Frontal circulation and submesoscale variability during the formation of a Southern Ocean mesoscale eddy

Observations made in the Scotia Sea during the May 2015 Surface Mixed Layer Evolution at Submesoscales (SMILES) research cruise captured sub-mesoscale, O (1-10 km), variability along the periphery of a mesoscale O (10-100 km) meander precisely as it separated from the Antarctic Circumpolar Current (ACC) and formed a cyclonic eddy ∼ 120 km in diameter. The me-ander developed in the Scotia Sea, an eddy-rich region east of the Drake Passage where the Subantarctic and Polar fronts converge and modiﬁcations of Subantarctic mode water (SAMW) occur. In situ measurements reveal a rich submesoscale structure of temperature and salinity and a loss of frontal integrity along the newly-formed southern sector of the eddy. A mathematical framework is developed to estimate vertical velocity from co-located drifter and horizontal water velocity time series, under certain simplifying assumptions appropriate for the current data set. Upwelling (downwelling) rates of O (100 m day -1 ) are found in the northern (southern) eddy sector. Favorable conditions for submesoscale instabilities are found in the mixed layer, particularly at the beginning of the survey in the vicinity of density fronts. Shallower mixed layer depths and increased stratiﬁcation are observed later in the survey on the inner edge of the front. Evolution in T-S space indicates modiﬁcation of water mass properties in the upper 200 m over 2 days. Modiﬁcations along σ θ 27 - 27.2 kg m − 3 have climate-related implications for mode and intermediate water transformation in the Scotia Sea on ﬁner spatiotemporal scales than observed previously.


Introduction
The Southern Ocean hosts the most energetic current system in the world, the Antarctic Circumpolar Current (ACC).Zonally unbounded by land, the ACC connects ocean basins and transports an estimated 173 Sv through the Drake Passage (Donohue et al. 2016).The ACC is predominantly in geostrophic balance with sea surface height (SSH) gradients and lateral density gradients, hereafter fronts.Large-scale instabilities in the balanced ACC flow cause mesoscale, O(10-100 km), meanders and eddies in the Southern Ocean.While the rich mesoscale structure of the ACC has been studied intensely, finer-scale variability along Southern Ocean fronts is less understood and observed.
Two of the most prominent fronts in the Southern Ocean are the Subantarctic and Polar fronts (hereafter, SAF and PF).Due to sparse data coverage in the Southern Ocean, altimetry-based frontal definitions have been developed; SSH SAF = -0.25 m and SSH PF = -0.70m are updated values from Sallée et al. (2008).North of the SAF, water masses such as Subantarctic Mode water (SAMW) and Antarctic Intermediate water (AAIW), subduct along isopycnals at specific locations in the Southern Ocean, such as the Scotia Sea (Sallée et al. 2010).The subducted pools of SAMW and AAIW observed north of the ACC contain high levels of anthropogenic CO 2 (Sabine et al. 2004;Pardo et al. 2014) and heat (Frölicher et al. 2015).Currently, SAMW is thought to be transformed by air-sea buoyancy fluxes (Cerovecki et al. 2013) and subsequently mixed and subducted with AAIW, σ θ 27.2 , to the South Atlantic (Sallée et al. 2010).In locations 'upstream' of the subducted SAMW/AAIW pools, mode water transformation occurs in the mixed layer at the SAF and has climatic implications.The large-scale, O(100-1000 km), physical processes, such as wind-driven and eddy-driven Ekman pumping, responsible for the subduction of heat and carbon in SAMW/AAIW pools have been discussed and documented, e.g.Sallée et al. (2010Sallée et al. ( , 2012)), but very little is known about subduction associated with smaller scales processes (Naveira Garabato et al. 2001).
A potentially important class of dynamics responsible for modulating the vertical exchange at fronts in the Southern Ocean occurs at the submesoscale, O(1-10 km).The oceanic submesoscale is instrumental in extracting energy from density fronts and transferring the energy from mesoscale to submesoscale and dissipative scales (Thomas and Taylor 2010;Capet et al. 2008).The downscale transfer of energy results in ageostrophic motions with large vertical velocities, O(100 m day −1 ) (Mahadevan and Tandon 2006;Capet et al. 2008;Thomas et al. 2008) capable of transporting heat and tracers across the base of the mixed layer.Where energetic submesoscale processes exist, the resulting vertical buoyancy fluxes may attain an importance equal to or greater than those forced by air-sea exchange.
