Eddy-driven meridional transport across the ACC upstream of Drake Passage

GEOPHYSICAL RESEARCH LETTERS, VOL.???, XXXX, DOI:10.1029/, 1 2 Eddy-driven meridional transport across the ACC upstream of Drake Passage Loïc. Jullion, 1,3 Kevin G. Speer, 1 Dhruv Balwada, 1 Raffaele Ferrari 2 and John Marshall, 2 1 Geophysical Fluid Dynamics Institute, Florida State University, Tallahassee, Florida, USA. 2 Earth, Ocean and Atmospheric Sciences, Florida State University, Tallahassee, Florida, USA. 3 Department of Earth, Atmospheric and Planetary Sciences, Massachusetts Institute of Technology,Cambidge, USA 4 Now at Mediterranean Institute of Oceanography, Aix-Marseille University, CNRS/INSU, IRD, UM110, Marseille, France

X - 2 JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 The magnitude and vertical (density) distribution of the eddy-driven transport in the Southeastern Pacific sector of the Southern Ocean is inferred from observations and recent estimates of isopycnal diffusivity from the Diapycnal and Isopycnal Mixing Experiments in the Southern Ocean (DIMES) experiment. A vertical profile of isopycnal diffusivity derived from numerical modelling, tracer release and accoustically-tracked isopycnal floats released during DIMES is combined with historical CTD profiles. Away from the surface layer and above the depth of the sill of the Drake Passage, the eddydriven flow is southward between γ n = 27.5 and γ n = 28.16 kg m 3 before reversing to become northward where the Potential Vorticity gradient reverses. The calculated eddy velocities are small compared to other regions of the Southern Ocean where complex topography promotes instability and eddy transfer. These results lend support to the idea that local dynamics, dictated by flow-topography interactions, strongly influence the magnitude of the tracer transport and of the overturning circulation in the Southern Ocean.

JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE X - 3 1. Introduction 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 In the Southern Ocean, the combination of a strong eastward Antarctic Circumpolar Current (ACC) and a cross-acc meridional circulation has far-reaching consequences for the closure of the Meridional Overturning Circulation (MOC). The role of the MOC in the global climate system has motivated, over recent decades, intense research on various aspects of its dynamics and associated water mass transformation [Rintoul and Garabato, 2013]. A combination of wind and air-sea-ice buoyancy forcing over the SubAntarctic and Antarctic domains drives the ACC and produces sharply rising isopycnals, providing an efficient isopycnal pathway to return deep water to the surface [Marshall and Speer, 2012]. Air-sea-ice interactions lead to water mass transformation of upwelling water where, north of the Antarctic Divergence ( 60 65 ), it gains buoyancy and is transformed into Antarctic Intermediate Water (AAIW) and Sub-Antarctic Mode Water (SAMW), and continues to circulate within the thermocline. This Upper Cell of the MOC is accompanied by a Lower Cell, in which water flowing close to the Antarctic margin is transformed into a variety of dense waters on the shelf and slope, some of which becomes Antarctic Bottom Water. Recent ideas about the MOC in the Southern Ocean emphasize the southward eddydriven flow of deep waters, balancing to some extent the northward Ekman transport in the upper layers, the so-called eddy compensation, e.g. Marshall and Radko [2003]. Quantifying the eddy fluxes remain challenging in both observations and models due to the small spatial scales of the eddies (10-100km), their inhomogeneity and intermittency. Recent studies, for example, have identified eddy fluxes situated on the lee side of to-

X - 4 JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE 39 40 41 42 43 44 45 46 pography, where the orientation of ACC fronts departs significantly from its quasi-zonal orientation, as major contributions to the cross-stream transport. In other regions, by contrast, strong ACC fronts can act as a barrier to mixing [Abernathey et al., 2010; Lu and Speer, 2010; Naveira-Garabato et al., 2011; Thompson and Sallée, 2012]. This zonal complexity emphasizes the need to evaluate regional contributions to the circumpolar meridional eddy fluxes. In this study we focus on a quiet region where the ACC is essentially zonal, the Southeast Pacific sector of the Southern Ocean just upstream of Drake Passage (Fig. 1). Here, 47 we seek to quantify the southward eddy-driven transport. The area comprises a zone 48 49 50 51 52 53 54 55 56 57 of weak eddy-activity between more active regions close to the East Pacific Rise and Drake Passage, where the ACC passes through topographic constrictions. The Diapycnal and Isopycnal Mixing Experiments in the Southern Ocean (DIMES) project measured isopycnal tracer mixing in this region with both a tracer and subsurface floats. The tracer was confined to a narrow range of densities, while the float data was analyzed over a relatively broad range of densities and depths; both approaches required averaging over the entire sector upstream of Drake Passage, from 105 W to Drake Passage. Through a combination of the observed regional-average of eddy-diffusivity coefficient, extrapolated to a vertical profile with a high-resolution numerical model, and in situ hydrographic profiles, we estimate an average meridional eddy flux and residual flow in this area.

JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE X - 5 2. Data and Methodology The residual momentum budget of the ACC below the surface, where momentum is input by the wind, and above topographic features that support pressure gradients is: ρ 0 fvz γ = ρ 0 z γ 2 v Q. (1) 58 59 60 61 (.) denotes time and along-stream spatial averages, primes are departure from those averages, Q = (f + ζ)/z γ is the isentropic Ertel potential vorticity, IPV [Vallis, 2006], f is the Coriolis parameter, ζ is the vertical component of the vorticity and z γ = z/ γ is the layer thickness. Assuming that the IPV flux is down its mean gradient we have ρ 0 fvz γ = ρ 0 z γ 2 κ P V y Q (2) 62 where κ P V is the IPV eddy diffusivity. If we further neglect the relative vorticity and the 63 64 65 meridional gradient in f in Equation 2, as appropriate in the Southern Ocean [Mazloff et al., 2013], we can relate the meridional mass flux to the product of the IPV diffusivity and the merdional thickness grandient, vz γ = κ P V y z γ (3) 66 Recognizing that the residual velocity is defined as v = vz γ /z γ, we can also write v = κ P V y z γ /z γ (4) A different approach to parameterize the effect of eddies on the residual circulation was suggested by Gent and McWilliams [1990] and relies on the energy cycle of baroclinic

X - 6 JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE instability. Eddies develop in the Southern Ocean as a result of baroclinic instability of the ACC and extract energy from the mean current. An expression of the residual velocity that is guaranteed to always extract energy from the mean flow was proposed v = γ (κ GM y z/z γ ). (5) 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 The reader is referred to Gent et al. [1995] for a demonstration that with this form, eddies extract energy from the mean flow regardless of the value of κ GM diffusivity. The IPV eddy diffusivity, κ P V, is obtained from observations and high-resolution modeling performed within the DIMES project. The observations came from 140 acousticallytracked RAFOS floats (shown in Fig. 1, Hancock and Speer [2013]) and 3 years of tracer surveys in the Southeast Pacific and the Scotia Sea [Watson et al., 2014]. Floats [Lacasce et al., 2014] and tracer data [Tulloch et al., 2014] are used to compute the isopycnal diffusivity at the depths where data are available. State of the art numerical simulations with MITgcm are used to ENHANCE(?) the observations. [Lacasce et al., 2014] obtained a vertical diffusivity profile by releasing numerical floats at 20 different depths while [Tulloch et al., 2014] used the simulations to assess the bias caused by undersampling the tracer patch and to obtain a vertical profile of diffusivity. The diffusivity estimated from the dispersion of Lagrangian particles have been shown to match the tracer diffusivity providing that these diffusivites are calculated over sufficiently long periods [Klocker et al., 2012]. Indeed, both tracer-based and float-based diffusivity are in good agreement, giving confidence in the results. In this study, we use vertical profiles of κ P V obtained by Tulloch et al. [2014] and Lacasce et al. [2014]. The eddy-driven meridional transport, v, is derived from κ P V using Eq. 4.

JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE X - 7 85 86 87 We derive the meridional thickness gradient, y (z γ ) across streamlines by constructing climatological sections of the thermohaline properties and pressure along isopycnals based on 783 historical full-depth CTD profiles between 80 W and 110 W (corresponding to the 88 DIMES region, Figure 1) extracted from the Southern Ocean Data Base. These are 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 complemented with additional recent profiles from the CLIVAR Repeat Hydrography program (S04P in 2011, P18 in 2008, AAIW05 and AAIW06 occupied in 2005 and 2006 respectively). We use a neutral density [Jackett and McDougall, 1997]-dynamic height coordinate system in order to differentiate true along-isopycnal variability from noise introduced by isopycnal heaving and ACC meridional meandering. Dynamic height (Φ) has been shown to be a useful proxy of the position of the ACC fronts delimiting the location of the ACC [Sun and Watts, 2002]. The sections are constructed following the approach used by Naveira-Garabato et al. [2011], that is for each CTD profile, the dynamic height at 400 m relative to 3000m, Φ 3000 400, is calculated. Figure 2a shows that Φ 3000 400 increases monotonically between 70 S and 50 S (corresponding to Φ 3000 400 = [8 18] m 2 s 2 ) thus supporting our use of Φ 3000 400 as a pseudo-geographical coordinate. Potential temperature, Θ(γ n ), pressure, P(γ n ), salinity, S(γ n ), and oxygen, O(γ n ), profiles are mapped on discrete neutral surfaces, γ n at 0.02 kg m 3 intervals. Then, we take advantage of the strong zonal coherence of the ACC water masses and make the assumption that in this region, characterized by relatively smooth topography, the along-streamline isopycnal properties are essentially homogeneous. This assumption is supported by figure 2b, which shows that the distribution of salinity along an isopycnal (γ n = 27.6 kg m 3 ) as a function of dynamic height is coherent throughout

X - 8 JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE 107 108 109 110 111 112 113 114 115 116 117 118 119 the region irrespective of zonal location. Therefore, we define a Φ 3000 400 grid ranging from 8 to 18 m 2 s 2 with an increment δφ = 0.4 and for each grid point, i, we average together all profiles: Θ Φ i = 1 m mp=1 Θ(γ n ), where m is the number of profiles with dynamic height = Φ i ± δφ/2. Calculating the cross-stream gradients requires computing the distance between the climatological streamlines. Given the quasi-zonal structure of the ACC in the region, we assume that the climatological section is meridional and we calculate the climatological latitude of each streamline by fitting a cubic spline to the Φ 3000 400 vs. latitude. Finally, the thickness gradient across the ACC is computed by fitting a cubic spline over each isopycnal pressure section (Figure 2d) in order to minimize the influence of small scale noise caused by data scarcity and the presence of small-scale features and emphasize the large-scale cross-acc gradients. In the following section,we average this thickness gradient over the ACC region between the SAF and PF. 3. Estimates of eddy-induced transport 120 121 122 123 124 125 126 127 The climatological sections of salinity, oxygen, isopycnal thickness and pressure alongside their corresponding cross-stream gradients as well as the position of the ACC fronts [Orsi et al., 1995] are shown in Figure 3. In the deep layers, the low-oxygen signal of the Upper Circumpolar Deep Water (UCDW) at γ n = 27.9 kg m 3 overlies Lower Circumpolar Deep Water (LCDW), characterized by a salinity maximum centered around γ n = 28.1 kg m 3. At shallower levels the low salinity AAIW (above γ n = 27.5 kg m 3 ) covers the region. The core of the ACC, between the SAF and the PF, is characterized by relatively large isopycnal gradients. The salinity (oxygen) gradient (Figure 3d,e) is largely positive

JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE X - 9 128 129 130 131 132 133 134 135 136 137 138 139 140 (negative) and maximum above γ n = 28.00 kg m 3. The thickness gradient (Figure 3f), while relatively noisy, is predominantly positive, indicative of a net southward residual flow above γ n = 28.10 kg m 3, before reversing in the deeper water masses, suggesting a deep northward residual velocity. We note that, below about γ n = 28.2 kg m 3 (corresponding to the sill depth of Drake Passage), isopycnals intercept topography, indicating that below this depth, topographic form stress cannot be neglected. In addition, we note that above γ n = 27.3 kg m 3, isopycnals are located within the top 500 m of the water column where other dynamical processes (e.g. Ekman transport, outcropping isopycnals and associated geostrophic flow) will dominate the meridional flow. Therefore, we limit the calculation of v to the UCDW-LCDW density range (27.5 < γ n < 28.2 kg m 3 ) where eddies are expected to be the dominant contributor to the meridional transport according to the simple balance described above. 141 Figure 4 shows the vertical profiles of κ P V from Lacasce et al. [2014] (solid black line) 142 143 144 145 146 147 148 149 and Tulloch et al. [2014] (dotted black line), v (light gray) calculated using Equation 4, κ GM (dark gray) calculated by numerically solving 5, salinity (red) and oxygen (brown) against density (right panel) and depth (left panel) in the Southeastern sector of the Southern Ocean (between 55 and 60 S). κ P V reaches a maximum be 900 m 2 s 1 between γ n of 28.00 and 28.1 kg m 3 (2000-3000m) corresponding to the salinity maximum of LCDW. Across the oxygen minimum UCDW layer, the float-derived diffusivity quickly decreases to about 500 m 2 s 1 at 27.8 kg m 3 while the tracer-based diffusivity gradually decreases to reach the same value at γ n = 27.65 kg m 3.

