Observational Evidence of Cold Filamentary Intensification in an Energetic Meander
of the Antarctic Circumpolar Current
M
AYA
I. J
AKES
,
a,b,c
H
ELEN
E. P
HILLIPS
,
a,c,d
A
NNIE
F
OPPERT
,
a,c
A
JITHA
C
YRIAC
,
e,a
N
ATHANIEL
L. B
INDOFF
,
c,a,d
S
TEPHEN
R. R
INTOUL
,
f,c
AND
A
NDREW
F. T
HOMPSON
g
a
Institute for Marine and Antarctic Studies, University of Tasmania, Hobart, Tasmania, Australia
b
ARC Centre of Excellence for Climate Extremes, Hobart, Tasmania, Australia
c
Australian Antarctic Partnership Program, Hobart, Tasmania, Australia
d
Australian Centre for Excellence in Antarctic Science, Hobart, Tasmania, Australia
e
CSIRO Environment, Perth, Western Australia, Australia
f
CSIRO Environment, Hobart, Tasmania, Australia
g
Environmental Science and Engineering, California Institute of Technology, Pasadena, California
(Manuscript received 19 May 2023, in
fi
nal form 13 December 2023, accepted 26 December 2023)
ABSTRACT: Eddy stirring at mesoscale oceanic fronts generates
fi
nescale
fi
laments, visible in submesoscale-resolving
model simulations and high-resolution satellite images of sea surface temperature, ocean color, and sea ice. Submesoscale
fi
laments have widths of
O
(1
–
10) km and evolve on time scales of hours to days, making them extremely challenging to
observe. Despite their relatively small scale, submesoscale processes play a key role in the climate system by providing a
route to dissipation; altering the strati
fi
cation of the ocean interior; and generating strong vertical velocities that exchange
heat, carbon, nutrients, and oxygen between the mixed layer and the ocean interior. We present a unique set of in situ and
satellite observations in a standing meander region of the Antarctic Circumpolar Current (ACC) that supports the theory
of cold
fi
lamentary intensi
fi
cation
}
revealing enhanced vertical velocities and evidence of subduction and ventilation asso-
ciated with
fi
nescale cold
fi
laments. We show that these processes are not con
fi
ned to the mixed layer; EM-APEX
fl
oats re-
veal enhanced downward velocities (
.
100 m day
2
1
) and evidence of ageostrophic motion extending as deep as 1600 dbar,
associated with a
;
20-km-wide cold
fi
lament. A
fi
ner-scale (
;
5kmwide)cold
fi
lament crossed by a towed Triaxus is asso-
ciated with anomalous chlorophyll and oxygen values extending at least 100
–
200 dbar below the base of the mixed layer,
implying recent subduction and ventilation. Energetic standing meanders within the weakly strati
fi
ed ACC provide an en-
vironment conductive to the generation of
fi
nescale
fi
laments that can transport water mass properties across mesoscale
fronts and deep into the ocean interior.
KEYWORDS: Frontogenesis/frontolysis; Ocean dynamics; Small scale processes; Vertical motion; In situ oceanic
observations; Satellite observations
1. Introduction
The Southern Ocean plays a central role in the climate sys-
tem, connecting the major ocean basins and allowing a global
overturning circulation to exist (Rintoul 2000
; Rintoul et al.
2001 ). The Antarctic Circumpolar Current (ACC)
fl
ows from
west to east around the Antarctic continent and consists of
three major fronts, from north to south
}
the Subantarctic
Front (SAF), the Polar Front (PF), and the Southern ACC
Front (SACCF) (
Klinck and Nowlin 2001
; Olbers et al. 2004
).
The fronts are often aligned with particular circumpolar sea
surface height (SSH) streamlines (
Sokolov and Rintoul 2009
)
and are associated with strong zonal jets that regulate poleward
heat transport (
Naveira Garabato et al. 2011
). Large-scale up-
welling of deep waters from the ocean interior is facilitated by
the steeply sloping isopycnals across the ACC (
Rintoul et al.
2001;
Olbers et al. 2004;
Thompson 2008
; Tamsittetal.2017
).
This large-scale upwelling and ventilation of the deep ocean is
fundamentally important for heat and carbon uptake in the
Southern Ocean, where
;
70% of global anthropogenic heat
(Fr
̈
olicher et al. 2015
)and
;
40% of anthropogenic carbon
(Caldeira and Duffy 2000
; Le Quéré et al. 2007
; Khatiwala
et al. 2009
) is absorbed into the ocean.
Southern Ocean circulation and dynamics vary around the
Antarctic continent, particularly in regions where the ACC in-
teracts with major topographic features. Flow
–
topography in-
teractions generate standing meanders and rich eddy
fi
elds
that are hotspots of poleward heat transport (
Phillips and
Rintoul 2000
; Foppert et al. 2017
), large-scale upwelling
(Tamsitt et al. 2017
; Rintoul 2018
), biological activity (
Siegelman
et al. 2019
), carbon sequestration (
Sallée et al. 2012;
Langlais et al.
2017), and ventilation of the ocean interior (
Dove et al. 2022).
Standing meanders form due to the westward propagation of to-
pographically generated Rossby waves that balance the eastward
mean
fl
ow of the ACC (
Hughes 2005
; Thompson and Naveira
Garabato 2014;
Meijer et al. 2022
; Zhang et al. 2023a). Complex
eddy
–
mean
fl
ow interactions within the meander generate bar-
otropic and baroclinic instabilities, resulting in rapid eddy ki-
netic
energy (EKE) generation (MacCready and Rhines
2001
; Youngs et al. 2017
; Foppert 2019
). Vertical motion at
Denotes content that is immediately available upon publica-
tion as open access.
Corresponding author
: Maya Jakes, maya.jakes@utas.edu.au
DOI: 10.1175/JPO-D-23-0085.1
Ó
2024 American Meteorological Society. This published article is licensed under the terms of the default AMS reuse license. For information regarding
reuse of this content and general copyright information, consult the AMS Copyright Policy (
www.ametsoc.org/PUBSReuseLicenses
).
JAKES ET AL.
717
M
ARCH
2024
Brought to you by Caltech Library | Unauthenticated | Downloaded 12/04/24 08:10 PM UTC
the mesoscale is tied to the phase of the meander (Phillips
and Bindoff 2014
), driven by the ageostrophic component of
the gradient-wind
fl
ow
fi
eld resulting from the meander cur-
vature (
Meijer et al. 2022
). At smaller temporal and spatial
scales, submesoscale dynamics in energetic meander regions
can drive localized subduction and tracer exchange between
the mixed layer and the ocean interior (Rosso et al. 2014
,
2015; Llort et al. 2018
; Balwada et al. 2018
; Taylor et al. 2018
;
Freilich and Mahadevan 2021
; Dove et al. 2021
; Morrison
et al. 2022
).
