Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
 High resolution hydrographic sections and maps of the gradient of sea surface height (SSH) reveal that the Antarctic Circumpolar Current (ACC) consists of multiple jets or frontal filaments. Here we use a 15 year time series of SSH observations to determine the circumpolar structure and distribution of the ACC fronts. The jets are consistently aligned with particular streamlines along the entire circumpolar path, confirming and extending the results of an earlier study restricted to the region south of Australia. The intensity of the fronts (as measured by the cross-front gradient of SSH) varies along the fronts and the individual branches merge and diverge, often in response to interactions with bathymetry. Maps of absolute velocity at 1000 m depth derived from Argo trajectories confirm the existence of multiple current cores throughout the Southern Ocean. High resolution hydrographic sections and profiles of temperature and salinity from Argo floats are used to show that the front locations derived from fitting SSH contours to maps of SSH gradient are consistent with locations inferred from the traditional criteria based on water mass properties, suitably modified to account for multiple frontal branches. Three regions are examined in detail: the Crozet Plateau, the Kerguelen Plateau and the Scotia Sea. These examples show how recognition of the multiple jets of the ACC can help resolve discrepancies between previous studies of ACC fronts.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The flow of the Antarctic Circumpolar Current (ACC) has long been known to be concentrated in a number of fronts characterized by enhanced meridional gradients of water properties [Deacon, 1937; Nowlin and Clifford, 1982; Sievers and Nowlin, 1984; Orsi et al., 1995; Belkin and Gordon, 1996]. Following Orsi et al. , it is now traditional to identify three primary Southern Ocean fronts – the Subantarctic Front (SAF), the Polar Front (PF) and the southern ACC front (SACCF). A fourth feature, the southern boundary of the ACC (SB), marks the southern limit of the circumpolar flow. Each of the fronts is circumpolar in extent and extends from the sea surface to the seafloor. The fronts are associated with particular water mass features, allowing simple criteria based on temperature and salinity to be used to determine the location of the fronts on hydrographic sections. While different authors have used different criteria in different regions, leading to some confusion in the literature, it is a remarkable fact that simple phenomenological criteria can be reliably applied to locate individual fronts along the entire 20,000 km long path of the ACC.
 The notion that the ACC consists of three, and only three, fronts probably reflects in part the fact that most of the early studies of the current were carried out in Drake Passage, where three fronts are commonly observed. The availability of high resolution hydrography from other parts of the Southern Ocean and circumpolar coverage of sea surface height (SSH) and sea surface temperature (SST) from satellites has made it clear that the frontal structure of the ACC is more complex than once thought. For example, maps of SST gradient from high-resolution satellite data indicate multiple filaments that are sufficiently robust to appear as distinct narrow features in the four year mean [Hughes and Ash, 2001]. Some, but not all, of these filaments coincide with the fronts identified using hydrographic climatologies [e.g., Orsi et al. 1995]. Additional jets beyond the classical three fronts are also clear in surface velocity measured by drifters [Niiler et al., 2003]. The rich jet structure revealed in high-resolution numerical simulations of the Southern Ocean has been difficult to reconcile with the hydrographer's view of a small number of continuous circumpolar fronts [e.g., Maltrud et al., 1998; Hallberg and Gnanadesikan, 2006].
 In two previous studies, we have considered the frontal structure of the ACC south of Australia in some detail. Sokolov and Rintoul  used repeat occupations of a high resolution hydrographic section near 140°E (WOCE/CLIVAR SR3) to demonstrate that each of the three primary ACC fronts consisted of multiple branches. Each frontal branch coincided with enhanced meridional gradients of temperature, salinity, density and SSH and with maxima in geostrophic flow. While the frontal structure did change with time, multiple branches of the primary fronts were observed on every occupation of the section. For example, the northern and southern branches of the Polar Front were always separated by at least 500 km and were clearly distinct features.
Sokolov and Rintoul  also found that each of the frontal branches coincided with a nearly constant value of dynamic height. In other words, although the fronts changed position and intensity with time, they tended to be consistently aligned with a particular streamline. Sokolov and Rintoul [2007, hereinafter SR2007] pursued this idea further. Using a 12 year sequence of weekly SSH maps, they showed that the distribution of large SSH gradients (i.e., fronts) was strongly peaked at particular values of SSH (large was defined as gradients exceeding 0.25 m/100 km). Three distinct branches of the SAF and PF and two branches of the SACCF were identified in the sector between 100°E and 180°E. The streamline corresponding to each frontal branch was nearly constant across the sector, despite the fact that frontal dynamics (e.g., baroclinic instability) and interactions with bathymetry drive energetic variability of the fronts in this region. While the intensity of the fronts (i.e., ∇SSH) varies along the streamline, and individual frontal branches frequently merge and diverge, each front remains associated with a particular streamline across the region. SR2007 argued that since the SSH contours approximate streamlines in an equivalent barotropic flow like the ACC [Killworth and Hughes, 2002], and streamlines also correspond to a particular water column structure in such a flow [Watts et al., 2001], this implies that front definition criteria based on water mass properties will also coincide with particular values of SSH.
 Finally, SR2007 showed that the front locations inferred from SSH agree closely with positions inferred from traditional water mass criteria. This work helps to reconcile the two distinct views of the current derived from classical hydrographic studies (3 continuous circumpolar fronts, associated with water mass features) and from high resolution observations and models (multiple, discontinuous filaments). A more in depth discussion of the different perspectives on ACC jets derived from dynamical theory, hydrographic data, remote sensing and numerical modeling can be found in Section 2 of SR2007.
 The goal of the present paper is to extend the analysis of SR2007 to the entire circumpolar belt and test whether the association between fronts and streamlines holds for the entire Southern Ocean. We find that the answer is yes, allowing the circumpolar structure and distribution of the ACC fronts to be determined in detail from maps of absolute SSH. We use the large data set from Argo floats to test the inferred front locations in two ways: the velocity field derived from Argo trajectories is used to confirm that the fronts are aligned with strong current jets, and the water properties measured by the floats are used to demonstrate that the front positions we derive are consistent with traditional criteria based on water mass properties. The frontal structure determined from SSH is also shown to agree well with that inferred from high quality hydrographic sections collected during WOCE. Three regional examples are considered in detail: near the Southwest Indian Ridge (28°E to 50°E), the Kerguelen Plateau (50°E to 100°E), and in the Drake Passage and Scotia Sea (280°E to 350°E). A companion paper considers the variability of the fronts in space and time.
