In December 2000 we conducted high-resolution hydrographic and towed undulator transects across the South Georgia shelf and into the deep waters of the Georgia Basin. The Southern Antarctic Circumpolar Current Front (SACCF) was observed flowing northwestward close to the base of the slope at 53.7°S, with a second manifestation around 53.4°S having flow in the reverse direction. Both crossings had a significant barotropic component aligned with the flow. The region of the SACCF was characterized by large-scale undulations of properties on horizontal scales of 5–10 km, similar to the local internal Rossby radius. We observe a distinct surface temperature gradient associated with the SACCF in the vicinity of South Georgia and demonstrate the usefulness of this by tracing the course of the front around the island with advanced very high resolution radiometer data. A distinct northward deflection of the front to the north of South Georgia is probably due to topographic steering by the North Georgia Rise. The signature of the SACCF is pronounced in the Circumpolar Deep Water, with strong isopycnal interleaving indicative of cross-frontal mixing. The salinity of the Weddell Sea Deep Water is marginally higher between the two manifestations of the SACCF, reflecting the different pathways from its source regions in the Weddell Sea. Remarkably strong vertical gradients of potential temperature and salinity exist in the bottom 100 m of the water column (at depths of around 3500 m) because of the vertical juxtaposition of waters that have taken different routes around the Northeast Georgia Rise.
 South Georgia lies on the North Scotia Ridge toward the eastern end of the Scotia Sea (approximately 54°S, 37°W; Figure 1). It is separated from the deep ocean by a shelf region of around 50–150 km width; this shelf is typically shallower than 300 m, but is frequently dissected by submarine canyons. Waters over the South Georgia shelf often have different characteristics from the water masses beyond the shelf break [Brandon et al., 1999, 2000; M. P. Meredith et al., Variability in oceanographic conditions to the east and northwest of South Georgia, 1996–2001, submitted to Progress in Oceanography, 2002, hereinafter referred to as Meredith et al., submitted manuscript, 2002], a consequence of retention processes combined with significant fluxes of heat and freshwater. Occasionally, the boundary between these different waters takes the form of a distinct shelf-break front [Brandon et al., 1999]; at other times, a much more gradual horizontal transition is apparent (Meredith et al., submitted manuscript, 2002). Significant interannual variability has been observed both over the South Georgia shelf and beyond the shelf break (Meredith et al., submitted manuscript, 2002), with both small-scale local processes and large- (basin-) scale processes contributing. At the very large scale, climate anomalies have been observed precessing to South Georgia, resulting in interannual variability in temperature around the island's shelf [Trathan and Murphy, 2002]. These anomalies can transfer from over the deep ocean environment onto the South Georgia shelf, and significantly perturb local conditions there, with consequences for the local ecosystem (Meredith et al., submitted manuscript, 2002). The mechanisms by which these climate anomalies are transferred from the Pacific into the South Atlantic sector of the Southern Ocean and toward South Georgia involve advection by the Antarctic Circumpolar Current (ACC).
 The ACC is a banded structure consisting of a number of bottom reaching fronts with high current speeds separated by broader, more quiescent zones of water. Much of our present-day understanding of the ACC derives from work undertaken during the International Southern Ocean Studies (ISOS) program in the 1970s and 1980s [e.g., Sievers and Nowlin, 1984]; this concentrated on Drake Passage, and the fronts observed here were termed (north to south) the Subantarctic Front (SAF), the Polar Front (PF), and the Continental Water Boundary (CWB). More recently, Orsi et al.  used a comprehensive set of historical hydrographic measurements to trace these features around Antarctica. They noted that the term CWB was somewhat illogical, since it was actually a circumpolar feature, and in some other areas was situated at a great distance from the Antarctic continent. The name Southern ACC Front (SACCF) was suggested as an alternative. A Southern Boundary to the ACC was also identified; south of this in the Scotia Sea lie the waters of the Weddell-Scotia Confluence (WSC), characterized by reduced vertical gradients of water mass properties because of the injection of dense shelf waters near the tip of the Antarctic Peninsula [Whitworth et al., 1994].
 Over the deep ocean environment around South Georgia, waters in the upper 200 m are generally termed Antarctic surface waters. Typical surface temperatures during the austral summer (when most in situ measurements are made) are around 2°–3°C over the South Georgia shelf, the slope, and the adjacent deeper ocean (Meredith et al., submitted manuscript, 2002), although values approaching 4.5°C have been observed. Salinities are generally around 33.6–33.8. It is extremely rare, but not completely unknown, for the ocean to freeze around South Georgia in winter. In the austral summer, the Antarctic surface waters feature a pronounced temperature minimum at around 100 m depth. This is the remnant of the previous winter's mixed layer, and is formed seasonally by deep convection. It is commonly termed Winter Water (WW) [Mosby, 1934]. Typical values for the temperature of the WW close to South Georgia are around 0°C although they can be as low as −1°C (Meredith et al., submitted manuscript, 2002). Salinities in the WW layer are generally around 33.9.