X - 10 JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE 150 151 152 153 154 155 156 157 158 159 The corresponding eddy-driven velocity in the UCDW and LCDW layers (γ n = 27.4 28.10 kg m 3 ), is of the order of 1 10 4 m s 1 southward before changing sign (at the depth where the PV gradient also reverses) to reach 3 10 4 m s 1 at γ n = 28.16 kg m 3 just below the salinity maximum. The eddy-driven velocity calculated from the tracerbased diffusivity differs by up to 40% in the UCDW density range, around γ n = 27.8 kg m 3 corresponding to the isopycnal range where the two diffusivities differ the most. We estimate the uncertainty related to the spline fitting by varying the smoothing coefficient used. The impact on v is less than 15% except for one isopycnal (γ n = 28.02kg m 3 ) where the error reaches 40%. This is due to the relatively weak thickness gradient (not shown) making it more sensitive to the choice of the fit. The similitude of v for different 160 values of κ (v 1 10 4 m s 1 for κ 500 m 2 s 1 near γ n = 27.7 kg m 3 and 161 for κ = 900 m 2 s 1 at γ n = 28.0 kg m 3 ) reflects the overarching importance of the 162 163 164 cross-stream thickness gradient (hence PV gradient) in setting the magnitude of the eddy fluxes. The κ GM varies significantly from the float-derived κ P V both in its vertical structure 165 and intensity (Figure 4). While κ P V peaks at 800 m 2 s 1 around γ n = 28.00 28.1 kg 166 167 m 3, κ GM is nearly vertically uniform, with a mean value of 460 m 2 s 1. Both estimates of kappa GM (from floats and tracers) are very close 4. Discussion 168 169 170 Extrapolated over the whole Southern Ocean ( 20000 km along 60 ), the eddy-induced transports found here of UCDW (γ n = [27.5 28.00] kg m 3 ) and LCDW (γ n = [28.00 28.20] kg m 3 ) equate to 1.6 Sv and 3 Sv respectively based on the float-derived diffusivity

JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE X - 11 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 and 3.1 and 3.0 Sv with the tracer-derived one. These transport estimates, based on the vertical profile of v in the SE Pacific, are expected to be weak compared with other sectors of the Southern Ocean. Speer et al. [2000] applying similar methodology using hydrographic data south of Australia found values of v roughly one order of magnitude larger than in this study. Assuming the eddy-driven upwelling associated with the meridional residual velocity is predominantly oriented along isopycnals, the upwelling rate, w, can be estimated from the slope of the isopycnals across the ACC defined as s γ = z γ / y = w /v, where z γ is the difference in the isopycnal depth between the northern and southern side of the ACC ( 1000 m within the UCDW layer) and y is distance between northern and southern limit of the ACC ( 860 km). In the UCDW layer, the average v 1.10 4 m s 1 so that we find the rate of isopycnal upwelling w 110 7 m s 1 ) (4 m yr 1 ). Naveira- Garabato et al. [2007] found an eddy-driven isopycnal upwelling of 330±110 m yr 1 (110 5 m s 1 ) in the Drake Passage/Scotia Sea region based on the cross-stream spreading rate of mantle helium injected in the ACC near Drake Passage within the UCDW density range. Our estimate of isopycnal upwelling just upstream of Drake Passage is about two orders of magnitude lower than the value found by Naveira-Garabato et al. [2007]. One possible explanation for the lowest upwelling rate might be the flatter isopycnal slope in the SE Pacific sector compared with Drake Passage (the 28.00 isopycnal rises from 900m and 1500m respectively). Our eddy-induced transports are significantly smaller than the 15 Sv of deep water geostrophically flowing southward across 32 S based on large-scale inverse models [Sloyan