2. Methods and Data
 The approach used to identify fronts in SSH data is summarized briefly here; readers are referred to SR2007 for more detail. We used 15 years (1992–2007) of weekly SSH maps to calculate gradients of SSH. We subdivided the circumpolar belt into 12 sectors of 30° of longitude. Regions of high gradient were defined as those that exceeded a threshold of 0.25 m per 100 km. In each 30° sector, we then find the sea surface height contours ζi that most efficiently represent the high gradient regions and all local maxima in the SSH gradient field (i.e., local maxima are included in the fit even if they do not exceed the threshold. This ensures the weak southern fronts are included). The calculation is made by minimizing the functional
where S is the total area where ∇SSH > 0.25 m (100 km)−1 and local maxima in SSH gradient field are observed, is the area of high gradients covered by SSH contour ζi, and n is the number of frontal branches used in the optimization. As the minimization problem is a function of area, each contour needs to be associated with a width. We define the width of each SSH contour as the distance over which ∇SSH exceeds the threshold value, estimated only in those areas where the SSH gradients were associated with a single front. See SR2007 for further details.
 We used the Collecte, Localization, Satellites (CLS)/Archiving, Validation, and Interpretation of Satellite Oceanographic Data (AVISO) “Mean Sea Level Anomaly” (MSLA) maps, which are produced by mapping data from several satellite altimeters [Le Traon et al., 1998]. The mean mapping error of the SSH anomalies is less than 10% of the signal variance (<2 cm in our region) [Ducet et al., 2000], much smaller than the 20 to 60 cm change in SSH across the SO fronts. To produce maps of absolute SSH, we added the anomalies to the mean surface dynamic height (relative to 2500 dbar) derived from the WOCE global hydrographic climatology [Gouretski and Koltermann, 2004]. The front locations derived using this approach are insensitive to the choice of climatology used, although the SSH value (or “label”) associated with each front does depend on the climatology, as shown in the Appendix to SR2007.
 We found that a reference level of 2500 dbar for the mean surface dynamic height field is the most suitable for mapping ACC fronts. This level is deep enough to capture most of the baroclinic signal associated with thermohaline changes across the Southern Ocean, while leaving only a few shallow regions where the mean dynamic height is not defined. We therefore use dynamic height at the surface relative to 2500 dbar as approximate streamlines for the Southern Ocean circulation. The ACC fronts are deep reaching [Sokolov and Rintoul, 2002] and rarely cross regions shallower than 2000 m. On encountering shallow topography, the fronts tend to wrap around the bathymetric features and their paths are well approximated by the 2000 m isobath. In regions shallower than 2500 m, the mean surface dynamic heights relative to 2500 dbar were extrapolated according to the changes in density field at shallower depths. The SSH over the shallow bathymetry was used to determine whether the frontal branch passed the shallow feature to the left or the right (i.e., the front path was chosen so that higher SSH was to the left of the front facing downstream).
 We found that 12 SSH contours explain on circumpolar average more than 90% of the total area occupied by local SSH maxima and SSH gradients exceeding the threshold and used 12 contours in the subsequent analysis. Nine contours were associated with the frontal structure of the ACC, as found in SR2007. Three of the contours lie in a SSH range outside that spanning the ACC, and reflect high SSH gradients associated with subtropical western boundary currents and their extension along the northern edge of the Southern Ocean. Use of additional contours did not improve the mapping of the ACC fronts, instead introducing additional contours coincident with transient jet-like features in the subtropical and subantarctic zones.
 To validate the front positions derived using SSH, we used the independent Argo data set. The Argo data were collected and made freely available by the International Argo Project (http://www.argo.ucsd.edu, http://argo.jcommops.org) and by the Coriolis project (http://www.coriolis.eu.org). A total of 70387 profiles reaching 1500 m depth were collected in the Southern Ocean prior to March 2008. The Argo data (a combination of real-time and delayed-mode data) were quality controlled in the following way. The first stage included a rejection of profiles with large gaps between neighboring depths and clearly dubious profiles with high levels of instrumental noise and sensor drift. At the next stage, we used the vertical gradient of neutral density to identify and remove profiles with unrealistic spikes and jumps. At the final stage, instead of comparing Argo T, S profiles to a regional hydrographic climatology (e.g., as done by Wong et al. ), we used dynamic heights estimated from the Argo profiles to bin the data and rejected the outliers. As a result, the number of Argo profiles used in the study was reduced to 66385. For those profiles that passed the quality control procedure, the mean difference between the independent estimates of SSH measured by altimetry and calculated from the float data was 3.8 cm, comparable to the mean mapping error of the SSH anomalies.
 We also used the trajectories of the Argo floats to provide a qualitative test of the inferred mean front positions derived using SSH gradient maps. Argo floats drift at a parking depth of typically 1000 m, descending to 2000 m and then rising to the surface every 10 days to record a temperature and salinity profile. The uncertainty of individual subsurface drift velocity estimates usually does not exceed 0.5 cm s–1 (Park et al.  show that the errors can be further reduced with careful attention to surface drift). These uncertainties in the drift velocities derived using Argo floats are one order of magnitude smaller than the middepth currents typical for the ACC deep-reaching frontal jets [Phillips and Rintoul, 2000].
 We also used high resolution hydrographic data (e.g., WOCE sections) to demonstrate that the multiple ACC fronts inferred from SSH agree with the locations inferred using traditional water mass criteria.
3.1. Circumpolar SSH Labels of the ACC Fronts and Their Variability
SR2007 demonstrated that the Southern Ocean fronts south of Australia and New Zealand consist of multiple branches or filaments, which often merge and diverge. Similar properties of the ACC fronts are observed elsewhere in the Southern Ocean. In Figure 1, time series of ∇SSH versus latitude are shown at three different locations: in the central South Atlantic and South Pacific sectors of the Southern Ocean and in the Drake Passage. Multiple jets (maxima in ∇SSH) are observed at each location. The intensity, temporal persistence, width and spacing of the frontal branches differ between the regions, reflecting local interactions of the current with bottom topography and channel width and changes in the dynamical stability of the fronts along the circumpolar path. Nevertheless, in each region about eight or nine maxima of ∇SSH are observed at any one time within the latitude range spanning the ACC.
