manuscript submitted to
Geophysical Research Letters
Enhanced ventilation in energetic regions of the
1
Antarctic Circumpolar Current
2
Lilian A. Dove
1
, Dhruv Balwada
2
, Andrew F. Thompson
1
, Alison R. Gray
3
3
1
Division of Geological and Planetary Sciences, Environmental Science and Engineering, California
4
Institute of Technology, Pasadena, CA, USA
5
2
Lamont-Doherty Earth Observatory, Columbia University, Palisades, NY, USA
6
3
School of Oceanography, University of Washington, Seattle, WA, USA
7
Key Points:
8
•
The relationship between apparent oxygen utilization (AOU) and eddy kinetic en-
9
ergy (EKE) is assessed in the Antarctic Circumpolar Current.
10
•
AOU has relatively reduced values below the mixed layer in high-EKE standing
11
meanders as compared to low-EKE regions.
12
•
Modification of the density structure and enhanced meso- and submesoscale mo-
13
tions enhance ventilation in standing meanders.
14
Corresponding author: Lily Dove,
dove@caltech.edu
–1–
Accepted
Article
This article has been accepted for publication and undergone full peer review but has not been through
the copyediting, typesetting, pagination and proofreading process, which may lead to differences between
this version and the
Version of Record
. Please cite this article as
doi: 10.1029/2021GL097574
.
This article is protected by copyright. All rights reserved.
Accepted
Article
This article is protected by copyright. All rights reserved.
manuscript submitted to
Geophysical Research Letters
Abstract
15
Flow-topography interactions along the path of the Antarctic Circumpolar Current (ACC)
16
generate standing meanders, create regions of enhanced eddy kinetic energy (EKE), and
17
modify frontal structure. We consider the impact of standing meanders on ventilation
18
based on oxygen measurements from Argo floats and the patterns of apparent oxygen
19
utilization (AOU). Regions of high-EKE have relatively reduced AOU values at depths
20
200-700 meters below the base of the mixed layer and larger AOU variance, suggesting
21
enhanced ventilation due to both along-isopycnal stirring and enhanced exchange across
22
the base of the mixed layer. Vertical exchange is inferred from finite-size Lyapunov ex-
23
ponents, a proxy for the magnitude of surface lateral density gradients, which suggest
24
that submesoscale vertical velocities may contribute to ventilation. The shaping of ven-
25
tilation by standing meanders has implications for the temporal and spatial variability
26
of air-sea exchange.
27
Plain Language Summary
28
The circulation of the Southern Ocean is dominated by the eastward-flowing Antarc-
29
tic Circumpolar Current (ACC). The characteristics of the ACC are not uniform around
30
the Southern Ocean. Rather, when the ACC encounters underwater mountain ranges
31
the flow is diverted, which causes these regions to be more energetic through the gen-
32
eration of ocean eddies in a process similar to atmospheric storm tracks. Numerical mod-
33
els have suggested that the exchange of properties, such as heat and carbon dioxide, be-
34
tween the atmosphere and the interior ocean is enhanced in these energetic regions. In
35
this study, data from freely-floating robotic floats in the Southern Ocean is used to ob-
36
serve the vertical structure of dissolved oxygen. Transfer of properties between the ocean’s
37
surface and the interior ocean preferentially occurs in high energy regions of the ACC.
38
Most previous work has relied on numerical models of the ocean that, due to computa-
39
tional limits, do not represent all aspects of the ACC’s energetic regions. This study has
40
implications for how the Southern Ocean’s ability to take up excess carbon dioxide from
41
the atmosphere will evolve in the future.
42
1 Introduction
43
The Southern Ocean is a key region for the ventilation and formation of interme-
44
diate and deep water masses. Tilted density surfaces associated with the Antarctic Cir-
45
cumpolar Current (ACC) allow for the adiabatic upwelling of Circumpolar Deep Water
46
(CDW) that has been sequestered from the surface for
O
(100-1000 years). At the sur-
47
face, CDW exchanges heat and gases with the atmosphere, outgassing natural carbon
48
stocks and acting as a sink for anthropogenic CO
2
(Gruber et al., 2019; Landschtzer et
49
al., 2015). Numerical models suggest that ventilation is spatially heterogeneous within
50
the ACC (Viglione & Thompson, 2016; Tamsitt et al., 2017). Interactions of the ACC
51
with underwater topography can result in the diversion and compaction of frontal cur-
52
rents, creating standing meanders (Sokolov & Rintoul, 2007) associated with enhanced
53
mesoscale eddy kinetic energy (EKE; Figure 1a,b) (Gille & Kelly, 1996). The ACC’s ma-
54
jor standing meanders are present at the Kergeulen Plateau, Campbell Plateau, East-
55
ern Pacific Rise, Crozet Plateau, and Drake Passage; these regions are thought to shape
56
uptake and sequestration of heat and carbon (Salle et al., 2012; Roach et al., 2016; Klocker,
57
2018; Brady et al., 2021).