 The most voluminous water mass in the region is Circumpolar Deep Water (CDW), derived from the North Atlantic Deep Water (NADW) that becomes incorporated into the ACC [Reid et al., 1977; Whitworth and Nowlin, 1987]. This is divided into two components, specifically Upper CDW (UCDW) and Lower CDW (LCDW). UCDW is characterized by a maximum in potential temperature (and a minimum in dissolved oxygen) at around 500 m depth, while LCDW is characterized by a maximum in salinity around 1000–1500 m depth. The densest water in the Scotia Sea and Georgia Basin is Weddell Sea Deep Water (WSDW), which forms through mixing of mid-depth Warm Deep Water (WDW) from the Antarctic subpolar gyres with cold, saline shelf water [Gill, 1973; Carmack and Foster, 1975; Foster and Carmack, 1976] and is part of the continuum that comprises Antarctic Bottom Water.
Orsi et al.  found the properties of some of these water masses to be useful in determining the location of the SACCF. Specifically, they located stations north of the SACCF as having: potential temperature greater than 1.8°C along the potential temperature maximum at depths greater than 500 m; salinity greater than 34.73 along the salinity maximum at depths greater than 800 m; and dissolved oxygen concentrations less than 4.2 mL l−1 along the oxygen minimum at depths greater than 500 m. Stations south of the SACCF were identified as having potential temperature less than 0°C along the potential temperature minimum at depths shallower than 150 m. They stated that each of these generalized criteria might not apply in a given region. An example of this is given by Thorpe , who noted that on the World Ocean Circulation Experiment section A23 close to South Georgia, the progression of temperature at the subsurface temperature minimum was actually reversed, a consequence of local mixed layer processes.
 Approximate mean positions of the ACC fronts across the Scotia Sea (derived largely from historical hydrographic data, though supplemented with some more recent measurements) are marked in Figure 1. The SACCF is believed to be especially important in dictating oceanographic conditions at South Georgia since it loops anticyclonically around the island from the south before retroflecting to the east, and thus remains in close proximity to an extended region of the slope. Trathan et al. [1997, 2000] examined hydrographic data from surveys conducted to the northeast of South Georgia, and observed some apparent variability in the location of the SACCF. They argued that a more easterly location of the SACCF was associated with warmer waters close to South Georgia. Thorpe et al.  used a high-resolution map of historical geopotential anomaly to identify the mean position of the SACCF retroflection north of South Georgia: this was found to be close to 36°W, approximately 400 km further east than it was depicted in the earlier climatological map of Orsi et al. . Thorpe et al.  also presented evidence for temporal variability associated with the SACCF in the South Georgia region, and used drifter data to demonstrate variability in the western extent of the SACCF retroflection north of South Georgia and for the presence of eddies in the region.
 There have been rather few calculations of the SACCF flux in the Scotia Sea region. Orsi et al.  derived a value of 15 Sv from a historical section crossing the central Scotia Sea, but this was solely for the baroclinic component relative to 3000 m or the deepest common level. The WOCE section A23 crossed the SACCF three times in the vicinity of South Georgia, and baroclinic transports of 16.5 Sv, 13 Sv and 14 Sv were obtained for these crossings [Thorpe et al., 2002].
 The pathways and fluxes of the SACCF have implications for the South Georgia ecosystem, since the local population of Antarctic krill (Euphausia superba) is not self-sustaining [Everson, 1984], and thus requires input from other regions. Krill is the key food species for the vast colonies of higher predators at South Georgia; variability in the reproductive success of these predators is connected with fluctuations in the local abundance of krill [Croxall et al., 1988, 1999]. It has been hypothesized that krill are advected by the SACCF across the Scotia Sea from their source regions near the Antarctic Peninsula and in the southern Scotia Sea [e.g., Hofmann et al., 1998; Murphy et al., 1998; Fach et al., 2002]. Direct observations of krill flux have been significantly lacking, however; this is discussed in detail by Murphy et al. .
 As with the other ACC fronts, the SACCF can have a strong effect on the deep water masses in the Scotia Sea and the Georgia Basin farther north [Naveira Garabato et al., 2002; Meredith et al., 2001]. Being a full depth feature, it can act as a hydrodynamic barrier for the northward spreading of water masses from southern sources, and may also act to increase abyssal mixing rates. Interaction of the front with the convoluted topography of the region produces a number of different pathways for water masses entering the region from Drake Passage to the west or the Weddell Sea to the south [Arhan et al., 1999; Naveira Garabato et al., 2002], with the different routes being discernible through the different mixing histories of the water.
 In this paper, we present results from a fine-scale hydrographic survey of the SACCF in the southern Georgia Basin close to South Georgia. This consists of a station-based section that profiles the full depth features of the waters encountered, and a towed undulator section that provides a characterization of the upper layers of the SACCF at spatial scales unattainable using conventional stations. Physical characteristics of the front and its impact on the various water masses encountered are discussed, with special emphasis on the role of the front (and its interaction with topography) in dictating the spreading and mixing of the waters. The potential for infrared remote sensing of the SACCF close to South Georgia is demonstrated, and we use this potential to trace the course of the front as it loops anticyclonically around the island, interacts with the bathymetry, and retroflects to the east. Physical fluxes in both the baroclinic and barotropic modes associated with the front are presented.