 Despite the fact the structure of the SSH gradient field shown in Figure 1 is complex – jets merge and split, and weaken and strengthen – the ∇SSH maxima tend to coincide with nearly constant absolute SSH values along the entire circumpolar path of the ACC. As an example, a snapshot of ∇SSH covering a quarter of the circumpolar path in the Atlantic and Indian sectors of the Southern Ocean is shown in Figure 2. The synoptic ∇SSH map is overlaid with the synoptic position of the 12 SSH contours found to coincide with the ∇SSH maxima observed in the 15 year time series of SSH maps (i.e., the values of the SSH streamlines are derived for the whole period of altimetry observations, not just fit to this particular snapshot). The map shows a large number of discontinuous filaments of elevated ∇SSH. However, the best fit SSH contours pass through the great majority of regions of elevated ∇SSH. While the patchy distribution of ∇SSH suggests little in the way of continuous fronts, the extrema of ∇SSH are consistently aligned along particular streamlines. The fact that the optimized SSH contours do a good job of tracking the high ∇SSH regions even in a particular synoptic map (see SR2007 for more examples) indicates that the association between streamlines and the fronts of the ACC is persistent and robust. Individual frontal branches merge and diverge as they traverse the region, intensifying or weakening the gradient of SSH, but each filament remains “locked” to a particular streamline. Where frontal branches merge, local maxima in ∇SSH are found at values of SSH that lie between those corresponding to the frontal branches, rather than on the best fit contour (see examples in Figure S1 in the auxiliary material). However, only about 10% of the local ∇SSH maxima are not coincident with the best fit frontal contours and are not described by our mapping. These transient filaments of the ocean currents are most common north of the ACC, where the circumpolar changes in the local optimal frontal labels are large.
 The streamlines found to coincide with maxima in ∇SSH are remarkably constant both in time and around the circumpolar path. For example, Figure 3 shows the SSH values corresponding to each front, derived from independent fits in each 30° longitude sector of the Southern Ocean. The error bars represent the temporal variability of the SSH values corresponding to each frontal branch (±1 standard deviation). The errors are typically a few cm or less, much smaller than the difference in SSH between adjacent frontal branches. The SSH values for each jet do not change significantly along the circumpolar path of the ACC, with the changes of a few cm observed between sectors generally falling within the uncertainty assessed from temporal variability within each sector. Nevertheless, there is some structure in the small deviations from the circumpolar average for each frontal branch: the SSH contours fit to the SAF and PF are consistently low in the sector downstream of the Agulhas Current (30°E to 60°E), while the PF and SACCF-N are higher than the circumpolar average value in the southeast Pacific. Large SSH gradients are observed north of the ACC, in particular near the poleward extension of the boundary currents of the subtropical gyres. (As discussed below, in some cases fronts of the ACC can merge with the western boundary current extensions, so the two regimes are not always distinct.) The variability of streamlines fit to these high gradient regions is much greater than for the ACC jets, both in time and in longitude.
 Temporal changes in the SSH values coincident with the fronts of the ACC are also small. In Figure 4 the time series of SSH labels corresponding to fronts in the western part of the Atlantic sector (30–60°W) of the Southern Ocean are shown as an example. The SSH frontal labels are most stable within the dynamic height range corresponding to the ACC, in particular for the PF (SSH between 1.0 and 1.4 m, with standard deviations of about 1.5 cm). The variability increases slightly for the southern fronts (note that the southern fronts are sometimes not visible in altimetry because they are obscured by sea ice). The temporal variability of the labels corresponding to the SAF fronts is larger than for the PF, consistent with the much greater SSH gradient across the SAF. The largest temporal variability occurs north of the ACC, where the association between elevated ∇SSH and particular streamlines is less robust. For all of the fronts, deviations of SSH from the mean value for each frontal branch are dominated by periods less than 4 months, with the standard deviation at longer time scales of order 1 cm (not shown).
Figure 5 shows, sector by sector, how well the best fit SSH contours explain the distribution of SSH gradients exceeding the threshold (∇SSH > 0.25 m/100 km) and local ∇SSH maxima. Within the SSH range spanning the ACC, generally less than 10% of the elevated ∇SSH values are missed. The three sectors downstream of the extension of the subtropical western boundary currents (30–60°E sector in the Indian Ocean, 180–210°E sector in the Pacific Ocean and 330–360°E sector in the Atlantic Ocean) show slightly larger values (<15% error). The elevated SSH gradients north of the ACC are not as strongly associated with particular streamlines, with 10–20% of SSH gradients not explained by the three contours used in this SSH range. Including additional contours in the fitting procedure reduces the fraction of unexplained SSH gradients north of the ACC, with little change to the streamlines associated with the ACC fronts.
3.2. Distribution of the Multiple Jets of the ACC
The mean location of the streamlines corresponding to the 9 frontal branches of the ACC is presented in Figure 6. The streamlines show the familiar sharp turn to the north through Drake Passage and the Scotia Sea and the overall southward trend of the ACC path from the mid-Atlantic eastward to the southeast Pacific. The path of the streamlines is often strongly deflected where the flow interacts with topography. Topographic interactions also affect the spread of the ACC fronts: the fronts tend to converge near large bathymetry (e.g., upstream of Southwest Indian Ridge, downstream of the Kerguelen Plateau, near the Southeast Indian Ridge between 145°E and 175°E, and the Eltanin and Udsinov Fracture Zones at 215°E) and become more widely spaced over the abyssal plains (e.g., the Enderby, Australian Antarctic, and Southeast Pacific Basins).