Submesoscale motions of
O
(10) km are distinct from meso-
scale motions of
O
(100) km in that they are not as strongly
constrained by Earth
’
s rotation
}
with a Rossby number close
to one, they have a strong ageostrophic component and can
develop much stronger vertical velocities in comparison to
mesoscale motions (
Thomas et al. 2008
; Gula et al. 2022
;
Taylor and Thompson 2023
). In addition to their relatively
small spatial scales, submesoscale features evolve on time
scales of hours to days, making them extremely challenging to
observe (
Siegelman et al. 2020a
; Archer et al. 2020
). Meso-
scale fronts provide an ideal environment for submesoscale
processes to occur due to the strong horizontal density gra-
dients that provide a source of potential and mean kinetic en-
ergy to fuel submesoscale instabilities (
Yu et al. 2019
). The
rich eddy
fi
elds in energetic meander regions act to stir the
large-scale density gradient and generate
fi
nescale fronts and
fi
laments (
Mahadevan and Tandon 2006
; Capet et al. 2008
;
Gula et al. 2022
). The strain
fi
eld between two counterrotat-
ing eddies in an eddy dipole structure can also generate
submesoscale fronts and instabilities that drive localized sub-
duction (
Klein and Lapeyre 2009
; Archer et al. 2020
).
Despite their relatively small scale, submesoscale processes
can in
fl
uence components of the large-scale ocean circulation
and Earth
’
s climate (
Su et al. 2018
; Taylor and Thompson
2023 ; Hewitt et al. 2022
; Swa
rt et al. 2023
). They provide an
important pathway for the cascade of energy and tracer vari-
ance from large scales to dissipative scales (
D
’
Asaro et al. 2011;
Gula et al. 2022) and are therefore essential to understanding
water mass variability and ocean mixing, as well as localized
subduction and ventilation. Submesoscale processes are particu-
larly prevalent in the
upper ocean and play a key role in the
vertical exchange of heat, nutrients, and carbon across the base
of the mixed layer
}
in
fl
uencing primary productivity and the
carbon cycle (
Lapeyre and Klein 2006
; Thomas et al. 2008
;
Mahadevan 2016;
Lévy et al. 2018
; Archer et al. 2020;
Siegelman
et al. 2020a
; Su et al. 2020). Submesoscale dynamics can alter the
depth of the surface mixed layer (
Lapeyre et al. 2006
; du Pleiss
et al. 2017
; Bachman et al. 2017;
du Pleiss et al. 2019;
Bachman
and Klocker 2020), controlling the connectivity between the at-
mosphere and the ocean interior, and thus in
fl
uencing the
strength of the global overturning circulation (
Fox-Kemper et al.
2011;
Swartetal.2023
).
A key process that alters the upper-ocean density structure
and connectivity to the ocean interior is frontogenesis.
Frontogenesis is well described in the atmospheric context
(Hoskins 1982
) and refers to the sharpening of horizontal density
gradients, driven by the mesoscale strain
fi
eld, generating an
along-front acceleration (McWilliams et al. 2009
; McWilliams
2021). During frontogenesis (
Fig. 1a
), the front can transition
from a state of thermal wind balance to turbulent thermal wind
balance, where turbulent mixing becomes a key component in
the momentum balance (
Hoskins and Bretherton 1972;
Gula
et al. 2014
; McWilliams et al. 2015
). Turbulent mixing reduces
the along-front vertical shear and leads to an unbalanced cross-
front pressure gradient. This drives a cross-front ageostrophic
secondary circulation (ASC) that acts to
fl
atten isopycnals and
restore the
fl
ow back to hydrostatic and thermal wind balance
(McWilliams et al.
2009;
Archer et al. 2020;
Gula et al. 2022).
This single-cell circulation results in downwelling on the dense
side of the front and upwelling on the light side (
Fig. 1a
).
The
ASC tends to be strongest where the magnitude of the horizontal
density gradient peaks (Gula et al. 2022). Vertical motion associ-
ated with an ASC can in
fl
uence vertical nutrient
fl
uxes and the
residence time of phytoplankton in the mixed layer (
Lévy et al.
2018), with consequences for primary production and marine
ecosystems.
Filamentogenesis is similar to frontogenesis but, instead of
a single cross-front circulation cell, there are two circulation
F
IG
. 1. Schematic of Southern Hemisphere (a) frontogenesis and (b),(c)
fi
lamentogenesis associated with cold and
warm
fi
laments, respectively. Blue arrows indicate the ageostrophic secondary circulation (ASC). Adapted from
McWilliams et al. (2009)
and Gula et al. (2022)
.
JOURNAL OF PHYSICAL OCEANOGRAPHY
V
OLUME
54
718
Brought to you by Caltech Library | Unauthenticated | Downloaded 12/04/24 08:10 PM UTC
cells
}
one on each side of the
fi
lament (
Figs. 1b,c
). In the case
of a cold (i.e., dense)
fi
lament (
Fig. 1b
), ageostrophic conver-
gence occurs at the surface in the core of the
fi
lament, aligned
with the con
fl
uence axis associated with the deformation
fl
ow
(McWilliams et al. 2009
). The
fi
lament narrows and there is
an intensi
fi
cation of downward velocities in the
fi
lament, with
weaker upwelling motions on either side. This process is
known as cold
fi
lamentary intensi
fi
cation (
McWilliams et al.
2009;
Gula et al. 2014). While warm
fi
lamentary intensi
fi
ca-
tioncanalsooccur(
Fig. 1c
)(Lapeyre and Klein 2006), the
surface divergence associated with the ASC acts to broaden
the
fi
lament and opposes the con
fl
uent
fl
ow
}
leading to
upwelling that has a slower intensi
fi
cation rate than the
downwelling associated with a cold
fi
lament (
McWilliams
et al. 2009).
During frontogenesis and
fi
lamentogenesis, submesoscale
instabilities can develop as they feed off the local potential
and kinetic energy of the intensifying front (
Boccaletti et al.
2007 ; Capet et al. 2008
; Fox-Kemper and Ferrari 2008
; Gula
et al. 2014
; Archer et al. 2020
). Submesoscale instabilities can
prevent further frontal sharpening (i.e., frontal arrest)
(McWilliams and Molemaker 2011
), enhance vertical motion
(Thomas et al. 2013
; Callies et al. 2016
; Yu et al. 2019
), restra-
tify the upper ocean (
Boccaletti et al. 2007
; Fox-Kemper and
Ferrari 2008
; Yu et al. 2019
),
drive localized subduction
(Thomas et al. 2013
; Freilich and Mahadevan 2021
), and facil-
itate a dynamical route to molecular diffusion by extracting
energy from the geostrophically balanced
fl
ow ( D
’
Asaro et al.