58
Ventilation in the ACC depends on the local density structure as well as advection
59
and stirring along isopycnals, and thus responds to a variety of processes and scales. Stand-
60
ing meanders lead to sloped isopycnals that store available potential energy (Bischoff &
61
Thompson, 2014; Chapman et al., 2015; Klocker, 2018), which is released by baroclinic
62
instability, producing a rich mesoscale
O
(100 km) eddy field approximately 100 km down-
63
stream of the standing meander (Thompson & Naveira Garabato, 2014; Rintoul, 2018).
64
–2–
Accepted
Article
This article is protected by copyright. All rights reserved.
manuscript submitted to
Geophysical Research Letters
These eddies then stir and strain the surface density field, leading to frontogenesis and
65
influencing submesoscale motions (Klein & Lapeyre, 2009; Rosso et al., 2015; Balwada
66
et al., 2018; Bachman & Klocker, 2020). Through both lateral (Abernathey & Marshall,
67
2013; Roach et al., 2018) and vertical (Klein & Lapeyre, 2009; Adams et al., 2017) mo-
68
tions, mesoscale and submesoscale eddies contribute significantly to ventilation in the
69
ACC. Throughout this work, we refer to “ventilation” as any process or combination of
70
processes that work to transfer surface waters and tracers into the pycnocline, which as
71
described above, can occur on a variety of temporal and spatial scales (Morrison et al.,
72
2022). Additionally, stirring refers to the advection of tracers by an eddying velocity field,
73
while mixing is an irreversible process that removes tracer variance; only the former con-
74
tributes directly to ventilation although mixing influences the interpretation of ventila-
75
tion from tracer distributions (Villermaux, 2019).
76
Numerical models demonstrate that regions with higher EKE have enhanced ca-
77
pacity for submesoscale transport of tracers across the base of the mixed layer (Lvy et
78
al., 2018; Balwada et al., 2018; Uchida et al., 2020) and can have an outsized impact on
79
ventilation (Naveira Garabato et al., 2011; Viglione & Thompson, 2016; Tamsitt et al.,
80
2016; Rintoul, 2018). Standing meanders have also been identified as regions where older
81
waters enriched in dissolved inorganic carbon are preferentially transported to the sur-
82
face (Tamsitt et al., 2017; Brady et al., 2021), which can potentially create local regions
83
of enhanced air-sea gas exchange. Observational studies are needed to validate these largely
84
numerical results.
85
Due to coarse, ship-based sampling, examination of spatial variations in ventila-
86
tion have focused on the basin (or ACC sector) scale (Salle et al., 2012; Morrison et al.,
87
2022). More recently, observations from floats have shown that air-sea fluxes of carbon
88
(Gray et al., 2018) and oxygen (Bushinsky et al., 2017) vary across the Southern Ocean.
89
Evidence for finer-scale variability in biogeochemical distributions comes from the anal-
90
ysis of Biogeochemical Argo (BGC-Argo) profiles, in which subsurface tracer anomalies
91
are found to be more prevalent in high-EKE regions, suggesting stronger ventilation and
92
export (Llort et al., 2018). High-resolution glider observations near the Southwest In-
93
dian Ridge also showed reduced vertical tracer gradients in the standing meander as com-
94
pared to the low-EKE region downstream (Dove et al., 2021). Although these observa-
95
tional studies have provided initial evidence for the importance of standing meanders in
96
ventilation, the physical processes in the ACC that set the dominant spatial and tem-
97
poral scales of variability in surface-interior exchange have not yet been fully explored.
98
This study uses the broad spatial coverage of subsurface dissolved oxygen measure-
99
ments collected by the BGC-Argo array, as well as remote sensing products, to consider
100
controls on apparent oxygen utilization (AOU) patterns in the Southern Ocean and its
101
relationship to ventilation of surface waters. Both vertical and isopycnal distributions
102
of AOU exhibit substantial variations along the path of the ACC that can be linked pri-
103
marily to enhanced ventilation in the ACC’s major standing meanders. We identify sev-
104
eral physical mechanisms that are consistent with these distributions. This work is a crit-
105
ical step for validating ocean models and observationally describing key regions of ven-
106
tilation of climatologically-important tracers in the Southern Ocean.
