1
The impact of open oceanic processes on the Antarctic
2
Bottom Water outflows
3
4
Shinichiro Kida1
5
Earth Simulator Center, Japan Agency for Marine-Earth Science and
6
Technology, Yokohama, Japan
7
8
(Resubmitted to the Journal of Physical Oceanography, April 2011)
9
10
11
12
13
14
15
16
17
18
19
____________________
20
Corresponding author address: Shinichiro Kida, Earth Simulator Center, Japan Agency for Marine-
21
Earth Science and Technology, 3173-25 Showa-machi, Kanazawa-ku, Yokohama 236-0001 Japan.
22
E-mail: [email protected]
23
1
24
Abstract
25
The impact of open oceanic processes on the Antarctic Bottom Water (AABW) outflows is
26
investigated using a numerical model with a focus on outflows that occur through deep
27
channels. A major branch of the AABW outflow is known to occur as an overflow from the
28
Filchner Depression to the Weddell Sea through a deep channel few hundred kilometers
29
wide and a sill roughly 500 m deep. When this overflow enters the Weddell Sea, it
30
encounters the Antarctic Slope Front (ASF) at the shelf-break, a density front commonly
31
found along the Antarctic continental shelf-break. The presence of an AABW outflow and
32
the ASF create a v-shaped isopycnal structure across the shelf-break, indicating an
33
interaction between the overflow and oceanic processes. Model experiments show the
34
overflow transport to increase significantly when an oceanic wind stress increases the depth
35
of the ASF. This enhancement of overflow transport occurs because the overflow transport
36
is geostrophically controlled by its ambient oceanic water at the shelf-break and the
37
presence of the channel walls allows a pressure gradient in the along-slope direction to
38
exist. Since the ASF is associated with a lighter water mass that reaches the depth close to
39
that of the channel, an increase in its depth increases the density gradient across the shelf-
40
break and therefore the geostrophic overflow transport. The enhancement of overflow
41
transport is also likely to result in lighter overflow water mass although such adjustment of
42
density likely occurs on much longer time scale than the adjustment of transport.
43
2
44
1. Overflows and its interaction with the open ocean
45
Overflows originate from marginal seas where topographic constraints enable the formation
46
of dense water mass through excess evaporation, ice formation, and heat-loss (Warren
47
1980). While the transport of each overflow is on the order of few Sverdrups (Sv),
48
overflows play a significant role on the deep oceanic circulation (Price and Yang, 1998).
49
Those from the Denmark Strait, Faroe Bank Channel, Kara Sea, Mediterranean Sea, Red
50
Sea, and Filchner Bank Channel are some of the overflows known to play such role.
51
Overflows enter the open ocean through sills and channels and then flow along the
52
isobaths of the continental slope as bottom friction (e.g. Smith, 1975) and topographic
53
features like canyons and seamounts (e.g. Wahlin, 2002) cause them to descend.
54
Thermobaricity (Killworth, 1977, Gordon et al. 1993) and baroclinic instability (e.g. Jiang
55
and Garwood, 1996, Tanaka and Akitomo, 2001, Kida et al., 2009) may further enhance the
56
descent. Overflows also entrain upper oceanic water locally near the shelf-break (Price and
57
Baringer, 1994) and induce an upper oceanic circulation (Kida et al. 2008). These past
58
studies of overflows on the continental slope have led to major improvements on its
59
parameterization in General Circulation Models (GCMs) (see Legg et al. 2010 and
60
reference there in).
61
62
a. Can open oceanic processes affect overflows?
63
Do open oceanic processes affect overflows? Overflows enter the open ocean through deep
64
channels but previous studies on the overflow-upper ocean interaction have mostly focused
65
on this dynamics when the overflow is on the slope. The ambient open ocean has often
3
66
been kept quiescent or without externally driven flows. The real open ocean is, however,
67
comprised of flows forced by external processes. Coastal currents and the oceanic
68
stratification are known to affect dense water cascades by changing its density and cross-
69
slope penetration distance (e.g. Chapman, 2000, Gawarkiewicz, 2000, Ivanov and Golovin,
70
2007). Dense water cascades are another form of dense water outflows which involves the
71
dense water entering the open ocean from the continental shelves where the bathymetry is
72
less bounded by sills and channels (see Ivanov et al., 2004, for more complete list of dense
73
water cascades). Dense water cascades are often directly affected by open oceanic flows but
74
overflows originate from marginal seas where open oceanic flows do not directly affect.
75
Thus while overflows are likely influenced by open oceanic flows once it is on the
76
continental slope, whether their initial properties at the channel are influenced by open
77
oceanic processes is still unclear. Studies on the overflow-upper ocean interaction have
78
mostly prescribed this initial property of overflows at the channel when the open oceanic
79
conditions are changed (e.g. Jiang and Garwood, 1998).
80
Our motivation for answering the question raised above comes from the
81
observations of v-shaped temperature and salinity fields along the continental slope of the
82
Weddell and Ross Seas (Figure 1a, b) (Gill, 1973, Jacobs, 1991, Fahrbach et al., 1992). The
83
in-shore side of the v-shape is associated with the outflow of the cold and salty Antarctic
84
Bottom Water (AABW, Orsi et al. 1999) from the shelf. The offshore side of the v-shape is
85
associated with the Antarctic Slope Front (ASF), which is a front that separates the warm
86
and salty Circumpolar Deep Water and the cold fresh surface water on the shelf. The v-
87
shaped structure suggests direct interaction between the AABW outflow and the open
4
88
ocean at the continental shelf-break (Gill, 1973). Although the presence of this v-shaped
89
structure has been known for some time, further details on how the two processes may
90
interact have remained an open question.
91
The ASF is observed almost continuously along the continental shelf-break of
92
Antarctica except for the region just west of Antarctic Peninsula (Whitworth et al. 1998,
93
Heywood et al., 2004). It is located well off-shore where the continental shelf is wide (Gill,
94
1973) but is located close to the coastline where the continental shelf is narrow. It is also
95
co-located with the Antarctic Coastal Current (Whitworth et al., 1998, Bindoff et al., 2000),
96
a westward flow on the continental shelf often found along the front of ice shelf (Jacobs,
97
1991). The ASF/Antarctic Coastal Current system is observed with a transport of about 14
98
Sv at 17W and the ASF is considered to account about half of this transport (Heywood et al.,
99
1998). A maximum velocity of about 10 cm s-1 is observed near the surface. The
100
geostrophic shear associated with the v-shape of the ASF suggests a river like flow at the
101
shelf-break even away from the coast (Gill, 1973, Whitworth et al., 1998). The slope
102
current in the Ross Sea that is associated with the v-shape at the shelf-break is observed
103
with a flow speed of more than few cm s-1. A slope front and an along-slope flow are
104
observed along the shelf-break of the western Weddell Sea (Muench and Gordon, 1995). It
105
is suggestive that a slope current with a flow speed of few cm s-1 or more exists along the
106
shelf-break of the Weddell Sea away from the coast. Baines (2009) suggests that mixing
107
between the overflow and the surface water may establish the ASF. The easterly wind that
108
is observed along the coast of Antarctica may also establish the ASF. The easterly wind has
109
traditionally been considered responsible for establishing the ASF that is observed along
5
110
the coastlines because an easterly wind will force a southward Ekman transport, accumulate
111
surface water near the coast, and establish a coastal front. Seasonal variability of the flow
112
speed in the ASF/Antarctic Coastal Current system observed along the coast near 11W is
113
found to correlate well with that of the easterly wind (Figure 1c) (Fahrbach et al. 1992).
114
The ASF along shelf-breaks away from coastlines may establish through the advection of
115
the coastal front along the isobaths of the shelf-break. The ASF/Antarctic Coastal Current
116
system is observed to divide into that along the coast and that along the slope where the
117
shelf widens about 27 W, east of the Filchner Bank Channel, although the transport of each
118
branch remains unclear (Foster and Carmack, 1976). The two mechanisms mentioned here
119
for the establishment of the ASF may very well co-exist and suggest that the ASF is
120
capable of being affected by open oceanic processes.
121
Are AABW outflows affected by oceanic processes through their impact on the ASF?