X - 12 JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE 193 194 and Rintoul, 2001; Lumpkin and Speer, 2007; Naveira-Garabato et al., 2013] suggesting that the SE Pacific sector of the Southern Ocean plays a relatively minor role in the 195 ageostrophic meridional overturning. Describing the dynamical processes setting such 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 small values of eddy transport is beyond the scope of this short paper; however, there is a growing body of evidence suggesting that zonal modulations in the ACC characteristics play a significant role in setting the strength of the meridional eddy transport (e.g. Abernathey et al. [2010]; Lu and Speer [2010]; Thompson et al. [2010]; Naveira-Garabato et al. [2011]; Sallée et al. [2011]; Thompson and Sallée [2012]. The emerging view points towards the existence of mixing hotspots, where local interactions between the ACC fronts and the topography enhance cross-frontal mixing, separated by more quiescent regions where the ACC fronts act as mixing barriers. Thompson et al. [2010] and Thompson and Sallée [2012] based on modelling and altimetry data also support the view that the SE Pacific is a region of little cross-frontal mixing due to the absence of prominent topographical obstacle in the path of the ACC. Abernathey et al. [2010], using a Southern ocean state estimate also found reduced isopycnal mixing in the 60 W-120 W region. Obtaining a circumpolarly integrated estimate of the eddy-driven tracer flux remains a key step toward closing the MOC in the Southern Ocean. However, there is no compelling way to extrapolate the measurements without accounting for the horizontal distribution of the PV (thickness) gradient and the zonal variations in the vertical profiles of κ P V. Here, we briefly discuss the impact of variations in the PV gradient for different regions 213 of the Southern Ocean on v. Although observations and models suggest that κ P V varies 214 circumpolarly [Abernathey et al., 2010; Lu and Speer, 2010; Naveira-Garabato et al., 2011],

JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE X - 13 215 216 217 we apply a similar methodology by combining climatological sections constructed from the Southern Ocean Data Base in the Drake Passage (Sr1b section), 30 E (corresponding to the WOCE I6S section) and 140 E (corresponding to the WOCE SR3 line) using the 218 vertical profile of κ P V characteristics of the SE Pacific (Figure 4). These back of the 219 220 221 envelope estimations (not shown) reveal significant zonal variations in the intensity and vertical structure of v with the Drake Passage hosting the strongest residual velocity ( 1 10 3 m s 1 just above γ n = 28.00) while south of Africa and Australia, the residual 222 velocities peak at 4 10 4 m s 1 just above γ n = 28.00, reflecting significant zonal 223 224 225 226 227 228 229 230 231 232 233 234 235 differences in the layer thickness gradient (and thus the large-scale PV gradient). Numerical simulations have shown significant differences both in the vertical structure and magnitude of the eddy PV and eddy buoyancy diffusivities [Treguier, 1999; Smith and Marshall, 2009; Abernathey et al., 2013] suggesting that isopycnal diffusivity is different from the diffusivity to be used as part of Gent-McWilliam parameterization (a difference already noted in Gent et al. [1995]). [Smith and Marshall, 2009] provided an equation (their equation 4.3) relating both diffusivities. Abernathey et al. [2013] suggested that it was possible to obtain an estimate of κ GM from observations by assuming that the v* from the PV parameterization (equation 4) should be the same as the v* from the GM parametrization (equation 5). This is the same as the ode presented in equation 4.3 from [Smith and Marshall, 2009]. Here we follow this approach and solve the ode using the boundary condition that the κ GM = κ P V at 500m (corresponding to γ n = 27.5 kg m 3 ). This choice of boundary condition will impact on the magnitude of κ GM but numerical

