Upper ocean temperature and the baroclinic transport stream function relationship in Drake Passage

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 109,, doi:10.1029/2003jc002010, 2004 Upper ocean temperature and the baroclinic transport stream function relationship in Drake Passage Serguei Sokolov, 1 Brian A. King, 2 Stephen R. Rintoul, 1 and Ricardo L. Rojas 3 Received 18 June 2003; revised 4 February 2004; accepted 13 February 2004; published 1 May 2004. [1] Repeat hydrographic sections across the Antarctic Circumpolar Current (ACC) in Drake Passage are used to derive an empirical relationship between upper ocean temperature and the baroclinic transport stream function. Cross validation shows this relationship can be used to infer baroclinic transport (above and relative to 2500 m) from expendable bathythermograph (XBT) temperature measurements with an error of a few per cent. Transport errors of less than 2 Sv are obtained if temperature at depths between 600 and 1600 m is used to define the relationship. Temperature at depths above 300 m provides an unreliable index of transport because of variability in temperature-salinity (T-S) properties produced by air-sea interaction. The scatter in the relationship between temperature and stream function from repeat observations along a single line is similar in magnitude to the scatter observed when data from the broader Drake Passage area are considered. In both cases, variability about the mean temperature-stream function relationship reflects advection of water with anomalous T-S properties. The tight relationship between temperature and stream function in Drake Passage and south of Australia suggests baroclinic transports can be inferred from XBT temperatures with high accuracy in the Southern Ocean, providing a cost-effective means of monitoring ACC variability. However, care must be taken at the end points, particularly in the Drake Passage where the strong flow of the Subantarctic Front sometimes lies over the continental slope. INDEX TERMS: 4207 Oceanography: General: Arctic and Antarctic oceanography; 4223 Oceanography: General: Descriptive and regional oceanography; 4512 Oceanography: Physical: Currents; KEYWORDS: Antarctic Circumpolar Current, baroclinic transport, Southern Ocean Citation: Sokolov, S., B. A. King, S. R. Rintoul, and R. L. Rojas (2004), Upper ocean temperature and the baroclinic transport stream function relationship in Drake Passage, J. Geophys. Res., 109,, doi:10.1029/2003jc002010. 1. Introduction [2] The strong eastward flow of the Antarctic Circumpolar Current (ACC) connects the ocean basins and permits a global-scale overturning circulation to exist. The transport of the ACC has therefore long been of interest to oceanographers. Repeat hydrographic sections and moored current measurements during the International Southern Ocean Studies (ISOS) experiment in the 1970s provided the first estimate of ACC transport: 134 ± 13 Sv (1 Sv = 10 6 m 3 s 1 ) [Whitworth, 1983; Whitworth and Peterson, 1985]. (The error bar on the mean transport is based on the standard deviation of net transport estimated from four hydrographic sections referenced to the same set of current meter moorings.) 1 Antarctic Climate and Ecosystems Cooperative Research Centre and CSIRO Marine Research, Hobart, Tasmania, Australia. 2 Southampton Oceanography Centre, Southampton, UK. 3 Centro Nacional de Datos Oceanogrficos de Chile (CENDOC) Servicio Hidrografico y Oceanografico de la Armada (SHOA), Valparaiso, Chile. Copyright 2004 by the American Geophysical Union. 0148-0227/04/2003JC002010 [3] One of the conclusions of the ISOS experiment was that the variability of the ACC was primarily barotropic. Subsequent repeat measurements carried out in Drake Passage during the World Ocean Circulation Experiment (WOCE) suggested that the baroclinic and barotropic variability were comparable [Cunningham et al., 2003]. This observation suggests that monitoring of ACC transport variability requires measurements of the density field as well as changes in the pressure field associated with the barotropic flow. The traditional way to observe the density field is with full depth hydrographic sections from research vessels. The resources required for this mean that we are unlikely to ever be able to conduct repeat hydrographic sections to monitor the interannual and shorter-term baroclinic variability of the ACC with sufficient temporal resolution to avoid aliasing. An alternative approach is to develop a proxy technique based on measurements which can be obtained more frequently because they are less expensive and do not require research vessels, such as expendable bathythermograph (XBT) observations. [4] In regions characterized by a single well-defined temperature-salinity (T-S) relationship, the T-S curve can be used to identify the salinity corresponding to each temperature observation. Density, dynamic height, and 1of14