2011 ; Gula et al. 2022). Different types of submesoscale in-
stabilities can be distinguished by their energetics, using
indicators such as potential vorticity (PV), the buoyancy
frequency (
N
2
), and the Richardson number (Ri) (
Taylor
and Thompson 2023
; Gula et al. 2022).
While submesoscale fronts and instabilities are often con-
fi
ned to the mixed layer, ageostrophic motion associated with
frontogenesis can extend down to 400
–
900 m in the weakly
strati
fi
ed Southern Ocean (
Siegelman 2020
; Siegelman et al.
2020a
). Submesoscale processes can also extend deeper in the
water column in the presence of steep topography (Rosso
et al. 2014
; Gula et al. 2022
), in strain regions on the periphery
of mesoscale eddies (
Yu et al. 2019
; Siegelman et al. 2020b
;
Dove et al. 2021
), and during wintertime when strati
fi
cation
at the base of the mixed layer is weak (
Thompson et al. 2016
;
Erickson and Thompson 2018
; Yu et al. 2019
).
Many studies over the past decade discuss the importance
of submesoscale dynamics in the climate system; however,
there is a lack of
fi
nescale observations to test current theory
and understanding, particularly in the Southern Ocean. We
present a unique set of
fi
nescale observations from an energetic
meander region of the ACC, south of Tasmania, that supports
the theory of cold
fi
lamentary intensi
fi
cation. The meander forms
between two topographic features
}
downstream of the South-
east Indian Ridge (SEIR) and just upstream of Macquarie Ridge
(MR) (
Fig. 2
). Zhang et al. (2023a)
provide a detailed analysis of
the dynamics governing the meander formation in this region.
This particular meander has been identi
fi
ed in previous studies
as a hotspot of EKE (
Thompson and Naveira Garabato 2014
),
cross-frontal exchange (
Thompson and Sallée 2012
), poleward
eddy heat
fl
ux ( Foppert et al. 2017) and turbulent mixing (
Cyriac
et al. 2022
). The observations we present here were collected on
a voyage on the R/V
Investigator
in October 2018, and highlight
theroleof
fi
nescale
fi
laments in generating strong vertical
F
IG
. 2. Map of the study domain. Trajectories of three EM-APEX
fl
oats during rapid sampling
are indicated by the colored circles. Triaxus tows are shown as black solid lines. Background
color indicates ocean depth (m), highlighting Macquarie Ridge (MR) near 160
8
E, Campbell Pla-
teau (CP) to the northeast, and part of the Southeast Indian Ridge (SEIR) to the west. White
contours show SSH in 0.1-m intervals from
2
0.8 to 0.2 m, averaged over the
fl
oat rapid sampling
period (21 Oct
–
5 Dec 2018). The ACC standing meander forms between the SEIR and MR,
with the crest (trough) being the poleward (equatorward) excursion of the meander. Black con-
tours show 20-yr (1998
–
2018) average SSH contours corresponding to the mean position of the
PF (
2
0.65 m; solid), SAF south (
2
0.4 m; dashed), and SAF north (0.2 m; dash
–
dot), as in
Cyriac
et al. (2022)
.
JAKES ET AL.
719
M
ARCH
2024
Brought to you by Caltech Library | Unauthenticated | Downloaded 12/04/24 08:10 PM UTC
motions, intensifying vertical exchange between the mixed layer
and the ocean interior, and transporting water mass properties
across mesoscale fronts. We
fi
nd enhanced vertical velocities and
evidence of ageostrophic motion extending as deep as 1600
dbar
}
well below the base of even the wintertime mixed layer. We
also
fi
nd evidence of localized subduction and ventilation associated
with a
fi
nescale cold
fi
lament. The vertical extent of these dynamics
has important consequences for ventilation of the ocean interior,
vertical heat and nutrient
fl
uxes, interior strati
fi
cation, and carbon
cycling in the Southern Ocean.
2. Data and methods
We draw together in situ observations from EM-APEX
fl
oats, a towed Triaxus and the shipboard acoustic Doppler
current pro
fi
ler (ADCP), as well as satellite observations of
SSH and high-resolution sea surface temperature (SST). De-
tails about these data sources, processing steps and derived
variables are detailed in their respective subsections.
a. EM-APEX floats
Electromagnetic autonomous pro
fi
ling explorer (EM-APEX)
fl
oats are autonomous pro
fi
ling
fl
oats that follow the current
while pro
fi
ling up and down through the water column, gather-
ing high-resolution vertical pro
fi
les of temperature, salinity,
pressure, and horizontal velocity. The
fl
oats behave similar to
an Argo
fl
oat but pro
fi
le more frequently, gathering data at
higher temporal and spatial resolution. In addition to a CTD
(conductivity
–
temperature
–
depth) sensor, the EM-APEX
fl
oats
are
fi
tted with an electromagnetic (EM) subsystem, containing
two pairs of electrodes that measure the electric current gener-
ated by the motion of seawater through Earth
’
smagnetic
fi
eld
(Sanford et al. 2005
). External
fi
ns rotate the
fl
oat as it ascends
and descends, to allow processing that removes the electric self-
potential between electrode pairs and leaves the potential dif-
ference induced by the ocean currents (Sanford 1971
; Phillips
and Bindoff 2014). The electric current voltages are converted
into velocity components, relative to a depth-independent refer-
ence velocity. Relative velocities are then converted to absolute
velocities by adding a velocity offset, so that the theoretical re-
surface position after each up-pro
fi
le matches the actual surface
position from the GPS (see
appendix
).
During the R/V
Investigator
voyage (16 October
–
15 November
2018), six EM-APEX
fl
oats were deployed in the standing mean-
der of the ACC upstream of Macquarie Ridge. Two of the six
fl
oats failed to collect any data and one failed to retrieve velocity
information. In this study, we use data from the three
fl
oats that
were fully functional
}
containing temperature, salinity, pressure,
and velocity information. The trajectories of these
fl
oats are
shown in
Fig. 2
.
The
fl
oats were programmed to pro
fi
le to 1600 dbar, with a ver-
tical resolution of 3
–
4 dbar, and complete one pro
fi
ling cycle per
day that consists of four vertical pro
fi
les, and a drift at 1000 dbar
(Fig. 3a
). In this analysis, we only use data during the
fl
oa
ts
’
as-
cent or descent; data collected during the
fl
oats
’
drift has been
removed. This pattern was designed to resolve velocities close
to the inertial frequency (Phillips and Bindoff 2014
)
}
with con-
secutive down-pro
fi
les and consecutive up-pro
fi
les separated by
approximately half an inertial period (
p
/
f
5
7.29 h) at the latitude
of
fl
oat deployment (see
appendix
for further information). In
this study, we only use data from the
fl
oats
’
rapid sampling
period (21 October
–
5 December 2018), providing the highest
temporal and spatial resolution. The lateral distance between
consecutive pro
fi
les (based on surface GPS positions) during
rapid sampling has a mean of
;
3.5 km, with individual separa-
tion distances ranging from
,
1to
;
15 km depending on the
speed of the background current (
Fig. 3b
).