 In situ data used in this paper were collected on RRS James Clark Ross cruise 57, during which a transect oriented northeast from South Georgia (Figure 2) was conducted. An initial outward leg was performed during daylight on 26 December 2000 using towed and underway instrumentation (described below), followed by a return occupation of the transect during which hydrographic stations were conducted. The locations of these stations are shown in Figure 2. At each station, data were collected with a SeaBird 911plus conductivity-temperature-depth (CTD) instrument. Station separation was approximately 10 km, close to the Rossby radius of deformation in this region [Houry et al., 1987]. The deviation from a straight line transect at station 81 was a diversion to avoid an iceberg. Discrete water samples were obtained from a SeaBird 12-position carousel water sampler carrying 10 L Niskin bottles. These were analyzed for salinity on a Guildline 8400B salinometer using ampoules of standard seawater (batches P133 and P137) prepared by Ocean Scientific International Ltd. CTD salinities were calibrated using these discrete measurements, and CTD downcast profiles were averaged on 2 dbar intervals for analysis. On the initial outward transect, a Chelsea Instruments NvShuttle undulating oceanographic recorder (UOR) was employed. This cycled between around 5 and 140 m, completing a full undulation approximately every 2 km; temperature, pressure and conductivity were among the suite of parameters logged. The UOR data was calibrated using collocated CTD casts. Since these were not conducted simultaneously (the UOR transect was conducted prior to the CTD section), there is a potential source of error here: the estimated accuracy of the UOR salinity is 0.007. The UOR transect does not cover as much of the South Georgia shelf region as the CTD section because of initial problems with the undulator data logging.
 Continuous measurements of water velocity in the upper part of the water column were made using a 153.6 kHz RD Instruments vessel-mounted Acoustic Doppler Current Profiler (VM-ADCP), sited in a sea chest to afford protection from ice. The ADCP was configured to record data in 64 8 m bins, and data were captured in ensembles of 2 minute duration. The “blank beyond transmit” was set to 4 m; coupled to the transducer depth (6 m), this gives the centre of the first bin depth at 14 m. Absolute velocities were obtained by using data from the Global Positioning System obtained with an Ashtech ADU-2 to correct for drift and error in the ship's gyrocompass, with bottom track data being used to correct for misalignment of the Ashtech antenna array relative to the ADCP transducers and for the inherent scaling factor associated with ADCP velocities.
 Satellite imagery was obtained from the AVHRR flown on the National Oceanic and Atmospheric Administration series of weather satellites. Images were captured at the British Antarctic Survey (BAS) receiving station at Rothera, with sea surface temperature (SST) being obtained using the split window retrieval algorithm on channel 4 (11 μm) and channel 5 (12 μm) following Llewellyn-Jones et al. . Images were contrast stretched to improve the apparent thermal resolution of the ocean surface for the temperature range of interest, and false colored for clarity.
3.1. Identification of the SACCF
 Full depth vertical fields of potential temperature (θ), salinity (S), neutral density (γn) [Jackett and McDougall, 1997] and baroclinic velocity are shown in Figure 3. These sections show a doming of isopleths centered near station 88. Clearly there is a recirculation, with perpendicular flow oriented to the northwest of the section between stations 91 and 101, and perpendicular flow to the southeast of the section between stations 72 and 85. This is the signature of a loop in the SACCF, and is particularly pronounced in the bidirectional structure of the baroclinic velocity field referenced to the deepest common level (Figure 3d). For the purposes of this paper, we term these manifestations of the SACCF on the section crossing 1 (closest to South Georgia) and crossing 2 (farthest from South Georgia). Crossing 1 is a narrow feature, being around 30 km total width. Crossing 2 is somewhat broader, with lower geostrophic velocities. It is important to note that the station pair farthest to the northeast on the section (stations 69 and 72) is consistent with flow in the reverse direction to that associated with the adjacent crossing 2. This seems to identify station 72 as being the centre of a second, less pronounced recirculation, though with only one station farther out from South Georgia than this, it is difficult to be more expansive from the hydrographic data alone.
 On traversing crossing 1, the temperature at the core of the WW changes from −0.32 to 0.41°C, the salinity at the core of the LCDW changes from 34.705 to 34.708, and the potential temperature at the core of the UCDW changes from 1.85 to 2.08°C. For traversing crossing 2, the temperature at the core of the WW changes from 0.04 to 0.44°C, the salinity at the core of the LCDW changes from 34.704 to 34.722, and the potential temperature at the core of the UCDW changes from 1.91 to 1.98°C. It is apparent that, in the locality under study, the large-scale criteria presented by Orsi et al.  do not provide reliable indicators for the position of the SACCF. Orsi et al.  state specifically that their criteria are intended solely as useful indices, and that regional variability in the circumpolar property fields [e.g., Gordon and Molinelli, 1982] implies that one or more of them need not apply at any given location. Indeed, it has been shown at other locations that the fronts change along-stream as they split and merge and are subject to seasonal cycles in heat and freshwater flux [e.g., Belkin and Gordon, 1996]. In addition to this, we note that strong isopycnal mixing is observed on our section (see below); this will change the property values at their core extrema, and thus further reduce the local applicability of large-scale criteria based on them.