122
This is the main question investigated in this paper. We will focus on the AABW outflow
123
that occurs through a deep channel, such as that observed from the Filchner depression
124
through the Filchner Bank Channel to the Weddell Sea (Figure 1a), which we will refer to
125
as the AABW overflow hereafter. We will also focus on the initial overflow properties that
126
enter the oceanic basin from the channel rather than that on the continental slope. Note that
127
our focus is not on the dense water cascades like that occurring on the western part of the
128
continental shelves of the Weddell Sea. Overflow properties have been suggested to vary
129
when the surface heat flux or the surface buoyancy change, since they affect the water mass
130
transformation rate. Changes in the local wind field have also been suggested to affect
131
overflows since they may change the barotropic flow field along a channel (Kohl et al.
6
132
2007). Unlike these past studies, this study will focus on the impact of a density front along
133
a shelf-break that is varied by the wind field out in the open ocean.
134
Partly due to difficulties associated with long-term direct observations around Antarctic
135
in the presence of strong tides and sea ice, observations have been limited. Gordon et al.
136
(2009) recently showed that the AABW overflow from the Drygalski trough in the Ross
137
Sea is associated with significant seasonal variability since more pulses of cold dense water
138
events were observed from Spring to Fall (November to May) than from Fall to Spring
139
(Figure 2). This seasonal variability may be induced by the seasonally varying wind field
140
observed in the open ocean which appears to correlate well (Figure 1c): the number of
141
pulses increases as the easterly wind increases towards summer. Such seasonal variability
142
of the easterly wind is found around the shelf-breaks of Antarctica. However, how these
143
varying winds affect the slope current is unclear and the seasonal variability of the slope
144
current may very well be different from that found in the coastal current at 11W. The
145
mechanism responsible for the seasonal variability of the AABW overflow is still an open
146
question and the seasonal variability of the water mass transformation rate is another likely
147
candidate. Current observations only suggest the possible influence that a wind field out in
148
the open ocean may have on the AABW overflow.
149
150
b. A mechanism for the oceanic influence on overflows
151
A mechanism showing how changes in the ASF may influence the AABW outflow was
152
first proposed by Tanaka and Akitomo (2000). They hypothesized that the slope current
153
associated with the ASF may increase the AABW outflow transport by enhancing its
7
154
bottom Ekman transport. This is based on the assumption that the overflow is two-
155
dimensional (constant along-slope properties) and steady, which therefore is more
156
appropriate for dense water cascades. Nonetheless, we will assume that it is still applicable
157
for the AABW overflows. The overflow transport at the shelf-break, Mo, is estimated by the
158
bottom Ekman transport, MEK, integrated in the across-channel direction:
159
M O  M EK 
160
where Lx and x are the width of the overflow and the bottom stress in the cross-channel
161
direction respectively. CD is the quadratic bottom drag coefficient. |U| is the flow speed,
162
U  u 2  v 2 , and u and v are the velocities in the cross-channel and along-channel
163
directions. Eq (1) indicates that the presence of a slope current increases the overflow
164
transport Mo at the shelf-break because its background presence will increase the flow
165
speed of overflows. The slope current is likely westward, which is the same direction as
166
that of the AABW overflow. Based on a numerical model, Tanaka and Akitomo (2000)
167
estimate that more than half of the AABW overflow transport may be caused by this
168
enhancement of bottom Ekman transport by the slope current.
CD U u
x
Lx 
Lx ,
f o
fo
(1)
169
The limitation of Eq (1) is the assumption of a two-dimensional along-slope flow.
170
For AABW overflows that occur through deep channels rather than uniformly along the
171
continental shelf-break such assumption may be inadequate. The Filchner Bank Channel
172
overflow is such case (Foldvik et al. 2004). In the presence of channels, the AABW
173
overflow may occur geostrophically because a pressure gradient can be maintained across
174
the channel. We suggest that the ASF will enhance the AABW overflow through the
8
175
enhancement of this geostrophic flow component. When assuming that the overflow
176
transport is mostly geostrophic and that the upper oceanic flow is absent, the overflow
177
transport Mo can be estimated by that suggested by Toulany and Garett (1984) and
178
generalized by Helfrich et al. (1999)
179
g' h2
MO 
,
2 fo
180
where g’ is the reduced gravity between the overflow and the open oceanic water, h is the
181
upstream height of overflow water above the sill. Eq (2) is the steady state solution of the
182
along-channel transport for the Rossby adjustment problem in a wide channel setting for a
183
one layer flow. Unlike the original solution (Helfrich et al. 1999), the reduced gravity is
184
used since we are currently focusing on a baroclinic flow. The Rossby adjustment problem
185
is a dam break problem between two density layers which we will consider as the overflow
186
layer and the open oceanic layer (Figure 3a). The familiar solution that establishes a flow
187
along the density front is based on a two-dimensional assumption which is equivalent to
188
having an infinite wide channel (Cushman-Roisin 1994). For a finite channel much larger
189
than the deformation radius, the lateral boundaries will create a flow in the along channel
190
direction and Eq (2) is such solution. We suspect that this physical setting resembles the
191
AABW overflow. The deep channel between the Filchner Depression and the Weddell Sea
192
is roughly 100-200 km wide (Foldvik et al. 2004) which is significantly larger than the
193
local deformation radius of about 5 km.
(2)
194
Eq (2) indicates that the overflow transport increases when the depth of the ASF
195
increases. When the ASF is weak (Figure 3a), g’ can be estimated by that between the
9
196
overflow and the deep water in the open ocean, (g, where  is the background
197
density, 1025 kg m-3. However, when the ASF is deepened by external processes, lighter
198
surface water is brought down to deeper depth and g’ becomes (g. Since 
199
(Figure 3b), this is an increase in g’ which may increase the overflow transport. We suspect
200
such mechanism possible because the flow speed of the AABW overflow is below the
201
gravity wave speed and is subcritical near the shelf-break. Once the overflow is on the
202
continental slope, the flow accelerates and becomes supercritical so that it is unlikely to
203
influence the dynamics upstream.
204
In this paper, we will investigate whether the AABW overflows are affected by
205
open oceanic processes through the changes in the ASF, with a focus on the Filchner Bank
206
Channel overflow. A numerical model is used and the paper is organized as follows. The
207
details of the model configuration are first described in Section 2. The dynamics of the
208
AABW overflow at the shelf-break is examined in Section 3. The sensitivity of the AABW
209
overflow to the changes in the ASF is examined in Section 4. Summary and discussion are
210
presented in Section 5.
211
212
2. A Shelf – Oceanic Basin model
213
A non-hydrostatic ocean model developed at the Earth Simulator Center of Japan Agency
214
for Marine-Earth Science and Technology (JAMSTEC) is used. This model is named the
215
Multi-scale Simulator for the Geoenvironment (MSSG) (Takahashi et al. 2006, Baba et al.
216
2010) and is an air-sea coupled model but here we only use its oceanic component. MSSG
10
217
is based on a C-grid using the Latitude-Longitude system for the horizontal coordinate and
218
the z-coordinate for the vertical.
219
220
a. Experimental Setups
221
The model is on an f-plane with f set to -1x10-4 s-1. The model domain has 200 grid points
222
zonally and 300 grid points meridionally with a horizontal resolution of 1 km (Figure 4).
223
Vertical resolution is 10 m and the rigid-lid approximation is used at the sea surface.
224
The bathymetry of the model (Figure 4a, b, and c) is set analogous to where the
225
Filchner Bank Channel overflow occurs (Figure 1a). The circular basin in the north, with a
226
maximum depth of 1000 m, represents the Weddell Sea, which we will refer to as the
227
oceanic basin hereafter. A continental slope, with a slope of 0.01, exists along the perimeter
228
of this oceanic basin. The squared basin in the south represents the Filchner depression with
229
a maximum depth of 700 m, which we will refer to as the shelf hereafter. A sill, 500 m high,
230
and a channel, 100 km wide and 50 km long, that connect the oceanic basin and the shelf
231
represents the Filchner Bank Channel. The channel walls are bounded from the bottom to
232
the surface unlike the actual Filchner Bank Channel where it is about 300 m deep, but we
233
have chosen to do so in order to simplify the location of the exchange flow. The model
234
domain is much smaller than the actual Weddell Sea and its continental shelf but the
235
Filchner Bank Channel is well represented and we consider it feasible for examining the
236
basic interaction between the overflow and oceanic processes.
237
Laplacian lateral viscosity and diffusion are used for the sub-grid scale viscosity and
238
diffusion with the coefficients set constant to Kh = 20 m2s-1 and Ah = 20 m2s-1, respectively.