The presence of fronts preconditions the mixed layer to the development of submesoscale processes, which are characterized by O(1) Rossby (Ro) and balanced Richardson (Ri B ) numbers (Thomas et al. 2008).Submesoscale dynamics are often associated with hydrodynamic instabilities including baroclinic mixed layer instability (MLI), symmetric instability (SI), inertial instability (II) and gravitational instability (GI) (Haine and Marshall 1998;Fox-Kemper et al. 2008;Thomas et al. 2008).These instabilities, with the exception of GI, grow at the expense of available potential energy associated with lateral density gradients (MLI) or thermal wind kinetic energy (II and SI).In each of these cases, instabilities are likely to develop at fronts and can significantly modify the mixed layer density structure (Boccaletti et al. 2007;Hosegood et al. 2008;Taylor and Ferrari 2009;Mahadevan et al. 2010).GI, conversely, is convectively-driven and generated by unstable vertical stratification.Mixing associated with GI leads to deeper mixed layers, while MLI and SI results in restratification.
Sampling submesoscale processes presents challenges due to the complex dynamics of the mixed layer and the short spatiotemporal scales of variability, from hours to days and meters to kilometers.Very few submesoscale-resolving measurements have been made in the Southern Ocean (Rocha et al. 2016), though a recent modeling study has demonstrated the dependence of submesoscale vertical velocities on an energetic mesoscale eddy and strain field (Rosso et al. 2015).An energetic submesoscale is, therefore, expected in a region with high mesoscale EKE, such as the Scotia Sea, a mesoscale eddy hot spot (Frenger et al. 2015).Large, high-Ro meanders of the SAF and PF fronts (Figure 2) are indicative of a highly energetic mesoscale field in the Scotia Sea region, suggesting the presence of a similarly energetic submesoscale field.
Here we present novel observations of submesoscale variability in the Southern Ocean from the SMILES (Surface Mixed Layer Evolution at Submesoscales) project, http://www.smiles-project.org.SMILES aims to (1) characterize submesoscale dynamics and (2) evaluate the role of submesoscales in mode water transformation in the Scotia Sea using a combination of observations and models.The observational component of the SMILES project consists of a single research cruise to the Scotia Sea in May 2015, just before the austral winter.During a drifter-following cross-front survey, a northward meander of the SAF and PF (Figure 2) separated from the ACC and formed a cold-core mesoscale eddy.
In this paper, we focus on the observed frontal circulation and submesoscale variability along the periphery of the newly-formed eddy.Data sources and processing methods are described in Section 2. Results from the drifter-following survey are presented as follows in Section 3: a) eddy formation, b) frontal circulation, c) cross-frontal variability, and d) water mass modification.
Section 4 presents an estimation of vertical velocity and a submesoscale instability analysis with implications for mode water modification.In Section 5, results are summarized and the implications of submesoscale processes during eddy formation in the Scotia Sea are discussed.

Ship-based data sources
The field component of the SMILES project consisted of a Scotia Sea research cruise, 22 April -21 May 2015, performed aboard the British Antarctic Survey RRS James Clark Ross (JCR).
Seasoar, a winged and towed body equipped with a Seabird-Electronics Inc. SBE911, collected temperature, conductivity, and pressure measurements at 16 Hz.Seasoar data is collected in a saw-tooth pattern (Figure 3) at 8 knots (∼ 4 m s -1 ) with a horizontal spacing between apogees of 2 km for 200-m dives.Temperature and salinity variables were binned to 0.5 dbar intervals.
Binned data were gridded using a 2-dimensional Gaussian interpolation scheme (Barnes 1964) with regular spacing, 0.5-km horizontal and 1-m vertical, and decorrelation radii of 1 km and 2 m (Figure 3c).
Horizontal currents were collected in 8-m depth bins over 22 to 600 m of the water column by the ship-mounted RDI Ocean Sciences 75-kHz acoustic doppler current profiler (ADCP).The collected data was cleaned, corrected for ship speed and heading, and ensemble averaged to 150second bins using Common Ocean Data Access System (CODAS) processing tools.North and east velocity components from 30 -200 m were gridded to the same grid as the Seasoar data then rotated into along-front and cross-front velocity components using the drifter trajectories as explained below.Error velocities reported from the ADCP processing software are used as estimates of velocity uncertainty in the calculations in Section 4a.