3.2. Impact of the SACCF on Water Mass Properties
3.2.1. Surface Waters
Figure 4 shows the characteristics of the water masses encountered on the section in potential temperature/salinity space. On our section, the WW properties clearly fall into two categories. Stations 78–97 (between the two crossings of the SACCF) exhibit water at the temperature minimum that is colder than 0.3°C, with a neutral density higher than 27.55 (σ0 > 27.35). All other stations on the section have much warmer potential temperature minima (θ > 0.4°C) that are less dense (γn < 27.55; σ0 < 27.35). The potential temperature minima of stations 78–97 are, in general, also marginally more saline. This is clearly the influence of the SACCF, with the WW properties between crossing 1 and crossing 2 being characteristic of waters of more southerly origin.
 The coldest WW on the section (θ < −0.5°C) is present at station 88, the centre of the SACCF loop. Cold WW is also found at the adjacent station 91 (θ < −0.2°C), and also at station 78 (θ < 0°C). This latter station also features a lens of cold water at around 300 m intruding into the deeper water below. The temperature minimum layer extends onto the South Georgia shelf (Figure 4); here it is capped by waters that are even warmer than northeast of crossing 2, and also become progressively fresher with decreasing distance from land. These are the effects of water mass retention on the shelf coupled to freshwater discharge and warming through insolation [Brandon et al., 1999; Meredith et al., submitted manuscript, 2002].
 The fine-scale (minimum order of 2 km) spatial characteristics of the two crossings of the SACCF can be seen in the vertical sections derived from the UOR covering the upper 150 m of the water column (Figure 5). In addition to the same broader-scale (minimum order of 10 km) structure of the SACCF observed in the CTD section, there is evidence for large-amplitude (approximately 50 m) undulations in properties beneath the seasonal halocline, the base of which is around 75 m depth between the two SACCF crossings. These undulations typically occur over horizontal scales of 5–10 km. The undulations of salinity and density are strongly correlated, as is expected for waters at low temperature.
 Initial concepts of the SACCF were that it did not separate surface water masses, and hence should not have a surface signature [Orsi et al., 1995]. In some regions, evidence has been presented to support this [e.g., Heywood et al., 1999], although some fine-scale surveys do seem to show a surface expression of the SACCF [e.g., Holliday and Read, 1998]. Figure 6 shows the CTD-derived temperature at the 1 dbar level across our transect; clearly there are large horizontal gradients between stations 91 and 107 (approximately −0.03°C/km) and stations 72 and 78 (approximately 0.05°C/km). These coincide spatially with the locations of crossing 1 and crossing 2 of the SACCF, and the sense of the temperature gradient (negative closest to South Georgia, positive farthest) is in accord with the different directions of flow in these two crossings. The presence of a surface temperature gradient across the SACCF close to South Georgia is more than just a curiosity, since it will have consequences concerning (for example) the solubility of biologically and climatically important gases, and the behavior of temperature-dependent ecosystems functions. It also offers the potential for tracing the course of the SACCF using satellite-borne infrared radiometers, as will be demonstrated below.
3.2.2. Deep and Bottom Waters
 UCDW typically inhabits the approximate ranges 1.6 < θ < 3.2°C, 34.00 < S< 34.71, 27.55 < γn < 28.00 [Naveira Garabato et al., 2002]. Previous definitions of this water mass based on potential density placed it in the range σ0 > 27.35, σ2 < 37.00 [Reid et al., 1977; Sievers and Nowlin, 1984]. LCDW generally occupies the neutral density range 28.00 < γn < 28.26 [Naveira Garabato et al., 2002], equivalent to σ2 > 37.00, σ4 < 46.04 [Sievers and Nowlin, 1984]. In addition to these, WDW also has an influence on the properties of the waters observed. WDW is derived from the LCDW of the ACC that becomes incorporated into the eastern Weddell Gyre through a broad discontinuity in the southwest Indian Ridge [Orsi et al., 1993], and circulates cyclonically within it before exiting over or around the South Scotia Ridge. Whilst having similar density characteristics to LCDW, WDW can be distinguished through the effects of its different mixing history: it has lower potential temperature (0.2 < θ < 0.6°C) and salinity (less than 34.69 at the maximum).
 On our transect, the potential temperature maximum of the UCDW was strongest between station 97 and the slope, where it exceeded 2°C (Figure 33), and northeast of station 78, where it again exceeded 2°C. Between stations 81 and 94, the potential temperature maximum of the UCDW was much lower, at just above 1.8°C (again, the influence of the SACCF). Very strong isoneutral interleaving is apparent for the density range of UCDW (Figure 4a). This is dominant for stations 78–97 (the same stations that featured the cooler type of WW), and much weaker for the other stations. The potential temperature/salinity curves of stations 78–97 are consistent with waters from equatorward of the SACCF in the ACC zonation mixing with cooler, fresher waters of a more southerly origin that spread north toward South Georgia within and south of the SACCF.
 None of the deep waters sampled along our transect in the density range 28.00 < γn < 28.26 fully meet the criteria for WDW, and thus are more properly termed LCDW. However, as with the UCDW, there is evidence of profound isoneutral interleaving of the LCDW sampled with cooler, fresher waters (Figure 4b); that is, the waters observed are the product of significant mixing between LCDW and WDW. This is the signature of mixing upstream with WDW. That the interleaving is so intense is probably related to the SACCF and its retroflection: the large property gradients, coupled with the strong vertical and horizontal shears, will act to enhance the mixing.