velocities can then be calculated in the usual manner [e.g., Stommel, 1947]. In the Southern Ocean, such a tight T-S curve does not exist. Temperature and salinity profiles generally cool and freshen to the south, and at some locations the relationship is multivalued: one temperature corresponds to several salinity values. However, the regional T-S curve is stable in the sense that the envelope of T-S curves does not change with time, and each streamline is associated with a particular T-S curve. This stability was exploited to derive a relationship between upper ocean temperature T h and the potential energy anomaly c or baroclinic transport stream function at the WOCE SR3 repeat section between Tasmania and Antarctica [Rintoul et al., 2002]. The relationship allowed transports to be estimated with a RMS error of <3 Sv. Each of the main ACC branches was well captured by the temperature-based transport estimate. The empirical relationship was used to infer baroclinic transport from summer XBT measurements obtained between 1993 and 1999 and was verified independently by baroclinic transport estimates derived using sea surface height (SSH) anomaly data measured by the altimeter. [5] Here we test the application of such a relationship to the Drake Passage region. First we investigate the sensitivity of the transport errors to the choice of depth interval over which to average the temperature measurements. We find that using temperature averaged over 100 m intervals centered on depths between 600 and 1600 m all give reliable baroclinic transport results. We then examine how well the relationship derived from observations at a single repeat transect across Drake Passage applies to the region as a whole. We investigate the spatial scales over which the relationship varies, how the variability in space compares to temporal variability along a single line, and identify the processes contributing to variability in the c T h relationship. [6] The Drake Passage results are compared to those from an earlier study south of Australia. The success of the method in two widely separated sectors of the Southern Ocean, which are characterized by different water mass properties and frontal structures, suggests the technique is likely to succeed in much or all of the Southern Ocean. Throughout the region, the deep reaching currents [e.g., Killworth, 1992] result in a one-to-one relationship between temperature at a given depth and an integral of the density field (e.g., dynamic height or the baroclinic transport stream function) at that location. Watts et al. [2001] extended this idea (originally stated by Watts and Rossby [1977]) to show that profiles of T, S and r could be inferred from integral proxy measurements such as acoustic travel time as measured by inverted echo sounders. Sun and Watts [2001] showed this approach could be extended to the circumpolar Southern Ocean. In our case, we use limited profile information (upper ocean temperatures) to infer a vertical integral of the circulation (baroclinic transport). [7] Estimates of geostrophic volume transport across a section depend solely on the end point stations (in the absence of intervening topography and neglecting changes in the Coriolis parameter; estimates of the transport of heat or other properties of course depend on correlation between velocity and concentration along a section, not just the end points). Therefore any attempt to use a proxy measurement for transport must pay careful attention to the end points. Figure 1. Location of WOCE repeat section SR1 (diamonds), repeat Chilean CTD section (squares), and historical hydrographic sections occupied in 1975 1976 carried out by the FDRAKE program (circles). The RMS sea surface height variability (larger than 0.04 m) from the merged TOPEX/Poseidon and ERS satellite altimeters (the AVISO MSLA product) is contoured. The 3000 m depth contour is shown. For example, south of Australia the section extends into subtropical water, with a distinct T-S relationship requiring separate c T h