Drogued drifters
A triplet of drogued drifters was used in the survey to estimate horizontal water velocities at 50m depth.The drifters consisted of a sealed buoy with GPS and satellite communications, a 'holey-sock' drogue 10-m long and 90-cm in diameter centered at 50-m depth, and 3.5 mm Dyneema line.This design provided a drag area ratio of 44 which is accurate to follow water parcels to within 1 cm s −1 (Sybrandy et al. 2009).Drifter location updates were received at 10-minute intervals.
The drifters were released in the northern portion of the meander just south of the maximum jet velocity and temperature gradient (Figure 3a) for the first Seasoar leg of the survey.In a current of ∼1.25 m s −1 , the three minute separation of the drifter releases yields an initial along-front drifter separation of ∼225 m.The trajectory of the first drifter released, D16, was chosen to define the along-front direction in the survey analysis, θ along (Table 1).The along-front reference frame assumes the drifter maintains its position in the front and jet, which is shown in Figure 4.The closest drifter crossing in time and space of each Seasoar leg defines the center of each section, with cross-frontal distance increasing outward, or away, from the eddy center.Each leg was rotated to a cross-front heading, θ cross , defined as the orthogonal direction to θ along for each respective Seasoar leg (Table 1).Similarly, measured horizontal water velocities were rotated into along-front and cross-front components for each leg.

Remote data sources
Satellite sea surface temperature (SST) and sea surface height (SSH) data were used for mesoscale frontal and eddy detection during the cruise and the analysis.Both data sets are available daily on a 0.25 • grid.Figure 2 is an example of the remote sensing data available during the SMILES cruise.The daily, gridded optimally interpolated microwave SST data (OISST) was obtained from Remote Sensing Systems, (http://www.remss.com).SSH, or absolute dynamic topography, and altimetrically-derived geostrophic surface current data were downloaded from AVISO Cnes (www.aviso.altimetry.fr)(Pujol et al. 2016).SAF and PF positions are defined using SSH contours of -0.25 m and -0.7 m, respectively, updated from the definitions in Sallée et al. (2008).

a. Eddy formation
A northward meander of the Subantarctic front (SAF) and Polar front (PF) developed along the Antarctic Circumpolar Current (ACC) (Figure 2) in late April 2015.This mesoscale, O(100 km), feature characterized by meridional changes of 4 • C SST and 0.5-m SSH over 50 km, formed just south of the North Scotia Ridge.Antarctic surface water, <2 • C south of the PF (Orsi et al. 1995), is observed in the center of the meander.The vorticity Rossby number, Ro = ζ f −1 , of the meander as calculated from altimetry-derived geostrophic surface currents from 20 April is ∼ 0.4.This moderate Ro value based on coarse altimetry data does not account for ageostrophic contributions from curvature of the flow, e.g.cyclogeostrophic flow.Although the moderate Ro estimate is high compared to previous submesoscale-focused process studies, e.g.Ro ∼ 0.1 in the North Pacific (Hosegood et al. 2013), it not uncommon for this region.
A triplet of drogued drifters released in the northwest sector of the meander on 08 May 2015 20:00 GMT was followed with the RRS JCR while towing the Seasoar CTD perpendicular to the drifter trajectories.The daily progression of SST, SSH, drifter trajectories and the ship track are presented in Figure 4 for 8-12 May 2015.At the time of the drifter release, 18 days after the SST and SSH observations presented in Figure 2, the meander had sharpened yet remained tethered to the ACC as observed by SST and SSH fields, Figure 4a.During the survey, the drifters initially traveled east (Figure 4b) and southeast (Figure 4c) around the meander and remarkably continued along a cyclonic trajectory precisely as the meander separated from the ACC and formed a cold closed-core eddy, Figure 4c-e.Initially, the cyclonic eddy measured approximately 120km in diameter with a dynamic height anomaly of 0.5 m (-0.2 to -0.7 m SSH).After the eddy formed, Figure 4e, the SAF and PF returned to a zonal orientation south of the eddy.Hereafter, the meander/eddy feature will be referred to as an eddy for the duration of the Seasoar survey.
The Seasoar survey, shown as the ship track in Figure 4 Horizontal water velocities measured at 50-m depth are included in Figure 5c-d where the alongfront and cross-front components are determined relative to a drifter trajectory direction (Table 1) for each Seasoar section.A ∼ 70% decrease (1.5 to 0.4 m s −1 ) in drifter and along-front water velocities is observed from the N to S legs.Geostrophic surface velocity vectors (Figure 4) also

b. Cross-frontal variability
Vertical cross-sections of potential density anomaly (σ θ , kg m -3 ), temperature ( • C), salinity, and horizontal water velocities (m s -1 ) are presented in Figure 7 for the Seasoar legs labeled in Figure 5.