 WSDW typically has characteristics around −0.7 < θ < 0.2°C, S < 34.703, and 28.26 < γn < 28.40 or 46.04 < σ4 < 46.16 [Arhan et al., 1999; Naveira Garabato et al., 2002]. It circulates cyclonically in the Weddell Gyre, and can escape through gaps in the South Scotia Ridge to flow through the Scotia Sea prior to exiting through Georgia Passage [Locarnini et al., 1993]. WSDW can also spread northward on the eastern side of the South Sandwich Arc. Once having traversed the Scotia Sea or circulated around the South Sandwich Arc, two routes are possible for WSDW to reach the sampling location, namely anticyclonically around the Northeast Georgia Rise (NEGR; Figure 1), or through the gap between the NEGR and the South Georgia shelf break. Naveira Garabato et al.  showed that water denser than around γn = 28.29 kg/m3 lies too deep in the water column to flow along this latter route.
 Striking in the potential temperature, salinity and density fields (Figure 33) is a remarkably intense abyssal layer, around 100 m thick, with very large gradients in properties (stations 78–94). This layer exists in waters around 3500 m deep, and possesses vertical gradients similar to those found in the near-surface thermocline. The most likely explanation is that vertical stacking of different types of WSDW is occurring in the Georgia Basin, with different layers having taken different routes from the Weddell Sea. Densities within the very bottom layer are greater than γn = 28.29 kg/m3, thus this abyssal layer will have reached the sampling location by circulating anticyclonically around the NEGR [Naveira Garabato et al., 2002]. Conversely, the layer immediately above this can enter the basin through the gap between the NEGR and the South Georgia slope. The vertical juxtaposition of these different types of WSDW within the Georgia Basin locally creates the strong abyssal gradients in properties.
 Within this dense abyssal layer, two separate types of WSDW exist: stations 85 and 88 show WSDW that is more saline (by around 0.002–0.003) than WSDW elsewhere along the transect (Figure 4b). These two stations are situated between the two crossings of the SACCF. This small difference in salinity is indicative of different pathways followed by the waters from their source regions in the south. One branch of WSDW crosses the South Scotia Ridge from the Weddell Sea to enter the Scotia Sea; this water subsequently exits the Scotia Sea through Georgia Passage [Locarnini et al., 1993]. A second branch of WSDW flows northward on the eastern side of the South Sandwich Arc, around 26°W (Figure 1). The former type of WSDW becomes freshened by interaction with dense waters that spill off the shelf near the tip of the Antarctic Peninsula [Whitworth et al., 1994], while the latter branch is not subject to this freshening. Arhan et al.  and Meredith et al.  used the resulting small difference in salinity to identify different pathways from the Weddell Sea into the Georgia Basin and Scotia Sea respectively. For the present section, the WSDW in the dense abyssal layer at stations other than 85 and 88 has clearly crossed the South Scotia Ridge, flowed through the Scotia Sea, exited via Georgia Passage and reached the location of the transect by circulating anticyclonically around the NEGR. Conversely, WSDW at stations 85 and 88 has flowed around the South Sandwich Arc and anticyclonically around the NEGR to reach the location of the transect. Thus the difference in WSDW salinity witnessed along the section is evidence of the SACCF interacting with the complex bathymetry of the region to control the spreading pathways of the abyssal waters.
3.3. Spatial Characteristics of the SACCF Close to South Georgia
 We have seen that the SACCF exhibits a temperature gradient close to South Georgia such that surface temperatures change by around 1.5°C over spatial scales around 50 km (Figure 6). These values compare to typical radiometric and spatial resolution of surface temperature imagery from AVHRR of around 0.6°C and just over 1 km respectively. This clearly offers a capability for tracing the course of the SACCF using satellite data, subject to the availability of cloud-free satellite passes.
 One such image (partially cloud contaminated) is shown in Figure 7a. The data were recorded on 26 December 2000 at 20:06 GMT, thus it is an image from the evening before our hydrographic transect commenced. The surface structure in the southern and western areas of the image is obscured by high cloud (white). The image is somewhat confused by the presence of low cloud, which has temperatures comparable to those of the sea surface. (This low cloud was identified using different contrast stretching and coloring to those used for Figure 8). The northwest limit of the low cloud has been approximately marked on the image with a black line; red pixels to the south and east of this line should not be interpreted as representing areas of warm sea surface. The formation of this low cloud is clearly associated with the mountainous topography of South Georgia, since it starts at the island as cooler, higher cloud (blue pixels), and warms as it sinks during its spreading to the east. If such cloud formation and eastward spreading were a common occurrence at South Georgia it could have significant consequences for the local ecosystem, since levels of incident solar radiation at the ocean surface would tend to be higher on the western side of the island than the east.