approximations north and south of the Subtropical Front (STF). The Drake Passage section is short and remains in Southern Ocean waters. However, we show that the proximity of the strong flow of the Subantarctic Front (SAF) to the continental slope can introduce significant errors into net transport estimates for Drake Passage if flow over the continental slope is not accounted for. 2. Data [8] The locations of the repeat CTD sections in Drake Passage are shown in Figure 1. The WOCE CTD line SR1 has been repeated six times between 1993 and 2000 [Cunningham et al., 2003]. Here we used data from the first four sections. The usual station spacing on the CTD sections is 38 km, with tighter spacing of 5 10 km over steeply sloping bathymetry. All stations reached to within about 10 m of the seafloor, and tight spacing over the continental slope allows accurate estimates of geostrophic transport to be made. A discussion of the major circulation features along the SR1 section and baroclinic and barotropic transport variability is given by Cunningham et al. [2003]. [9] Additional repeat CTD sections across Drake Passage were carried out by Chile during WOCE (Figure 1). The Chilean sections are located to the west of the SR1 repeat sections. The Chilean line was repeated five times between 1993 and 1998; however, the first occupation was shallow 2of14

Table 1. Baroclinic Volume Transports on Four Occupations of the WOCE SR1 Line a Cruise U dcd U 2500 U 2500 /U dcd U N 2500 U deep U c U S 2500 1993 133.8 90.6 67.7 8.8 80.3 79.0 1.4 1994 142.2 95.6 67.2 1.5 92.5 90.8 1.5 1996 125.9 87.2 69.3 8.3 73.7 72.9 0.4 1997 147.4 97.3 66.0 3.6 91.7 91.1 1.7 mean ± std 137.3 ± 9.5 92.7 ± 4.6 67.6 ± 1.3 5.6 ± 3.6 84.6 ± 9.1 83.5 ± 9.0 1.1 ± 1.0 a The third column gives the ratio of the transport above and relative 2500 dbar (U 2500 ) to the total baroclinic volume transports (U dcd ) across SR1 relative to the deepest common depth (dcd). U N 2500 (U S 2500 ) is transport inshore of the 2500 m isobath at the north (south) of the section. U deep (U c ) is measured (estimated from c T h relationship) transport across the part of the section where the ocean depth is more that 2500 m. All transports are given in Sv. (to 1500 m) and located further west relative to other repeats, and was not used here. [10] We also used historical hydrographic sections in Drake Passage occupied in 1975 1976 by the FDRAKE program. Though the spacing between the stations at those sections was sparse, eight sections were spread across the Drake Passage region, and five of them are coincident with the Chilean CTD line. [11] The results obtained in Drake Passage were compared to the WOCE SR3 section occupied between Tasmania and Antarctica. The SR3 section was repeated six times between 1991 and 1996, with at least one section in each season of the year [Rintoul and Sokolov, 2001]. The usual station spacing on the CTD sections is 56 km, with tighter spacing across the sharp fronts of the ACC and over steeply sloping bathymetry. All stations reached to within about 10 m of the seafloor. A discussion of the major circulation features and water masses along the SR3 section is given by Rintoul and Bullister [1999] and Rintoul and Sokolov [2001], and frontal structure variability along the SR3 is discussed by Sokolov and Rintoul [2002]. [12] The Drake Passage climatology presented here is based on Olbers et al. [1992] Hydrographic Atlas of the Southern Ocean. We also used satellite altimeter measurements of sea surface height (SSH) anomalies (the Mean Sea Level Anomaly (MSLA) product from CLS/AVISO, which provides sea level anomalies every 10 days mapped on a 1/4 grid using the TOPEX/POSEIDON, ERS-1 and ERS-2 altimeters [Le Traon et al., 1998]). 3. Results 3.1. Testing the C T h Relationship in Drake Passage [13] Our interest is in baroclinic transport and so we define an empirical relationship between temperature, T, and the potential energy anomaly, c, asbyrintoul et al. [2002]. c is a stream function for the baroclinic transport: Up ð o Þ ¼ 1 r o f cðp o Þ ¼ 1 g Z po 0 pddp; @ @y c ð p oþ; Vðp o Þ ¼ 1 r o f ð1þ @ @x c ð p oþ ð2þ where d is the specific volume anomaly, g is gravity, p is pressure, p o and r o are the reference pressure and density respectively, and U (V) is the zonal (meridional) geostrophic volume transport per unit width, above and relative to p 0.In the following, we use c integrated to 2500 dbar (c 2500 ). At the SR3 line, Rintoul et al. [2002] used the 