The five sections span approximately two days and 180 degrees of heading of the drifter-following survey.Each section is referenced in a similar manner with respect to the front; the left (right) -hand side of the sections will be referred to as inner (outer) with negative (positive) cross-frontal distance.Since the sections are centered using the drifter trajectories, a cross-frontal distance of zero is not an explicit definition of the frontal center with respect to density.
In Leg N, σ θ increases laterally away from the eddy core except for a dense filament ∼ 5 km in width located in the center of the leg (Figure 7a).The filament, with temperatures < 1.5 • C, is observed between two outcropping isopycnals with a potential density anomaly of 27.0 kg m -3 (hereafter σ θ 27 ).The inner density gradient, 0.09 kg m -3 in 5 km, is nearly twice the magnitude of the warm, outer density front, 0.04 kg m -3 in 5 km.In Leg E, the σ θ 27 is observed subsurface.
By Leg S, the depth of the σ θ 27 is much shallower on the inner side of the leg.
Mixed layer depth, MLD, defined as the level of a 0.01 kg m -3 density increase from 5-m depth, is included in Figure 7a.This strict MLD definition was chosen to highlight the lateral density gradients in the current dataset.Values of MLD are O(100 m) for most of Leg N. In each leg, the mixed layer is shallowest within the density fronts, <50 m, and deepest within the dense filament at 130 m.The MLD shoals similarly to σ θ 27 in Leg S, suggestive of restratification of the inner front along the newly-formed sector of the eddy.The shallower MLD may be the result of temporal variability, e.g., restratification from submesoscale instabilities, or spatial variability.
Temperature and salinity fields vary similarly across the sections, Figure 7b-c, due to strong density compensation, characteristic of ACC fronts.In Leg N, the warm, salty outer region lies adjacent to a cold, dense filament at a cross-front distance of 0 km.Leg E, in the east sector of the survey, contains a small subsurface cold water intrusion at 120-m depth and 10-km cross-front distance.Intrusions of cold, fresh water on the outer side and warm, salty water on the inner side are observed in all legs collected in the east and southeast sectors of the survey.In leg SE, the intrusion is larger in vertical and horizontal extent and outcropped.In Leg S a loss of frontal integrity is observed compared to the well-organized, separated cold-fresh inner and warm-salty outer regions present in Leg N.
Vertical cross-sections of along-front and cross-front velocities, Figure 7d-e

c. Frontal circulation
The frontal circulation at the center of each Seasoar leg can be described using the co-located drifter and horizontal water velocity datasets.As shown in Figure 8a, the drifter and along-front water velocities at 50-m depth are in strong agreement.Drifters initially deployed in the northern sector of the cyclonic eddy decelerated around the eastern side toward the southern sector where the along-front velocity is minimum, after which the drifters accelerated around the western edge.Similar trends were observed in the measured along-front velocity.The cross-frontal gradient of cross-frontal velocity, was positive (diffluent) during the along-front deceleration and negative (confluent) during the along-front acceleration as shown in (Figure 8b).

Analysis a. Estimation of vertical velocity
Vertical velocities, w, were not directly measured in the SMILES Seasoar survey.However, the co-located drifter and ADCP datasets allow for the following mathematical framework which yields a solvable expression for w at a specific depth and a cross-frontal location in each Seasoar leg.Assumptions made in the following derivation are tested in Appendix A.
Let x D (t), and u D (t) be the measured drifter position and velocity vectors at time t where where (3) Here we set z 1 = 0 at the surface and z 2 = 50 m, the drifter drogue depth.This assumes the drogued drifter is moving with the depth-integrated Eulerian velocity in the top 50-m of the water column.
Justification for this assumption is presented in the Appendix.Expanding the derivative in (2), where From ( 4), the rate of change of the along-front drifter velocity is Substituting ( 7) in ( 6) gives Re-arranging (8) yields an expression for the difference of vertical velocity from An expression for w E at the drogue depth, z 2 , is obtained by setting w E (z 1 = 0) = 0, We can make a steady-state assumption, if local accelerations are smaller than drifter accelerations on time scales greater than a day, the filtering window of the drifter velocities.This is tested in the Appendix making use of ship track intersections during the survey.We then have an expression, that allows for the calculation of vertical velocity in the center of each Seasoar leg at z 2 , the drifter drogue depth of 50 m, (Figure 8d).Velocity components ūE and vE are first calculated by averaging velocities from the first good ADCP bin, 30-m, to 50-m, as in Figure 8a.Extrapolations

b. Submesoscale instabilities
Although direct measurements of submesoscale instabilities were not made during the Seasoar survey, it is possible to diagnose whether conditions were favorable for submesoscale instability growth and which specific instabilities were possible (Thomas et al. 2013;Thompson et al. 2016).