 Despite the partial cloud contamination, broad areas of ocean surface are visible. These show the frequent mesoscale meanders and eddies characteristic of the baroclinically unstable ACC fronts. The PF can be distinguished beneath the patchy cloud cover at the northwest side of the image, looping around the southern side of the Maurice Ewing Bank [Moore et al., 1997; Arhan et al., 2002]. Marked on the image is the location of the hydrographic transect (red line), and the positions and directions of the SACCF crossings derived from the hydrographic data. Crossing 1 is partially obscured by low cloud, so it is difficult to make assertions concerning the spatial structure of the SACCF here. Crossing 2 is cloud free, and it is immediately obvious that it represents the southwestern edge of a detached warm-core ring of SACCF water situated at approximately 34.5°W, 53.2°S. This is consistent with the observation from the hydrographic section of reverse direction flow for the station pair at the very northeastern end of the transect (stations 69 and 72); clearly this station pair was situated to the northeast of the centre of this warm-core ring. We note that the surface waters in Figure 8 just north of the SACCF (around 52°S, 35°W) are similar in temperature to the waters at the PF close to the Maurice Ewing Bank. This would appear to be an indication that the PF and SACCF come very close at the point, and raises the possibility that they occasionally merge.
 The warm-core ring with which crossing 2 is associated can be traced back in the AVHRR image to the eastward flowing retroflected manifestation of the SACCF. Modelling studies tracing particles in the output of Ocean General Circulation Models (OGCMs) have indicated that while flow in the northwest flowing SACCF and its retroflection is relatively rapid, water is often transferred into the region between them from the retroflected SACCF in the form of rings (Figure 7b) [adapted from Thorpe et al., 2002]. Our observation of a warm-core ring shed to the south of the eastward flowing SACCF retroflection seems to verify this OGCM-predicted process. The modeled rings typically feature high velocities, but tend to remain roughly stationary for extended periods, and are thus relatively ineffective at transferring material into or out of the vicinity. The implication is that material contained in the warm-core ring may have been present in the region significantly longer than material present in crossing 1. This has profound consequences for ecosystem operations and the interpretation of biological measurements [Ward et al., 2002].
 A second example of the use of SST imagery in tracing the course of the SACCF close to South Georgia is shown in Figure 8. This is an AVHRR image from 12 July 2000, and thus is not coincident in time with the occupation of the hydrographic transect. However, it is notable for showing the course of the SACCF around South Georgia with comparatively little interference from cloud cover. (Such cloud contamination as exists around the edges of the image is depicted with white pixels.) It is clear from the image that the SST signature of the SACCF broadly follows the predicted course of the front, looping anticylonically around the island from the south, and subsequently retroflecting to the east on the northern side of the island. Noteworthy however is that at the time of image capture the SACCF appears to first come close to the South Georgia slope at around 40°W, 55°S, i.e., close to the southwestern edge of the island's shelf. This is contrary to the general depiction in climatological maps of the SACCF, which tend to have the front first approaching the slope much farther east (typically around 34°W, i.e., the southeastern edge of the shelf; see Figure 1). There is a large approximately meridional bathymetric ridge (around 2200 m depth) running along 39°W in the Scotia Sea between around 55.5 and 58°S (Figure 1); our SST image from July 2000 suggests that (at the time of image capture) the SACCF was lying north of this feature rather than south. The region of this bathymetric feature is obscured by cloud in the AVHRR image; further data are needed to gain more insight into this, and into possible variability of the front's position. In either case, the proximity of the SACCF to the southern side of the South Georgia shelf is potentially important for the local ecosystem. It has been argued that the SACCF is a major advective instrument in the transfer of krill across the Scotia Sea from upstream near the Antarctic Peninsula [Hofmann et al., 1998; Fach et al., 2002]. Regions where the front is closest to the shelf may be areas where krill transport into the shelf waters is facilitated. Although clearly influenced by the topography of the South Georgia slope, the SACCF also shows some large mesoscale meanders during its loop around the island. Particularly prominent is the large southward excursion of warm water at around 36°W to the south of South Georgia, and an eddy spinning off this excursion at around 57°S, 35°W.
 To the north of the island, the SACCF appears to be deflected away from South Georgia at around 37°W. A very similar deflection was observed previously at this location in the track of an iceberg that had been apparently following the path of the SACCF close to the bottom of the slope adjacent to South Georgia [Trathan et al., 1997, Figure 12]. The deflection of the iceberg was clearly associated with the presence of the North Georgia Rise (NGR; an approximately 1500-m-high topographic feature centered on 37°W, 52.5°S). This iceberg subsequently moved closer to South Georgia having rounded the NGR, and followed a cyclonic course around the periphery of the basin and north to the vicinity of the PF. The presence of the deflection in the satellite image of the SACCF again suggests the role of the NGR in steering the flow at the north side of South Georgia. After its deflection to the north, the SACCF appears to adopt a hammerhead eddy configuration, and sheds a number of rings to the west. It then retroflects to the east, and can be seen to meander again greatly as it flows north.
3.4. Fluxes Associated With the SACCF
 Baroclinic volume transports associated with crossings 1 and 2 of the SACCF can be obtained by spatially integrating the velocity field shown in Figure 3d. To account for bottom triangles, we extrapolate the baroclinic velocity profiles downward from the deepest common layer of each CTD pair. The resultant fluxes perpendicular to the section are shown in Figure 9a. The two crossings of the SACCF are clearly visible, with the highest magnitudes of flux between individual station pairs being −4.5 Sv (stations 94 and 97; 1 Sv = 106 m3/s), and 3.2 Sv (stations 75 and 78). At the northeastern end of the section (stations 69 and 72), flow in the reverse direction to the adjacent station pair is observed; this is the manifestation in the velocity field of the SACCF ring observed previously (e.g., Figure 7).