The data from each
fl
oat went through a routine quality
control procedure (
Phillips and Bindoff 2014
; Cyriac et al.
2021 ). This involves a pressure drift correction, where the sur-
face pressure value is subtracted from all pressure values in
that pro
fi
le, to reset the surface pressure to zero. Temperature
and salinity pro
fi
les below the seasonal thermocline (200 dbar)
are compared with the CSIRO Atlas of Regional Seas (CARS)
(Ridgway et al. 2002
) and satellite gravest empirical mode
(SatGEM) (
Meijers et al. 2011
) climatologies to identify errone-
ous data and spikes that are then manually removed. Velocity
spikes are identi
fi
ed using a depth-dependent cutoff based on the
RMS error of velocity, ranging from 0.5 cm s
2
1
below 900 dbar
to 1.5 cm s
2
1
between 100 and 220 dbar.
1
Temperature, salinity,
and velocity pro
fi
les are vertically interpolated to an even grid
of 2 dbar. Following the initial QC, 3 (out of 1072) pro
fi
les with
discontinuities in the temperature and salinity were identi
fi
ed
and manually removed. Similarly, several pro
fi
les of relative ve-
locities were manually removed after the initial QC due to large
inconsistencies with surrounding pro
fi
les. The white vertical
F
IG
.3.(a)EM-APEX
fl
oat pro
fi
ling pattern showing two daily
cycles that each produce four vertical pro
fi
les from 0 to 1600 dbar.
The drift at 1000 dbar has been removed for this analysis. (b) Histo-
gram of separation distance between consecutive pro
fi
les in 1-km bins.
Depth-averaged speed for the pro
fi
les in each 1-km bin is indicated by
the black line and corresponds to the RHS
y
axis.
1
For velocities above 100 dbar, where the surface waves domi-
nate, we do not apply the cutoff based on RMS error. Instead, we
exclude velocities with a magnitude greater than 2 m s
2
1
.
JOURNAL OF PHYSICAL OCEANOGRAPHY
V
OLUME
54
720
Brought to you by Caltech Library | Unauthenticated | Downloaded 12/04/24 08:10 PM UTC
lines in the
fl
oat sections are the erroneous pro
fi
les with data
removed.
With consecutive down-pro
fi
les and consecutive up-pro
fi
les
being approximately half an inertial period apart, we remove in-
ertial oscillations from the
fl
oat data by averaging half-inertial
pairs and interpolating back onto the original time grid. Abso-
lute velocities are rotated to along and cross trajectory as a
proxy for along and cross stream
fl
ow. Further details on the
EM-APEX velocity processing pipeline and half-inertial pair
averaging can be found in the
appendix
.
D
ERIVED QUANTITIES
Potential density referenced to the surface (
r
), gravitational
acceleration
g
, and the square of the buoyancy frequency
[
N
2
52
(
g
/
r
)(
D
r
/
D
z
)
, where
z
is the vertical depth coordinate]
are calculated for EM-APEX pro
fi
les using the Thermodynamic
Equation of Seawater (TEOS-10) Oceanographic Toolbox
(McDougall and Barker 2011
). Mixed layer depth (MLD) is
calculated using a density difference criterion of 0.05 kg m
2
3
from a 10 dbar (near-surface) reference level.
The curvature of the
fl
oat track
k
(m
2
1
) is calculated fol-
lowing
Bower and Rossby (1989)
,
k
5
(
x
y
2
y
x
)
(
x
2
1
y
2
)
3
/
2
,
(1)
where
x
and
y
are the surface GPS positions of each pro
fi
le
in Cartesian coordinates. First and second derivatives with re-
spect to time are denoted by prime and double prime, respectively.
Initial surface GPS positions and times are smoothed using a
Savitzky
–
Golay
fi
lter,
fi
tting a third-degree polynomial to a moving
window of 9 pro
fi
les.
2
The smoothing
fi
lter is applied again after
the
fi
rst derivative to reduce the error introduced by large curva-
ture values. After the computation, outliers were detected using
Tukey
’
s boxplot method, visually checked, and manually removed.
Along-isopycnal vertical velocity for the
fl
oat pro
fi
les is cal-
culated following
Phillips and Bindoff (2014)
,
w
total
5
2
r
t
r
z
︸︷︷︸
w
tchng
1
f
r
g
U
V
z
2
V
U
z
()
r
z
︸︷︷︸
w
rot
1
k
c
r
g
U
V
z
2
V
U
z
()
r
z
︸︷︷︸
w
curv
1
2
c
r
g
U
z
U
s
2
V
z
V
s
()
r
z
︸︷︷︸
w
accel
:
(2)
The four terms in the equation, from left to right, correspond
to the vertical velocity due to the time rate of change of the
density
fi
eld at a
fi
xed point in space (
w
tchng
), rotation of hori-
zontal velocity with depth (
w
rot
), curvature of the
fl
oat trajec-
tory (
w
curv
), and acceleration of the
fl
ow along the
fl
oat
trajectory (
w
accel
). Here,
U
and
V
are the along-trajectory and
cross-trajectory components of absolute velocity, respectively.
The
U
and
V
for each pro
fi
le are vertically smoothed using
a Savitzky
–
Golay
fi
lter with a centered rolling window of
75 points (150 dbar). The computation takes three consecu-
tive pro
fi
les at a time, with horizontal derivatives (
t
and
s
)
estimated using a central differencing scheme. The term
c
is
the speed of translation of the
fl
oat along its trajectory, esti-
mated from the GPS positions of the
fi
rst and third pro
fi
les;
s
is distance along the
fl
oat trajectory, representing the along-
stream coordinate;
f
is the Coriolis parameter;
g
is gravitational
acceleration; and
r
is potential density. For more information
on the assumptions and corrections that are required for this
computation using EM-APEX
fl
oat data, see
Phillips and
Bindoff (2014)
.
The Richardson number, Ri, gives an indication of the pres-
ence of ageostrophic dynamics and the susceptibility of the
fl
ow to shear-driven overturning (
Miles 1961
; Thomas et al.
2008 ). It can be calculated as the ratio of the vertical strati
fi
ca-
tion to the vertical shear,
N
2
/
(
u
2
z
1
y
2
z
)
(Mahadevan and
Tandon 2006
; Thomas et al. 2008
; Siegelman et al. 2020b
).
Here,
u
and
y
are the eastward and northward components,
respectively, of absolute velocity after half-inertial pair aver-
aging. We then apply a Savitsky
–
Golay
fi
lter with a rolling
window of 30 dbar to remove some of the noise associated
with taking vertical derivatives at small (2 dbar) intervals.
