The meridional heat flux required to balance the heat lost by ocean to atmosphere at high latitudes must be accomplished by some mechanism other than mean advection and the heat flux by eddies crossing the Antarctic Circumpolar Current (ACC) may be a candidate. In this study, the positions of the main ACC fronts are determined based on 23 expendable bathythermographs (XBT) transects collected from 1994 to 2010 and are compared with those detected through satellite altimetry. Then, cold core anomalies in XBT sections are identified and altimetry is used to follow the spatial-temporal evolution of these cold, low sea level anomalies. Mean values of main parameters, such as speed (0.35 km/h), lifetime (79 weeks), and diameter (105 km), are estimated. Moreover, estimations of rotational speed (0.9–76.8 cm/s), ocean surface layer heat content along temperature sections and eddy available heat anomaly (mean value −9.74 × 109 Jm−2) give a wider description of the detected eddies. In our study area, the spawning of eddies is found to occur downstream of the Southeast Indian Ridge and in correspondence of the polar front (PF) with regard to the ACC frontal structure. The contribution of eddies to the global heat budget is not only linked to their ability to cross the ACC fronts but also to the capacity of keeping partially unaltered the properties of water inside them. Analysis of the relation between the translation and rotational speeds shows that a typical eddy may effectively be a significant part (0.8%) of the net meridional heat transport across the PF with a mean heat content/anomaly of −7.65 × 1019 J.
 The Southern Ocean is one of the focal points of the global overturning circulation thanks to the presence of dense water formation sites and the existence of the Antarctic Circumpolar Current (ACC). The absence of continental barriers in the latitude band of Drake Passage permits the ACC to exist and to act as the main conduit for oceanic exchanges between the three major ocean basins. Moreover, with its 26,000 km long continuous path around Antarctica, the ACC allows a global scale overturning circulation to exist [e.g., Speer et al., 2000; Rintoul et al., 2001; Morrow et al., 2004; Naveira Garabato et al., 2011; Marshall and Speer, 2012].
 The ACC flow is geostrophically supported by sloping of isopycnals and is known to be concentrated in several jets associated with regions of strong horizontal gradients in water mass properties and sea surface height [Deacon, 1937; Orsi et al., 1995; Belkin and Gordon, 1996; Sokolov and Rintoul, 2007]. It is common to distinguish, from north to south, three main circumpolar Southern Ocean fronts extending from the sea surface to the deep ocean: the subantarctic front (SAF), the polar front (PF), and the southern ACC front (sACCf), while a fourth feature, the Southern boundary of the ACC (Sbdy), marks the southern limit of the current following [Orsi et al., 1995; Sokolov and Rintoul, 2007].
 The lack of continental barriers along the ACC path also has implications for the meridional circulation. The absence of land to support zonal pressure gradients implies that there can be no mean meridional geostrophic flow across the channel at depths shallower than the sea floor topography [Morrow et al., 2004]. In this situation, the meridional heat flux required to balance the approximately 0.3 × 1015 W lost to the atmosphere in the Antarctic Zone, south of the PF [Keffer and Holloway, 1988], must be provided through some mechanism other than mean advection [Morrow et al., 2004]. Therefore, it has been demonstrated [de Szoeke and Levine, 1981] and confirmed by different studies based on current-meter measurements [Nowlin et al., 1985], altimetry [Keffer and Holloway, 1988], and high-resolution ocean general circulation models [e.g., Holland and Lin, 1975a, 1975b; Holland, 1978; McWilliams et al., 1978; Cox, 1985; Jayne and Marotzke, 2002] that the movement of mesoscale eddies across the ACC fronts is the most likely oceanographic mechanism capable of achieving the heat transport required to balance the loss to the atmosphere south of the PF. In particular, cold-core eddies transport cool, low salinity polar water, across the PF and the SAF into the Subantarctic Zone where mode and intermediate waters form, contributing to cool and fresh surrounding subantarctic waters [Morrow et al., 2004].
 In spite of this crucial role in the global system, the mesoscale variability of ACC and its fronts does not occur uniformly along the length of the current [e.g., Lutjeharms and Baker, 1980; Gille, 2003; Morrow et al., 2004]. Eddy shedding is strongly dependent on underlying large-scale bottom bathymetry, and regions of the most intensive activity are observed downstream of large-scale topographic features for all ACC fronts [Sokolov and Rintoul, 2009a].
 The meridional heat transport is strictly connected to the strength of the water mass anomalies associated with an eddy as well as eddy size and frequency [Swart et al., 2008]. Moreover, the heat transport is associated with the effective possibility that these eddies have to cross the ACC fronts and interact with them in the so-called “heat scavenging” behavior [Lutjeharms, 1988; Koshlyakov et al., 1985]. Sokolov and Rintoul [2009b] identified a limited number of areas characterized by high mesoscale variability, and they associated with each of them the variability of particular ACC fronts and the related cross-frontal water exchange. The Southern Ocean sector extending from about 90°E to 180° is characterized by strong meandering of ACC fronts due to topographic features and is involved in eddy spawning from the ACC as well as the Drake Passage, the Greenwich Meridian area, and the southeastern corner of the Pacific sector. In particular, the formation of eddies linked to PF is observed downstream of the Southeast Indian Ridge and of the Macquarie Ridge between 150°E and 170°E, while eddy-induced cross-frontal exchange linked to the SAF is strong downstream of the SAF interaction with the Campbell Plateau south of New Zealand [Sokolov and Rintoul, 2009b]. This sector of the Southern Ocean has been investigated since 1994 in the frameworks of the Climatic Long-term Interaction for the Mass balance in Antarctica (CLIMA) and the Southern Ocean Chokepoints Italian Contribution (SOChIC) projects of the Italian National Research Program in Antarctica (PNRA). During these projects, in situ temperature sections of the surface layer of the Southern Ocean along the track New Zealand-Ross Sea (Figure 1) have been realized almost every summer season by means of expendable bathythermographs (XBT).