 In general, the correspondence between the mean frontal positions derived from SSH and the results of previous studies of the ACC fronts [Belkin and Gordon, 1996; Orsi et al., 1995; Moore et al., 1999] is similar to that observed south of Australia and New Zealand (not shown) and discussed in detail in SR2007. The Belkin and Gordon SAF (coincident with a temperature of 6°C at 400 m depth) usually closely follows the northern branch identified here, while the Orsi et al. SAF (defined using a criterion of 4–5°C at 400 m depth) oscillates between the middle and southern branch. In respect to the PF, the Belkin and Gordon PF (defined as the northern extent of the subsurface temperature minimum cooler than 2°C) typically lies close to the northern branch; the Orsi et al. PF (defined as the northern limit of the 2°C water in the temperature minimum and the southern limit of water warmer than 2.2°C in the temperature maximum layer) lies consistently to the south of the Belkin and Gordon position and corresponds to either the northern or middle branch of the PF; and the Moore et al. PF (defined as the more poleward zone of enhanced surface temperature gradients observed in the temperature range of the PF) is found even further south, between the middle and southern branches. The front positions derived from ∇SSH suggest the different criteria used by previous authors selected different branches of the primary fronts.
 Having identified a robust association between each frontal branch and a particular SSH contour (or approximate streamline), we can use the 15 year sequence of SSH maps to assess variability of the front locations. Figure 7 shows an example for the middle branches of the SAF and PF, the northern branch of the SACCF and the SB (similar maps for every branch of the ACC fronts are included in the auxiliary material). For each front, the probability distribution of the front position is shown along the circumpolar path. The width of the “meander envelope” of the SAF is relatively constant at about ±1 degree of latitude. The path shows a number of standing meanders as the front passes through complex topography (e.g., a northward turn between the Del Caño Rise and Crozet Plateau at 48°E, a southward deflection further east and another turn to the north to round the Kerguelen Plateau, and a turn to the north to cross the Southeast Indian Ridge near 90°E). The meander envelopes of the PF and SACCF are more variable, i.e., narrow when translations of the fronts are inhibited by topography (e.g., along the Pacific-Antarctic Ridge (165°E to 215°E) and approaching Drake Passage and the Scotia Sea) and broad over the abyssal plains.
3.3. Validation of SSH-Derived Front Locations: Comparison to WOCE Hydrography
 We next use hydrographic data to demonstrate that the ACC fronts identified in SSH correspond to the classical fronts identified using traditional water mass criteria. Full depth, high spatial resolution hydrographic sections like those occupied during the WOCE program are the “gold standard” for defining fronts, providing detailed information on property distributions and gradients (and therefore geostrophic velocity).
 First we consider the WOCE P16S section crossing the South Pacific along 210°E. Figure 8 shows the mean SSH gradient in the vicinity of the section, averaged over the two week period required to occupy this part of the section. The SSH contours corresponding to each frontal branch (with frontal SSH labels obtained from the fit to 15 years of ∇SSH maps) are overlaid on the gradient map. The synoptic ∇SSH field shows a complex pattern of meandering jets, which correspond well with the SSH contours associated with each front. The most intense SSH gradients are observed when one or more fronts merge or interact with detached eddies.
 How do the fronts inferred from the SSH field compare to those identified in the hydrographic section using traditional temperature and salinity criteria? Figure 9 shows absolute SSH interpolated to the time and location of each station. The blue bars indicate the location of each frontal branch inferred from the best fit SSH contours. Each of the blue bars coincides with enhanced meridional gradients of SSH, temperature, salinity and density. The SAF is commonly associated with the northward descent of the salinity minimum associated with Antarctic Intermediate Water (AAIW) [Whitworth and Nowlin, 1987, Orsi et al., 1995]. This definition is difficult to apply to the P16S section, where the salinity minimum descends in three steps over 5 degrees of latitude, each of which coincides with an enhanced horizontal density gradient: which of these features is “the” SAF? Our analysis suggests that each of these three steps corresponds to a branch of the SAF. Figure 8 shows that the contours fit to the ∇SSH field are associated with three distinct features between 200°E and 220°E, which merge and diverge but can be separated by as much as 700 km. Each of these frontal branches has characteristics of the SAF (e.g., coincident with a deepening of the salinity minimum and the presence of a mode water pycnostad to the north) and is therefore best interpreted as a branch of the SAF.
 Poleward of the SAF, the isopycnals continue to shoal to the south, with the strongest slopes between 56.5°S and 58.5°S. While the half-degree station spacing is barely adequate to resolve them, there is a hint of enhanced gradients within this two degree band, where the best fit SSH contours (Figures 8 and 9a) suggest the three branches of the PF and the northern branch of the SACCF are close together. These enhanced gradients agree well with the traditional water mass indicators for these fronts. For example, the PF-N is commonly associated with the northern extent of the temperature minimum cooler than 2°C near 200 m depth, placing the front near 56.5°S as also indicated by the best fit SSH contours. Sokolov and Rintoul  found that the southern branch of the PF was consistently associated with the southern limit of temperature maximum water warmer than 2.2°C south of Australia. On P16S, this places the PF-S at 58°S, in agreement with the SSH analysis. The SACCF-N was found by Sokolov and Rintoul  to correspond to the southern limit of temperature maximum water warmer than 2.0°C, which lies at 58.4°S on P16S consistent with the latitude of the SSH contour corresponding to this front.
 The synoptic map of ∇SSH (Figure 8) is entirely consistent with this interpretation of the frontal structure at P16S. For example, the branches of the SAF are separated by four degrees of latitude; the three branches of the PF appear to converge with the SACCF-N near P16S, enhancing the SSH gradient before diverging again downstream; and the section runs roughly along the SAF-S between 54 and 55°S, consistent with the broader meridional gradients across this frontal branch.
 An analysis of WOCE sections P18S, A23, and I6 shows similar agreement between the SSH and water mass indicators of the ACC fronts, as discussed in the auxiliary material.
3.4. Validation of SSH-Derived Front Locations: Comparison to Argo Profiles
 The WOCE hydrographic sections represent quasi-synoptic snapshots of the ocean structure. As a second demonstration that the front positions derived from SSH are consistent with those derived from traditional water mass criteria, we used the large number of hydrographic profiles collected by Argo floats. In each sector, we averaged the Argo profiles along neutral surfaces and along baroclinic streamlines. (In this case we used the surface dynamic height relative to 1500 m to define the streamlines, as the Argo profiles do not extend as deep as 2500 m. We repeated the front mapping optimization procedure using 0–1500 m dynamic height. The position, mapping errors and variability of the fronts are very similar to those obtained using 0–2500 m dynamic height, as expected given the equivalent barotropic nature of the ACC.) We then compared the front locations inferred from SSH to those inferred from the hydrographic data. We repeated the analysis using temperature and salinity profiles from the WOCE hydrographic climatology and found similar results, although the sections are smoother (not shown).
