary-topographic Rossby waves with the same zonal wavenumber as the
topography; 𝛽 is a scaled planetary coefficient 𝑑𝑓 /𝑑𝑦. There are terms (𝑛- and 𝑚-terms)
arising from nonlinearities (advection) and diffusion. Note that (3.80) is the balance of
vertically integrated momentum. We can identify the barotropic and baroclinic contribu-
tions to the bottom formstress, 𝛿𝑇𝑐 and 𝛿ℎ𝑆𝑐 , respectively.

𝜏0 = 10−4 m2 s−2 𝐴ℎ = 104 m2 s−1 𝑌 = 1000 km 𝐻1 = 1000 m


𝐵0 = 6.0 × 10−7 m3 s−3 𝐾ℎ = 800 m2 s−1 𝐿 = 4000 km 𝑑𝑓 /𝑑𝑦 = 2 × 10−11 m−1 s−1
𝑓0 = −1 × 10−4 s−1 𝐻2 = 3000 m 𝑅 = 20.8 km 𝑔 ′ = 5.7 × 10−3 ms−2
𝑊0 = 1.67 × 10−4 𝑄0 = 2.19 × 10−5 𝜖 = 1.24 × 10−2 𝜅 = 9.92 × 10−4
𝑏 = 7.96 × 10−2 𝜆 = 6.52 × 10−2 𝛽 = 0.40 ℎ = 0.25

Table 3.2.: List of parameters and standard values. The 𝐵0 value corresponds to a heat flux of 2.5 Wm−2 .
Furthermore, the coupling coefficients are ℓ = 3.70, 𝑚 = 1.33, 𝑛 = 5.33.

Considering the steady state of the above model we recover the most important phys-
ical mechanisms which we have outlined in the previous discussion of this chapter,
namely:

The barotropic formstress drag A barotropic state is obtained for 𝜆 = 𝑄0 = 0. Then


all 𝑆-fields are identically zero and a barotropic solution as described by (3.77) and (3.79)
emerges in which the transport decays quadratically with increasing topography height
due to the drag of the barotropic formstress.
104 Wave-induced bottom formstress

Johnson-Bryden dynamics If lateral friction is small and the influence of topography


and standing eddy IFS are ignored we regain from (3.81) the Johnson-Bryden state (??),

𝜅 𝑊0 𝑄0
2
𝑆 = + 2 (3.86)
𝜆 ℎ 𝜆 (1 − ℎ)

in which the transport contained in the shear current is governed by windstress and in
extension of the Johnson-Bryden model by buoyancy flux.

The compensation of barotropic and baroclinic formstresses For strong stratification


(large 𝜆) or vanishing nonlinearities (cornered bracket terms neglected, i.e. no advection)
we learn from the zonal mean baroclinic balance (3.81) that if the formstress terms 𝛿𝑇𝑐 and
𝛿ℎ𝑆𝑐 are individually increasing with topography height, they must compensate since the
remaining terms in the balance do not increase. The reason for the compensation of the
barotropic and baroclinic formstresses, discussed in detail in section ??, is thus found
in the compensation of the vertical lifting or lowering of the background stratification
(given by 𝜆2 ) by the pumping induced by the Ekman velocity and the barotropic and
baroclinic vertical velocities of the flow (the first and second terms on the rhs of (3.81)).
The latter two velocities are generated by the flow passing across the topography (they
are proportional to 𝛿). It is not clear yet, however, which of the two formstresses drives
and which brakes the zonal flow and under what circumstance they would increase with
the height of the barriers.

Breaking of ’Hidaka’s dilemma’ by stratification and eddies Adding (3.80) and (3.81)
in a suitable way to eliminate the formstress parts we find the balance of upper layer
transport 𝑇 + (1 − ℎ)𝑆, given by

𝑊0 𝑄0 1−ℎ
+ 2 − 𝜖 (𝑇 + (1 − ℎ)𝑆) = (𝜅𝑆 − ℓ [𝑇𝑐 𝑆𝑠 − 𝑇𝑠 𝑆𝑐 ]) (3.87)
ℎ 𝜆 𝜆2

and eliminating the wind forcing between (3.80) and (3.81), we get the balance

ℎ𝑄0 𝛿 ℎ
− + (𝑇𝑐 − ℎ𝑆𝑐 ) − 𝜖 (𝑇 − ℎ𝑆) = − 2 (𝜅𝑆 + ℓ [𝑇𝑐 𝑆𝑠 − 𝑇𝑠 𝑆𝑐 ]) (3.88)
(1 − ℎ)𝜆2 1 − ℎ 𝜆

for the lower layer transport 𝑇 − ℎ𝑆. On the right-hand side we find the IFS due to
transient and standing eddies.