 In this paper, we use the XBT data set and altimetry data provided by Archiving Validation and Interpretation of Satellite Oceanography (AVISO) to describe and compare the frontal structure of ACC from in situ and satellite data, also updating the results reported by Russo et al.  and Budillon and Rintoul  on the position of ACC fronts from in situ data.
 Moreover, we also focus on cold-core anomalies that were recognized in the meridional distribution of temperature from XBT data. Anomalies in temperature sections through the ACC have been classified as eddies or ACC meanders on the basis of altimetry data. The cold-core eddies identified in the XBT sections are studied and the spatial-temporal evolution, from their generation to decay, is described. Furthermore, an estimation of the heat content that is transported across the ACC fronts by these cold, low sea-level anomalies is provided.
2.Data and Methods
 The XBT data originate from 23 transects between New Zealand and the Ross Sea realized by the R/V “Italica” during the 1994–2010 austral summers, mainly in January and February, in the framework of the CLIMA and SOChIC projects.
 Data are usually collected using Sippican T7 probes providing temperature profiles with a vertical resolution of 65 cm and a maximum nominal depth of 760 m, even if ship speed lower than 15 kn allows the probe to reach about 900 m depth. The majority of transects were completed in 5 days to offer a synoptic picture of the thermal structure of the upper Southern Ocean. A regular 20 km sampling rate was adopted with increased sampling frequency over the frontal regions of the ACC. Successively, in order to remove expected errors, all temperature profiles have been extensively analyzed through quality control procedures as comparisons between adjacent profiles and spike editing.
 The temperature sections along latitude obtained from XBT data are then studied to identify the location of the main ACC fronts and to compare their position with the position obtained from satellite altimetry data.
 The temperature-based criteria used for positioning the ACC fronts from in situ data (Table 1) follow Budillon and Rintoul , summarizing the usual hydrographic definition [Botnikov,1963; Belkin, 1990; Orsi et al., 1995] and the specific indication [Rintoul et al., 1997] for the position of the northern (NSAF) and southern (SSAF) branches of the SAF south of New Zealand.
Table 1. Antarctic Circumpolar Current's Front Names, Acronyms, and Temperature-Based Criteria Used for Their Location
Temperature decrease in the range 4°C–7°C at 300 m toward south
 The Southern Boundary (Sbdy) of the ACC, usually described as the maximum southern extent of vertical maximum of T > 1.5°C at about 200 m [Orsi et al., 1995], is not described in this work as its position is coincident with the sACCf position in the majority of the available temperature sections. Moreover, the extremely southern position occupied by the sACCf in the study area makes this front not always detectable in the satellite altimetry data set.
 After the location of ACC fronts, XBT data are used to determine the presence of anomalies in the meridional distribution of temperature (i.e., isolated cold or warm water masses) that could be associated to ACC meandering or to the presence of mesoscale eddies separated from the main flow. Unfortunately, basing exclusively on data from vertical sections of temperature along latitude, eddies and meanders generate similar anomalies in the temperature distribution and discrimination between them is not possible.
 Two main satellite altimetry products, provided by CLS (Collecte Localisation Satellites)/AVISO through the Ssalto/Duacs system, the sea-level anomaly (SLA) and the geostrophic velocity maps are used to distinguish between temperature anomalies that are associated with ACC meanders and those associated with eddies. In this case, satellite data are also used to follow their path and evolution from generation to decay.
 Weekly maps of SLA, simultaneous with the CLIMA/SOChIC XBT cruises, extending from 160°E and 160°W and between 45°S and 75°S are used to discriminate between meanders and eddies, as detached cold (warm)-core eddies constitute isolated lower (higher) SLAs in the satellite data map. The anomalies characterized by clearly defined boundaries, in both longitude and latitude, which are separated by the nearest fronts, have been first associated with the existence of cold- or warm-core eddies detached by the ACC. The second step in the eddy identification procedure adopted in this work is a reference level Δ = 20 cm, where Δ is the difference between the value of SLA in the center of the eddy and the surrounding area.
 The definition of the boundary of eddies and of the reference level Δ is a crucial point, as it allows us to discriminate eddies from the surrounding waters and also to measure their extent. The value of Δ has been chosen according to previous studies [Joyce et al., 1981; Sprintall, 2003] that determined a sea surface signature of 20 cm in SLA to be associated with cold- or warm-core eddies in the ACC latitude band. After both in situ and satellite data identification procedures, eddies are tracked on the SLA maps from spawning to decay, excluding all those eventually reabsorbed into the ACC. Their principal characteristics are studied merging information from in situ and satellite data. Successively, absolute geostrophic velocity maps by AVISO, based on the CNES-CLS09 mean dynamic topography and referenced to the geoid, are used to quantify the rotational speed of the eddies as sea surface inclination measured by satellite altimeter is directly related to the geostrophic velocity at the surface (http://www.aviso.oceanobs.com). Both altimetric data sets derive from the merging process of data from all altimeter missions (Jason-1&2, T/P, Envisat, Geosat Follow On, European Remote-Sensing Satellites-1 & 2, and Geosat) performed by Ssalto/Duacs to provide a consistent and homogeneous catalogue of satellite data on a regular 1/3° Mercator grid referenced to a 7-year (1993–1999) mean. Maps of SLA and absolute geostrophic velocity are provided by AVISO in “near real time” or “delayed time mode” and, in this second category, in “up-to-date” (UPD) or “referenced” (REF) mode.