First, instability development is favored when Ertel potential vorticity (EPV), is the opposite sign of f (Hoskins 1974;Haine and Marshall 1998;Thomas et al. 2008); the absolute vorticity, ω a , is the sum of planetary and relative vorticity and buoyancy is b = −gρ ρ −1 0 .
The perturbation density, ρ , is the measured density, ρ, minus the average leg density, ρ 0 .This EPV criterion has been shown to hold even in flow regimes where ageostrophic processes such as down-front winds (Thomas and Taylor 2010;Thomas et al. 2013), inertial shear (Thomas et al. 2016) and surface-wave driven shear (Haney et al. 2015) drive symmetric instability.
Expanding (13) gives where subscripts indicate a partial derivative and x and y are the along-front and cross-front directions.Neglecting ∂ x terms in (14) assumes along-front gradients cross-front gradients.This simplification yields, an approximation for EPV dependent on cross-front and vertical gradients in the along-front velocity and buoyancy.The 2-dimensional approximation of EPV ( 15) is shown in Figure 9 below the cross-frontal buoyancy gradient at 10-m depth which identifies density fronts in each leg.Regions with positive EPV ( f < 0) are favorable for the instabilities described above and are observed on either side of the lateral buoyancy gradients, or fronts, and mostly above the MLD.The band of negative EPV in each leg is stable to instabilities due to the strong vertical stratification, b z , of the ML base.
The EPV calculation in (15), expressed in the local Cartesian coordinate system for each Seasoar leg, neglects effects due to the curved flow around the eddy.We now consider EPV in cylindrical coordinates (Shakespeare 2016), where R is the curvature of the flow and the azimuthal velocity, u θ , is negative in a clockwise (cyclonic) rotational sense.The curvature term, u θ R −1 b z , is negative in stably stratified conditions (b z > 0) and, therefore will decrease EPV from the estimate in (15).The EPV calculation was repeated using ( 16) for legs N-S using an eddy radius R = 50 km (not shown).The average percent difference in Leg N (Leg S) is 23% (7%), however the inclusion of the curvature term has an indiscernible change on the EPV panels in Figure 9b.The number of locations with EPV > 0 decreased by 0.75% (Leg N) -0.15% (Leg S).Therefore, the total EPV is slightly lower when curvature effects are considered.This could result in a slight overestimation in the number of locations identified as favorable for inertial and symmetric instabilities below.
Throughout the survey, the mixed layer was consistently more susceptible to submesoscale in-

Discussion & Summary
Here we have presented high-resolution observations across the ACC as a cyclonic eddy formed in the Scotia Sea.The novel observations reveal submesoscale frontal variability and two distinct dynamic regimes along the periphery of the eddy as depicted in Figure 11.
In the northern to eastern regime of the survey, diffluent flow and deceleration were observed in the cross-front and along-front directions, respectively.Along the newly-formed southern edge of the eddy, along-front acceleration and cross-front confluent flow is observed coincident with a complex T-S structure, similar to submesoscale features found in other studies, e.g.filaments and streamers (Gula et al. 2014;Klymak et al. 2016).A submesoscale instability analysis identified regions across each cross-frontal section prone to the development of gravitational, mixed, symmetric and inertial instabilities.Favorable conditions for mixed and symmetric instabilities were found near large cross-frontal density gradients in the mixed layer throughout the survey.Despite the loss of frontal integrity observed in the southern regime, the eddy discussed here maintained a distinct signature in SST and SSH over the following two months as evidenced by remote sensing imagery.
The Scotia Sea hosts an especially high abundance of mesoscale eddies (Frenger et al. 2015) in the eddy-rich Southern Ocean.Eddy kinetic energy (EKE) in this region, calculated from timemean removed, altimetry-derived geostrophic surface currents (AVISO;1993-2015) is O(0.1 -1 m 2 s −2 ).Recent submesoscale-resolving modeling results indicate a strong correlation between mesoscale EKE and submesoscale vertical velocity in the Southern Ocean (Rosso et al. 2015) implicating a downscale energy transfer.Although the Scotia Sea EKE values and w esimates presented here are much higher than the domain-averaged magnitudes reported in Rosso et al. (2015), the trend of high EKE and high w is consistent.