 An example is shown in Figure 10 for the sector between 90 and 150°E in the Australian sector. The horizontal axis is SSH (as a result, the fronts are not expected to appear as regions of enhanced horizontal gradients, as horizontal gradients of SSH are by definition reduced when plotted in this way). The front positions derived from SSH are indicated by black vertical lines. These positions agree well with the water mass criteria traditionally used to identify the fronts.
 For example, the three branches of the SAF correspond well with the descent of the salinity minimum in Figure 10. The individual branches of the SAF coincide with the 2.9, 4.7 and 7.5°C isotherm crossings of depths between 300 and 400 m, in close agreement with the criteria derived from analysis of repeat hydrographic sections on the WOCE SR3 line near 140°E [Sokolov and Rintoul, 2002]. The northern branch of the PF coincides with the northern extent of the subsurface temperature minimum cooler than 2°C, the classical definition of the front [e.g., Orsi et al., 1995]. The middle and southern branches of the PF correspond to water warmer than 2.3° and 2.2°C in the temperature maximum layer, respectively. Likewise, the two branches of the SACCF coincide with the poleward limit of water warmer than 2.0 and 1.8°C in the temperature maximum layer. Figure 10 confirms the close agreement between the SSH values associated with each front and the traditional criteria based on water properties.
 By carrying out such an analysis in each sector of the Southern Ocean, we can derive a set of circumpolar mean temperature indicators that correspond to each frontal branch identified in SSH (Table 1). The relatively small error bars confirm why the simple temperature-based criteria of the ACC fronts work well around the circumpolar belt.
Table 1. Temperature Criteria Corresponding to Each ACC Frontal Branch Identified in SSH, Derived From Argo Floats and the WOCE Global Hydrographic Climatologya
Northern, middle, and southern branches of each front, where applicable, are indicated by -N, -M, and -S, respectively. SSH frontal labels are referenced to 1500 dbar. The error bars represent circumpolar changes in temperature criteria and SSH frontal labels (±1 standard deviation).
θ at p = 400 dbar
6.06 ± 0.79
6.32 ± 0.71
1.44 ± 0.02
θ at p = 400 dbar
4.06 ± 0.35
4.06 ± 0.34
1.23 ± 0.03
θ at p = 400 dbar
2.78 ± 0.15
2.65 ± 0.21
1.03 ± 0.02
θ at p = 200 dbar
1.98 ± 0.14
1.99 ± 0.23
0.90 ± 0.01
θ in θmax
2.25 ± 0.07
2.15 ± 0.03
0.80 ± 0.01
θ in θmin
1.15 ± 0.16
1.22 ± 0.30
0.80 ± 0.01
θ in θmax
2.11 ± 0.09
2.01 ± 0.06
0.72 ± 0.01
θ in θmin
0.98 ± 0.22
0.48 ± 0.37
0.72 ± 0.01
θ in θmax
1.93 ± 0.16
1.87 ± 0.12
0.64 ± 0.02
θ in θmin
0 ± 0.23
−0.29 ± 0.33
0.64 ± 0.02
θ in θmax
1.59 ± 0.28
1.59 ± 0.19
0.57 ± 0.01
θ in θmin
−0.56 ± 0.43
−0.71 ± 0.34
0.57 ± 0.01
θ in θmin
−0.95 ± 0.29
−1.11 ± 0.09
0.51 ± 0.01
3.5. Regional Examples
 Next we consider three regional examples to demonstrate the correspondence between the classical ACC fronts defined using hydrograhic data and the multiple ACC jets identified using satellite altimetry. In the first two regions, the Southwest Indian Ridge (SWIR, 28° to 50°E) and the Kerguelen Plateau (50° to 100°E), the position of the ACC fronts has been subject to considerable debate. Recognition of the existence of multiple branches of the primary ACC fronts, along with the ability to track the position of the fronts continuously over broad areas using altimetry, helps to clarify the structure of the ACC in these regions. The third region, Drake Passage and the Scotia Sea, is the most well studied in the Southern Ocean, and the large amount of historical data there provides a good test of our approach and conclusions.
3.5.1. SAF Conundrum in the Vicinity of the Southwest Indian Ridge
Pollard and Read  and Pollard et al. [2002, 2007] studied in detail the circulation pathways of the ACC and Agulhas Return Current around the SWIR and Del Caño Rise. They observed that in regions of complex bathymetry the ACC was fragmented into many frontal jets. They concluded that in such regions it is inappropriate to think of the ACC fronts as circumpolar features since the strength and number of frontal jets vary across the region, controlled by the underlying bathymetry. They also questioned if traditional phenomenological criteria used to define the fronts are useful after all as these proxies are not everywhere associated with the major current jets. Here we reconsider the flow in the vicinity of the Del Caño Rise and show that the multiple fronts inferred from satellite altimetry are entirely consistent with in situ data from ships, drifters and Argo floats, helping to reconcile some discrepancies in earlier studies.
 The path of the SAF in this region has been a topic of debate. Some earlier studies [e.g., Belkin and Gordon, 1996] placed the SAF north of the Del Caño Rise, while others [e.g., Orsi et al., 1995] showed the SAF passing well south of the Del Caño Rise. Pollard and coauthors focused their attention on a strong flow they observed passing south of the Del Caño Rise, then turning sharply north to pass through the gap between Del Caño Rise and the Crozet Plateau, and finally turning back to the east. They called this feature the SAF. However, they noted that hydrographic, drifter and satellite data revealed two other current cores between the Agulhas Return Current and the PF. One passed to the north of the Del Caño Rise and Crozet Plateau, while the second continued flowing eastward to the south of both bathymetric features. Pollard et al.  did not connect these current features with any particular front.
 All three of these jets appear in maps of ∇SSH (Figure 11). The branch identified as the SAF by Pollard et al.  agrees closely with the middle branch of the SAF identified here, including the “S bend” made by the current as it passes through the Del Caño Rise – Crozet Plateau gap. The frontal branches identified but not named by Pollard et al.  coincide closely with the northern and southern branches of the SAF inferred by fitting SSH contours to maps of ∇SSH.