 In addition to the baroclinic component, we are able to quantify the barotropic (depth-independent) component of velocity across the section. To do this, we use direct velocity measurements from the VM-ADCP. 2-minute ensemble ADCP vectors from 22 m depth are shown in Figure 9b, with the most obvious features again being the two crossings of the SACCF. Note that the vectors associated with crossing 2 are at a more obtuse angle to the section than those of crossing 1, and veer across to reverse direction at the northeastern end of the section (the ring). The angle of the vectors associated with crossing 2 indicates that the section cut across the ring to the south of its core.
 To derive the barotropic component, we average the ADCP current vectors between station pairs, resolving perpendicular to the ship's track to enable comparison with baroclinic velocity. The tidal components of the VM-ADCP velocities were subtracted using version 3.1 of the Oregon State University global tidal model [Egbert et al., 1994]. Tidal velocities removed were always less than 2 cm/s for the meridional component, and always less than 0.5 cm/s for the zonal component. Below the ageostrophic layer (typically around 40–50 m depth), the baroclinic component (from the CTD data; e.g., Figure 3d) and the total velocity (from the averaged ADCP data) become close to parallel. The offset between the different velocity profiles represents the barotropic component of velocity. For some of the station pairs used here, the baroclinic and total velocity profiles were closer to parallel when adjacent stations were also considered, i.e., using two or more sets of stations pairs in one averaging procedure. In such circumstances, the barotropic component is more reliably derived than by considering each station pair separately, hence we adopt the averaged velocities and make the assumption that the average velocity applies to each separate station pair. The uncertainty in the offset (barotropic) velocity is estimated as ±2 cm/s.
Figure 9c shows the full depth fluxes associated with the barotropic component of the velocity field. The vectors appear less spatially variable than the baroclinic vectors in Figure 9a because of the averaging of more than two adjacent stations. The barotropic flux vectors are generally aligned in the same direction as the baroclinic flux vectors, i.e., northwestward in crossing 1 and southeastward in crossing 2. We interpret this as indicating that the SACCF has a barotropic flux component that enhances the magnitude of the baroclinic component. The one location where barotropic and baroclinic fluxes are notably oriented in opposing directions is directly in between the two crossings (stations 85 to 91), for which we have no clear explanation. Barotropic fluxes at the slope are smaller than the fluxes near the centre of the two SACCF crossings (Figure 9c). This is a reflection of the shallower water depth rather than lower barotropic velocity however, since velocities are actually substantially higher at the slope (see below).
 Total geostrophic fluxes are obtained by summing the barotropic and baroclinic components. For crossing 1, the total geostrophic flux is −14.1 ± 2.4 Sv. Approximately −9.8 Sv of this are in the baroclinic mode, and −4.3 Sv in the barotropic mode. For crossing 2, the total geostrophic flux is 10.2 ± 3.6 Sv, with approximately 5.7 Sv in the baroclinic mode and 4.5 Sv in the barotropic mode.
Thorpe et al.  deduced baroclinic fluxes of 13–16.5 Sv for the SACCF near South Georgia from an approximately meridional section undertaken in 1995. Our baroclinic fluxes are close to this, with that of crossing 2 being marginally smaller. At the western end of the Scotia Sea, Cunningham et al.  used combined hydrography and lowered ADCP measurements to deduce a transport for the SACCF at Drake Passage of 9.3 ± 2.4 Sv, again broadly consistent with our measurements. Whilst our measurements of transport across the section appear generally consistent with most previous estimates of SACCF transport, it is interesting to note that the maximum velocities associated with the SACCF in our measurements seem very high. Maximum baroclinic velocities at crossing 1 were above 35 cm/s; addition of the barotropic component raises this velocity still further. By comparison, Thorpe et al.  describe maximum baroclinic velocities associated with the SACCF close to South Georgia between 7.5 and 10 cm/s. Heywood et al.  show no values higher than 15 cm/s for the region of the SACCF and SB near 85°E in the Southern Indian Ocean. Typical values for the surface velocity of the SACCF at Drake Passage lie in the range 10–20 cm/s [e.g., Whitworth et al., 1982; Roether et al., 1993]. The consistency of our transport measurements with previous estimates, combined with the extremely high velocities we measured, is a consequence of crossing 1 being a very narrow, intense jet, even compared to other crossings of the SACCF. It is possible that local topography acts to constrict and accelerate the flow of the SACCF as it loops anticyclonically around South Georgia. Given that the SACCF is generally found in waters with depths approximately 2500–3000 m, it could well be that the gap between the South Georgia shelf and the NEGR (Figure 2) is responsible for this, with the SACCF effectively being funneled through.
Figures 9d and 9e show the baroclinic and barotropic fluxes associated with the upper 250 m of the water column, derived as above. The same general directions of flow are present, with the barotropic and baroclinic parts being generally coaligned for both crossing 1 and crossing 2. In addition to these geostrophic components, we can quantify also the surface ageostrophic flux caused predominantly by direct wind forcing and inertial effects. For this, we compared the geostrophic (baroclinic plus barotropic) velocity profiles averaged between stations with the directly measured total velocity from the VM-ADCP. For each comparison, the part of the water column closest to the surface (generally shallower than 40 to 50 m depth) showed significant deviation of total velocity from geostrophic velocity. In each case, this deviation was oriented to the northwest of the transect. We subtracted the geostrophic component from the total velocity in these layers, and integrated the residual velocity to obtain ageostrophic volume fluxes perpendicular to the transect (Figure 9f). The fluxes are very much smaller than the geostrophic fluxes, thus any error in them will have minimal impact on the total flux calculated. The ageostrophic flux being oriented to the northwest along the transect is consistent with the general wind forcing of the region; predominantly westerly winds will cause a northward Ekman transport.