When Ri
..
1, strati
fi
cation dominates and turbulent mixing
is generally suppressed, but when Ri
,
0.25, velocity shear
can overcome the strati
fi
cationandleadtodynamicorcon-
vective instabilities (
Miles 1961). Richardson numbers of
O
(1) are indicative of an ageostrophic regime associated
with strong vertical motions (
Mahadevan 2006;
Siegelman
2020).
b. Triaxus and ADCP
In addition to the EM-APEX
fl
oat deployment, eight high-
resolution Triaxus tows to a depth of 300 dbar were con-
ducted across the ACC meander during the R/V
Investigator
voyage (
Fig. 2
). The Triaxus is an undulating CTD system
that is towed behind the ship. It was equipped with a dual-
sensor Seabird SBE911 CTD unit as well as sensors to mea-
sure dissolved oxygen (SBE43), nitrate, and chlorophyll
(ECO Triplet). Quality control and processing was under-
taken by CSIRO Marine National Facility (MNF). This in-
cluded spike removal, identi
fi
cation of water entry and exit
times, conductivity sensor lag corrections and determination
of pressure offsets. Data were gridded in 1-dbar vertical bins,
using a linear least squares
fi
t as a function of pressure to in-
terpolate the value for the bin midpoint. Vertical casts were
created from the vertically gridded data, using linear interpo-
lation to a maximum of two casts
’
distance. An estimate of
the Triaxus
’
average position for each vertical cast was
2
The curvature values are insensitive to the temporal discretization.
JAKES ET AL.
721
M
ARCH
2024
Brought to you by Caltech Library | Unauthenticated | Downloaded 12/04/24 08:10 PM UTC
calculated using the wire out, pressure, and ship location. Fur-
ther information can be found in the processing reports on
the MNF website (
marine.csiro.au/data/trawler/survey_details.
cfm?survey=IN2018_V05). We plot the Triaxus data against
along-track distance, using the estimated Triaxus positions for
each vertical cast. The average lateral distance between vertical
castsis1.16
6
0.18 km.
In this study, we use the fourth Triaxus tow that crosses the
ACC meander at 54.5
8
S, between the crest (poleward excur-
sion of the meander) and trough (equatorward excursion).
This transect, occupied from west to east, is roughly perpen-
dicular to SSH streamlines and crosses a
fi
nescale cold
fi
la-
ment that
fl
ows northward near 152.2
8
E, in between an eddy
dipole structure (
Fig. 4a
). The cold
fi
lament measured by the
Triaxus appears in almost the same location as the larger cold
fi
lament sampled by the
fl
oats a few days later. With the
fl
oats
roughly traveling along-stream and the Triaxus being almost
perpendicular to the
fi
lament, we are able to extract both an
along-stream and cross-stream perspective of the cold
fi
la-
ments generated in this region.
Vertical pro
fi
les of absolute velocity from near-surface
(16 dbar) to
;
400 dbar were obtained from the 150-kHz ship-
board ADCP mounted on the ship
’
s hull. The velocities are
gridded in 8-dbar vertical bins and 5-min time intervals. Fur-
ther details on the ADCP processing can be found in the
voyage reports on the MNF website. We use the ADCP ve-
locities corresponding to the start and end times of the Tri-
axus transect, minus 10 min to account for the spatial offset
between the ship and the towed Triaxus. A 10-min lag was
found to closely align the GPS positions of the ship (and
ADCP measurements) with the estimated positions of the
towed Triaxus. For consistency with the Triaxus measure-
ments, we only show velocities for the upper 300 dbar of the
water column.
Zonal (
u
) and meridional (
y
) velocity components from the
ADCP are rotated to along and cross track to avoid the ambi-
guity of a cross-front/along-front framework in a region rich
in
fi
nescale features.
3
The edges of the Triaxus transect are as-
sociated with the anticyclonic and cyclonic eddies of an eddy
dipole structure (
Fig. 4b
), with
fl
ow to the northeast on the
western side, and to the northwest on the eastern side. At the
center of the transect, SSH contours and the SST
fi
lament in-
tersect the track roughly perpendicularly. Therefore, in the
context of the
fi
lament, the rotation to along track is an ap-
propriate proxy for cross front.
D
ERIVED QUANTITIES
Potential density,
N
2
, and MLD along the Triaxus transect
are calculated as described above for the
fl
oat pro
fi
les. The
Richardson number is calculated using the gridded ADCP ve-
locities, interpolated onto the
N
2
pressure and distance grid.
Apparent oxygen utilization (AOU) is calculated as the
difference between the oxygen solubility of a water parcel
(dependent on temperature and salinity) and its measured ox-
ygen concentration (
[
O
2
sol
]
–
[
O
2
obs
]
), giving an indication of
how much respiration and/or primary production has taken
place since the water mass was at the surface (
Ito et al. 2004
).
Low AOU values below the surface mixed layer are indicative
of recent ventilation. Negative AOU values at the surface in-
dicate waters supersaturated in oxygen that may result from
high primary productivity.
Flow becomes susceptible to various instabilities when PV
takes the opposite sign to the Coriolis parameter (
f
)(D
’
Asaro
et al. 2011
; Archer et al. 2020
; Chrysagi et al. 2021
; Taylor and
Thompson 2023
). Combining both the Triaxus and ADCP
measurements, we estimate PV following
Naveira Garabato
et al. (2019)
and Archer et al. (2020)
,
F
IG
. 4. (a) Sea surface temperature (SST) map on 29 Oct 2018
}
the day of the Triaxus tow (black line). The locations
of the EM-APEX
fl
oats on this day are shown as blue circles. (b) Triaxus tow in relation to the EM-APEX
fl
oat trajecto-
ries (blue lines). Thin gray contours show SSH on the same day (29 Oct 2018), in intervals of 0.1 m from
2
0.8 to 0.4 m.
Anticyclonic and cyclonic eddies associated with the eddy dipole are marked as AE and CE, respectively. The thick gray
line underneath the Triaxus transect in each panel is the ship track associated with the ADCP measurements.
3
We also tested rotating the velocity components with respect to
SSH contours and with respect to finite-sized Lyapunov exponent
(FSLE) eigenvectors. The coarse resolution of the SSH product and
the flow curvature associated with the eddy dipole structure led to
unrealistic and rapidly varying frontal orientations.