 According to literature [Sokolov and Rintoul, 2007, 2009a, 2009b; Swart et al., 2008] and to the small-scale structure that characterizes the ACC, we chose to employ the UPD data set that makes use of up to four satellites at a given time.
 The varying number of used satellites implies that the quality of the data set in time is not homogeneous. Anyway for long periods, they provide an improved resolution and data accuracy compared with the classical REF dataset that is homogeneous in time but based on two satellites with the same groundtrack.
 The necessary corrections needed to minimize errors in the altimetry (tides, atmospheric interference, satellite orbital error, etc.) are applied to AVISO data prior to publication on the Live Access Server, and more details can be found in various papers [Le Traon et al., 1998; Le Traon and Ogor, 1998; Lagerloef et al., 1999; Ducet et al., 2000].
 In this paper, we describe the main characteristics of eddies detaching from the ACC but great information about them is held in the relative position of eddies, from their generation to decay, and ACC fronts. In order to study the relations between eddies and ACC fronts, we identified the position of the main fronts from satellite data on weekly basis and compared the results with the in situ position of the fronts obtained through XBT data.
 Positions of the fronts from satellite altimetry are determined on the basis of nearly constant sea surface height (SSH) values associated to each front of the ACC [Sokolov and Rintoul, 2002, 2007, 2009a, 2009b]. SSH values used in this work to identify the position of the different fronts and branches are reported in Table 2. Weekly absolute-SSH maps are obtained by adding SLA values to the mean surface dynamic height relative to 2500 dbar derived from the world ocean circulation experiment (WOCE) global hydrographic climatology [Gouretski and Koltermann, 2004]. The locations of fronts derived through this approach are insensitive to the choice of the climatology as shown in the appendix of the study by Sokolov and Rintoul . Reference level of 2500 dbar was chosen to capture most of the baroclinic signal associated with thermohaline changes across the Southern Ocean [Sokolov and Rintoul, 2009a]. Moreover, in the New Zealand sector of the Southern Ocean, the 2500 dbar reference level allows us to obtain dynamic height values over all the study area leaving only the Campbell Plateau region out of the estimation. Anyway, as the ACC fronts are deep reaching features they tend to wrap around bathymetries shallower than 2000 m [Sokolov and Rintoul, 2009a] and their positions are not subject to variations when estimation of dynamic height values at depth shallower than 2500 m is performed.
Table 2. Values of SSH (m) Relative to 2500 dbar Associated With the Main Fronts of the ACC and Their Branches Adapted From Sokolov and Rintoul [2009b]a
SSH (m) Values (Relative to 2500 dbar)
Errors represent standard deviations of front SSH isolines as in the study by Sokolov and Rintoul [2009b].
1.97 ± 0.02
1.47 ± 0.02
PF Northern branch
1.29 ± 0.01
PF Southern branch
1.04 ± 0.02
sACCf northern branch
0.94 ± 0.02
sACCf southern branch
0.84 ± 0.01
Southern ACC boundary
0.75 ± 0.01
 The ability to cross the ACC fronts is a focal point in determining how eddies can contribute to the heat transport required to balance the heat loss to the atmosphere south of the PF [Swart et al., 2008]. For this reason, we estimated how eddies influence the heat content of the surface layer of the ocean, calculating the along-track heat content from XBT data in the layer 0/500 m depth (equation (1)).
 The upper ocean heat content is derived as:
where ρ = 1027 kg m−3 is taken as a constant mean water density for the region, the specific heat cp is taken as 3930 J kg−1 K−1 and T(z) is the XBT temperature in degrees Kelvin at every depth between 0 and 500 m.
 Successively, we also calculated the available heat anomaly (AHA) associated with each eddy and an indicative amount of Joules associated to a typical eddy to better estimate the contribution given by eddies to the global heat transport.
3.Results and Discussion
3.1.Fronts From In Situ and Satellite Data
 Through the analysis of 23 temperature sections from 1994 to 2010 between New Zealand and Antarctica, we identified the position of the main ACC fronts. The temperature-based criteria, previously described, allowed us to determine the position of two branches of the SAF, the PF, and the SACCf shown in Figures 2a–2c, respectively. Assuming that the ship track can be considered constant during the 23 CLIMA/SOChIC cruises, we studied the variation of fronts positions during the Antarctic summer from 1994 to 2010. This assumption is acceptable for all the cruises except for December 2001. During this particular cruise, the ship followed a more westerly route reaching about 160°E and passing on the western edge of the Campbell Plateau. In this area, the ACC fronts occupy lower latitudes being not yet pushed south by the presence of the plateau [Orsi et al., 1995; Sokolov and Rintoul, 2009a]. For this reason, data from the December 2001 cruise are not discussed and included in the calculation of mean positions and standard deviations of the fronts even if fronts locations are represented in Figures 2 and 8.