107
2 Data and Methods
108
2.1 Biogeochemical-Argo floats
109
The Argo program has deployed over 10,000 profiling floats across the global ocean
110
since 1999 (Riser et al., 2016) with the Southern Ocean Carbon and Climate Observa-
111
tions and Modeling (SOCCOM) program playing a vital role in increasing the BGC-Argo
112
population of the Southern Ocean (Claustre et al., 2020; Johnson et al., 2017). Argo floats
113
–3–
Accepted
Article
This article is protected by copyright. All rights reserved.
manuscript submitted to
Geophysical Research Letters
sample the upper 2000 m of the ocean every 10 days. In between profiles, the floats drift
114
at 1000 m and follow a quasi-Lagrangian trajectory. (Roemmich et al., 2009).
115
This study uses 21,941 profiles of dissolved oxygen, along with the associated tem-
116
perature and salinity profiles, that were collected within the boundaries of the ACC (de-
117
fined in section 2.3) during the period January 15, 2003 to May 16, 2021 (Figure 1c-e).
118
Only data that have undergone delayed-mode quality control procedures and have been
119
flagged as ”good” are used in this analysis. All profile data were obtained from the “Sprof”
120
files provided by the Argo Global Data Assembly Center (GDAC), which merge biogeo-
121
chemical samples that are measured at slightly different vertical positions onto a single
122
common pressure axis.
123
KP
CrP
EPR
DP
CP
(a)
(d)
(e)
(b)
(c)
Longitude
Latitude
CP
EPR
DP
CrP
KP
Figure 1.
(a) Bathymetry and major fronts of the Southern Ocean. Gray contours show
1000 m, 2000 m, and 3000 m isobaths. Fronts are the Subantarctic Front (SAF; blue), Polar
Front (PF; orange), Southern ACC Front (SACCF; green), and the Southern Boundary (S Bdy;
red). (b) Base-10 logarithm of eddy kinetic energy [log
10
m
2
s
−
2
]. Black solid lines show the
ACC boundaries used in this study. Black dotted lines denote regions of high EKE. Standing
meanders are labeled by the corresponding bathymetric feature: Crozet Plateau (CrP), Kerguelen
Plateau (KP), Campbell Plateau (CP), East Pacific Rise (EPR), and Drake Passage (DP). (c)
Spatial distribution of float profiles containing oxygen data across the Southern Ocean within
the ACC; ∆latitude = 1.25
◦
, ∆longitude = 2.5
◦
. Black dotted lines show the ACC boundaries
used in this study. (d) Histogram of the number of float profiles as a function of longitude within
the ACC boundaries in panel (b). Profiles categorized as low-EKE are in orange, with high-EKE
profiles are in blue. Standing meanders are labeled the same as in panel (b). (e) Histogram of the
number of float profiles at a given latitude within the ACC boundaries in panel (b). Colors are
the same as in panel (d).
–4–
Accepted
Article
This article is protected by copyright. All rights reserved.
manuscript submitted to
Geophysical Research Letters
2.2 Derived Variables
124
Apparent oxygen utilization (AOU) is the difference between oxygen saturation con-
125
centration and observed dissolved oxygen concentrations (AOU =
O
sat
.
2
−
O
obs
.
2
), where
126
the oxygen saturation is a function of observed conservative temperature and absolute
127
salinity. AOU in the surface ocean is typically close to 0 due to equilibration with the
128
atmosphere. Bushinsky et al. (2017) showed that AOU
≈
0 is generally true for the ACC,
129
but small variations of
±
5-10
μ
mol kg
−
1
exist due to biological activity, surface heat fluxes,
130
or rapid entrainment of thermocline waters (Ito et al., 2004). Lower AOU values are used
131
as a proxy for younger age, signaling recent ventilation, since respiration in the ocean
132
interior is a persistent oxygen sink. AOU is a non-conservative tracer with its value de-
133
termined by several processes,
e.g.
remineralization, along-isopycnal stirring, cross-isopycnal
134
mixing, and the non-conservative nature of solubility. AOU has been used to trace path-
135
ways between the surface and interior (Llort et al., 2018), and both vertical and along-
136
isopycnal variations provide insight into ventilation dynamics.