Suppose for the moment that the IFS are small or the Rossby radius is large (strong
stratification). We then arrive at the physically intuitive statement that the transport in
the surface layer, 𝑇 + (1 − ℎ)𝑆, is decoupled from the topography: it is given by the
windstress and buoyancy flux acting in the above combination against lateral friction.
The surface layer transport is then in a ’Hidaka’-type state (inversely proportional to the
3 The Circulation of the Southern Ocean 105

eddy viscosity 𝜖 and thus large for a reasonably sized 𝜖). The transport of the lower layer,
𝑇 − ℎ𝑆, is governed by the magnitude of the bottom formstress and external heating 𝑄0 .
This Hidaka dilemma must be resolved by weak stratification and large IFS from tran-
sient or/and standing eddies. Notice that the crucial term [𝑇𝑐 𝑆𝑠 − 𝑇𝑠 𝑆𝑐 ] of the standing
eddies is nullified for the flat bottom wave solution of the equations (3.82) to (3.85). It is,
however, supported by linear topographic waves. The wave equations yield after some
manipulations

( / )2 𝜅 (𝑇 − ℎ𝑆)2
𝑇𝑐 𝑆𝑠 − 𝑇𝑠 𝑆𝑐 = −4 𝛿 𝛽 (3.89)
𝛽𝜆2 ((𝜖/𝛽)2 + 1)((𝜗/𝛽)2 + 1)

The standing wave IFS is – in this approximation – negative if the topography is undu-
lated and the bottom flow non-vanishing, it must transfer eastward momentum down-
ward. It is worth noticing that it depends marginally on the viscosity 𝜖 but is entirely
supported by the interfacial friction 𝜅.

The bottom formstress terms 𝛿𝑇𝑐 and 𝛿ℎ𝑆𝑐 arise by the response of the barotropic and
baroclinic wave system, described by the four equations (3.82) to (3.85), to the zonal flow
crossing the topography (there is no direct forcing because the external heating function
was assumed strictly zonal). A reasonable solution of the model should yield transports
in the range 𝛽 ≫ 𝑇, 𝑆 ≳ 𝛽𝜆2 , meaning that the flow velocities are much less than the
speeds of barotropic Rossby waves but supercritical with respect to the baroclinic waves,
as in realistic ACC conditions. Then the nonlinearities in (3.82) and (3.84) are small and
these equations represent a linear barotropic planetary-topographic wave. The barotropic
formstress 𝛿𝑇𝑐 can thus be explained by long linear Rossby waves generated by the deep
current, 𝑇 − ℎ𝑆 in the above model, crossing the large ridges blocking the circumpolar
path of the ACC. In contrast, the baroclinic formstress 𝛿ℎ𝑆𝑐 might be governed by a non-
linear response, according to 𝑇, 𝑆 ≳ 𝛽𝜆2 in (3.83) and (3.85). We discuss the consequences
further below.

A linear model Still it is worth considering the linearized version of the above model.
We thus neglect the standing eddy IFS in (3.81) and the advection terms in the wave
equations. The formstress terms become

( / )2 𝜖(𝑇 − ℎ𝑆) ( / )2 𝜗(𝑇 − ℎ𝑆)


𝛿𝑇𝑐 = −2 𝛿 𝛽 and 𝛿ℎ𝑆𝑐 = 2 𝛿 𝛽 ℎ̃ (3.90)
(𝜖/𝛽)2 + 1 (𝜗/𝛽)2 + 1

with 𝜗 = 𝜅/𝜆2 + 𝜖, ℎ̃ = ℎ/(1 − ℎ). The barotropic formstress is thus supported by lateral
friction, the baroclinic by interfacial friction and stratification. However, both extract
eastward momentum from the flow if the deep transport is eastward: the linear wave
model thus does not imply the afore mentioned formstress compensation effect. The
total transport 𝑇 and the shear transport 𝑆 are readily evaluated as lengthly expressions
of 𝛿, 𝜖, 𝜅 and 𝜆,
106 Wave-induced bottom formstress

𝑇 = 𝒜 𝑊0 + ℬ (𝑊0 + ℎ̃𝑄0 /𝜆2 ) ℎ𝑆 = 𝒞 𝑊0 + 𝒟 (𝑊0 + ℎ̃𝑄0 /𝜆2 ) (3.91)

where the direct wind effect and the baroclinic forcing have been separated. The factors
attached to the forcing amplitudes are due to friction and formstress by topographically
induced Rossby waves. These transport response functions are