The strong vertical circulation found at the SAF, suggests that submesoscale processes might be critical in transforming and subducting mode and intermediate waters, although such processes have been mostly ignored in previous studies.Water mass properties across the frontal region were initially observed as a cold, fresh eddy region and a warm, salty outer region.The rapid spread in T-S space suggests mixing occurred during the eddy formation.Enhanced vertical circulation and mixing, prompted by submesoscale processes, have the potential to transform mode and intermediate water density classes and contribute to the uptake of anthropogenic heat and carbon to the Southern Ocean.A quantification of the net water-mass subduction associated with the observed circulation will be part of a future study.

Cyclonic mesoscale eddies have been observed with high chlorophyll signatures in the Scotia
Sea (Kahru et al. 2007), implicating their importance on primary production in the region.Studies resolving submesoscale dynamics in mesoscale eddies have shown that strong vertical velocities, like those presented here, may drive the vertical exchange in the upper ocean with important effects on nutrient supply to the photic zone (Lévy et al. 2001;Mahadevan et al. 2008;Lévy et al. 2012; cruise are a focus of a future study. Acknowledgments

APPENDIX
The derivation of the vertical velocity presented in Section 4a requires two key assumptions.The first relates to the depth range over with the drifter acceleration is valid, and the second requires that the local Eulerian acceleration is much smaller than the drifter acceleration.Both assumptions have critical implications for the estimate of vertical velocity and we thus expand on the justification for making these assumptions below.
The assumption presented in ( 2   The along-front direction, θ along , is defined by the drifter, D16, trajectory.The cross-front direction, θ cross , is θ along -90.The mean true heading of Seasoar legs, θ leg , are calculated with cross-front distance increasing away from the eddy center.Legs are projected onto a crossfrontal axis through a rotation of θ rot = θ cross − θ leg .The axis projection alters the horizontal spacing of survey measurements by the multiplication factor, cos(θ rot ).Legs labeled N-S correspond to section labels in Figure 5.
, consisted of 25 sections around the edge of the eddy ranging from 25 -40 km in length.Maps of 10-m depth temperature and salinity from these 25 sections are presented in Figure 5a-b.The beginning northern sector of the survey is characterized by sharp temperature and salinity fronts (2 • C, 0.2 psu in 2 km at 4-m depth) with warm, salty water outside and cold, fresh waters inside the eddy.A region characterized by a loss of temperature and salinity frontal integrity is observed along the southern portion of the survey.The repeat observation of temperature and salinity intrusions in consecutive sections suggests the presence of a 3-dimensional structure such as a submesoscale streamer or filament, only a few kilometers across, in the newly-formed southern eddy sector.Note that the filaments occur in a region that was previously an open meander characterized by weak lateral gradients in temperature and salinity.
show weaker currents in the southern portion of the eddy compared to the north.A sign change in cross-frontal velocities on either side of the drifters indicates diffluent flow during the majority of the survey with confluent cross-frontal flow in the southern portion of the survey.The ageostrophic component of the curved flow around the eddy can be estimated from the along-frontal velocities by comparing the centripetal acceleration term with the Coriolis acceler-ation, C = u(R f ) −1 .Assuming an eddy radius, R = 50 km, C is maximum along the northern portion of the survey (0.25) and minimum (0.10) in the southern eddy sector.This indicates a larger cyclogeostrophic component to the flow in the north.Wind forcing during the Seasoar survey was unusually calm for April in the Southern Ocean with wind speeds < 10 m s −1 and winds from SE to NW rather than the expected westerlies.A partial infrared SST image of the eddy was captured during the Seasoar survey by an AVHRR sensor aboard the Metop-a satellite on 11 May 2015 at 12:42 GMT (Figure 6a).The highresolution (1 km) SST data show strong gradients along the northern eddy boundary and weaker gradients to the southeast, similar to Figure 5a.Unfortunately, clouds mask the southern and western sectors of the eddy.The ship's underway temperature data at 4-m depth is overlaid on the infrared SST data in Figure 6b.The noticeable offset in temperatures is due to the northward movement of the eddy in the 2.5 days between the beginning of the survey and the satellite measurements.The ship's temperature data is also plotted atop optimally-interpolated microwave SST data for 11 May 2015.The eddy boundary, defined by the 3 • C isotherm in Figure 6a and c is drastically different between the 1-km infrared and coarser microwave SST data.