137
We study the distribution of AOU in both density and depth coordinates. Addi-
138
tionally, to account for temporal and spatial variations in mixed layer depths, vertical
139
variations in AOU are also considered as deviations from observed values at the base of
140
the mixed layer in each profile. Depth below the base of the mixed layer is given by ∆h,
141
and ∆AOU refers to the difference in AOU between the value at ∆h and at the mixed
142
layer depth. The mixed layer depth (MLD) was defined by a density difference criterion
143
of 0.03 kg m
−
3
from the surface (Montgut et al., 2004). Other derived variables, such
144
as potential density, were calculated from temperature and salinity using the Thermo-
145
dynamic Equation of Seawater 2010 (McDougall & Barker, 2011).
146
2.3 Satellite Data
147
Eddy kinetic energy (EKE) was calculated as EKE=
1
2
√
u
′
2
+
v
′
2
, where
u
′
and
148
v
′
are the zonal and meridional eddy geostrophic velocities estimated from the time-varying
149
sea surface height (SSH) anomaly field, and
(
.
) represents a time average calculated over
150
1993-2016. Regions with EKE greater than 250 cm
2
s
−
2
were considered “high-EKE”
151
(Figure 1b, Figure S1), and individual float profiles were tagged as “high” or “low” EKE
152
based on their surfacing locations. Previous studies have identified distinct dynamical
153
regimes within individual standing meanders (Youngs et al., 2017; Barthel et al., 2017),
154
but we do not distinguish these here.
155
The ACC boundaries were defined using absolute dynamic topography (ADT) with
156
the northern and southern boundaries given by the -0.1 m and the -1.05 m ADT con-
157
tours, respectively. These boundaries were selected in part to avoid inclusion of the Ag-
158
ulhas Retroflection, which is a region of enhanced EKE but is not considered in this study.
159
Several definitions of the northern and southern boundaries of the ACC were tested, in-
160
cluding hydrographic definitions of frontal boundaries (shown in Figure 1a) as opposed
161
to sea level anomaly (Kim & Orsi, 2014), but these led to minimal differences in the re-
162
sults.
163
Finite-size Lyapunov Exponents (FSLEs) describe the orientation and timescale
164
of strain fields by quantifying stretching and compression (d’Ovidio et al., 2004; dOvidio
165
et al., 2010). They are a Lagrangian diagnostic, and for a given flow field are defined as
166
the separation growth rate for particle pairs,
λ
(
d
0
,d
f
) =
1
τ
log
(
d
f
d
0
), where
d
0
and
d
f
are
167
the initial and final separation distances and
τ
is the first time where separation distance
168
d
f
is reached. Here we use FSLE estimates provided by AVISO+ that were computed
169
from satellite-derived geostrophic velocities. We use the FSLEs from January 1, 2018 to
170
December 31, 2020, but the exact choice of period does not impact the results.
171
–5–
Accepted
Article
This article is protected by copyright. All rights reserved.
manuscript submitted to
Geophysical Research Letters
3 Results
172
3.1 Subsurface signatures of ventilation
173
Variations in AOU with depth and density may arise from various mechanisms, some
174
reflecting differences in advection and stirring at scales smaller than the standing me-
175
ander (discussed in section 3.2), and others related to the larger-scale density structure
176
of the ACC. A simple partitioning of individual float profiles between high- and low-EKE
177
regions reveals striking differences in the vertical structure of AOU between these two
178
regimes (Figure 2). Just below the mixed layer,
e.g.
∆h = 100 m, high- and low-EKE
179
regions both have low ∆AOU values with similar distributions (Figure 2a), although low-
180
EKE regions have a longer tail. At ∆h = 300 m, high- and low-EKE regions have dis-
181
tinct peaks with the high-EKE region having a lower median; the difference in the dis-
182
tributions’ medians is 27
μ
mol kg
−
1
and the difference in the modes is 68
μ
mol kg
−
1
(Fig-
183
ure 2b,d). For values of ∆h
≥
700 m, the two regions have approximately the same dis-
184
tribution, with a difference in medians of only 2
μ
mol kg
−
1
(Figure 2c,d). The largest
185
disparity in ∆AOU between the high and low-EKE regions is present for 200
<
∆h
<
186
700 m. (Figure 2d). This ∆AOU structure is set, in part, by meanders of the ACC that
187
horizontally transport lighter waters southward into energetic regions downstream of to-
188
pography. At the level of individual standing meanders, the high-EKE regions associ-
189
ated with Kerguelen Plateau, Campbell Plateau, and Eastern Pacific Rise have distri-
190
butions of ∆AOU that most closely align with the median distributions for the entire
191
ACC (Figure 2f.iii, iv, v). The distinction between high- and low-EKE regions is weak-
192
est at the Crozet Plateau (Figure 2f.ii), although data availability is reduced here.