X - 14 JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE 236 simulations seem to suggest that in the ACC, near 500m κ GM κ P V (c.f. for example 237 Figure 12 of Smith and Marshall [2009] or Figure 13 of Abernathey et al. [2013]). 238 The profile of κ GM obtained has a different vertical structure than the κ P V as shown 239 240 241 242 243 244 245 246 247 248 249 250 in Figure 4. κ GM has no mdi-depth maxima and just slightly increases with depth. This structure is the same as the one obtained by Smith and Marshall [2009] and Abernathey et al. [2013] from numerical simulations but the magnitudes are different (2 to 4 times smaller in our study). This is to be expected as Smith and Marshall [2009] used a QG model with mean hydrography based on a climatology of the Australian sector of the Southern Ocean and Abernathey et al. [2013] used an idealized primitive equation channel model, whereas our results are for the Southeastern Pacific sector of the Southern Ocean. This study provides evidence inferred from observations supporting the difference in the vertical structure of κ GM and κ P V. This profile of buoyancy diffusivity could be directly used in a numerical model of the region in order to capture both the slumping of isopycnals (characteristic of the GM parameterization) as well as providing the correct (within observational uncertainty) the residual velocity across the ACC. 5. Conclusion 251 252 253 254 255 We evaluate the strength of the eddy-driven fluxes in the SE Pacific sector of the Southern Ocean by combining a vertical profile of eddy-diffusivity derived from Lacasce et al. [2014] with climatological hydographic sections constructed from historical CTD profiles. Large meridional eddy fluxes were not expected in this region and the calculated eddydriven velocity v does appear to be small when compared with previous, indirect esti-

JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE X - 15 256 257 258 259 260 261 262 263 264 265 266 267 268 269 mates from inverse calculations [Speer et al., 2000] and natural tracer release experiment [Naveira-Garabato et al., 2007]. In addition, both thickness gradients and the vertical structure of κ vary significantly around the Southern Ocean [Lu and Speer, 2010; Naveira-Garabato et al., 2011] and extrapolating eddy flux on the basis of either diffusivity or thickness gradient separately is not likely to be accurate since both are likely to play a significant role in the magnitude of the eddy-driven tracer fluxes and the density range over which these apply. The relatively (compared with previous studies focusing on other sectors of the Southern Ocean) weak residual velocity found in the SE Pacific sector of the Southern Ocean lend support to the idea that other regions, where local dynamics dictated by flow-topography interactions strongly enhances the magnitude of the southward tracer transport, must play a disproportionately large role in order to close the Southern Ocean MOC. This variability needs to be incorporated in models of the Southern Ocean overturning circulation to properly reflect meridional physical and chemical transports. 270 271 272 273 274 Acknowledgments. LJ acknowledges financial support from NSF (OCE-1231803) and from the European Union via a Marie Curie fellowship (FP7-PEOPLE-IEF-2012 no 328416). We thank all the crew, scientists and officers who participated to the collection of the data used in this study, in particular during the DIMES cruises on board RRS James Clark Ross, RRS James Cook, RV Revelle, RV Thompson.

X - 16 JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE References 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 Abernathey, R. P., J. Marshall, M. Mazloff, and E. Shuckburgh (2010), Enhancement of mesoscale eddy stirring at steering levels in the Southern Ocean, J. Phys. Oceanogr., 72, 170 184. Abernathey, R. P., D. Ferreira, and A. Klocker (2013), Diagnostics of isopycnal mixing in a circumpolar channel, J. Phys. Oceanogr., 72, 1 16. Gent, P. R., and J. C. McWilliams (1990), Isopycnal Mixing in Ocean Circulation Models, J. Phys. Oceanogr., 20 (1), 150 155. Gent, P. R., J. Willebrand, T. J. McDougall, and J. C. McWilliams (1995), Parameterizing Eddy-Induced Tracer Transports in Ocean Circulation Models, J. Phys. Oceanogr., 25, 463 474. Hancock, C., and K. Speer (2013), Diapycnal and isopycnal mixing experiment in the southern ocean rafos float data report, Tech. rep., Marine Field group, Florida State University. Jackett, D., and T. McDougall (1997), A neutral density variable for the world s oceans, J. Phys. Oceanogr., 27 (2), 237 263. Klocker, A., R. Ferrari, J. H. Lacasce, and S. T. Merrifield (2012), Reconciling floatbased and tracer-based estimates of lateral diffusivities, Journal of Marine Research, 70, 569 602. Lacasce, J. H., R. Ferrari, J. Marshall, R. Tulloch, D. Balwada, and K. G. Speer (2014), Float-Derived Isopycnal Diffusivities in the DIMES Experiment, J. Phys. Oceanogr., 44 (2), 764 780.

JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE X - 17 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 Lu, J., and K. G. Speer (2010), Topography, jets, and eddy mixing in the Southern Ocean, J. Mar. Res., 68 (3-4), 479 502. Lumpkin, R., and K. G. Speer (2007), Global ocean meridional overturning, J. Phys. Oceanogr., 37 (10), 2550 2562. Marshall, J., and T. Radko (2003), Residual-Mean Solutions for the Antarctic Circumpolar Current and Its Associated Overturning Circulation, J. Phys. Oceanogr., 33, 2341. Marshall, J., and K. G. Speer (2012), Closure of the meridional overturning circulation through Southern Ocean upwelling, Nature Geoscience, 5 (3), 171 180. Mazloff, M. R., R. Ferrari, and T. Schneider (2013), The Force Balance of the Southern Ocean Meridional Overturning Circulation, J. Phys. Oceanogr., 43 (6), 1193 1208. Naveira-Garabato, A. C., D. P. Stevens, A. J. Watson, and W. Roether (2007), Shortcircuiting of the overturning circulation in the Antarctic Circumpolar Current, Nature, 447 (7141), 194 197. Naveira-Garabato, A. C., R. Ferrari, and K. L. Polzin (2011), Eddy stirring in the Southern Ocean, J. Geophys. Res., 116 (C9), C09,019. Naveira-Garabato, A. C., A. P. Williams, and S. Bacon (2013), The three-dimensional overturning circulation of the Southern Ocean during the WOCE era, Prog. Oceanogr., pp. 1 38. Orsi, A., T. Whitworth, and W. D. Nowlin (1995), On the meridional extent and fronts of the Antarctic Circumpolar Current, Deep-Sea Research I, 42 (5), 641 673. Rintoul, S., and A. N. Garabato (2013), Dynamics of the southern ocean circulation, in Ocean Circulation and Climate: A 21 st Century Perspective, edited by J. G. G. Siedler,

X - 18 JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 S. Griffies and J. Church, pp. 471 492, Academic Press. Sallée, J.-B., K. G. Speer, and S. R. Rintoul (2011), Mean-flow and topographic control on surface eddy-mixing in the Southern Ocean, J. Mar. Res., 69 (4-6), 753 777. Sloyan, B. M., and S. R. Rintoul (2001), The Southern Ocean limb of the global deep overturning circulation, J. Phys. Oceanogr., 31 (1), 143 173. Smith, K. S., and J. Marshall (2009), Evidence for Enhanced Eddy Mixing at Middepth in the Southern Ocean, J. Phys. Oceanogr., 39 (1), 50 69. Speer, K. G., S. R. Rintoul, and B. M. Sloyan (2000), The diabatic Deacon cell, J. Phys. Oceanogr., 30 (12), 3212 3222. Sun, C., and D. R. Watts (2002), Heat flux carried by the Antarctic Circumpolar Current mean flow, J. Geophys. Res., 107, 3119. Thompson, A. F., and J.-B. Sallée (2012), Jets and Topography: Jet Transitions and the Impact on Transport in the Antarctic Circumpolar Current, J. Phys. Oceanogr., 42 (6), 956 972. Thompson, A. F., P. H. Haynes, C. Wilson, and J. Richards Kelvin (2010), Rapid Southern Ocean front transitions in an eddy?resolving ocean GCM, Geophys. Res Letters, 37, L23,602,, doi:10.1029/2010gl045386. Treguier, A.-M. (1999), Evaluating eddy mixing coefficients from eddy-resolving ocean models: A case study, J. Mar. Res., 57, 89 108. Tulloch, R., et al. (2014), Direct Estimate of Lateral Eddy Diffusivity Upstream of the Drake Passage, J. Phys. Oceanogr., 44, 2593 2616, doi:10.1175/jpo-d-13-0120.1. Vallis, G. (2006), Atmospheric and Oceanic Fluid Dynamics, Cambridge University Press.

JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE X - 19 340 341 342 Watson, A. J., J. R. Ledwell, M.-J. Messias, B. A. King, N. Mackay, M. P. Meredith, B. Mills, and A. C. Naveira-Garabato (2014), Rapid cross-density ocean mixing at mid- depths in the Drake Passage measured by tracer release, Nature, 501 (7467), 408 411.