 During the 23 cruises, the positions of the NSAF and SSAF have a mean value of 53.4°S and 58.6°S with a standard deviation of 1.4° and 1°, respectively. Mean positions agree with the values present in literature, even if a northern shift is evident in the period 2003–2006 for NSAF and 2004–2006 for SSAF. This shift justifies the increase in standard deviations with respect to those calculated on the same data set for the period 1994–2001 by Budillon and Rintoul .
 The most northern position reached by the NSAF is detected at 50.3°S over the Campbell Plateau, during the January 2006 cruise (Figure 3), while the most southern position is reached by the NSAF in January 2003 at 56.7°S.
 The most northern position reached by the NSAF is the only case in our data set of front identification at depth shallower than 2000 m. This is the result of a particular large extension of the thermal gradient between 7°C and 4°C at 300 m depth, leading the zone between the NSAF and the SSAF to occupy the entire latitude band between 50.3°S and 58°S. In any case, the ACC fronts are known to bypass depths shallower than 2000 m [Sokolov and Rintoul, 2009a], and the identification of the NSAF at depth shallower than the Campbell Plateau is in complete discordance with actual knowledge about ACC fronts. For these reasons, the January 2006 NSAF in situ position is considered an anomaly in our data set; even if it is used to calculate the mean NSAF position and standard deviation on the basis of water properties criteria, we would tend to consider the NSAF position for this cruise coincident with the SSAF.
 The SSAF most northern and southern positions are detected in November 1994 (56.9°S) and December 2003 (60.5°S), respectively. The difference in dates between the northern and southern maximum extent NSAF and SSAF underlines the fact that the meridional shift of the two branches is not synchronous. In fact, the distance between them is not constant, ranging between 1.1° (January 2003) and 7.8° (December 1994) with probably no particular pattern in the variation of distance with time.
 The PF positions are characterized by the minimum (northest position) value of 60.2°S reached during November 1994. The mean position of the PF is slightly southern (61.7°S) than previous estimation (61.5°S) from Budillon and Rintoul , while the standard deviation (0.9°) remains unchanged. Also, the SACCf mean latitude (63.7°S) calculated on the 23 XBT transects is characterized by a slightly southern value than in previous studies, with a standard deviation of 0.6°.
 The positions of ACC fronts from XBT data have then been compared with the along-cruise track locations derived using altimetry. An example of this process is given in Figure 4 for the January 2010 cruise. Color scale indicates the SSH (cm), while continuous colored lines indicate the ACC fronts or their northern and southern branches. The SAF (black), the PF (gray), and the sACCf (red) cross the ship track at different latitudes while the Sbdy (yellow) is not defined in the area occupied by the ship route, then not intersecting its track. Red dots indicate XBT casts while along-track yellow stars indicate the position of the front from in situ data. The edge of Campbell Plateau is also indicated by the thin gray line associated with the 2000 m isobath.
 As evident from the given example, every single SSH isoline may episodically cross the ship track, and the XBT section, in more than one point. This could be caused by front and isoline meandering, presence of eddies or by the spatial resolution of satellite data and the connected interpolating procedures. This is evident, for example, for the northern and southern branches of the PF (gray lines) in Figure 4. These isolines intersect several times in the right part of Figure 4, probably due to meandering processes, while at about 60°S and 165°E, they underline the presence of an isolated low SLA.
 In order to correctly compare the satellite detection of fronts through the 23 cruises, we decided to identify, for each cruise, the along-track most northern and southern positions occupied by the different SSH values belonging to each front. In the example shown in Figure 4, this procedure gives for the SAF a latitude band ranging from 53.5°S to 60.2°S, while the PF ranges between 60.5°S and 63.5°S and the sACCF occupies the latitudes between 63.5°S and 66.7°S. All the latitude bands associated with fronts are indicated by black arrows and labels in Figure 4. These latitude intervals, for every CLIMA/SOChIC cruise, are shown for each front through the vertical gray bars in Figure 2.
 The comparison between in situ and satellite detection of ACC fronts positions shows a good correspondence between the two identification methods, even if each front seems to be characterized by a different kind of result. The NSAF seems to be quite well represented by the upper limit of the SAF altimetric range. No more than eight in situ values on 23, including the already discussed anomalous position of January 2006, differ by more than 1° from the satellite SAF latitude band and all of them are located north of it.
 Better results are given by the SSAF satellite positions with only one in situ position falling south of the satellite latitude band (February 2001) while all the others are very well represented by the lower limit of the SAF satellite band.
 The PF latitude band is characterized by smaller amplitude with respect to the SAF, with values ranging most of the times between 60°S and 64°S. Figure 2b also shows the presence of isolated latitude band associated to the PF-SSH values not linked to any in situ front position. These isolated signals, often characterized by a very small latitudinal extension, can be due to the presence of water masses with the same physical properties of PF water in an unusual northern position.
 The explanation for the northern migration of water with PF characteristics can be linked to the presence of eddies detaching from the PF and moving northward to the SAF.
 The good correspondence found between in situ and satellite data for the NSAF, the SSAF, and the PF is not confirmed for the positions associated with the sACCf. For this front, 19 in situ positions on 23 are located northern than the latitude band associated with sACCf through altimetry data, and distances between the in situ and satellite-based sACCf positions are not constant.
3.2.ACC Eddies From In Situ and Satellite Data
 ACC fronts and associated water masses distribution in the Southern Ocean usually generate a series of different gradients in surface height [Sokolov and Rintoul, 2007] and temperature [Deacon, 1937; Orsi et al., 1995; Belkin and Gordon, 1996] decreasing from north to south. Nevertheless, we identified seven sections presenting a standard temperature distribution and 16 sections characterized by anomalies in the meridional distribution of temperature among the 23 available XBT transects collected during the CLIMA and SOChIC projects.