193
Changes in hydrographic properties along the path of the ACC provide insight into
194
the origin of subsurface low-AOU waters found in high-EKE regions. CDW is distinguished
195
by high salinity (
>
34.6 g kg
−
1
) and low temperature (
∼
2
◦
C). Comparatively, Antarc-
196
tic Intermediate Water (AAIW), a more-recently ventilated water mass, is characterized
197
by lower salinity as a result of sea ice melt. Differences in hydrographic properties are
198
particularly distinct around ∆h = 300 m, consistent with large differences in ∆AOU me-
199
dians between the high- and low-EKE regions (Figure 3). In both the high- and low-EKE
200
regions at ∆h = 300 m, the distributions of mean AOU as a function of temperature and
201
salinity are similar (Figure 3a-c), suggesting that ∆AOU is predominantly tied to the
202
relative contributions of water masses below the mixed layer, with variations due to bi-
203
ology secondary. Stronger differences between the two regions are found, however, when
204
considering the frequency distribution in conservative temperature-absolute salinity space
205
(Figure 3d-f). In the low-EKE regions, CDW properties dominate, with a temperature
206
of 2
◦
C and high salinity (34.4 - 34.8 g kg
−
1
; Figure 3d). In the high-EKE regions a greater
207
fraction of the observations have lower values of salinity (34.0 - 34.2 g kg
−
1
) and also
208
warmer temperatures (3-5
◦
C; Figure 3e,f), consistent with intermediate waters that have
209
been subducted from the surface. The increased presence of waters consistent with AAIW
210
at these depths in the high-EKE regions suggests more intermediate water is subducted
211
in high-EKE regions of the ACC as compared to low-EKE regions.
212
This hydrographic analysis indicates that mixing of old CDW and recently venti-
213
lated AAIW at the basin-scale contributes to the patterns in ∆AOU described in Fig-
214
ure 2. Yet, coupled processes on the submesoscale-mesoscale spectrum may still play a
215
role in setting these subsurface ∆AOU distributions, as described in previous observa-
216
tional work in standing meanders (Dove et al., 2021). Using an oxygen utilization rate
217
(OUR) for the upper mesopelagic zone of 40
μ
mol kg
−
1
year
−
1
(Hennon et al., 2016),
218
low ∆AOU waters with a median of O(70
μ
mol kg
−
1
) in high-EKE regions would have
219
an age of
∼
2 years, suggesting there may be recent injection from the mixed layer. How-
220
ever, estimates of OUR in the Southern Ocean are sparse and there is a good deal of un-
221
certainty in the estimate of this time scale. Specifically, an OUR of 40
μ
mol kg
−
1
year
−
1
222
represents a regional, near-surface value that may not be representative of values at greater
223
depths or over the broader Southern Ocean. Therefore this OUR value should be con-
224
–6–
Accepted
Article
This article is protected by copyright. All rights reserved.
manuscript submitted to
Geophysical Research Letters
(e)
(
i
)
(
v
)
(
iii
)
(
iv
)
Δh
=
400 m
Δh
= 100 m
Δh
= 300 m
Δh
= 700 m
(a)
(b)
(c)
(
f
)
(d)
(
ii
)
Figure 2.
Probability density functions across the full ACC of ∆AOU [
μ
mol kg
−
1
] where (a)
∆h = 100 m, (b) ∆h = 400 m, (c) ∆h = 700 m. (d) Median in ∆AOU [
μ
mol kg
−
1
] at values of
∆h. (e) Variance of AOU [
μ
mol
2
kg
−
2
] on potential density surfaces. (f) Locations of profiles
used to create probability density functions of ∆AOU [
μ
mol kg
−
1
] at ∆h = 300 m at (i) Drake
Passage, (ii) Crozet Plateau, (iii) Kerguelen Plateau, (iv) Campbell Plateau, and (v) Eastern Pa-
cific Rise. In all panels, blue colors denote high-EKE regions and orange colors denote low-EKE
regions.