Figure 5c-d.Along-front velocities decrease whereas cross-front velocities switch from confluent to the surface are used to approximate ūE from z = 0 to 50 m, as detailed in the appendix.The cross-frontal velocity gradients, ∂ ūE ∂ y and ∂ vE ∂ y , are averaged +/-1 km from the center of each Seasoar leg (Figure 8b).Error velocities reported by the ADCP processing and propagated through the w calculation are shown as error bars in Figure 8d.Vertical velocities calculated from (12) are presented in Figure 8d with negative (positive) values during the N-SE (S) eddy survey sectors.Upwelling velocities are calculated during the N, NE and E sectors of the survey, when diffluent cross-front flow and drifter deceleration is observed.Subduction is indicated in the southern survey sector when drifters accelerated.There is a strong dependence on ∂ vE ∂ y in our calculation indicating the cross-frontal flow is related to the vertical circulation.The estimated magnitudes of w E | 50m , O(100 m day −1 ), are similar to reported values forsubmesoscale processes, however, we can not discern the relative contributions of the mesoscale and submesoscale vertical motions here.
using the balanced Richardson number, Ri B = f 2 b 2 z b −4 y .The criteria presented in Thomas et al.
stabilities than the deep, stable regions where EPV<0.Gravitational instability is most likely early in the survey and away from density fronts where MLD are large.The criteria for mixed and symmetric instabilities are met within density fronts in Legs N -E.Conditions conducive for inertial, or centrifugal, instability are located on the outer (right-hand) side with Ro g = ζ g f −1 < 0.Regions where conditions are conducive to the development of submesoscale instabilities are shown as a fraction of the mixed layer in Figure9d.There is a general decrease between the N and S legs, indicating a greater proportion of the ML is more prone to instabilities earlier in the survey versus in the legs collected in the southern sector of the eddy.Throughout the survey, the majority of the instability indications are for gravitational with conditions favorable for symmetric or mixed gravitational and symmetric concentrated near lateral density gradients.c.Water mass modification The sharp temperature and salinity fronts across the eddy boundary indicate the presence of different water masses.T-S histograms for Seasoar sections N-S, Figure 10a, show the prevalence of measurements in 0.15 • C and 0.015 salinity bins.In Leg N the T-S measurements largely populate two separate regions in T-S space, with cold, fresh inner waters in the bottom left of the diagram and the warm, salty (spicy) outer region measurements in the top right.The two regions in T-Sspace are connected via σ θ 27 , the isopycnal that outcrops on either side of the dense filament at the front center in Leg N, previously presented in Figure7.A similar connection along deeper isopy-cnals, such as σ θ 27.2 , is not observed in Leg N (Figure10).This is due to an unequal isopycnal upheaval across the Seasoar leg and the 200-m depth limit of the dataset.A cross-front exchange is observed in Legs NE-E as cool, fresh measurements σ θ 27 -27.2extend into warmer and saltier T-S space.By Leg S, the T-S space is fully populated indicating mixing or advection of new water masses, not previously observed at the start of the survey.Locations previously identified as susceptible to submesoscale instabilities in Section 4b are shown in T-S space (Figure 10b).Instabilities are mostly favored along the σ θ 27 , supporting an along isopycnal exchange across the frontal region.The exchange or modification along σ θ 27.1 -27.2 suggests that water mass properties below the MLD are also affected on timescales of O(1 day) and horizontal length scales of O(1-10 km) during the formation of this mesoscale eddy.
) sets the drifter acceleration equal to the depth-averaged Eulerian acceleration from the drogue depth of 50 m to the surface.If the drifter has a sufficient drag ratio (see Section 2), this assumption is justified and u D ∼ ūE .A comparison of u D and the ūE from 30 -50 m, presented in FigureA1a, shows very strong agreement.Due to the blanking distance of the 75 kHz ADCP, measurements of u E are only available for depths below 30 m. Slab and linear extrapolations from 30 m to the surface are used to approximate ūE from 0 to 50 m.The depth-integrated 30-50 m Eulerian velocities and depth-integrated extrapolated 0-50 m Eulerian velocities are compared with the drifter velocities in FigureA1b and A1c.The slab-extrapolated approximation is a better fit than the linear extrapolation.The calculation of vertical velocity in Section 4a is carried out with both extrapolation approximations of u E (Figure8d).