11
239
The slip boundary condition is used for lateral boundaries. Vertical viscosity (Kv) and
240
diffusivity (Av) coefficients are constant in the interior but increases exponentially near the
241
surface with a length scale of 100 m and near the bottom with a length scale of 50 m to
242
represent the surface and the bottom boundary layers:
243
Kv  Av  1  10 5  2  10  2 exp(
244
where zs is the depth from the sea surface (z=0) and zb is the height from the bottom. This
245
idealized vertical mixing parameterization allows the overflow to simulate entrainment of
246
oceanic water in the interior while mixing at the bottom and surface boundaries are
247
parameterized. Although an order higher resolution is required to fully resolve the Kelvin-
248
Helmholtz instability, Legg et al. (2010) has shown that some of the entrainment process is
249
permitted with similar resolution using a non-hydrostatic oceanic model. The thickness of
250
the surface enhancement in Eq. (3) is chosen to roughly match with the mixed layer of the
251
initial temperature profile. The thickness of the bottom enhancement is based on
252
observations showing the bottom boundary layer of roughly 50 m at the shelf-break of
253
Antarctica near the Adelie Depression (Hirano et al. 2010). A maximum value of 2x10-2 m2
254
s-1 is chosen close to that observed near the surface and the bottom and to ensure that the
255
model is reasonably resolving the bottom Ekman layer ( 2KV f ), which is about 35 m.
256
Quadratic drag is used for the bottom stress with a drag coefficient, CD, of 2x10-3.
2
2
zs
z
 b2),
2
100
50
(3)
257
The flow field is initially set quiescent. The temperature field is horizontally
258
uniform and decreases exponentially from the sea surface to the bottom with a length scale
259
of 200 m:
12
2
zs
)
200 2 .
260
T ( z )  2.0 exp(
261
Salinity is set constant to 34.5 psu. These idealized temperature and salinity vertical
262
profiles have the density field close to that observed around the Weddell Sea.
(4)
263
264
b. The dense overflow
265
The source water of the AABW overflow is created by cooling the water near the surface
266
within 20 km from the southern boundary of the shelf:
267
Q  Qo exp(
268
where Qo is set to 2x10-5 oC s-1 and y is the distance from the southern boundary in
269
kilometers. Ice does not form in the model and a minimum temperature is set to -1.8oC. The
270
magnitude of the prescribed surface cooling is significantly larger than observations
271
(Tamura et al. 2008) locally but its area-integrated value is similar. To avoid the formation
272
of fast descending cold chimney plumes that requires extremely small time-step to resolve,
273
lateral and vertical viscosity/diffusivity coefficients are enhanced to 40 m2s-1 and 1x10-2
274
m2s-1, respectively, where cooling is applied.
2
zs
y2 ,
)

exp(

)
50 2
20 2
275
The only connection between the shelf and the oceanic basin is the channel and so
276
the simulated dense overflow will be inevitably associated with a surface flow from the
277
oceanic basin to the shelf. The flow at the channel is therefore a two-layer exchange flow
278
and not a one-layer outflow. A surface inflow is not observed directly above the AABW
279
overflow at the Filchner Bank Channel so this is a major difference between the simulated
280
overflow and observations. The impact of having a surface inflow directly above an
13
281
overflow will be discussed later. The exact pathway of the surface inflow that eventually
282
becomes the Filchner Bank Channel overflow is not fully understood yet. The Antarctic
283
Coastal Current is observed to intrude onto the shelf along the eastern boundary of the
284
continental shelf (Foster and Carmack, 1976) but such coastal current is not included in this
285
study in order to focus on the impact of oceanic processes through the slope front.
286
287
c. The wind stress and the formation of a shelf front
288
A slope front is created along the coastlines of the oceanic basin by applying a cyclonic
289
wind stress over the northeast part the oceanic basin. The wind stress magnitude increases
290
sinusoidally from the interior towards the coast (Figure 5a):
291
r y  200

(r  70)
 o sin(140 ) 
r
x  
y  200

o 
(r  70)
r
,

r x  100

(r  70)
 o sin(140 )  r
y 
x  100

(r  70)
o 

r
x  1002  ( y  200) 2 ).
292
where r is the distance from the center of the oceanic basin ( r 
293
The wind stress magnitude is kept constant over the continental slope so that Ekman
294
upwelling is not forced there.  o is set to -0.10 N m-2, a representative value of the easterly
295
wind observed along the coasts of Antarctica and that observed in NCEP reanalysis
296
(Fahrbach et al. 1992). The cyclonic wind stress forces coastal downwelling, tilts the
297
isotherm, and establishes a coastal front. Such coastal front is associated with a cyclonic
298
flow which will advect the coastal front along the shelf-break and establish a slope front.
14
299
Note that this cyclonic flow does not directly enter the shelf since its geostrophic contours
300
are along isobaths and thus in the cross-channel direction. The wind stress is also set zero
301
near the channel so that the wind will not directly force an along-channel flow there. This
302
condition is similar to that in the Weddell Sea where sea ice exists above the shelf-break
303
most time of the year and prevents direct wind forcing.
304
The temperature field in the interior of the oceanic basin is restored to its initial
305
profile to maintain the basic stratification near the center of the oceanic basin. The restoring
306
time-scale is 2 days at the center but increases away from the center (Figure 5b). Restoring
307
is moderately enhanced in the northwestern part of the oceanic basin to prevent the slope
308
front and the overflow water from circulating around the oceanic basin multiple times.
309
310
3. The model flow field and the overflow through the channel
311
The simulated flow field reaches a quasi steady state in about 100 days (Figure 6). A slope
312
front establishes in the oceanic basin (Figure 7a) and is associated with a westward flow
313
along the coast and the shelf-break (Figure 7c). The dense overflow enters the oceanic
314
basin from the channel and flows cyclonically along the continental slope (Figure 7b). This
315
model experiment will be referred to as the control experiment (CTRL) hereafter and the
316
dynamics of its simulated overflow between 120 and 180 days is examined in this section.
317
318
a. The flow field
319
The wind, forced at the northeastern part of the oceanic basin, establishes a coastal front
320
along the coast of the oceanic basin. Coastal downwelling tilts the depth of the thermocline
15
321
above the continental slope roughly within 30 km from the coast. The cyclonic circulation
322
that establishes in the interior of the oceanic basin is forced by Ekman upwelling. The
323
cyclonic current that is associated with the coastal front has a transport of about 3.4 Sv
324
above the slope at y = 200 km. This cyclonic current is largely barotropic with a velocity of
325
about 10 cm s-1, advects the front along isobaths, and establishes a slope front at the shelf-
326
break of the channel. As a result, the isotherms form a v-shape structure at the shelf-break
327
(Figure 7c). The establishment of this v-shape structure suggests that the flow field of
328
CTRL is reasonably close to that observed near the Filchner Bank Channel.
329
The cyclonic circulation in the shelf region is induced by the surface cooling applied
330
along the southern boundary. It is observed with a maximum velocity of about 20 cm s-1
331
near the surface. The cold dense water occupies the western half of the shelf near the
332
surface but occupies the rest of the shelf below.
333
334
b. The transport of the overflow at the strait
335
The overflow enters the oceanic basin through the channel with a time-averaged transport
336
of about 0.36 Sv and velocity weighted temperature of about -1.3 oC. The overflow water is
337
assumed that below 0.0 oC, which is the value of the thermocline that divides the overflow
338
and the oceanic water at the strait. The overflow transport and temperature are estimated at
339
the shelf-break (y = 100 km). The time-average is that taken between 120-180 days of
340
simulation. These definitions of overflow transport and temperature and time average are
341
used throughout this study. g’ is estimated as 8.5x10-4 m2 s-1 when taking the time-averaged
342
density difference between the overflow and upper oceanic water that exist within 20 km
343
from the strait and above the channel depth (500 m). The time averaged thickness of the
16
344
overflow h is about 200 m at the shelf-break. So Eq (2) estimates the magnitude of the
345
overflow transport as 0.17 Sv. This estimate is somewhat smaller but reasonably close to
346
the model result. Time averaged flow component ( u  h ) contributes to most of the total
347
transport and the eddy component ( u ' h' ) is negligible. The observed perturbation at the
348
shelf-break appears not to contribute much. The bottom Ekman transport across the shelf-
349
break (y = 100 km) is 0.03 Sv and represents only 8% of the total transport. The overflow
350
transport is therefore likely to be mostly geostrophically controlled.