–7–
Accepted
Article
This article is protected by copyright. All rights reserved.
manuscript submitted to
Geophysical Research Letters
(a)
(
f
)
(e)
(d)
(c)
(b)
Figure 3.
Absolute salinity (S
A
)-conservative temperature (CT) diagrams. Average AOU for
each S
A
-CT position at ∆h = 300 m in (a) the low-EKE regions and (b) the high-EKE regions.
(c) Difference in AOU between the low and high-EKE regions. Joint histogram of profile loca-
tions at at ∆h = 300 m in (d) the low-EKE regions and (e) the high-EKE regions. (f) Difference
in joint histograms between the low- and high-EKE regions. Grey contours are potential density
[kg m
−
3
], with the black contour at 27.2 kg m
−
3
. In all panels, only where there were more than
five points at a given CT-S
A
value that could be averaged are shown. ∆CT = 0.2
◦
C, ∆S
A
=
0.025 g kg
−
1
.
sidered an upper bound, and the low ∆AOU waters observed in high-EKE regions likely
225
include waters that have been subducted below the surface boundary layer for periods
226
longer than 2 years.
227
It is important to also consider AOU variations on density surfaces because the mean
228
density structure between the high and low-EKE regions is different: the lower EKE re-
229
gions host denser isopycnals linked to deeper depths and higher AOU values in the mid-
230
and low-latitude basins to the north (Figure S2a,b). These variations along the path of
231
the ACC are related to changes in outcropping density classes as well as the steepening
232
of lateral density gradients within standing meanders (Thompson & Naveira Garabato,
233
2014; Chapman et al., 2015), which may enable recently-ventilated surface waters to be
234
displaced downward in the water column. Despite the different density ranges between
235
the regions, the vertical stratification, measured by the vertical buoyancy gradient
N
2
,
236
is similar (Figure S2c). Considering the Argo observations in density space shows that
237
the heavier isopycnals have relatively homogeneous mean AOU distributions along the
238
path of the ACC, which is likely a result of the rapid along-ACC circulation (Figure S3,
239
S4). However, lighter isopycnals and regions where isopycnals are shallower show inho-
240
mogeneities in mean AOU along the path of the ACC, due in part to the outcropping
241
–8–
Accepted
Article
This article is protected by copyright. All rights reserved.
manuscript submitted to
Geophysical Research Letters
of denser isopycnals in low-EKE regions. Some of the signal of lower mean AOU con-
242
centrations, particularly at deeper depths, may be attributed to adiabatic heaving rather
243
than ventilation by advection and mixing. In the next section, we offer evidence that high-
244
EKE regions are subject to more energetic stirring, leading to enhanced along-isopycnal
245
variance of AOU, suggesting that AOU variations do not result from isopycnal heaving
246
alone.
247
3.2 Mesoscale and submesoscale contributions to ventilation
248
Variations in AOU due to the ACC’s density structure occur at standing meander
249
and larger scales (
≥
1000 km); below these scales, mesoscale and submesoscale motions
250
can impact ventilation through a number of different mechanisms. These include (1) in-
251
creased along-isopycnal stirring as a result of enhanced EKE; (2) frontal subduction as
252
a result of frontogenesis in the standing meander; and (3) enhanced vertical transport
253
by submesoscale motions. Here, we investigate how these processes shape along-ACC dif-
254
ferences in AOU distributions.
255
Differences in isopycnal AOU variance between high- and low-EKE regions offers
256
insight into how along-isopycnal stirring contributes to ventilation within the ACC. To
257
remove the effects of vertical isopycnal displacement (i.e. heave; Figure S3), we consider
258
deviations from a longitude-dependent (10-degree longitude bins), along-isopycnal mean
259
AOU value. The variance in AOU on density surfaces
<
27.4 is up to 18% larger in high-
260
EKE regions than in low-EKE regions, with a peak in variance at 27.3 kg m
−
3
in both
261
regions. The observed enhanced AOU variance in high-EKE regions is consistent with
262
along-isopycnal stirring bringing low-∆AOU waters to depth, as opposed to this signal
263
solely occurring due to variations in the ACC’s large-scale density structure. Enhanced
264
variance in the high-EKE regions may arise from both stirring processes and injection
265
of tracer anomalies from the surface layer onto density surfaces below the mixed layer.