⎛ ⎞ ⎛
( / )2 ⎞
𝒜 𝜗 + 2 𝛿 𝛽 ℎ̃𝜂
⎜ 2 (𝛿 /𝛽 )2 𝜂 ⎟
⎜ ⎟ ⎜ ⎟
⎜ℬ ⎟ 1
⎜ ⎟ = (3.92)
⎜ ⎟ ⎜ ⎟
( / )2 ⎜ ( / )2 ⎟
⎜𝒞 ⎟
⎝ 2 𝛿 (𝛽/ ℎ̃𝜂
𝜖𝜗 + 2 𝛿 𝛽 𝜂(𝜗 + ℎ̃𝜖) ⎜ ⎟
⎝ ⎠ )2 ⎠
𝒟 𝜖+2 𝛿 𝛽 𝜂

with 𝜂 = 𝜖/(1+(𝜖/𝛽)2 )+ ℎ̃𝜗/(1+(𝜗/𝛽)2 ). There are three sources of transport according to


the relation (3.91): 𝑇 and ℎ𝑆 are generated by direct wind forcing, indirect wind forcing
via Ekman pumping on the mean stratification, and buoyancy flux. For strictly barotropic
conditions where 𝜆 → 0 we get 𝜗 → ∞ and the Charney-DeVore result (3.79) is regained.
The response functions 𝒜, ℬ, 𝒞 and 𝒟 are displayed in Figure 3.33 as function of the
height and otherwise typical parameters. All functions flatten out to a plateau at large
topography heights but have different critical heights above which this happens. The
critical height of the barotropic response is clearly 𝛿 ∼ 𝛽 (see also the Charney-DeVore
model (3.79)). The baroclinic response scale depends very much on the sizes of 𝜖, 𝜅 and 𝜆.
The figure also elucidates the shares of the different transport sources for typical 𝑊0 and
𝑄0 in the total and the shear transports. It becomes clear that only below quite moderate
heights of the topography we find the direct wind effect dominating. At larger heights
and for the present parameters the Ekman pumping acting on the stratification is the

2
response functions −2
total transport −3
shear transport
10 10 10

1 −3
10 10
−4
10
0 −4
10 10
−5
10
−1 −5
10 10

−2 −6 −6
10 10 10
0 0 05 01 0 15 02 0 0 05 01 0 15 02 0 0 05 01 0 15 02
δ/π δ/π δ/π

Figure 3.33.: Left: height dependence of the response functions 𝒜 [full], ℬ [dashed], 𝒞 [dashdotted] and
𝒟 [dotted] of the linear transport model. Middle: the contributions to the total transport 𝑇 of
direct wind forcing [full], Ekman pumping [dashed] and heating [dashdotted]. Right: same
for shear transport ℎ𝑆. Parameters are: 𝑊0 corresponds to 0.1 Nm−2 , 𝑄0 to 10 Wm−2 and
𝜆 to a Rossby radius of 10 km, furthermore ℎ = 0.75, 𝐴ℎ = 104 m2 s−1 , 𝐾 = 500 m2 s−1 .
3 The Circulation of the Southern Ocean 107

most important driving agent of transport. Only if ℎ̃𝑄0 /𝜆2 becomes of the order of 𝑊0
the buoyancy forcing is getting effective for the transport.

Finally we consider a nonlinear extension of the above model. We want to improve on


the baroclinic bottom formstress and thus still neglect the standing eddy IFS. Including
now all terms in the baroclinic wave equations (3.83) and (3.85) the baroclinic bottom
formstress becomes

( )
𝜗 ( / )2 / 𝑛ℎ𝑆
𝛿ℎ𝑆𝑐 = 2 𝛿 𝛽 ℎ̃(𝑇 − ℎ𝑆) + (𝛿 𝛽)𝑇𝑐 (3.93)
(𝜗/𝛽)2 + 1 − (𝑛𝑇 /𝛽𝜆2 )2 𝜖𝜆2

and it is evident that supercritical conditions, 𝑇 > 𝛽𝜆2 , can indeed lead to a negative
baroclinic bottom formstress which then drives the current eastward if the bottom flow
𝑇 −ℎ𝑆 is eastward and the last term on the rhs is small – actually requiring a large viscos-
ity 𝜖. We conclude that the supercriticality of the ACC with respect to baroclinic Rossby
waves is essential for the system to achieve the observed balance of zonal momentum in
which wind and baroclinic bottom formstress drive and barotropic formstress decelerates
the current.

You might also like