 The mean ∇SSH field in the region is dominated by the Agulhas Return Current near 40°S (Figure 11a). Further south, the largest ∇SSH is observed where the ACC interacts with the SWIR. Upstream of the ridge, the middle and southern branches of the SAF, the PF and the SACCF all converge into a single broad front at 28°E spanning more than 5 degrees of latitude between 48 and 53°S. East of the ridge, the fronts fan out across the Enderby Basin. The mean ∇SSH field is dominated by the extremely large gradients in the ARC and merged ACC fronts near 30°E, and is somewhat smoothed, so that the individual frontal branches do not in all cases stand out as distinct features in Figure 11a.
 The mean absolute drift velocities obtained from Argo floats passing through the region reveal multiple jet-like features, each of which is aligned with one of the ACC frontal branches identified using SSH (Figure 11b). In the mean SSH gradient field the northern branch of the SAF is adjacent to the Agulhas Return Current north of the Del Caño Rise and does not appear as a distinct maximum. However, the Argo float trajectories clearly indicate the existence of a separate current core (Pollard et al.  associated this deep-reaching jet with the STF). The middle and southern branches of the SAF and three branches of the PF, merged together in the west of the region, each take a separate path through the complex bathymetry: the middle branch of the SAF escapes through the narrow passage between the Del Caño Rise and the Crozet Plateau; the southern branch of the SAF continues eastward between the Crozet Plateau and the Conrad Rise; the northern PF first parallels the southern branch of the SAF and then deviates to the south toward the other PF branches once east of the Conrad Rise; and the middle and southern branches of the PF pass to the north and south of the Conrad Rise, respectively (Figure 12a).
 Water mass properties measured by the Argo floats are consistent with the circulation pattern described above (Figure 12b). The profile positions are indicated by dots, color coded according to their location relative to the ACC fronts, as inferred using the water mass criteria similar to Table 1. For example, profiles with temperatures at 300–400 m that are between the temperature associated with the middle branch of the SAF, plus or minus half the temperature difference to the neighboring frontal branch, are plotted in the same color as the SSH contour corresponding to the SAF-M. (Because the water mass properties vary slightly along the fronts, here we have used criteria derived for this sector rather than the circumpolar mean values of Table 1.) Each of the fronts identified in SSH passes through a cloud of points of the same color, illustrating the consistency between the front positions inferred from SSH and the independent water property data.
 The profiles collected by the Argo floats reflect both meandering of the fronts and eddy shedding. As a result, water masses associated with each frontal branch can appear at a distance from the mean front location indicated by the contours in Figure 12b. For example, water profiles typical of the SAF-M are sometimes observed north of the Del Cano Rise. Trajectories of the individual floats (not shown) indicate that the main jet of the middle branch of the SAF usually passes between Del Caño Rise and Crozet Plateau. When the front follows a typical path between Del Caño Rise and Crozet Plateau, it often forms a meander to the west which can break off to leave eddies north of Del Caño Rise (as also shown by Pollard et al. ). As a result, the water profiles typical of the SAF-M are well spread in this region of strong interaction of the current with complex bottom topography.
 The circulation pattern derived by Pollard and Read  and updated by Pollard et al.  is very similar to the circulation pattern inferred from satellite altimetry at the time of their cruise. The baroclinic transport estimates obtained by Pollard and Read  from hydrographic data referenced to 4000 m are also consistent with the transports carried by individual frontal branches of the ACC and mapped using SSH data. (Here we estimate transport from SSH using the relationship between the cumulative ACC volume transport and dynamic height reported by Rintoul et al. ). For example, the volume of water transported by the northern and middle branches of SAF is estimated to be about 25 Sv each, similar to the 20 Sv north of the Del Caño Rise and 30 Sv in the gap between the Del Caño Rise and Crozet Plateau reported by Pollard and Read . Their estimate of 60 Sv carried by the rest of the ACC between 50 and 55°S is also comparable to the altimeter-based transport of 60 Sv carried by the southern branch of the SAF and the PF.
 In summary, the multiple fronts identified by fitting SSH contours to maxima in the ∇SSH field agree well with the complex pattern of current jets observed in hydrographic, drifter, Argo float, and satellite data in the vicinity of the SWIR. Pollard et al.  were led by this complexity to question whether the concept of circumpolar fronts was really appropriate for the ACC. The robust association between streamlines and fronts identified in this study allows the multiple fronts in the region to be understood as the local expression of features that merge and diverge but remain coherent and identifiable throughout the circumpolar path of the ACC. When these multiple jets are observed in a single region, they are often interpreted as transient or local features. Instead we find, for example, that the same streamlines associated with the jets that pass to the north, between, and south of the Del Caño Rise and Crozet Plateau are also associated with a strong current flow in most other locations around the circumpolar belt.
3.5.2. ACC Fronts Near the Kerguelen Plateau
 The path of the ACC fronts near the Kerguelen Plateau has also been subject to considerable debate. The Kerguelen Plateau provides the largest obstruction to the circumpolar flow of the ACC and has a strong influence over the regional circulation. However, the lack of historical hydrographic observations and the complex interactions with bathymetry have meant that the circulation in the region is still poorly understood. Several recent studies have started to fill in the picture [e.g., McCartney and Donohue, 2007; Roquet et al., 2009; Park et al., 2008; Rintoul et al., 2008] but considerable uncertainty remains.
 The Kerguelen Plateau region was probably the first where the PF was shown clearly to be composed of two separate branches [Sparrow et al., 1996]. Sparrow et al.  showed the two branches of the PF are separated by as much as 8 degrees of latitude as the frontal jets navigate around the plateau. The northern branch of the PF in this region was often called the “subsurface expression” of the PF, since it coincides with a subsurface water mass feature, the northern limit of the temperature minimum (or Winter Water) layer. The southern branch of the PF was called the “surface expression” of the front, since it often corresponds to an enhanced SST gradient. However, as shown by Sokolov and Rintoul  downstream of the Kerguelen Plateau, the two “expressions” of the PF both extend throughout the water column and remain distinct features over zonal distances of 1000 s of kms.