266
With regard to exchange out of the mixed layer, seasonal or along-stream changes in mixed
267
layer properties may be expected to modify ventilation. However, the float data indi-
268
ciate that MLD and stratification at the base of the mixed layer are similar in high- and
269
low-EKE regions and therefore do not contribute to the disparity in subsurface ∆AOU
270
distributions (Figure S5).
271
In addition to being regions of energetic mesoscale eddies, ACC standing mean-
272
ders are regions of strong surface frontogenesis that give rise to large mixed layer lateral
273
density gradients. These gradients are reservoirs of potential energy that can give rise
274
to instabilities that lead to intense submesoscale vertical motions and increase the ef-
275
ficiency of tracer transport between the surface and interior ocean, contributing to ven-
276
tilation (Klein & Lapeyre, 2009; Mahadevan, 2016; Lvy et al., 2018). While these insta-
277
bilities typically occur on spatial and temporal scales consistent with the submesoscale,
278
they are shaped by the mesoscale flow field (Rosso et al., 2015; Balwada et al., 2018, 2020).
279
We investigate the potential of enhanced ventilation occurring via frontogenesis and sub-
280
mesoscale subduction by considering the relative magnitude of lateral density gradients
281
between low- and high-EKE regions. Measuring lateral density gradients can be achieved
282
with high temporal and spatial resolution measurements, but such observations are sparse
283
in the Southern Ocean. Siegelman et al. (2020) empirically showed that maximum stretch-
284
ing FSLEs (hereafter FSLEs) calculated from satellite-derived flow fields can be used to
285
approximate the magnitude of lateral density gradients and derived a relationship be-
286
tween the two quantities. Specifically, density anomalies are physically aligned with FSLEs,
287
so larger magnitude FSLEs are correlated with stronger lateral density gradients. While
288
FSLEs have previously been linked to mixed-layer density gradients, Siegelman et al. (2020)
289
demonstrated that this relationship may extend below the mixed layer in the Southern
290
Ocean, particularly in energetic regions.
291
–9–
Accepted
Article
This article is protected by copyright. All rights reserved.
manuscript submitted to
Geophysical Research Letters
(a)
(d)
(c)
(b)
Probability Density
Probability Density
Figure 4.
(a) Snapshot of FSLEs from March 1, 2020 centered on the Crozet Plateau region
of the ACC. Blue regions are high-EKE while orange are low-EKE, using the EKE definition de-
fined in the methods. (b) Probability density function of maximum stretching FSLE in the high-
(blue) and low- (orange) EKE regions across the full ACC. Gray lines are expanded in (c). (c)
Probability density function of maximum stretching FSLE at in the Kergulen Plateau (red), the
Crozet Plateau (black), the Campbell Plateau (yellow), Drake Passage (green), and the Eastern
Pacific Rise (EPR; blue). Vertical lines represent 75th percentiles with colors the same as above.
(d) Plot of the mode of the FSLE in the high-EKE region versus the differences in the high and
low medians of the ∆AOU probability density functions (
δ
AOU) at ∆h = 250 m. Error bars are
standard deviations of
δ
AOU and colors are the same as in (c).
Consistent with the heterogeneous distribution of EKE in the ACC, lateral den-
292
sity gradients (as inferred from FSLEs) undergo abrupt transitions in standing mean-
293
der regions (Figure 4a). The probability distribution of FSLE has a log-normal distri-
294
bution within both low- and high-EKE regions (Figure 4b). However, in the high-EKE
295
region, the median value is shifted to larger magnitudes and the distribution has a longer
296
tail, which we link to stronger and more frequent small-scale surface density gradients.
297
The FSLE probability density function also differs for each individual standing mean-
298
der (Figure 4c). The standing meander at Kerguelen Plateau has the most negative (strongest)
299
FSLE values, implying an increased frequency of strong lateral density gradients and po-
300
tentially enhanced vertical transport. The standing meanders associated with the Crozet
301
Plateau and Eastern Pacific Rise have the least negative (weakest) mode of FSLE prob-
302
ability, with the Campbell Plateau and Drake Passage falling between the extremes.