 For the upper 250 m, the total volume transport associated with crossing 1 is −2.9 ± 0.3 Sv; around −2.0 Sv of this is in the baroclinic mode, −0.8 Sv in the barotropic mode, and −0.1 Sv due to ageostrophic components. For crossing 2, the total volume transport is 1.2 ± 0.3 Sv; around 1.1 Sv of this is in the baroclinic mode, 0.3 Sv in the barotropic mode, and −0.2 Sv is due to ageostrophic components. Of particular note is the large barotropic flux between stations 101 and 104 (over the South Georgia slope) that is closer to South Georgia than the highest baroclinic velocities (Figure 3d). This is suggestive (but by no means proof) of a topographically steered barotropic mode at the South Georgia slope; further measurements are needed to better examine this possibility. The impact of these upper layer fluxes on the horizontal advection of krill is examined by Murphy et al. .
 We have observed several important effects associated with the SACCF in the southern Georgia Basin. The impact of the front on the deep waters is profound: it interacts with topography to control the northward spreading of abyssal waters, with WSDW between the two crossings of the SACCF having been taken along a different transport path from WSDW outside the SACCF. The WSDW in the Georgia Basin flows north to become a component of the Antarctic Bottom Water of the global thermohaline circulation, thus restriction and steering of the flow in the Georgia Basin and farther upstream has potential consequences on a much broader scale than simply local circulation. The WSDW in the Georgia Basin also exhibits extremely large vertical gradients of properties in the abyssal layer, most likely because of different routes taken by waters of different densities around the North East Georgia Rise. The SACCF seems to be a region of enhanced deep mixing, with very large isoneutral intrusions at the density ranges of CDW.
 In the upper ocean beneath the seasonal halocline, we have observed fine-scale (5–10 km) undulations of properties across the SACCF. Such “streakiness” has been observed previously across ACC fronts [e.g., Pollard et al., 1995], and has been attributed to differential advection due to the large velocities of the front. Water mass properties are transferred along the axis of the front more rapidly than either side; this can act to sharpen cross-frontal gradients. Pollard et al.  argued that such differential advection can lead to phytoplankton patchiness, and it is well-known that mesoscale upwelling at fronts can also lead to chlorophyll patchiness on the scale of the local internal Rossby radius [Strass, 1992].
 We have shown that, at the time of our section, a surface temperature signal was associated with the SACCF close to South Georgia. We have used this surface temperature signal in satellite imagery to identify an isolated ring of SACCF water at the end of the transect (half of which coincided with crossing 2). Such rings were predicted by OGCMs to form in this region; this is good observational verification of this process. We also used satellite imagery to trace the course of the SACCF right around South Georgia. We note that a surface temperature signal associated with the SACCF close to South Georgia is also apparently present in the monthly composite images produced during the ongoing global survey of SST fronts conducted at the University of Rhode Island (I. M. Belkin, et al., Global survey of ocean fronts from Pathfinder SST data: 1. Atlantic Ocean, manuscript in preparation, 2003). This suggests that the presence of the SST gradient is not specific solely to the times of the data presented in this paper, but a more persistent feature, and offers scope for future remote sensing investigations of the SACCF close to South Georgia.
 We were able to detect some important features of the SACCF in the region, such as its close proximity to the southern slope of South Georgia, and its apparent northward deflection by the North Georgia Rise. Whether these features are persistent or relevant solely to the time of image capture requires further investigation. One of the images presented shows the SACCF appearing to shed rings to the west at the point of its retroflection. These rings seem to move south from the point of their generation, and hence they may be important in transferring biological material to the northwestern end of the island. Such circulations often feature enhanced local upwelling; this would tend to raise nutrient levels in the near-surface layers, thereby promoting primary production. The region around Bird Island, at the western tip of South Georgia, is one of the most biologically active areas of the entire Southern Ocean, with large colonies of predators that depend on krill as a food source. It has been argued that upwellings induced by mesoscale circulations may be critical in determining the characteristics of Southern Ocean ecosystems [e.g., Barth et al., 2001], thus the presumed persistence of the SACCF retroflection, and the possible frequency of the ring shedding, may be crucial. If so, the western extent of the retroflection, and the ring generation and translation processes, deserve further attention. Satellite infrared remote sensing offers the capability to study this, subject to the availability of cloud-free imagery.
 We thank the captain, crew and scientific party of RRS James Clark Ross cruise 57 for their support of this work. We are grateful to Steve Colwell, Jose Xavier and Steve Hart for assistance with the satellite data. We are especially grateful to Igor Belkin for valuable advice concerning the SACCF, and specifically its surface temperature signal close to South Georgia. Our thanks to the two reviewers whose comments greatly improved the original manuscript.