 As summarized by Roquet et al. , previous studies of the regional circulation and water mass distribution placed the mean path of the PF at different locations. For example, most of the studies, which used full depth hydrographic sections in the region and adopted the criteria of the northern limit of the subsurface Winter Water (WW), indicated the PF passing north of the Kerguelen Plateau [e.g., Orsi et al., 1995; Belkin and Gordon, 1996]. Sparrow et al.  demonstrated the existence of a southern branch of the PF that crossed the plateau through the Fawn Trough, the gap between the northern and southern parts of the plateau. Studies that used surface indicators of the front placed the front just south of Kerguelen Island, [e.g., Moore et al., 1999; Park et al., 1998; Roquet et al., 2009], who used position of the 2°C isotherm in the WW, also placed the front just south of Kerguelen Island; several sections shown by Gamberoni et al. , on the other hand, clearly show a temperature minimum cooler than 2°C over the continental slope to the north and northwest of the island. A complex pattern of multiple fronts was found in the Enderby Basin, upstream of the Kerguelen Plateau, by Roquet et al. , using hydrographic sections collected by elephant seals.
 In Figure 13a the mean position of the ACC fronts mapped using SSH labels is overlaid on the SSH gradient field in the vicinity of the Kerguelen Plateau. Immediately west of the plateau, the three branches of the SAF are merged to form a single strong front, contributing to the largest SSH gradients in the region. The SAF passes north of the Kerguelen Plateau. Further downstream over the deep Australian-Antarctic Basin, the SAF frontal branches diverge, with the middle branch of the SAF closely aligned with the mean position of the SAF depicted in the Orsi et al.  study.
 Upstream of the plateau in the Enderby Basin, each of the PF branches is distinct and separated from neighboring fronts. On encountering the plateau, the northern branch of the PF deviates to the north, following the 2000 m isobath, and merges with the SAF at the north. (It is possible that a shallow limb of the northern branch of the PF crosses the plateau immediately south of Kerguelen Island, as shown by Park et al.  and in the surface drifter velocities shown by Park et al. , but the deeper part of the front carrying most of the transport must follow a deeper path along isobaths to the north of the island.) The remaining two branches of the PF and the northern branch of the SACCF form a single intense front and escape through the Fawn Trough, a flow called the Fawn Trough Current by Roquet et al. . The southern jet of the SACCF passes to the south of the southern Kerguelen Plateau, as shown by Orsi et al. . East of the Fawn Trough and the southern plateau, each of the fronts deviates equatorward, consistent with the presence of a boundary current along the eastern slope of the Kerguelen Plateau [e.g., Speer and Forbes, 1994; Rintoul, 2007; McCartney and Donohue, 2007].
 The absolute velocity field derived from the Argo floats is in general very consistent with the frontal pattern described above (Figure 13b) (the smaller number of Argo floats in this region causes gaps and some noise in the velocity field maps). The strongest flow in the northwest of the region is associated with the merged branches of the SAF and the extension of the ARC. The flow escaping through the Fawn Trough is also pronounced. The trajectories of the individual floats, displayed in Figure 14a, are coherent with the baroclinic stream function field, depicted by the frontal pathways. The strong current cores, revealed by the float trajectories, are reminiscent of the multiple branch frontal structure of the ACC, particularly east of the plateau, where the coverage of drifting floats passing through the region is high.
 The water properties monitored by the Argo floats are also in accordance with the circulation pattern described above (Figure 14b). The profiles measured by the floats and consistent with the water properties transported by the northern branch of the PF indicate that the front passes north of the plateau. The middle and southern jets of the PF stream through the Fawn Trough. The southern fronts of the ACC are not covered by the Argo floats, particularly west of the Kerguelen Plateau, owing to a lack of floats and heavy ice conditions in the region.
3.5.3. Path of the ACC Through Drake Passage and the Scotia Sea
 The regions of Drake Passage and the Scotia Sea are the most studied regions of the Southern Ocean, a “cornerstone” in our knowledge of the ACC frontal structure and its variability. The relatively narrow width of Drake Passage and strong bathymetric control of the ACC by complex bottom topography in the Scotia Sea ensured a large degree of consistency between different studies. The Drake Passage is about 7 degrees of latitude wide, providing a condensed view of the dramatic changes in the thermohaline structure occurring across the entire Southern Ocean. The ACC fronts in the passage are intense, closely spaced, and current cores associated with the fronts are characterized by strong velocities. In the Scotia Sea the frontal jets are more spread out than in the Drake Passage, but still form a series of intense currents navigating through the narrow deep channels formed by complex bottom topography.
 In these conditions the classical four front concept of the ACC was developed and refined, and phenomenological proxies for the fronts were established and confirmed elsewhere in the Southern Ocean. However, this picture was developed largely from historical hydrographic sections with low spatial resolution, and little information about the structure of the fronts up- and downstream of Drake Passage. More recent observations with greater spatial resolution have revealed that even in Drake Passage the main ACC fronts may be split into one or more branches, as observed for the PF in the early study of Sievers and Nowlin . When such features were observed in oceanographic sections, they were often dismissed as transient or local features.
 A synoptic ∇SSH map illustrates the complicated frontal structure in the southwest Atlantic (Figure 15). Even in the narrow Drake Passage, where the ACC fronts are closely spaced, each frontal branch on every SSH gradient map can be tracked as a separate feature despite the fact that they are constantly merging with and diverging from neighboring frontal jets, meandering and shedding eddies. In many regions within the Drake Passage and Scotia Sea two, three and even four frontal branches can merge together (e.g., south of the Burdwood Bank and just north of the Northeast Georgia Rise, Figure 16a), but as the meanders and eddies propagate, these high-gradient zones migrate, change configuration, and fade, to reappear once again in another region.
 Even in the narrowest part of the Drake Passage the intense ACC jets are not fixed and continue to meander and shed eddies. The mean drift velocities at 1000 m measured by Argo floats once again are in fairly good agreement with the ∇SSH field (Figure 16b), confirming that the bulk of the volume transport carried by the ACC in the Drake Passage is concentrated in the northern half of the passage and the ACC exits the Scotia Sea by three main routes, as well documented in the literature. The middle branch of the SAF escapes through the narrow and relatively shallow gap between the Burdwood Bank and the western extent of the North Scotia Ridge, which is the main SAF pathway indicated by Orsi et al. . The southern branch of the SAF and two branches of the PF exit further east through the deeper passages in the North Scotia Ridge, which is a path of the PF described by Moore et al.  and Orsi et al. . The southern branch of the PF and the SACCF pass through the passage between South Georgia and the South Sandwich Islands.