303
To consider the relationship between FSLEs and ∆AOU within individual stand-
304
ing meanders, we define localized low-EKE regions that surround each high-EKE stand-
305
ing meander, defined between the north-south ACC boundaries and extending 5 degrees
306
of longitude to either side of the meander. For each of the five major standing meander
307
regions, the median difference in ∆AOU between the high- and
localized
low-EKE re-
308
gions at ∆h = 250 m is calculated; we refer to this as
δ
AOU. A large magnitude of
δ
AOU
309
represents large differences in ∆AOU distributions between the high-EKE standing me-
310
–10–
Accepted
Article
This article is protected by copyright. All rights reserved.
manuscript submitted to
Geophysical Research Letters
ander and the surrounding low-EKE region, while a negative
δ
AOU indicates a greater
311
volume of low-AOU water in the high-EKE region. In other words, a large, negative value
312
of
δ
AOU suggests that the high-EKE region experiences enhanced ventilation as a re-
313
sult of the stirring and submesoscale subduction processes described above, as compared
314
to the surrounding low-EKE region. Differences in FSLE distributions between mean-
315
ders are correlated with differences in
δ
AOU (Figure 4d) for depths of ∆h up to 500 m.
316
The standing meander that has the largest FSLE mode magnitude (implying strongest
317
stirring), Kerguelen Plateau, is associated with the largest
δ
AOU. Standing meander re-
318
gions with smaller magnitude FSLE modes, the Eastern Pacific Rise and Crozet Plateau,
319
have
δ
AOU values closer to zero. While five meanders dominate the high-EKE regions
320
in the ACC, this analysis suggests that contributions of low ∆AOU waters to depth may
321
be localized to only one or two intense standing meanders, Kerguelen and Campbell plateaus,
322
indicating these standing meanders may play the dominant role in ventilation of the ACC.
323
4 Discussion
324
Ventilation of surface properties and tracers can arise from a combination of large-
325
scale circulation features,
e.g.
shaping of density surfaces through flow-topography in-
326
teractions, as well as smaller-scale stirring by mesoscale and submesoscale motions. There
327
is increasing evidence from both observational and numerical studies that motions oc-
328
curring in the mesoscale and submesoscale range are tightly coupled. Mesoscale strain,
329
through the process of frontogenesis, is responsible for the generation of near-surface lat-
330
eral density gradients that are precursors for strong vertical submesoscale velocities (Archer
331
et al., 2020; Su et al., 2020). A possible scenario is that these enhanced submesoscale
332
velocities in high-EKE regions rapidly inject surface properties to depths of 300-m or even
333
deeper. Llort et al. (2018) did indeed find evidence of deep, unmodified waters (anoma-
334
lously low AOU values), which only occurred in high-EKE standing meander regions.
335
Yet, these deep anomalies were found in
<
1% of all float profiles. An alternate scenario,
336
in line with Balwada et al. (2018, 2021) and Freilich and Mahadevan (2021), and one that
337
is more consistent with the observed ∆AOU values, is that submesoscale motions play
338
the essential role of efficiently carrying surface properties across the base of the mixed
339
layer. After this, stirring along isopycnals, by the same eddies that create the surface den-
340
sity gradients, enhances the transfer of these surface properties to depth. Thus, while
341
attributing ventilation to different physical processes is important for ensuring that they
342
are represented accurately in climate models, the coupling of motions across scales likely
343
makes this task challenging. Accordingly, numerical models that do not fully resolve mesoscale
344
and submesoscale processes may misrepresent the formation of intermediate waters, as
345
well as the concentration of oxygen in thermocline.
346
While this study has focused on ventilation pathways of oxygen in the ACC, these
347
results likely have important implications for the spatial variability of air-sea CO
2
fluxes.
348
Oxygen has an equilibrium timescale that is at least an order of magnitude shorter than
349
that of CO
2
, which has an equilibration timescale of
O
(6 months). Combining this study
350
with evidence that ACC standing meanders are also sites of enhanced upwelling (Tamsitt
351
et al., 2017; Brady et al., 2021) suggests that recently-ventilated deep waters in these
352
regions may have short surface residence times, and therefore full equilibration with at-
353
mospheric CO
2
may not be reached (Jones et al., 2014). This provides further motiva-
354
tion for exploring how localized high-EKE regions impact exchange of waters between
355
the surface and interior and the larger Southern Ocean carbon cycle.
356
Various estimates of air-sea exchange of CO
2
in the Southern Ocean have identi-
357
fied interannual to decadal-scale variations in the region’s ability to provide an atmo-
358
spheric carbon sink (Landschtzer et al., 2015; Gruber et al., 2019). Notably, these es-
359
timates are obtained after some form of interpolation or mapping,
e.g.
neural network
360
(Landschtzer et al., 2016), of CO
2
measurements from repeat shiptracks that typically
361
do not sample the strongest and most variable EKE regions. Decadal-scale variations
362
–11–