351
The bottom flow field within the channel shows that the overflow begins to flow
352
along the eastern wall from the shelf, turns westward (across the channel) near the shelf-
353
break, and then enters the oceanic basin along the western wall (Figure 7b). This pathway
354
of the overflow is consistent with the past studies on overflows over a sill (e.g. Helfrich and
355
Pratt, 2010). The time-averaged flow speed in the along-channel direction is strongest near
356
the western and eastern walls, both for the overflow and the oceanic inflow (Figure 7d).
357
The on-shore flow observed at about x = 120 km reflects the flow along the shelf-break
358
which is slightly anticyclonic because the oceanic basin is circular. The observed flow
359
pattern is strongly associated with the tilt of isotherms across the channel, suggesting the
360
importance of the geostrophy. The time variability observed at the strait is weak and has a
361
frequency of about 4-5 days.
362
Once the overflow enters the oceanic basin, it flows cyclonically along the
363
continental slope and descends. A downward sloping of isopycnals is observed within the
364
overflow layer which reflects the adjustment of the overflow to the slope and entrainment
365
of oceanic water. Entrainment on the slope can also be recognized by the weakening of the
366
cold temperature signal associated with the overflow water as it flows away from the
17
367
channel. The overflow loses most of its cold temperature signal upon entering the enhanced
368
restoring region at the northwest part of the oceanic basin. The overflow also appears to
369
become baroclinically unstable on the slope and detach from the coastline. The flow field
370
on the slope is observed with more time variability than that near the shelf-break. The time
371
variability observed on the slope is about 2-3 days which is twice as short as that observed
372
at the shelf-break. While overflow is subcritical in the channel, it appears to become
373
supercritical on the continental slope. The Froude number of the overflow layer on the
374
slope is above one since its flow speed is about 0.15 m s-1 downstream where the thickness
375
is about 100 m and the gravity speed ( g' h ) is about 0.10 m s-1. This indicates that the
376
observed variability of the overflow on the slope is unlikely to influence the dynamics at
377
the channel.
378
379
c. Sensitivity to the width of the strait
380
To further examine whether Eq (2) is a useful tool for estimating the overflow transport, we
381
test the sensitivity of the overflow transport to the width of the strait. The dependence of
382
the overflow transport on the channel width is one of the significant differences between Eq
383
(1) and (2). While Eq (1) suggests linear dependence of the overflow transport to the
384
channel width, Eq (2) suggests that it is insensitive to the channel width.
385
When the channel width is narrowed to 10, 20, and 50 km, the overflow transport
386
became 0.40, 0.37, and 0.39 Sv, similar to that observed in CTRL. All model settings are
387
kept the same as that of CTRL except for the width of the channel. 10 km is an order
388
smaller than the 100 km wide channel used in CTRL but is still wider than the deformation
18
389
radius (about 5 km). The experiment suggests that Eq (2) is indeed a useful tool for
390
overflows that occurs through channels wider than the deformation radius.
391
392
4. The overflow – open ocean interaction
393
CTRL suggests that the AABW overflow enters the oceanic basin as a geostrophic flow.
394
The question is then; are AABW overflows affected by changes in oceanic processes? To
395
answer this question, we will examine the sensitivity of the overflow properties to the
396
magnitude of the wind stress applied in the oceanic basin.
397
398
a. The impact of the wind on the overflow transport
399
When the wind stress is zero (o=0.0), the overflow transport reduces to 0.25 Sv, which is
400
69% of CTRL (Figure 8). This decrease indicates that open oceanic processes do affect the
401
AABW overflow. All model setups in this model experiment except for o are the same as
402
CTRL and we will refer to the experiment as NoWND hereafter. The cyclonic circulation
403
that establishes in the shelf region in NoWND remains similar to CTRL because the shelf is
404
still driven by the surface cooling (Figure 8a).
405
An anti-cyclonic circulation establishes at the surface of the oceanic basin (Figure
406
8b). This circulation is forced by the upper oceanic branch of the exchange flow at the strait,
407
which is toward the shelf, and establishes in the direction of the topographic Rossby wave.
408
The circulation at the surface is opposite from CTRL so the surface flow and the overflow
409
in NoWND are now flowing in the opposite directions on the slope. The v-shape at the
410
shelf-break has also weakened significantly compared to CTRL (Figure 8c). The overflow
411
appears to be more unstable compared to CTRL since more eddies are observed and they
19
412
form much closer to the shelf-break. A slower advection speed in the along-slope direction,
413
due to the absence of a wind-induced cyclonic flow, may have led to this faster growth of
414
instability near the strait.
415
Further model experiments show that the overflow transport increases roughly
416
linearly as the wind stress magnitude increases (Figure 9). This sensitivity is also found to
417
roughly match with the sensitivity of the overflow transport to g’ as expected from Eq (2):
418
 h2 
M S ~ 
g '  ,
2f 
419
where is the difference from CTRL. g’ is found to increase as the magnitude of the wind
420
stress increases. The upstream interface height, h, is set constant to 250 m, which is half the
421
depth of the channel. A weaker dependence is found when h based on that estimated at the
422
shelf-break is used and moreover, it shows a decrease when o =0.25. This occurs because
423
the h there decreases as the wind stress increases. A better fit to model results based on a
424
prescribed channel depth implies that the overflow transports are also affected by the
425
external topographic parameter that constrains the flow between the shelf and the oceanic
426
basin.
(5)
427
The observed changes in g’ are mainly associated with the changes in the depth of
428
the slope front near the channel since the overflow temperature remains roughly -1.3 oC.
429
The depth of the slope front, , increases as the magnitude of the wind stress increases. The
430
changes in the depth of the slope front are found almost identical to that of the coastal front
431
along the coastlines, which depends linearly on the magnitude of the coastal upwelling
432
forced by the wind:
20
433
  
  ( w  t )  w  t   S   t ,
 f 0 RD 
434
where t is the time scale of the flow within the wind-forced area. The downwelling
435
velocity, w, is estimated by assuming that the convergence of surface Ekman transport
436
(  s  f 0 ) occurs within the deformation radius (RD) from the coast. g’ can be expressed as
437
g '    T , where  is the thermal expansion coefficient of seawater, 7x10-5 oC-1, and T is
438
the time-averaged temperature difference between the overflow and the oceanic water that
439
are located within 20 km from the strait (y = 100 km) and above the depth of the channel.
440
Thus based on Eq (5), Eq (6), and that an increase in the depth of the slope front is
441
associated with a warming of the upper oceanic water on the oceanic basin side of the strait,
442
 ( T )    , we find the overflow transport likely to increase linearly as the wind stress
443
increases:
444
h2
h2
M S 
g ' 
(T )     S .
2f
2f
445
This sensitivity of the overflow transport to the wind stress does appear to explain the
446
model results well (Figure 9). Note that we have again assumed h as a constant.
(6)
447
Since Eq (2) (and 5) assumes a one-layer overflow, we will also discuss how the
448
presence of an upper layer inflow may affect the overflow transport. Hunkins and
449
Whitehead (1992) suggests that for a two-layer exchange flow, Eq (2) is modified to
450
M S ~ 0.156
451
where H is the depth of the channel. This equation is based on an assumption that the
452
interface between the two layers is anti-symmetric about the horizontal center of the
g' H 2
f ,
(7)
21
453
channel and that the exchange flow is fully developed. Eq (7) shows the same linear
454
dependence of overflow transport on g’ as Eq (2) but uses a thickness parameter based on
455
the full depth of the channel and a coefficient of 0.156. The changes in the overflow
456
transports estimated from Eq (7) is found similar to Eq (2) and matches well with the model
457
results (Figure 9). The absolute value of transports estimated from Eq (7) also matches well
458
with that observed with the estimates further improving when bottom Ekman Transport is
459
added. The anti-symmetric assumption of the interface is also reasonably met, supporting
460
the assumption of the formula. The underestimate observed for o =0.25 may be due to the
461
gradual breakdown of the linearity and the assumption of an anti-symmetric interface
462
height as the exchange rate increases with the wind. While there are differences in Eq (2)
463
and (7), we find both equations showing a similar sensitivity of the overflow transport to
464
g’ which matches well with the model experiments. This supports our suggestion that the
465
overflow transport is affected by the changes in the depth of the slope front because it
466
changes g’. Expressed in terms of energy balance, the overflow transport is sensitive to the
467
changes in the slope front because it changes the oceanic stratification and the effective
468
potential energy of the overflow water at the shelf-break.