 On the basis of thermal characteristics of sea water and on the existent knowledge about ACC fronts, we defined as “standard” those temperature distributions characterized by temperature mean values at given depths decreasing southward and by rapid changes in water properties in correspondence to the ACC fronts. An example is provided in Figure 5a. From north to south, cruise data show the usual distribution of ACC fronts, from the NSAF to the sACCf, while the thermal properties of water between the fronts do not show any other variation than the expected temperature decrease toward south.
 On the other hand, what we defined as “anomalous” temperature distribution is the presence, in the inter frontal zones, of isolated water masses characterized by significant temperature differences with the surrounding waters, usually extending from the surface layer to the maximum available depth. Temperature values of this isolated water masses can be higher or lower than the surrounding and generate additional temperature gradients of different signs along latitude. These additional gradients can have implication in the positioning of ACC fronts on the basis of thermal criteria. The temperature section derived from February 2007 XBT data (Figure 5b) shows the presence of a typical cold-core anomaly in the SAF zone generating a temperature difference of about 3.5°C between the water inside the anomaly and the surrounding at 300 m depth. Looking at the temperature distribution along latitude, this anomaly duplicates the latitudinal temperature gradient associated to the NSAF and generates a further gradient with anomalous increase in water temperature going south in correspondence to the southern edge of the anomaly.
 Mainly, the anomalies can be caused by two processes: the meandering of ACC fronts or the detachment of mesoscale eddies from them [Budillon and Rintoul., 2003]. As described in section 2, we used satellite altimeter data to discriminate between the different kinds of anomalies and to eventually identify the boundary of eddies applying objective criteria for their definition. In this way, we identified nine anomalies associated with sea level drops larger than the selected threshold. Eddies were identified during the cruises of November 1994, February 1995, January 1996, February 1996, January 2000, January 2001, January 2003, February 2007, and January 2008, but we decided to exclude the cruise of January 1996 from our estimations of the mean parameters of eddy because the same temperature anomaly is better sampled and more evident during the February 1996 cruise.
 Figures 6a–6f illustrate the spawning and evolution of the cold-core eddy identified during the February 2007 cruise, also showing the detachment of the eddy from the PF and its migration through the Polar Frontal Zone across the SSAF and the NSAF. Figure 6c shows the SLA map coincident with the in situ XBT section and the presence of the low sea-level signal associated to the cold-core anomaly described in the temperature data (Figure 5b). Figure 6 also shows the positions of main ACC fronts as depicted on the basis of the SSH data. The presence of various eddies characterized by low SLA values located south of the SSAF at about 165°E, south of the Campbell Plateau, and north of the NSAF from 180°W eastward is evident, as well as the spawning of eddies from the NSAF (yellow arrows). The edges of some eddies are characterized by SSH values coincident with those normally associated with the NSAF and SSAF and consequently duplicate the SSH signal associated to the fronts. An example is given in Figure 6c where the black closed isolines between the two branches of the SAF and north of the NSAF reproduce the altimetric footprint of the SAF fronts far from their usual position. Panel 6d focuses on the SLA values of the area around the cold-core anomaly identified at about 55.5°S in the XBT section (Figure 5b) and whose SLA values reach about −80 cm. SSH isolines associated with ACC fronts are not showed in panel 6d to make more evident the large drop in sea level associated to the central area of the eddy.
 The ratio between the number of anomalies successfully associated with eddies and the number of CLIMA/SOChIC temperature sections confirms the importance of the study area for monitoring the properties, path, and decay of eddies detaching from the ACC, especially in the case of cold-core eddies. Even if some warm-core anomalies are present in the XBT temperature sections and in the SLA maps of the study area, we did not identify, through satellite data, any warm-core isolated eddy crossing the XBT track in coincidence with the cruise. The warm-water anomalies detected in the temperature sections seem instead to be the result of the isolation of water masses due to the presence of cold anomalies north of them. For example, February 2007 section (Figure 5b) shows the already described cold-water anomaly isolating a warm-water mass between the SSAF and the NSAF. This asymmetry in eddy generation, with a preference for the generation and detection of cold-core eddies in our study area, is in agreement with previous studies mainly based on altimetric data [Sokolov and Rintoul, 2009b], even if a more recent publication [Chelton et al., 2011] shows no preference for the generation of cyclonic (cold) or anticyclonic (warm) eddies in the Southern Ocean. This difference could be linked to the limited number of eddies identified in the CLIMA and SOChIC XBT sections that are the object of our study.
 Merging information from in situ and satellite data, we described the mean characteristics of eddies crossing the XBT sections, such as their dimensions, speed of rotation, and heat content (Table 3).
Table 3. Main Parameters of the Eight Cold-Core Eddies Characterized by a Sea-Level Drops Larger Than 20 cm Detected During the CLIMA/SOChIC Cruises
Diameter From Altimetry Data (km)
Δ SLA (cm)
Rotational Mean Speed Along Eddy Section (U) (cm/s)
 The mean diameter of the cold core has been estimated to be 105 km with a minimum extent of 60 km registered in January 2008 and a maximum of 190 km associated with the February 2007 cruise. The absolute minimum value of SLA associated with the center of eddies crossing the ship track is −80 cm and the maximum difference between the center of an eddy and its edge is 53.5 cm.