JOURNAL OF PHYSICAL OCEANOGRAPHY
V
OLUME
54
722
Brought to you by Caltech Library | Unauthenticated | Downloaded 12/04/24 08:10 PM UTC
PV
’
f
1
y
x
()
N
2
2
y
z
b
x
,
(3)
where
y
is the along-front (cross-track) velocity;
x
corre-
sponds to the cross-front (along-track) direction; and
b
is
buoyancy [
g
(1
2
r
/
r
0
)]. In this estimate of PV, we only have
measurements along one horizontal axis, so the vertical com-
ponent of relative vorticity [
z
5
(
y
/
x
)
2
(
u
/
y
)
] is approxi-
mated by the
fi
rst term. Similarly, we are not able to estimate
along-front buoyancy gradients (
b
/
y
). This simpli
fi
cation is
justi
fi
ed when the transect is perpendicular to the front, mak-
ing the assumption that cross-front variations of velocity and
buoyancy are larger than along-front variations (
Thompson
et al. 2016
; Naveira Garabato et al. 2019
). We acknowledge
further limitations associated with using a cross-track and
along-track framework that does not fully resolve the along-
front and cross-front directions; however, this is less of a prob-
lem in the center of the transect where the
fi
lament intersects
the transect perpendicularly.
c. Satellite products
Daily estimates of SSH, provided as absolute dynamic to-
pography, and derived surface geostrophic velocities in Carte-
sian coordinates were obtained from the 0.25
8
delayed-time
L4-gridded satellite altimeter product, provided by Coperni-
cus. To gain insight into the spatial distribution of submeso-
scale fronts and
fi
laments in the meander region, we use the
0.02
8
daily L3-gridded multisensor sea surface temperature
(SST) product provided by Australia
’
s Integrated Marine Ob-
serving System (IMOS). We remove data with a quality level
fl
ag of less than 3. Further information on the quality control of
the SST dataset is described in
Grif
fi
netal.
’
s (2017)
appendix
A(imos.org.au/facilities/srs/sstproducts/sstdata0/sstdata-
references/
). Due to the presence of clouds, we take a temporal
average over 2
–
3 days to capture the approximate location of
the
fi
laments and improve spatial coverage.
Surface EKE is calculated as 0
:
5
[(
u
2
u
)
2
1
(
y
2
y
)
2
]
,where
u
and
y
are satellite-derived zonal and meridional geostrophic
velocity components, respectively, and the overline represents a
3-yr temporal average at each grid point, starting two years
prior to the
fl
oat deployment (October 2016
–
19).
4
The Okubo
–
Weiss (OW) parameter (
Okubo 1970
; Weiss
1991 ) is a useful metric to determine the relative importance
of strain and vorticity in a two-dimensional
fl
ow
fi
eld. It is
given by
OW
5
s
2
n
1
s
2
s
2
v
2
5
u
x
2
y
y
()
2
1
y
x
1
u
y
()
2
2
y
x
2
u
y
()
2
,
(4)
where
u
and
y
are the surface geostrophic velocity compo-
nents derived from SSH. The
fi
rst two terms make up the
strain component, comprising normal strain (
s
n
) and shear
strain (
s
s
), and the third term is relative vorticity (
v
). OW is
often used to track eddies (
Henson and Thomas 2008
; Denes
et al. 2022
), as OW
,
0 indicates a vorticity-dominated
fl
ow
fi
eld generally associated with eddy cores, while OW
.
0 indi-
cates a strain-dominated
fl
ow
fi
eld associated with eddy
peripheries.
Daily
fi
nite-sized Lyapunov exponents (FSLEs), computed
backward-in-time from Copernicus L4-gridded geostrophic
velocities, were obtained from the 0.04
8
gridded Aviso
1
data
product. FSLEs are calculated by measuring the separation
rate of pairs of particles in a given three-dimensional velocity
fi
eld, with
fi
xed initial and
fi
nal separation distances (
Aurell
et al. 1996
; Artale et al. 1997;
Boffetta et al. 2001
). They provide
a direct measure of local stirring and can be used to characterize
the strain
fi
eld and identify
fi
nescale structures of the
fl
ow, oth-
erwise called Lagrangian coherent structures (
D
’
Ovidio et al.
2004;
Hern
́
andez-Carrasco et al. 2012
; Cotté et al. 2015). FSLEs
computed backward-in-time yield negative FSLE values that
provide a measure of horizontal con
fl
uence
}
strongly negative
values indicate regions where particles that were initially far
apart have converged rapidly. The opposite is true for FSLEs
computed forward-in-time, where strongly positive values indi-
cate a high separation rate of particles (i.e., strong stretching)
(D
’
Ovidio et al. 2004
). Containing both spatial and temporal in-
formation from satellite altimetry, FSLEs can provide informa-
tion about the growth rate and orientation of submesoscale
fronts (Siegelman et al. 2020a
). Large absolute FSLE values
are often observed around the periphery of mesoscale eddies,
associated with submesoscale fronts (
Siegelman et al. 2020a),
and in high EKE regions in the Southern Ocean (
Dove et al.
2022).
3. Results
a. Surface characteristics of the study region
During the
fl
oat sampling, there was a highly energetic
eddy
fi
eld within the meander, indicated by strong surface
EKE just upstream of MR (
Fig. 5a
), as well as the highly con-
torted
fl
oat trajectories (
Fig. 2
). EKE weakens where the
fl
ow
encounters MR, as the time-mean current travels through two
main gaps in the ridge at 53
8
and 56
8
S. The rich mesoscale
eddy
fi
eld in the meander resulted in strong surface FSLE val-
ues in the region (
Fig. 5b
), implying high submesoscale activ-
ity. The strongest EKE and FSLE signals within the red box
(where the observations were taken) are located within the
meander trough at 152.5
8
Eand53.5
8
S, where there was a persis-
tent cyclonic eddy with a strong negative Okubo
–
Weiss parame-
ter ( Fig. 5c). The OW parameter (
Fig. 5c
) also highlights the eddy
dipole present in the meander during the
fl
oat sampling
}
with
the two vorticity-dominated eddies (negative OW), at 150
8
and
152.5
8
E, separated by a strain-dominated region (positive OW).
In the high-resolution SST images over the
fl
oat sampling
period, a
;
20-km-wide (
Fig. 6b
) cold
fi
lament was observed
(Fig. 6a) and sampled by the
fl
oats in various locations
(Figs. 6c
–
e). The Rossby radius of deformation at this latitude
is
;
10
–
20 km (
Chelton et al. 1998
), indicating that the
fi
lament
4
Different time periods of 1, 3, and 10 years were tested for the
mean field; however, the spatial pattern and magnitude of EKE
were insensitive to the choice of time period, suggesting low inter-
annual variability of mean kinetic energy in the region.
JAKES ET AL.
723
M
ARCH
2024
Brought to you by Caltech Library | Unauthenticated | Downloaded 12/04/24 08:10 PM UTC
is on the upper end of the submesoscale range. Due to the pres-
ence of clouds, the snapshots in
Fig. 6
have been averaged over
3 days, introducing some smoothing, and show the approximate
location of the
fi
lament in relation to the EM-APEX
fl
oats.