X - 20 JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE SAF PF S. America SBdy Ant. Peninsula Figure 1. Bathymetry (ETOPO) of the South East Pacific/Drake Passage. The trajectories of the DIMES floats are shown in green. The CTD profiles used in this study are shown as blue dots and the climatological positions [Orsi et al., 1995] of the SubAntarctic Front (SAF), Polar Front (PF) and Southern Boundary of the ACC (SnBdy) are shown in red.

JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE X - 21 20 a) 15 φ 10 5 70 65 60 55 50 45 40 Latitude (S) 34.6 b) 34.4 S 34.2 34 33.8 6 8 10 12 14 16 18 20 Dynamic height (φ) 250 200 c) h 150 100 50 0 70 68 66 64 62 60 58 56 54 52 Latitude (S) Figure 2. Scatter plot of a) the CTD stations dynamic height, Φ (at 400m referenced to 3000m) versus latitude; b) salinity on γ n = 27.6 kg m 3 and c) thickness of the layer centered around 27.6 kg m 3 versus latitude. In each panel, the thick line is the spline fit.

X - 22 JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE γ n (kg m -3 ) γ n (kg m -3 ) γ n (kg m -3 ) 27 27.5 28 28.2 27 27.5 28 28.2 27 27.5 28 28.2 SACCF 68 66 64 62 60 58 56 Clim. Lat. PF SAF 34.7 34.5 34.3 34.1 33.9 33.7 33.5 350 300 250 200 150 800 700 600 500 400 300 200 100 0 SACCF PF SAF 27 27.5 S 28 28.2 27 27.5 O 2 28 28.2 27 27.5 H 28 28.2 68 66 64 62 60 58 56 Clim. Lat. 0.8 0.4 0 S (x 10-7 ) 0.4 0.8 0.8 0.4 0 O 2 (x 10-4 ) 0.4 0.8 0.8 0.4 0 H (x 10-4 ) 0.4 0.8 Figure 3. Climatological sections in gamma/latitude coordinates constructed from CTD profiles in the Southeast Pacific (80 110 W) of a) salinity, b) oxygen, c) layer thickness, d) y S, e) y O 2, y H. The thin black lines and white line represent the 150 500 1000 2000 4000 isobars and the shallowest density interfering with the topography at each streamline respectively. The vertical dashed lines correspond to the position of the SAF (SubAntarctic Front), PF (Polar Front), and SACCF (Southern ACC Front).

JULLION ET AL.: CROSS-STREAM EDDY TRANSPORT UPSTREAM OF DRAKE PASSAGE X - 23 Neutral density (γ n ) κ (m 2 s -1 ) 0 27 500 1000 κ (m 2 s -1 ) 0 500 1000 0 Depth (m) 27.2 500 AAIW UCDW 27.4 27.6 27.8 V* O 2 κ PV S 1000 1500 2000 2500 AAIW UCDW 28 κ GM 3000 LCDW 28.2 3500 LCDW 2 1 0 1 2 3 4 v* (x 10-4 m s -1 ) 2 1 0 1 2 3 4 v* (x 10-4 m s -1 ) 160 180 200 220 240 260 280 300 320 160 180 200 220 240 260 280 300 320 Oxygen (μm kg -1 ) Oxygen (μm kg -1 ) 33.4 33.6 33.8 34 34.2 34.4 34.6 33.4 33.6 33.8 34 34.2 34.4 34.6 Salinity Salinity Figure 4. Vertical profiles of eddy diffusivity, GM parameterization (dashed dark gray) and v (light gray) based on the DIMES floats (solid lines) [Lacasce et al., 2014] and the tracer spreading [Tulloch et al., 2014] (dotted lines) in density coordinates (left panel) and depth coordinates (right) averaged over the 55 60. Profiles of oxygen (brown) and salinity (red) are shown to relate eddy terms to the water mass structure. The zero crossing for v and neutral density boundaries of the water masses (AAIW = γ n = [27 27.5]; UCDW = γ n = [27.5 28]; LCDW = γ n = [28 28.2]) are shown as vertical and horizontal dotted lines respectively.