Fig. 8 .Fig. 11 .FIG. 1 .
Fig. 8. (a) Time series of along-front drifter velocity, u D (m s -1 ), for the three drifters released and followed during the Seasoar survey from 9 to 12 May 2015.Along-front, x, and cross-front, y, water velocity components, u E and v E , measured within 1 km of the frontal center are shown for the drifter drogue depth of 50 m.Water speed (*) is also included.(b) Cross-front gradients of u E (gray) and v E (black) at 50-m depth and averaged +/-1 km across the front.Negative ∂ vE ∂ y (black) indicates confluent flow.Error bars indicate uncertainty of the ADCP measurements.(c) Estimation of terms in Equation 9 after making steady-state assumption.(d) Vertical velocity at the drogue depth of 50 m, w 50m (m day -1 ), with ∂ vE ∂ y (s -1 ), from panel (b) shown in color.Error velocities of the ADCP are propagated through the calculation of w and are shown as error bars.Additional estimates of w are included for the slab (black) and linear (gray) extrapolations of u E and v E to the surface.Vertical velocities and ∂ vE ∂ y < 0 indicate subduction and confluence, respectively.The duration of Seasoar legs is shaded in each panel. . . . . . . . . . . . . . . . . . . . . . . .40 Fig. 9. (a) Cross-front buoyancy gradient, b y , (s -2 ) calculated at 10-m water depth for Seasoar legs N to S. Legs are oriented with the inside of the meander and eddy on the left-hand side of each panel.(b) A 2-dimensional estimate of Ertel potential vorticity (s −3 ) is shown with the zero contour in white and the MLD, defined as a 0.01 (0.1) kg m -3 density difference from the surface, as a thick (thin) black line.(c) Submesoscale instability analysis results based on the Ri B criteria.(d) Instances of instabilities identified in (c) shown as a fraction of the 0.01 kg m −3 density difference MLD. . . . . . . . . . . . . . . . .41 Fig. 10.T-S diagram histograms for Seasoar legs N to S. (a) Color indicates number of measurements in 0.15 • C and 0.015 salinity bins and (b) instability types in the mixed layer as diagnosed in Section 3d.The cold, fresh observations inside the meander and eddy occupy the bottom left 'hot spot' of measurements in T-S space in Leg N.An exchange along isopycnals σ θ 27 (bold) and σ θ 27.2 (gray) occurs over this series. . . . . . . . . . . . . .42 Fig.11.Cartoon summarizing frontal circulation during eddy formation.The two cross-frontal sections represent the northern and southern sectors of the survey, legs N and S. . . . . .43 Fig.A1.(a) Drogued drifter velocities (u D ) compared with depth-averaged Eulerian velocities between 30 and 50 m (u E , •) and approximations of u E from 0 to 50 m using slab (•) and linear (+) extrapolations.(b) Comparison of the measured u E (30 to 50 m) to the extrapolated approximations (0 to 50 m).(c) Same as (b) for v E .Linear regression fits and respective skills, var(fit) / var(data), are reported in each panel. . . . . . . . . . . . .44 Fig. A2.Estimates of Eulerian local acceleration during the Seasoar survey calculated from ship track intersections (map inset) during the Seasoar survey.Gray bars show the duration of each Seasoar leg; the N-S legs (dark gray) are labeled. . . . . . . . . . . . . .45 . The SMILES project is funded through the National Environmental Research Council, standard grant NE/J009857/1.JR311 data collection and technical support were received from the British Antarctic Survey, the crew of RRS James Clark Ross, and the NEODAAS and PML remote sensing groups.Seasoar operations were led by National Marine Facilities technicians Paul Provost, Dougal Mountifield, Julie Wood, and Candice Cameron, from the UK National Oceanographic Centre.Peter Ganderton designed and built the drifter electronics package.The altimeter products were produced by Ssalto/Duacs and distributed by Aviso, with support from Cnes (http://www.aviso.altimetry.fr/duacs/).Microwave OI SST data are produced by Remote Sensing Systems and sponsored by National Oceanographic Partnership Program (NOPP) and the NASA Earth Science Physical Oceanography Program.
are larger than the average Eulerian acceleration estimate.In the NE sector of the eddy and a few southern eddy legs, this is not true and the steady state assumption cannot be made from this ship intersection estimate alone.