 The use of satellite altimetry data also offered the opportunity to estimate the rotational speed of eddies while crossing the XBT section tracks. Rotational speed on the edge of the eddies is obtained through geostrophic velocities and ranges between 0.9 cm/s and 76.8 cm/s, while in Table 3, a mean value along the section of each eddy is reported.
3.3.Eddies Tracking and Heat Transport
 As described in section 1, eddies are supposed to contribute in great measure to global heat balance. For this reason, we estimated the heat content of the surface layer of the ocean (0–500 m) along the CLIMA/SOChIC tracks using in situ XBT data. Figure 7 shows the upper ocean heat content for January 2005 and February 2007 cruises. As for the temperature sections of the same cruises (Figures 5), the two heat content profiles along latitude will represent the heat content values in absence (Figure 7a) or presence (Figure 7b) of eddies. Black thin line and dots indicate heat content and launch position of each XBT, while red thick line represents heat content values when filtered with a running mean 3 filter to slightly smooth differences between adjacent profiles. The surface-layer heat content along latitude for the January 2005 cruise (Figure 7a) shows a regular decrease of the heat value from 8.78 × 1011 Jm−2 to 8.50 × 1011 Jm−2 going south, with marked variations in correspondence of the ACC fronts, whose positions, determined from XBT data, are indicated by vertical black lines. The heat content values along the February 2007 track (Figure 7b) do not show the same pattern because of the presence of some additional variations to the expected north-south decrease. The heat content reduction from 9.09 × 1011 Jm−2 at 48°S to 8.85 × 1011 Jm−2 at 72°S is characterized by the presence of a large latitude interval of small heat content values between 54.75°S and 56.25°S that coincides with the presence of a cold-core ring evident in both XBT (Figure 5b) and satellite data (Figure 6c). In the ring area, the heat content decreases from 9.07 × 1011 Jm−2 at 54.75°S to 8.93 × 1011 Jm−2 at 55.49°S, then it slightly increases up to 9.04 × 1011 Jm−2 at 56.25°S. In this area, the heat content is also influenced by the presence of the NSAF at 54.7°S, but the increase between 55.49°S and 56.25°S is a clear indication that the heat content variation must be mainly associated to eddy presence.
 The estimates of the available heat content in the upper layer (0–500 m) for all the anomalies identified as mesoscale eddies in the XBT cruises show clear signatures (heat content decrease) of the cold-core rings presence with heat content dropping of 0.8 × 1011 Jm−2 (heat content maximum decrease) in correspondence with the cold core rings.
 We also calculated eddy available heat anomalies relative to a temperature profile collected during the same cruise but outside the eddy in order to better represent the influence of eddies on the external environment.
 AHA calculation is based on the formula:
where ρ is a constant mean sea water density for the region, cp is the specific heat, h is the thickness of the layer influenced by eddy presence, and Te and Tr are the temperature profiles inside and outside the eddy, respectively. The vertical temperature profile outside the eddy is represented by the first available XBT profile near 48°S. We chose the 48°S profile as reference to provide a unique term of comparison for the entire period from 1994 to 2010 as temperature profiles at this latitude are available during all the cruises and do not show any particular variation in time. Moreover, due to the presence of the Campbell Plateau, no eddy reaches such northern latitude and then temperature profiles can always be considered free from eddy influences. Equation (2) is a variation of the method used by Joyce et al.  and Morrow et al.  to estimate the available heat content inside cold-core eddies along potential density layers. In this work, we will exclusively use temperature data from XBT to calculate AHA from the surface to 760 m depth for each of the eight eddies with SLA larger than −20 cm. Reference depth of 760 was chosen, because it is the maximum available depth from XBT data and also on the basis of the paper by Morrow et al. , which indicates 800–1000 m depth as the limit between the layers influenced by the input of fresh, cold water connected to eddies and the underlying waters that are simply influenced by heaving of the isopycnals. The calculated AHA ranges between −4.56 × 109Jm−2 (January 2000) and −1.45 × 1010Jm−2 (February 2007) with a mean value of −9.74 × 109Jm−2. Successively, a second series of reference temperature profiles for AHA estimation was chosen on the basis of the specific radius of each eddy to test these values. The first temperature profile outside the northern edge of each eddy was used as a reference and the derived heat anomaly fields only showed minor variations compared with previous estimations. Both series of AHA show high values of standard deviations (3.49 × 109Jm−2 with reference profile at 48°S), and large variability is observed between different years and eddies. We think that this could be largely due to sampling, since some of XBT casts crossed only the edge of the rings, missing their colder and lower sea-level core.
 In any case, we calculated the impact of the standard deviation value on each single eddy to give indications of how this variability affects the AHA estimations. These values give a measure of accuracy and range between 3.5 × 109Jm−2 and 5.2 × 109Jm−2, reflecting the discussed large variability (about 28%) associated to the mean estimated value.
 In order to evaluate the total heat anomaly of the eddies, we assumed cold-core rings to have a cylindrical shape with a mean radius of 50 km and a limited vertical extent of 760 m. Integrating the heat content anomalies of the rings over the radial surface πr2, we estimated that the “mean” heat content anomaly of a ring is −7.65 × 1019 J. The reliability of this value can be reduced by the impact of several assumptions (i.e., eddy shape and vertical extent) and errors (i.e., temperature and radius measurements). However, some tests were performed changing the mean radius of the eddies, probably the most crucial parameter, of about 10% of its value. Results show a percentage variability in the heat estimation (20%) lower than the standard deviation estimated on AHA.