469
We find similar sensitivity of the overflow transport to the oceanic winds when the
470
upper oceanic inflow occurs not directly above the overflow. This is tested by setting the
471
model bathymetry west of the channel to 250 m rather than a wall so that the upper oceanic
472
inflow occurs along the western boundary (x = 0) while the overflow still enters the oceanic
473
basin through the channel (not shown). We find the overflow transport to increase from
474
0.33 to 0.43 Sv when o increases from zero to 0.10, a similar sensitivity as that found
22
475
between NoWND and CTRL. The presence of an upper layer inflow is, therefore, unlikely
476
to be a major factor affecting the sensitivity of the overflow transport to oceanic processes.
477
478
b. The importance of oceanic stratification
479
We have suggested that oceanic processes affect overflow transports when they change the
480
depth of the slope front and the oceanic stratification near the shelf-break. This mechanism
481
implies that without oceanic stratification, changes in the wind stress are unlikely to change
482
the overflow transport. To test this significance of oceanic stratification, two model
483
experiments similar to CTRL and NoWND are pursued. The initial temperature field is set
484
constant to 2 oC, i.e., no stratification but all other model settings are kept the same as
485
CTRL and NoWND. When o is 0.10, the overflow transport is 0.67 Sv. When o is zero,
486
the transport is 0.69 Sv, which is almost the same. The experiment shows that the oceanic
487
stratification is indeed a key element that enables oceanic processes to enhance the
488
overflow transport and not the direct input of momentum from the wind. Note that the
489
increase in transport from CTRL (or NoWND) likely occurs because the temperature
490
difference across the sill increases when the oceanic basin is 2oC at all depth.
491
492
c. The impact of the slope front on the overflow temperature
493
Since the minimum temperature is set to -1.8 oC in the model, the overflow temperature in
494
the sensitivity experiments remain roughly -1.3 oC. This enables us to examine the impact
495
of the oceanic changes in isolation from the changes that may occur in the overflow
496
temperature. In reality, however, the overflow temperature is likely to change as the
23
497
overflow transport because the residence time on the shelf changes. Here we will discuss
498
how such changes in the overflow temperature may occur.
499
A bulk estimate of the overflow temperature can be solved from the basin integrated
500
mass and heat balance equations of the shelf region: Mo + Mi  0 and Mo To + Mi Ti  Q .
501
M is the transport between the shelf and the oceanic basin at the shelf-break, Q is the
502
surface heat flux received within the shelf, and T is the temperature. Subscripts o and i
503
represent the overflow and the surface inflow respectively. Note that we have assumed that
504
the heat balance is in a steady state and that diffusion is negligible. The overflow
505
temperature can then be expressed as
506
To  Ti 
Q
,
Mo
(8)
507
which shows that the overflow temperature will increase as the overflow transport increases
508
assuming that the temperature of the oceanic inflow and surface buoyancy loss remain
509
constant.
510
To examine the sensitivity of the overflow temperature to the changes in the
511
overflow transport, we first express Eq (8) using the temperature difference between the
512
overflow and the oceanic inflow:
513
T  (To  Ti ) 
Q
.
Mo
(9)
514
The overflow water is defined as that below 0.0oC and the time-averaged temperature of the
515
overflow and the oceanic inflow, To and Ti respectively, are estimated from the water mass
516
located within 20 km from the shelf-break (y = 100 km) as done previously. The model
517
experiments show an increase in the magnitude of T as the wind stress increases (Figure
518
10), which is a reflection of the warming in Ti since To remains roughly constant. However,
24
519
this increasing trend in the magnitude of T is opposite to that expected from Eq (9). This is
520
found to occur because the magnitude of surface cooling (Q) changes in the model. Any
521
excessive cooling that is applied beyond -1.8oC is neglected and so results in less surface
522
cooling than that prescribed. As the o increases from 0.0 to 0.05, 0.10, 0.15, and 0.25 N m-
523
2
, the ratio of the actual Q compared to that prescribed increases from 0.28, to 0.34, 0.37,
524
0.45, and 0.52, respectively. When this differences in Q are adjusted, the magnitude of T
525
show a decreasing trend as the wind stress increases (Figure 10). Note that the magnitude of
526
Q is adjusted to that estimated when o is zero and that changes in T are assumed to occur
527
linearly with Q.
528
The observed T match well with that estimated directly from Eq (9) using Q, Mo,
529
and the residual tendency and diffusion terms (Figure 10). Slight difference arises from the
530
use of spatial averaging when estimating the overflow and oceanic inflow temperatures
531
rather than just at the shelf-break. This match confirms that Eq (9) is a reasonable and
532
useful tool for examining the sensitivity of the overflow temperature to the exchange rate.
533
The model experiments also indicate that overflow temperatures are likely to take
534
longer time to adjust compared to overflow transports. The time series of the overflow
535
temperature and transport (Figure 6) show the overflow transport quickly adjusting to its
536
quasi-steady state value in about 50 days once it starts entering the oceanic basin while the
537
overflow temperature reaches its steady state gradually in about 100 days. This difference
538
reflects the difference between the dynamical time scale of the channel and the resident
539
time scale of the shelf. A topographic Rossby wave takes about 40 days, about a month, to
540
propagate across the channel. The resident time scale on the shelf is about 200 days
541
assuming a constant exchange transport of 0.36 Sv, which is about half a year. Model
25
542
experiments, therefore, suggest that the overflow transport will likely adjust to the seasonal
543
variability of the wind field in a month or two while the overflow temperature is likely to
544
take more than a season to adjust. Since the actual shelf in the Weddell Sea is much larger
545
than that used in our model, this adjustment in overflow temperature may take more than
546
half a year. Gill (1973) suggests resident time scale of about few years on the shelf.
547
548
5. Summary and closing remarks
549
The main goal of this paper was to answer whether the AABW outflows are affected by
550
oceanic processes through their influence on the ASF. We have focused on overflows
551
through channels, not cascades, and utilized a numerical model to answer this question.
552
Here we summarize and address the implication of our study.
553
554
a. Summary
555
The AABW overflow transport is found to increase significantly when the wind stress
556
increases the depth of the ASF. We find this sensitivity arising from two dynamical aspects
557
present at the strait:
558
(1) The transport of overflows is geostrophically controlled at the strait. The simulated
559
overflow had transport that matches well with the hydraulic formula derived by
560
Toulany and Garrett (1984), Helfrich et al. (1999), and Hunkins and Whitehead
561
(1992). The transport is also found insensitive to the width of the strait as long as it
562
is much wider than the local deformation radius, further supporting the use of the
563
formula.
26
564
(2) When the wind stress increases, it deepens the coastal front and the ASF. This
565
decreases the oceanic stratification near the shelf-break by bringing more light
566
surface water to the depth of the channel. This enables the overflow to interact with
567
lighter oceanic water.
568
Because the overflow transport is geostrophically controlled, it is sensitive to the changes in
569
the density gradient across the shelf-break. So when the oceanic wind increases the depth of
570
the ASF, the density gradient across the shelf-break increases and enhances the overflow
571
transport. In absence of oceanic stratification, such process is absent and the impact of
572
oceanic winds on the overflow transport is absent.
573
An increase in the oceanic wind stress was also found likely to warm the overflow
574
temperature if the magnitude of surface cooling remains constant. This is because enhanced
575
overflow transports lead to shorter resident time on the shelf. However, the adjustment of
576
overflow temperatures depends on the size of the shelf and is likely to take few years. This
577
is much longer than the time-scale expected for the adjustment in transport, which depends
578
on the width of the channel and likely takes 1-2 months.
579
580
b. Closing remarks
581
The entrainment of oceanic water to the overflow may have been insufficiently simulated in
582
our model because the horizontal resolution of 1 km is incapable of fully resolving the
583
Kelvin-Helmholtz instability. While entrainment is crucial part of the overflow dynamics
584
once it is on the slope, we suspect this insufficiency not to significantly affect the results of
585
this study since our focus is on the transport through the channel and not the dynamics after
586
the shelf-break. For the Filchner Bank Channel overflow, entrainment is also thought to
27
587
occur not right after the overflow passes the shelf-break but when it accelerates and
588
increase the velocity shear at much depth (Foldvik et al. 2004). High resolution model
589
simulations of the Filchner Bank Channel overflow by Wilchinsky and Feltham (2009) and
590
Matsumura and Hasumi (2010) show entrainment occurring on the slope away from the
591
shelf-break.