2500 dbar level because it is the deepest depth which still lies above the height of the mid-ocean ridge, so c 2500 is defined along the entire section. They showed that the transport above and relative to 2500 m is a large and nearly constant fraction of the total baroclinic transport across the section. The ratio of the transport above and relative to 2500 m to the total baroclinic transport relative to the deepest common depth at each station pair was estimated as 65.8 ± 2.1% derived from six occupations of the SR3 section. From four occupations of the SR1 section this ratio is also stable and very close to that estimated for the SR3 line: 67.6 ± 1.3% (or ±1.8 Sv; Table 1). [14] Figure 2 shows two examples of the relationship between c 2500 and T h, the temperature averaged over a 100 m interval centered on the depth h. For T 650 the largest scatter is observed at the southern (cold) and northern (warm) ends of the section. A tighter relationship is found when a temperature below 1000 m is used, although many XBTs do not extend beyond 800 m so a shallower relationship may need to be used in practice. A c 2500 reversal at low temperature (T <0.5 C) indicates weak westward baroclinic flow off the Antarctic continental slope. [15] The scatter in the relationship between temperature at depth h (T h ) and c 2500 depends on the choice of h (Figure 3). Root mean square (RMS) errors in the relationship were estimated in temperature bins from the scatter about a smoothing spline fit to the data (as shown in Figure 2). Three indications of the error are plotted: the maximum error in any temperature bin, the median error across all temperature bins, and the error in the northernmost temperature bin which determines the error in the net or total transport. (Note that, as shown later, the baroclinic transport variability is low at the southern end of the section and does not exceed 1 Sv). The respective RMS errors in volume transport (above and relative to 2500 m, in Sv) are shown in the right panel. The transport errors have been evaluated from the scatter in the c T h relationship. In general, the errors decrease with increasing depth h. If the temperature is taken near the surface, the median RMS error in baroclinic transport is about 8 Sv. At 600 m the median error is reduced by half. Below 1000 m the median transport error between station pairs is less than 1 Sv, and the error in net baroclinic transport is about 2 Sv. [16] The cumulative transport (integrated from south to north) above 2500 dbar predicted from the empirical relationship between c 2500 and T 1250 is compared to transports calculated from the full CTD data in Figure 4. The two estimates of transport are very similar, as expected from the error analysis described above. The RMS error in station pair transports ranges from 1.0 to 1.5 Sv for the four sections, or about 1% of the mean transport above and relative to 2500 m. In particular, each of the main transport 3of14

Figure 2. The relationship between baroclinic transport stream function c 2500 and temperature averaged between (left) 600 and 700 m depth and (right) 1200 and 1300 m depth. Data come from four occupations of the SR1 section. The curves are a smoothing spline fit to the data. Cool (warm) temperatures correspond to the southern (northern) end of the section. features (i.e., fronts) is well captured by the temperaturebased transport estimate. The errors are larger near the SAF, where the transports themselves are larger, and south of the Polar Front at 59 60 S, but otherwise vary randomly with latitude, with no sign of a systematic bias. The maximum error in the net transport across each section is 1.7 Sv, comparable to that in individual station pairs. [17] If we include all the available data in the Drake Passage region, the total errors become even smaller, the median errors remain about the same, and the maximum errors rise in the upper 400 m (Figure 5). The error in total transport decreases because the much larger set of hydrographic stations allows more statistically significant estimates of RMS error to be obtained, especially in the marginal bins. Maximum errors increase because the use of historical data from a wider area introduces additional scatter due to spatial and possibly temporal variability in T-S properties. The median and total errors are now about 2 Sv, if the temperature T h is taken below 700 m, and the maximum RMS error for h > 1000 m is around 3 Sv. 3.2. Comparison to WOCE