 The integrated AHA values we obtained are greater, but of the same order of magnitude, than previous estimation by Swart et al. , Morrow et al. , and Joyce et al. , although their calculations were performed in different ways. Moreover, heat content of the eddies can be compared with the total heat that is lost south of the Antarctic PF to the atmosphere. Considering 0.3 × 1015 W [Keffer and Holloway, 1988] as a typical required southward flux, then 9.46 × 1021J needs to be transferred across the Antarctic PF per year [Swart et al., 2008] and the contribution for a mean eddy is the 0.8% of the required annual poleward flux. The importance of this contribution is obviously linked to the number of eddies spawning in our study area in correspondence of the PF.
 Even if an estimation of the total number of cold core eddies generated along the ACC path is difficult, several papers based on altimetry data [Sokolov and Rintoul, 2009b; Chelton et al., 2011] have shown a large number of eddies near the southern edge of the Campbell Plateau and around 180° longitude. About 1 or 2 eddies per year are found south of New Zealand by Chelton et al.  over a 16 year period, while the frequency of occurrence of eddy positions next to the PF, the NSAF, and the sACCf in the same area is about 5% of the 15 years weekly altimetry data set analyzed by Sokolov and Rintoul [2009b]. According to these values, eddy generation seems to be not a rare event and the importance of the heat content transferred by a typical eddy becomes even more relevant. Moreover, the role of eddies is more evident when considering that totally enclosed synoptic eddies such as the one we observed, represent the so called divergent component of the total eddy heat flux [Jayne and Morotzke, 2002] and that they are an effective mechanism for transporting heat poleward, different from meandering processes that, transport equal amounts of heat in northward and southward directions, do not contribute to the poleward heat flux [Swart et al., 2008].
 Once that the cold anomalies in temperature sections have been associated to low sea-level cold-core eddies and after that all their mean properties at the moment of crossing the ship track have been described through the merging of in situ and satellite data, we used the weekly SLA maps to estimate their lifetime, path, and speed. This was done tracking eddies from their spawning from the ACC to the moment they crossed the XBT section and then following their path until dissipation.
 The technique used for tracking eddies is based on the visual check of the weekly SLA data starting from the maps of SLA coincident with the XBT cruises. The origin of each eddy was located in the place where altimetry isoline of the eddy was closed first, while eddy decay has been located where eddy was detectable for the last time before the loss of its signal. As sea-level values associated to eddies are not constant during their life, the tracking method was independent from the leveling of SLA that eddies experience during their life due to the expected loss of properties. The resulting mean lifetime of the rings is 79 weeks with a minimum of 46 weeks for the eddy of November 1994 and a maximum of 105 weeks for the eddy of February 1996, while translational speed has a mean value of 0.35 km/h with a maximum of 1.65 km/h registered for January 1996 eddy.
 The mean path followed by the cold-core rings is shown in Figure 1. The black arrow represents a simplification of the path of the eight eddies identified in the CLIMA and SOChIC XBT sections. Along their eastward track, eddies move north-east until about 162°E, then they follow the Campbell Plateau edge southward until 165°E and 55°S and finally northward until their decay.
 This behavior of the rings, never reaching depth shallower than −2000 m, is similar to the change in ACC fronts position due to the presence of shallow bathymetry. As for the ACC fronts, the deep reaching structure of eddies and conservation of potential vorticity is the reason for their behavior with respect to bathymetry [Fu, 2009; Lu and Speer, 2010; Sallée et al., 2011; Thompson and Sallée, 2012]. Moreover, bathymetry seems to play an important role in triggering the spawning of eddies from the ACC. Our results on eddies generation sites show that these are located downstream of the Southeast Indian Ridge at about 55°S and 150°E, just after that the ACC main streamlines and fronts cross the ridge. This result is in agreement with existent literature for which flow instabilities, most notably the baroclinic or shear flow instability mechanism [Best et al., 1999; Wolff, 1999], and consequent eddy generation are known to be closely correlated with areas where the ACC crosses prominent bottom topography, such as mid-ocean ridges [Swart et al., 2008; Sokolov and Rintoul, 2009a]. In particular, Sokolov and Rintoul [2007, 2009b] showed the existence of three areas of intense mesoscale activity in the sector between 150°E and 150°W. Two of these areas, located at latitude between 60°S and 50°S from West to East, are coincident with the origin and the northward turning of our eddies along their path.
 During their lives, eddies cross the ACC fronts and actively transport cold water and its properties northward. Merging information about the position of the ACC fronts from satellite altimetry, we found that a large number of eddies is generated in correspondence of the PF between 55°S and 60°S at about 160°W. These eddies cross the Polar Frontal Zone, reach the SSAF and eventually cross it while another series of eddies is sometimes generated in correspondence to the SSAF at the southern edge of the Campbell Plateau.
 This second group of eddies, together with some of the eddies coming from the Polar Frontal Zone, can still move north and cross the NSAF at about 50°S/175°W where other eddies seem to be generated in correspondence to this front. After this, eddies are out of the ACC main streamlines and dissipate. An example of this process can be found in Figure 6, where eddy spawning from the PF, the SSAF, and the NSAF is evident.