592
Our study suggests that the parameterization of AABW overflows in GCMs needs
593
to incorporate the additional impact from oceanic processes through their influence on the
594
ASF. The observed seasonal and inter-annual variability in the oceanic winds are likely to
595
induce variability in overflow transport. We suspect the seasonal variability to affect the
596
overflow transport the most: stronger transport when the wind is also stronger. Inter-annual
597
variability in the export of AABW from the Weddell Basin to the Atlantic has been
598
observed to correlate well with basin scale interannual climate variability signals at the
599
surface and the role of the wind-driven Weddell gyre has been suggested (Meredith et al.
600
2008, Jullion et al. 2010).
601
The significance of the oceanic impact on overflows may differ with the location of
602
the outflow. The overflow through the Drygalski trough in the Ross Sea has similar depth
603
and width as the Filchner Bank Channel overflow. However, the v-shape structure of the
604
ASF may be absent here (Gordon et al, 2009). A v-shape structure is observed at a location
605
close to the Drygalski trough (Visbeck and Thurnherr, 2009) but the basin view shows that
606
the structure is much narrower and weaker compared to the eastern side of the shelf-break
607
in the Ross Sea (Orsi and Wiederwohl, 2009). The external currents may thus have less
608
influence on the slope front near the Drygalski trough. Entrainment, on the other hand, may
609
be significant near the shelf-break since the flow speed is more than twice as fast as the
28
610
Filchner Bank Channel overflow (Gordon et al. 2009). Entrainment may have large control
611
on the slope front and the oceanic stratification near the shelf-break (Baines, 2009) which
612
suggests that this overflow may be less influenced by the changes in the external oceanic
613
processes compared to what was shown in this study. Gordon et al. (2010) recently
614
suggested that strong winds may also shift of the location of the ASF in-shore and prohibit
615
the escape of AABW outflow. Such mechanism is possible for an AABW outflow from the
616
wide western Weddell shelf where outflows occur through cascades and the location of
617
ASF may be less constrained to the slope. Changes in overflow properties may originate
618
from variability in surface buoyancy loss. Strong tides are observed along the shelves of
619
Antarctic and its importance on export of dense water from the shelf has been suggested
620
(Guan et al. 2009, Muench et al. 2009, Ou et al. 2009, and Padman et al. 2009). The
621
significance of the variability induced by winds, buoyancy losses, and tides on the
622
overflow-open oceanic interaction is likely to depend on the adjustment time scale of the
623
overflow, shelf, ambient oceanic basin, and the ASF and we plan to investigate this
624
problem next.
625
626
Acknowledgments.
627
The author thanks the anonymous reviewers, Drs. Bo Qiu, Jiayan Yang, and Kiyoshi
628
Tanaka for many useful comments on this paper. Assistance from Dr. Keiko Takahashi, Mr.
629
Koji Goto, and Hiromitsu Fuchigami with the MSSG code is also acknowledged.
630
631
References
29
632
Baba, Y., K. Takahashi, T. Sugimura, and K. Goto, 2010: Dynamical Core of an
633
Atmospheric General Circulation Model on a Yin-Yang Grid, Monthly Weather Review,
634
138, 3988-4005.
635
Baines, P G., 2009, A model for the structure of the Antarctic Slope Front, Deep Sea Res. II,
636
56, 859-873
637
Bindoff, N. L., M. A. Rosenberg, M. J. Warner, 2000: On the circulation and water masses
638
over the Antarctic continental slope and rise between 80 and 150oE, Deep Sea Research
639
Part II, 47, 2299-2326
640
Chapman, D. C. (2000), The influence of an alongshelf current on the formation and
641
offshore transport of dense water from a coastal polynya, J. Geophys. Res., 105(C10),
642
24,007–24,019, doi:10.1029/2000JC000296.
643
Cushman-Roisin, B., 1994, Introduction to Geophysical Fluid Dynamics, Prentice-Hall,
644
183-189
645
Fahrbach, E., G. Rohardt, and G. Krause, 1992: The Antarctic Coastal Current in the
646
southeastern Weddell Sea., Polar Biol., 12, 171-182.
647
Fahrbach, E., G. Rohardt, M Schroder, V. Strass, 1994: Transport and structure of the
648
Weddell Gyre., Ann .Geophysicae., 12, 840-855.
649
Foldvik, A., T. Gammelsrød, S. Østerhus, E. Fahrbach, G. Rohardt, M. Schröder, K. W.
650
Nicholls, L. Padman, and R. A. Woodgate, 2004: Ice shelf water overflow and bottom
651
water formation in the southern Weddell Sea, J. Geophys. Res., 109, C02015,
652
doi:10.1029/2003JC002008.
653
Gill, A. E., 1973, Circulation and bottom water production in the Weddell Sea, Deep Sea
654
Research and Oceanographic Abstracts, 20, 111-140
30
655
Gordon, A. L., B.A. Huber, H.H. Hellmer, and A. Ffield (1993), Deep and bottom water of
656
the Weddell Sea's western rim, Science, 262, 95-97,
657
Gordon, A. L., A. H. Orsi, R. Muench, B. A. Huber, E. Zambianchi, M. Visbeck, 2009:
658
Western Ross Sea continental slope gravity currents, Deep Sea Research Part II, 56, 796-
659
817
660
Gordon, A. L., B. Huber, D. McKee, and M. Visbeck, 2010: A seasonal cycle in the export
661
of bottom water from the Weddell Sea, Nature Geoscience, 3 (8). pp. 551-556.
662
doi: 10.1038/ngeo916
663
Guan, X., H. W. Ou, and D. Chen, 2009: Tidal effect on the dense water discharge, Part 2:
664
A numerical study, Deep Sea Research Part II, 56, 884-894
665
Gawarkiewicz, G. (2000), Effects of ambient stratification and shelfbreak topography on
666
offshore transport of dense water on continental shelves, J. Geophys. Res., 105(C2), 3307–
667
3324, doi:10.1029/1999JC900298.
668
Helfrich, K. R., A. C. Kuo, and L. J. Pratt, Nonlinear Rossby adjustment in a channel, J.
669
Fluid Mech., 390, 187-222
670
Helfrich, Karl R., Lawrence J. Pratt, 2003: Rotating Hydraulics and Upstream Basin
671
Circulation. J. Phys. Oceanogr., 33, 1651–1663.
672
Heywood, K.J., R. A. Locarnini, R. D. Frew, P. F. Dennis, and B. A. King, 1998, Transport
673
and water masses of the Antarctic Slope Front system in the eastern Weddell Sea, In S. S.
674
Jacobs & R. F. Weiss (Eds.), Interactions at the Antarctic Continental Margins (Vol. 74, pp.
675
1–27), Antarctic Research Series. Washington, DC: American Geophysical Union.
31
676
Heywood, K. J., A. C. Naveira Garabato, D. P. Stevens, and R. D. Muench (2004), On the
677
fate of the Antarctic Slope Front and the origin of the Weddell Front, J. Geophys. Res., 109,
678
C06021, doi:10.1029/2003JC002053.
679
Hirano, D., Y. Kitade, H. Nagashima and M. Matsuyama, 2010: Characteristics of
680
Observed Turbulent Mixing across the Antarctic Slope Front at 140°E, East Antarctica, J.
681
Oceanography, 66, 95-104.
682
Hunkins, K., and J. A. Whitehead (1992), Laboratory Simulation of Exchange Through
683
Fram Strait, J. Geophys. Res., 97(C7), 11,299–11,321, doi:10.1029/92JC00735.
684
Ivanov, V.V., G.I. Shapiro, J.M. Huthnance, D.L. Aleynik, and P.N. Golovin, 2004:
685
Cascades of dense water around the world ocean, Progress In Oceanography, 60, 47-98,
686
doi: 10.1016/j.pocean.2003.12.002.
687
Ivanov, V. V., and P. N. Golovin (2007), Observations and modeling of dense water
688
cascading from the northwestern Laptev Sea shelf, J. Geophys. Res., 112, C09003,
689
doi:10.1029/2006JC003882
690
Jacobs, S. S., 1991: On the nature and significance of the Antarctic Slope Front, Mar.