 Despite the fact the mean SSH gradient field is smooth across this passage, and the Argo mean drift velocities are patchy (with higher deep current velocities observed to the west of the passage close to the continental slope of South Georgia), the individual SSH gradient maps persistently indicate that there are several elongated high SHH gradient zones in the passage corresponding to at least two or three separate jets. On some occasions the southern branch of the PF does not pass south and east of South Georgia, but rather to the west closely aligned with the middle branch of the PF, as shown in Figure 15. In these cases the northern branch of the SACCF is shifted equatorwards and lies adjacent to the eastern slope of the South Georgia, replicating the path of the ACCF observed by Meredith et al. .
 The SB position depicted by Orsi et al.  is different to ours, with Orsi's front penetrating through the same passage between the South Georgia and South Sandwich Islands and looping northward as far as the Northeast Georgia Ridge. We found that persistent enhanced gradient zones, aligned with the SSH labels corresponding to the SB, do not deviate that far north and are usually adjacent to the 2000 m isobath wrapped around the South Sandwich Islands. The SB often forms eddies in this region (as discussed in a companion paper) with a typical example given in Figure 15 (feature at approximately 54°S, 329°E). Furthermore, Orsi's circumpolar map of dissolved oxygen concentrations within the core of the UCDW shows the 5 and 7 ml × l−1 contours (used as a proxy for the SB by Orsi et al. ) are far apart in this region, indicative of a less certain and probably more southern position of the front. However, we note that mapping of the SB using satellite altimetry data is less accurate compared to the other ACC fronts because of persistent sea ice cover.
 Another important issue, in comparing this study to previous work, is the fact that only two branches of the SAF penetrate the passage and carry waters from the Pacific to the Atlantic sector. Historical hydrographic data and recent T/S profiles measured by Argo floats confirm that the waters transported by the northern branch of the SAF west of the Drake Passage do not appear in the passage itself (Figure 17b). Rather, waters similar in properties to the South Pacific northern branch of the SAF reappear in the Atlantic sector much further north of the Falkland Return Current and its eastward extension, the classically defined SAF, and are transported by the Brazil Current (Figure 16a).
 The northern branch of the SAF reappears in the Atlantic sector as a continuation of the Brazil Current with the pathway reminiscent of the position of the STF in the Orsi et al.  study. However, this current is clearly a deep baroclinic current characterized by high SSH gradients and strong velocities even at a depth of 1000 m (Figure 16). Separated from the middle branch of the SAF by a long distance within the South Atlantic with typical intensive meandering and eddy formation, the SAF-N has somewhat weaker signature in the SSH gradient field here than elsewhere in the Southern Ocean. Still the current is associated with the canonical thermohaline structure associated with the SAF, with a clear signature of the AAIW deepening across the front (see discussion of WOCE section A23 in the auxiliary material).
 Finally, we note that when mapping ACC fronts in the Drake Passage region we used the 1800 m isobath to determine the path of the current around shallow bathymetry, instead of 2000 m depth applied elsewhere in the Sothern Ocean. The main reason for that is to map correctly the middle branch of the SAF escaping the Scotia Sea through the shallow passage between the Burdwood Bank and the western extent of the North Scotia Ridge. While this change in topographic control depth does not change much the position of other ACC fronts, the Falkland Current over the Falkland Plateau further north, appears to penetrate the slope even at more shallower depths as indicated by the trajectories of the individual Argo floats (Figure 17a).
 We conclude that while the new mapping using satellite altimetry reveals a richer and more complex frontal structure of the ACC, the paths of the fronts are very similar to those obtained in previous studies. The recognition of the multibranched frontal structure of the ACC more adequately explains the mesoscale SSH, SST or ocean color gradient fields observed from satellites, but, more importantly, the water properties observed along the ACC frontal branches indicate the frontal SSH labels (and corresponding thermochaline structure of the current) do not change much despite the strong interaction of the current with the bathymetry and subsequent intense mixing in such energetic regions of the Southern Ocean as the Drake Passage and Scotia Sea. The only regions where the SSH labels do change are in the SAZ, downstream of the main injection points of subtropical waters, such as the Agulhas Return Current, the Brazil Current and the Tasman outflow.
 We conclude that the ACC consists of multiple jets aligned along particular streamlines throughout the circumpolar path of the current. The SSH value (approximate streamline) associated with each frontal branch was found to be nearly constant, both in time and around the circumpolar path. The frontal branches inferred by fitting SSH contours to 15 years of SSH gradient maps agree very well with front positions inferred from synoptic sections of water mass properties. For example, the WOCE hydrographic sections typically have many more bands of enhanced gradient than expected from the classical three front view of the ACC. These features have often been dismissed as transient features of limited interest. Here we have shown that these multiple fronts are robust and persistent features of the ACC that can be identified around the entire circumpolar path of the current. High resolution hydrographic sections (e.g., WOCE lines), detailed regional studies (e.g., near Crozet, Kerguelen and in the Drake Passage), and both vertical profile and trajectory information from Argo floats were shown to be consistent with the conclusion of multiple jets coincident with streamlines.
 Recognition of the multiple jet structure of the ACC helps reconcile the disparate views of the ACC developed from traditional analysis of water mass features on hydrographic sections and from geostrophic turbulence theory and high resolution numerical simulations, as discussed in detail in SR2007. This new view of the ACC also helps reconcile discrepancies between earlier studies, where slightly different criteria led to differences in the inferred position of the fronts [e.g., Orsi et al., 1995; Belkin and Gordon, 1996], or where the complex synoptic structure led authors to question the existence of continuous fronts in the ACC [e.g., Pollard et al., 2002].
 The fact that the multiple ACC jets are associated with particular streamlines, and can therefore be traced using contours of SSH, means that the variability of the fronts in space and time can be determined with unprecedented resolution. A companion paper uses this approach to examine ACC variability in detail.