 Moving cold water north, the eddies also duplicate at lower latitude the SSH signal associated with water masses typical of southern latitudes. This is evident in Figure 2b, where for years 1995, 2000, 2001, 2002, 2007, and 2008, we find two separate latitudinal band for the SSH label associated with the PF. The southern band is almost always coincident with the in situ position of the front while the northern anomalous bands can be associated with the presence of eddies. These anomalies are even more evident when plotting together the latitudinal bands of SSH associated with the three main fronts of the ACC (Figure 8).
 In Figure 8, as for Figure 2, vertical bar represents the latitude interval where the SSH values associated with the fronts are found along the track of each CLIMA/SOChIC cruise. Bands for the SAF (gray), the PF (green), and the sACCf (red) are represented together with their mean position from in situ data.
 An overlapping of the SSH height signal of the different fronts is evident during many cruises; in particular, isolated SSH values associated with the PF are evident in the SAF band during 1995, 2000, 2001, 2002, and 2007 cruises all characterized by the presence of eddies. Some overlapping between sACCf and PF is also evident for the cruises of 1998, 2000, and 2003.
 The presence of the peculiar SSH values normally associated with southerly waters and the temperature profiles inside the eddies from XBT data suggest that eddies can effectively move cold waters of the Polar Frontal Zone up to the northern limit of the ACC system and cross the NSAF changing very little their own physical properties.
 The potential of moving cold water northward without any significant mixing can find confirmation when looking at the ratio between the rotational and the translational speed of eddies. Following recent paper by Chelton et al. , we calculated the advective nonlinearity parameter for our eddies. This parameter is defined as U/c, where U is the maximum circum-average geostrophic speed within the eddy interior and c is the translation speed of the eddy. The value of U/c > 1 implies that there is trapped fluid within the eddy interior [Chelton et al., 2011].
 We used geostrophic velocity derived from altimetry at the moment eddies cross the XBT sections and combined them with all the range of translational speed experienced by the eddies during their life. Results show that the advective nonlinearity parameter for eddies detected in the XBT sections is always greater than 1 even when combining the smaller rotational speeds with the larger translational velocities. Some typical values of U/c obtained using mean rotational and translational speeds, and so underestimating the advective nonlinearity parameter, are reported in Table 3 and range between 2.4 and 5. On this basis, we can assess that the eddies we identified can effectively trap water inside them and prevent this water from mixing with the surrounding until they dissipate.
 In the first part of this work, we determined the positions of the main ACC fronts based on 23 XBT transects collected by the CLIMA and SOChIC projects during the austral summers from 1994 to 2010 and compared them with those determined through satellite altimetry. This comparison between the in situ and satellite detection of ACC fronts positions shows a good correspondence between the two identification methods also if it gives different results for each front.
 In order to identify the possible presence of mesoscale eddies in the XBT sections, we identified the temperature sections presenting anomalies in the latitudinal distribution of temperature and analyzed them through satellite altimeter data. Temperature anomalies in the XBT data have been then associated to mesoscale eddies if:
 1. linked to an isolated low sea-level signal, larger than −20 cm, in the satellite altimetry maps
 2. successfully tracked from their spawning to decay in the satellite SLA maps.
 This approach, merging satellite and in situ data, allowed us to identify a limited number of cold-core eddies crossing the XBT sections and to estimate their mean path and parameters. Moreover, using the values of SSH associated to each ACC front, we determined the position of eddies with respect to the fronts.
 Results confirm that the spawning and paths of eddies are strongly influenced by the floor topography, in particular by the Southeast Indian Ridge and the Campbell Plateau. Regarding the ACC frontal structure, the spawning of eddies is found to occur in correspondence to the PF, at about 160°W, and of the SSAF at the southern edge of the Campbell Plateau. After generation, both groups of eddies can still move north and cross the NSAF where other eddies are eventually produced.
 To evaluate the contribution of eddies to the global heat budget, we estimated the surface layer heat content along each XBT track and the AHA that can be associated with a “typical” eddy, then compared them with the heat that is lost to the atmosphere at high latitude.
 The effective contribution of eddies to the global heat budget is obviously linked to their number, ability to cross the ACC fronts, and capacity of reaching the area north of the NSAF exporting cold water from the Antarctic Zone to the Subantarctic Zone.
 We observed cold-core anomalies in about one third of our XBT sections. This ratio, in agreement with existent bibliography [Sokolov and Rintoul, 2009b; Chelton et al., 2011], confirms the importance of our study area in the generation of eddies, while also providing an indication of the significance of the contribution from the eddies to the global heat balance. Moreover, analyses based on the ratio between the rotational and translational speed of each eddy prove that the mixing of the water masses inside our eddies with the surrounding is inhibited. In the case of some eddies generated at the PF, this water entrapment is also evident through the duplication, at lower latitudes, of the SSH signal associated to waters typical of the Polar Frontal Zone.
 Even if based on a reduced number of eddies identified in the CLIMA/SOChIC XBT sections, our results confirm the importance of eddies in the global heat balance and also underline the need of merging satellite data sets with high resolution in situ data.
 This study was performed in the framework of Climatic Long-term Interactions for the Mass-balance in Antarctica (CLIMA) and Southern Ocean Chokepoints Italian Contribution (SOChIC) projects as part of the Italian “National Program for Research in Antarctica” (PNRA) and was financially supported by ENEA through a joint research program. The authors thank two anonymous referees whose suggestions helped improving the paper. The altimeter products were produced by Ssalto/Duacs and distributed by Aviso, with support from Cnes (http://www.aviso.oceanobs.com/duacs/).