691
Chem., 35, 9-24
692
Jiang, L., and R. W. Garwood, 1996: Three-Dimensional Simulations of Overflows on
693
Continental Slopes. Journal of Physical Oceanography, 26, 1214-1233
694
Jiang, L., and R. W. Garwood Jr. (1998), Effects of topographic steering and ambient
695
stratification on overflows on continental slopes: A model study, J. Geophys. Res., 103(C3),
696
5459–5476, doi:10.1029/97JC03201.
32
697
Jullion, L., S. C. Jones, A. C. Naveira Garabato, and M. P. Meredith (2010), Wind-
698
controlled export of Antarctic Bottom Water from the Weddell Sea, Geophys. Res. Lett., 37,
699
L09609, doi:10.1029/2010GL042822.
700
Kida, S., J.F. Price, and J. Yang, 2008: The Upper-Oceanic Response to Overflows: A
701
Mechanism for the Azores Current. J. Phys. Oceanogr., 38, 880-895
702
Kida, S., J. Yang, and J.F. Price, 2009: Marginal Sea Overflows and the Upper Ocean
703
Interaction. J. Phys. Oceanogr., 39, 387–403.
704
Killworth, P. D., 1977: Mixing on the Weddell Sea continental slope. Deep-Sea Res., 24,
705
427–448.
706
Kohl, A. R. H. Kase, D. Stammer, and N. Serra, 2007, Causes of Changes in the Denmark
707
Strait Overflow., J. Phys. Oceanogr. 37, 1678-1696
708
Legg, S., R. W. Hallberg, and J. B. Girton, 2006: Comparison of entrainment in overflows
709
simulated by z-coordinate, isopycnal and non-hydrostatic models, Ocean Modelling, 11,
710
69-97, doi: 10.1016/j.ocemod.2004.11.006.
711
Legg
712
REPRESENTATION IN CLIMATE MODELS: The Gravity Current Entrainment Climate
713
Process Team, 2009, Bull. Amer. Meteor. Soc., 90, 657-670
714
Matsumura, Y., and H. Hasumi, 2010: Modeling ice shelf water overflow and bottom water
715
formation
716
doi:10.1029/2009JC005841.
717
Meredith, M. P., A. C. Naveira Garabato, A. L. Gordon, G. C. Johnson, 2008: Evolution of
718
the Deep and Bottom Waters of the Scotia Sea, Southern Ocean, during 1995–2005. J.
719
Climate, 21, 3327-3343
S.,
and
in
Coauthors,
the
southern
2009:
IMPROVING
Weddell
33
Sea, J.
OCEANIC
Geophys.
OVERFLOW
Res., 115,
C10033,
720
Muench, R. D., and A. L. Gordon (1995), Circulation and transport of water along the
721
western Weddell Sea margin, J. Geophys. Res., 100, 18503–18515, doi:10.1029/95JC00965.
722
Muench, R., L. Padman, A. Gordon, and A. Orsi, 2009: A dense water outflow from the
723
Ross Sea, Antarctica: Mixing and the contribution of tides, J. Mar. Sys., 77, 369-387
724
NCEP Daily Global Analyses data provided by the NOAA/OAR/ESRL PSD, Boulder,
725
Colorado, USA, from their Web site at http://www.esrl.noaa.gov/psd/
726
Orsi, A. H., G. C. Johnson, J. L. Bullister, 1999: Circulation, mixing, and production of
727
Antarctic Bottom Water, Progress In Oceanography, 43(1), 55-109
728
Orsi, A. H. and C. L. Wiederwohl, 2009: A recount of Ross Sea waters, Deep Sea Res. II,
729
56, 778-795, DOI: 10.1016/j.dsr2.2008.10.033.
730
Ou, H. W., X. Guan, and D. Chen, 2009: Tidal effect on the dense water discharge, Part 1:
731
Analytical model, Deep Sea Research Part II, 56, 874-883
732
Padman, L., S. L. Howard, A. H. Orsi, and R. D. Muench, 2009: Tides of the northwestern
733
Ross Sea and their impact on dense outflows of Antarctic Bottom Water., Deep-Sea
734
Research Part II, 56, 818-834
735
Price, J. and M. Baringer, 1994: Outflows and deep water production by marginal seas.
736
Prog. Oceanog., 33, 161–200.
737
Price, J. F., and J. Yang, 1998: Marginal sea overflows for climate simulations. Ocean
738
Modelling and Parameterization, E. Chassignet and J. Vernon, Eds., Kluwer Academic,
739
155–170.
740
Smith, P. C., 1975: A streamtube model for bottom boundary currents in the ocean. Deep-
741
Sea Res., 22, 853–873.
34
742
Takahashi, K. et al. 2007: High Resolution Numerical Modelling of the Atmosphere and
743
Ocean, Kevin Hamilton and W. Ohfuchi, Eds, Springer-Verlag New York, LLC
744
Tamura, T., K. I. Ohshima, and S. Nihashi, 2008, Mapping of sea ice production for
745
Antarctic coastal polynyas, Geophys. Res. Lett., 35, L07606, doi:10.1029/2007GL032903.
746
Tanaka K. and K. Akitomo, 2000: Density Current Descending along Continental Slope
747
and the Associated Deep Water Formation: Two-Dimensional Numerical Experiments with
748
a Non-hydrostatic Model J. Oceanogr., 56, 117-130.
749
Tanaka, K., and K. Akitomo (2001), Baroclinic instability of density current along a
750
sloping bottom and the associated transport process, J. Geophys. Res., 106, 2621–2638,
751
doi:10.1029/2000JC000214.
752
Toulany, B., C. Garrett, 1984: Geostrophic Control of Fluctuating Barotropic Flow through
753
Straits. J. Phys. Oceanogr., 14, 649-655
754
Visbeck, M. and A. M. Thurnherr, 2009: High-resolution velocity and hydrographic
755
observations of the Drygalski Trough gravity plume, Deep Sea Res. II, 56, 835-842
756
Warren, B. A., 1980: Deep circulation of the world ocean. Evolution of Physical
757
Oceanography: Scientific Surveys in Honor of Henry Stommel, B. A. Warren and C.
758
Wunsch, Eds., MIT Press, 6–41.
759
Wahlin, A. K., 2002: Topographic steering of dense currents with application to submarines
760
canyons. Deep-Sea Res. I, 49, 305–320.
761
Wilchinsky, A. V., and D. L. Feltham (2009), Numerical simulation of the Filchner
762
overflow, J. Geophys. Res., 114, C12012, doi:10.1029/2008JC005013.
763
Whitworth III, T., A. H. Orsi, S.-J. Kim, W. D. Nowlin, Jr., and R. A. Locarnini, 1998,
764
Water Masses and Mixing near the Antarctic Slope Front., In S. S. Jacobs & R. F. Weiss
35
765
(Eds.), Interactions at the Antarctic Continental Margins (Vol. 74, pp. 1–27), Antarctic
766
Research Series. Washington, DC: American Geophysical Union.
767
768
36
769
List of Figures
770
FIG. 1. (a) The bathymetry of the Weddell Sea and its shelf contoured from -2000 m to 0 m.
771
The shelf is roughly 500 m deep and an overflow enters the Weddell Sea from the Filchner
772
Depression through the Filchner Bank Channel (black arrow). The ASF is observed along
773
the shelf-break (white dash). (b) Vertical sections of the observed potential temperature
774
(left) and salinity (right) observed across the shelf break at about 35W (Whitworth et al.
775
1998, see the close up of the region shown in (a)). Temperature and salinity (solid lines) as
776
well as density (dotted lines) are observed with a v-shape near the shelf-break. (c) A
777
comparison of the flow speed along the ASF near 11W and the surface wind speed
778
(Fahrbach et al., 1992). Both show similar seasonal variability and correlate well especially
779
from 1988 to 1989.
780
781
FIG. 2. The number of cold and fresh water anomaly events detected at the bottom of the
782
shelf-break in the Ross Sea (Gordon et al. 2009). Seasonal variability is observed with more
783
events occurring from November to May than from May to November. Inter-annual
784
variability is also observed.
785
786
FIG. 3. A schematic of an overflow, a shelf front and the open ocean viewed across the sill
787
(left) and at the sill looking toward the open ocean (right). Cooling in the shelf leads to a
788
formation of cold dense water that eventually spills over the sill and become overflows. (a)
789
Without a shelf front: The overflow encounters the deep water of the open ocean at the sill.