The
fi
lament originates from south of the front, travels
northward at 151.5
8
E, then turns eastward and appears to
feed into the cyclonic eddy at 154.5
8
E(Fig. 6a
). The
fi
lament
forms in between the eddy dipole structure, where you would
expect the mesoscale strain
fi
eld to drive
fi
lamentogenesis
(Lapeyre and Klein 2006
; McWilliams et al. 2009
). Float EM-
8492 travels northward in the
fi
lament and then crosses over
on the western side (
Fig. 6d
), while EM-8493 follows the
fi
la-
ment into the cyclonic eddy (Fig. 6e
) and EM-8489 travels out
of the
fi
lament to the east (
Fig. 6c
)
}
making a clockwise loop
before traveling northward. The
fi
lament is completely
eroded a few days later when
fl
oat EM-8489 travels north
(not shown). It is important to note that the EM-APEX
fl
oats
are not fully Lagrangian because of their pro
fi
ling
}
they occa-
sionally cross SSH streamlines and do not necessarily follow
the
fi
lament either.
b. Subsurface structure along the EM-APEX float
trajectories
The EM-APEX
fl
oats were deployed upstream of the me-
ander crest, close to the mean position of the PF based on the
SSH gradient and the position of the
2
0.65-m SSH contour
on the day of
fl
oat deployment (
Cyriac et al. 2022
). For the
fi
rst
;
200 km, the
fl
oats sample a temperature minimum
(
T
min
) layer of
,
2
8
C between 200- and 300-m depth (
Fig. 7a
)
}
a
characteristic signature of Antarctic Winter Water (WW) at (or
just south of) the PF (Orsi et al. 1995;
Belkin and Gordon 1996;
KlinckandNowlin2001
; Naveira Garabato et al. 2001). The
fl
oats then diverge into different trajectories, steered by the lo-
cal currents in the rich EKE and FSLE
fi
eld. Floats EM-8489
and EM-8492 transition into warmer waters north of the PF
(Fig. 7a
), with a deepening of low-salinity waters (Fig. 7b
) indic-
ative of the
fl
oats approaching the SAF (
Kim and Orsi 2014
).
Float EM-8493 travels into a cyclonic (cold-core) eddy within
the meander trough (Fig. 6e
)
}
where surface waters are
;
3
8
C
warmer than at the beginning of the
fl
oat trajectory but the
WW signature is retained between 200 and 300 dbar (
Fig. 7a).
There are a range of scales (from submesoscale to mesoscale)
captured in the
fl
oat data. Between 200 and 300 km in all
fl
oats,
there is a deepening of the WW layer (to 300
–
500 dbar) as the
water mass subducts along isopycnals (
Fig. 7a). This deepening
of
T
min
waters occurs as the
fl
oats travel northward (
Figs. 6c
–
e)
and may be a signal of the
fl
oats crossing the PF. However, the
fl
oats are also crossing submesoscale fronts and sampling the
cold
fi
lament that
fl
ows from the meander crest to the trough
(Figs. 6c
–
e). Float EM-8492 crosses the western side of the cold
fi
lament (
Fig. 6d
) at the location of WW subduction between
200 and 300 km (
Fig. 7a), while
fl
oat EM-8493 continues to fol-
low the cold
fi
lament into the cyclonic eddy (
Fig. 6e). Float
EM-8489 does not travel northward in the cold
fi
lament like the
other
fl
oats, but crosses the eastern side of the
fi
lament at its
base near 54.8
8
S.
F
IG
. 5. (a) Eddy kinetic energy (EKE), (b) FSLE, and (c) Okubo
–
Weiss parameter, calculated from satellite geostrophic velocities and
averaged over the
fl
oat rapid sampling period (21 Oct
–
5 Dec 2018). The thick gray contour in each panel shows the 1500-m depth level of
the bathymetry, highlighting Macquarie Ridge and Campbell Plateau. SSH streamlines averaged during the
fl
oat sampling period are
shown as thin contours; black contours show the approximate location of the PF (solid) and southern (dashed) and northern (dash
–
dot)
branches of the SAF, averaged over the same period. The red box marks the map area in
Figs. 4
and 6, where the observations were
taken.
JOURNAL OF PHYSICAL OCEANOGRAPHY
V
OLUME
54
724
Brought to you by Caltech Library | Unauthenticated | Downloaded 12/04/24 08:10 PM UTC
The subsurface velocity structure along the
fl
oat tracks helps
to distinguish between mesoscale and submesoscale dynamics.
A striking feature in the subsurface velocities is the reversal in
sign of the cross-track velocities with depth (
Fig. 8b
). This over-
turning signal is particularly prominent where the along-track
velocities are strongest (
Fig. 8a
)
}
between 200 and 450 km in
EM-8492 and between 200 and 350 km in EM-8493
}
and is asso-
ciated with intense downward velocities of
O
(10
2
3
)ms
2
1
extend-
ing from the surface to 1600 dbar (
Fig. 8c
). The inferred vertical
velocities are greater than 100 m day
2
1
}
an order of magnitude
F
IG
. 6. (a) Satellite sea surface temperature (SST) snapshot in the meander region,
averaged over three days
(4
–
6 Nov 2018). The 4.4
8
C isotherm (black line) highlights the cold
fi
lament and two cold-core eddies east of the
fi
la-
ment. (b) SST along the dashed line in (a). Shading represents the bounds marked by the 4.4
8
C isotherm. (c)
–
(e) The
same 3-day SST snapshot overlaid with each
fl
oat trajectory (white circles), the location of the
fl
oat during the 3 days
(black dots) and the 3-day mean SSH contours in 0.15-m intervals from
2
0.6 to 0.3 m. Along-trajectory distance at
the
fl
oat locations corresponding to the black dot
s are displayed at the top left of each panel.
F
IG
. 7. Subsurface sections of (a) Conservative Temperature and (b) Absolute Salinity in the upper 1000 dbar along each of the
fl
oat
tracks. Inertial oscillations have been removed by half-inertial pair averaging, and the data are interpolated onto a regular 3-km distance
grid. The thick white contour on (a) is the 2
8
C isotherm and the black bars at 0 dbar indicate the locations of the
fl
oats in
Figs. 6c
–
e.The
thin black line marks the mixed layer depth (MLD) and thin white contours show potential density in 0.2 kg m
2
3
increments.
JAKES ET AL.
725
M
ARCH
2024
Brought to you by Caltech Library | Unauthenticated | Downloaded 12/04/24 08:10 PM UTC
larger than expected from mesoscale motions. This pattern in the
velocity structure is coincident with the descent of
T
min
waters
(Fig. 7a
) and is localized where
fl
oats EM-8489 and EM-8492
sample the cold
fi
lament traveling northward, in between the
eddy dipole structure (Figs. 6d,e
). We interpret these obser-
vations as evidence of cold
fi
lamentary intensi
fi
cation. The
overturning signal in
fl
oats EM-8489 and EM-8492 is consis-
tent with an ASC acting to
fl
atten isopycnals on the western
side of the cold
fi
lament.