790
(b) With a shelf front: Warm surface water is brought down to the depth of the sill and the
791
interface between the surface water and the deep water tilts downward.
792
37
793
FIG. 4. (a) The side view of the model showing the bathymetry and the temperature field.
794
Temperature changes exponentially in the vertical with warm water (shaded) near the
795
surface. (b) The bird’s eye view of the bathymetry. The model domain is roughly 200 km
796
zonally and 300 km meridionally and has two regions separated by a sill 500 m deep and
797
channel 100 km wide. The northern and southern regions represent the oceanic basin and
798
the shelf, respectively. Cooling source is located at the southern boundary.
799
view of the channel that connects the shelf and the oceanic basin. The location of this cross-
800
section is shown in dotted lines in (a).
(c) The side
801
802
FIG. 5. (a) The wind stress forced over the oceanic basin. The direction of the wind stress is
803
shown in vectors and its magnitude is shown in gray. The wind stress is cyclonic and
804
increases closer to the coast. (b) The restoring time scale applied in the oceanic basin. The
805
minimum restoring time scale is 2 days and increases exponentially away from the center.
806
Restoring is enhanced in the northwest part of the oceanic basin.
807
808
FIG. 6. The overflow transport and temperature simulated in CTRL with time. The
809
overflow begins to enter the oceanic basin around day 30 and then reaches its quasi steady
810
state in about 50 days. Model analysis is based on the data output between days 120-180.
811
812
FIG. 7. Simulation results from CTRL. (a) A snapshot of the surface temperature and
813
velocity fields at day 120. A cyclonic flow establishes in the oceanic basin. (b) A snapshot
814
of the bottom temperature and velocity fields at day 120. Overflow descends the slope
815
cyclonically as it warms and splits into eddies. (c) A cross-section of the time averaged
38
816
temperature and zonal velocity fields at x = 120 km. The isotherms tilt downward at the
817
shelf-break and create a v-shape. (d) A cross-section of the time averaged temperature and
818
meridional velocity fields across the channel near the shelf-break (y = 100 km). Outflows
819
(positive) are observed along the western and eastern walls.
820
821
FIG. 8. Simulation results from NoWND. (a) A snapshot of the surface temperature and
822
velocity fields at day 120. An anti-cyclonic flow establishes in the oceanic basin. (b) A
823
snapshot of the bottom temperature and velocity fields at day 120. The overflow shows
824
more variability on the slope than CTRL (Figure 7b). (c) A cross-section of the time
825
averaged temperature and zonal velocity at x = 120 km. The v-shape is less pronounced
826
than CTRL. (d) A cross-section of the time averaged temperature and meridional velocity
827
across the channel near the shelf-break (y = 100 km). Outflows (positive) are observed
828
along the western and eastern walls similar to CTRL.
829
830
FIG. 9. The sensitivity of the overflow transport to the wind stress magnitude. Model
831
results (circle) show transport increasing linearly as the wind stress increases. The changes
832
in the transports compared to CTRL is well captured by Eq (5) when h is kept constant (250
833
m) (square) compared to when h estimated at the strait is used (dot). Changes expected
834
from Eq (7) (plus) are similar to Eq (5). The magnitude of the transport estimated from the
835
sum of Eq (7) and (1) (triangle) is also close to the model results.
836
837
FIG. 10. The dependence of the temperature difference between the overflow and the
838
oceanic inflow to the magnitude of the wind stress (circle). The overflow temperature
39
839
matches well with that estimated from Q, Ms, and Eq (16) (square). Model results show an
840
increase in the difference as the wind stress increase (circle) because of the changes in Q
841
but show a decrease (plus) when accounted for the difference in Q.
842
40
FIG. 1. (a) The bathymetry of the Weddell Sea and its shelf contoured from -2000 m to 0 m. The shelf is
roughly 500 m deep and an overflow enters the Weddell Sea from the Filchner Depression through the Filchner
Bank Channel (black arrow). The ASF is observed along the shelf-break (white dash). (b) Vertical sections of
the observed potential temperature (left) and salinity (right) observed across the shelf-break at about 35W
(Whitworth et al. 1998, see the close up of the region shown in (a)). Temperature and salinity (solid lines) as
well as density (dotted lines) are observed with a v-shape near the shelf-break. (c) A comparison of the flow
speed along the ASF near 11W and the surface wind speed (Fahrbach et al., 1992). Both show similar seasonal
variability and correlate well especially from 1988 to 1989.
41
FIG. 2. The number of cold and fresh water anomaly events detected at the bottom of the shelf-break in the
Ross Sea (Gordon et al. 2009). Seasonal variability is observed with more events occurring from November to
May than from May to November. Inter-annual variability is also observed.
FIG. 3. A schematic of an overflow, a shelf front and the open ocean viewed across the sill (left) and at the sill
looking toward the open ocean (right). Cooling in the shelf leads to a formation of cold dense water that
eventually spills over the sill and become overflows. (a) Without a shelf front: The overflow encounters the
deep water of the open ocean at the sill. (b) With a shelf front: Warm surface water is brought down to the depth
of the sill and the interface between the surface water and the deep water tilts downward.
42
FIG. 4. (a) The side view of the model showing the bathymetry and the temperature field. Temperature changes
exponentially in the vertical with warm water (shaded) near the surface. (b) The bird’s eye view of the
bathymetry. The model domain is roughly 200 km zonally and 300 km meridionally and has two regions
separated by a sill 500 m deep and channel 100 km wide. The northern and southern regions represent the
oceanic basin and the shelf, respectively. Cooling source is located at the southern boundary. (c) The side
view of the channel that connects the shelf and the oceanic basin. The location of this cross-section is shown in
dotted lines in (a).
43
(a)
(b)
FIG. 5. (a) The wind stress forced over the oceanic basin. The direction of the wind stress is shown in vectors
and its magnitude is shown in gray. The wind stress is cyclonic and increases closer to the coast. (b) The
restoring time scale applied in the oceanic basin. The minimum restoring time scale is 2 days and increases
exponentially away from the center. Restoring is enhanced in the northwest part of the oceanic basin.
FIG. 6. The overflow transport and temperature simulated in CTRL with time. The overflow begins to enter the
oceanic basin around day 30 and then reaches its quasi steady state in about 50 days. Model analysis is based on
the data output between days 120-180.
44
FIG. 7. Simulation results from CTRL. (a) A snapshot of the surface temperature and velocity fields at day 120.
A cyclonic flow establishes in the oceanic basin. (b) A snapshot of the bottom temperature and velocity fields at
day 120. Overflow descends the slope cyclonically as it warms and splits into eddies. (c) A cross-section of the
time averaged temperature and zonal velocity fields at x = 120 km. The isotherms tilt downward at the shelfbreak and create a v-shape. (d) A cross-section of the time averaged temperature and meridional velocity fields
across the channel near the shelf-break (y = 100 km). Outflows (positive) are observed along the western and
eastern walls.
45
FIG. 8. Simulation results from NoWND. (a) A snapshot of the surface temperature and velocity fields at day
120. An anti-cyclonic flow establishes in the oceanic basin. (b) A snapshot of the bottom temperature and
velocity fields at day 120. The overflow shows more variability on the slope than CTRL (Figure 7b). (c) A
cross-section of the time averaged temperature and zonal velocity at x = 120 km. The v-shape is less
pronounced than CTRL. (d) A cross-section of the time averaged temperature and meridional velocity across
the channel near the shelf-break (y = 100 km). Outflows (positive) are observed along the western and eastern
walls similar to CTRL.
46
FIG. 9. The sensitivity of the overflow transport to the wind stress magnitude. Model results (circle) show
transport increasing linearly as the wind stress increases. The changes in the transports compared to CTRL is
well captured by Eq (5) when h is kept constant (250 m) (square) compared to when h estimated at the strait is
used (dot). Changes expected from Eq (7) (plus) are similar to Eq (5). The magnitude of the transport estimated
from the sum of Eq (7) and (1) (triangle) is also close to the model results.
FIG. 10. The dependence of the temperature difference between the overflow and the oceanic inflow to the
magnitude of the wind stress (circle). The overflow temperature matches well with that estimated from Q, Ms,
and Eq (16) (square). Model results show an increase in the difference as the wind stress increase (circle)
because of the changes in Q but show a decrease (plus) when accounted for the difference in Q.
47