5
A reversal in sign of the cross-
trajectory velocities is also visible in the
fi
rst 200 km of the
fl
oat trajectories (Fig. 8b
), upstream of the
fi
lament, but the
strength of the velocities is weaker. The spatial pattern and
magnitude of vertical velocity in
Fig. 8c
is dominated by
w
rot
[Eq.
(2)], associated with the rotation of horizontal velocity
with depth. Without information in both the along-stream and
cross-stream directions, it remains dif
fi
cult to unambiguously
distinguish features related to the
fi
lament from those related
to the mesoscale meander. However, the observations reveal
an intensi
fi
cation of downward velocities associated with a
cross-track overturning circulation, localized within the cold
fi
lament, that cannot be explained by mesoscale motions
alone.
S
UBMESOSCALE INSTABILITIES
Along the
fl
oat tracks, we identify regions of the water col-
umn that are susceptible to submesoscale instabilities. Nega-
tive values of
N
2
indicate the potential for gravitational
instability. Richardson numbers of
O
(1) and lower indicate an
ageostrophic regime associated with intense vertical currents
(Thomas et al. 2008
; Siegelman et al. 2020b
).
The WW layer at the beginning of the
fl
oat tracks is associ-
ated with strong vertical strati
fi
cation (
Fig. 9a
), which weak-
ens as the WW descends along isopycnals to deeper depths.
Clusters of negative
N
2
values emerge below the mixed layer
(between 100 and 500 dbar), particularly in EM-8489 and
EM-8492 from 300 km onward, indicating the potential for
gravitational instabilities and convective overturning that
could facilitate vertical mixing and subduction (
Freilich and
Mahadevan 2021
). Although these negative values are ob-
served well below the mixed layer, they appear to coincide
with a reduction in MLD
}
notably at 400 km in EM-8489,
and 300 and 600 km in EM-8492. While this may re
fl
ect an-
other buoyancy-driven or mechanical process, or the move-
ment of the
fl
oat across a
fi
nescale front, submesoscale
instabilities can reduce the MLD (Boccaletti et al. 2007
;
Fox-Kemper and Ferrari 2008
; Yu et al. 2019
) and may be
partially responsible for this restrati
fi
cation. There are no neg-
ative
N
2
values along the trajectory of
fl
oat EM-8493,
F
IG
. 8. Subsurface (a) along-trajectory, (b) cross-trajectory, and (c) vertical velocity along each of the
fl
oat tracks. Inertial oscillations
have been removed by half-inertial pair averaging after the rotation and prior to the
w
calculation (see
appendix
). The data are interpo-
lated onto a regular 3-km distance grid. Gray contours show potential density in 0.2 kg m
2
3
increments. The MLD is marked by the black
solid line.
5
Positive cross-trajectory velocity is to the left of the float track.
JOURNAL OF PHYSICAL OCEANOGRAPHY
V
OLUME
54
726
Brought to you by Caltech Library | Unauthenticated | Downloaded 12/04/24 08:10 PM UTC
implying a lack of gravitational instabilities. This
fl
oat travels
within the cold
fi
lament into the core of the cyclonic eddy,
while the other two
fl
oats get ejected from the
fi
lament (
Figs.
6c
–
e) and travel through the strain-dominated region sur-
rounding the cyclonic eddy. This suggests that generation of
instabilities and mixing tends to occur on the periphery of the
fi
lament and the eddy, in the strain-dominated regions.
Richardson numbers of
O
(1) and lower are observed in
the surface mixed layer in all
fl
oats, where strati
fi
cation is
weak, but also appear in vertically coherent bands extending
from below the mixed layer down to
;
1500 dbar (
Fig. 9c
)
}
indicating the potential for ageostrophic motion in the ocean
interior. The spatial distribution of low Ri below the mixed
layer is dominated by the velocity shear (Fig. 9b
), which may
be associated with the development of an ASC during
fi
la-
mentogenesis. The regions where Ri
;
1 are similar to those
with elevated vertical velocities shown in
Fig. 8c. The magni-
tude of the inferred vertical velocities (
.
100 m day
2
1
), in ad-
dition to the low Richardson number, provide evidence of
submesoscale-driven ageostrophic motion in the vicinity of
the cold
fi
lament.
c. Cross-filament perspective
While the Triaxus does not sample the same
fi
lament that
the
fl
oats do, or at least not at the same stage in its develop-
ment, it crosses a
fi
nescale cold
fi
lament (
;
5 km wide) that
forms in almost the same location several days before the
fl
oats arrive. The Triaxus and shipboard ADCP observations
provide further evidence of enhanced ageostrophic motion, as
well as localized subduction and ventilation, associated with
fi
nescale cold
fi
laments in this region.
From the SST map (
Fig. 10a
), the
fi
lament again appears to
be drawn up from the southern side of the front and travels
northward in between the eddy dipole structure. The
fi
lament
thins as it travels north and is sampled by the Triaxus at its
northernmost point (cyan dot on
Fig. 10a
) before it is mixed
away. The cold
fi
lament is visible in the temperature anomaly
section (
Fig. 10e
) approximately halfway along the transect
(
;
40 km) and extends from the surface to 300 dbar
}
the
maximum depth of the Triaxus measurements. The cold tem-
perature anomaly is less pronounced in the mixed layer and
stronger at depth
Fig. 10e
). The
fi
lament is also associated
with a fresh anomaly (
Fig. 10f
) and elevated oxygen (
Fig. 10g
)
and chlorophyll (
Fig. 10h
) signatures that extend down to
300
–
200 dbar below the base of the mixed layer. The low AOU
values (
Fig. 10g
), indicative of oxygen-saturated waters, in addi-
tion to the elevated chlorophyll observed below the mixed layer
(Fig. 10h
), provide strong evidence of recent subduction and ven-
tilation associated with the cold
fi
lament. There are several
weaker cold intrusions with relatively fresh, oxygenated and chlo-
rophyll-rich signatures between 5 and 30 km that extend from
the base of the mixed layer down to 200
–
250 dbar. These features
are most pronounced in the interior, below the mixed layer, and
do not have as strong an anomalous signature at the surface.
F
IG
. 9. Subsurface sections of (a) buoyancy frequency (
N
2
) (upper 700 dbar), (b) vertical velocity shear (
u
2
z
1
y
2
z
), and (c) Richardson
number (Ri) from 0 to 1500 dbar, interpolated onto an even 3-km distance grid. Potential density contours and MLD are as in
Figs. 7
and 8.
JAKES ET AL.
727
M
ARCH
2024
Brought to you by Caltech Library | Unauthenticated | Downloaded 12/04/24 08:10 PM UTC