SR3 Section [18] The c T h relationship at Drake Passage is similar, but not identical, to that found at SR3 (Figure 6).When temperature at a deep level (e.g., 1250 m) is used to define the relationship, the curves are very similar at SR1 and SR3 (Figure 6 (right)). When a depth of 650 m is used, the two curves diverge for temperatures warmer than about 2.5 C (Figure 6 (right)). The similarities and differences between the c T h relationships can be explained by comparing the potential temperature-salinity properties at the two locations (Figure 7). The deep water has relatively similar q S properties at SR1 and SR3. The q S curves diverge for temperatures warmer than 2.5 C and salinities less than about 34.40: at SR1, the water column becomes fresher at temperatures above 2.5 C, while at SR3 the salinity changes are more modest. As a result, density decreases by a larger amount for a given temperature change at SR1 than at SR3, and the slope of the c T 650 relationship is therefore larger as observed in Figure 6 (right). The point where the envelope of q S curves diverge (2.5 C, 34.40) coincides roughly with the SAF, reflecting the fact that the Subantarctic Mode Water (SAMW) and Antarctic Intermediate Water (AAIW) found north of the SAF in Drake Passage are significantly cooler and fresher than the SAMW and AAIW varieties found south of Australia. [19] A major difference between SR3 and Drake Passage is that two smoothing splines are required to fit the observations at SR3 (Figure 6). The curve covering the complete temperature range applies to all casts south of the Subtropical Front (STF). North of the STF the water masses have a different T-S structure, resulting in a different relationship between c and T h. The location of the STF, defined by temperatures at 150 m depth greater than 11 C 4of14

Figure 3. (left) Dependence of the RMS error in the c 2500 T h relationship on the depth h used to define T h. Estimates were obtained for the WOCE SR1 section. RMS errors were estimated in temperature bins from the scatter about the smoothing spline fit. Three estimates of the error are shown: the maximum error (corresponding to the maximum scatter in any temperature bin), the median error over all temperature bins, and the total or net error that depends on the scatter in the warmest or northernmost temperature bin. The 90% confidence intervals are provided for the median error estimates. (right) The corresponding volume transport errors (above and relative to 2500 m, in Sv) are shown. The transport errors have been evaluated from differences in c 2500 over a 1 change in latitude. [Rintoul and Bullister, 1999], is easy to determine from temperature observations alone, and so shifting from one curve to the other when crossing the front is straightforward. The STF does not pass through Drake Passage, so a second curve fit to subtropical waters is not required there. [20] For comparison we conducted a similar analysis (not presented by Rintoul et al. [2002]) of how the RMS fit changes with T h on the SR3 line. The dependence of the estimation error on depth h is broadly similar to that obtained in the Drake Passage region (Figure 8). However, the median and maximum RMS errors reduce with depth more sharply at SR3, and the median error is less than 2 Sv for temperatures taken at depths below 600 m. However, the total baroclinic transport error is larger than in Drake Passage and varies with depth: a minimum in the error (<3 Sv) for T h taken between 600 800 m, an increase to 4 Sv between 1000 and 1400 m, <1 Sv error at between 1500 and 1700 m, and an increase to >3 Sv below 1800 m. [21] This vertical structure in the total error distribution reflects the circulation pattern south of Tasmania: the westward flow at the northern end of the section is associated with two flows, an anticyclonic recirculation in the Subantarctic Zone (SAZ) and an outflow of water from the Tasman Sea [Rintoul and Sokolov, 2001; Sokolov and Rintoul, 2000]. The Tasman outflow is highly variable, and its T-S properties distinctly different from the water recirculating just north of the ACC in the SAZ. The largest differences occur at the core level of the AAIW situated at depths of 1000 to 1400 m, and the Circumpolar Deep Water (CDW) found at around 2800 m depth. The AAIW in the Tasman outflow is warmer, saltier, and its core level is located at greater depth (1300 1400 m) than the newer, fresher and cooler AAIW recirculating further south in the SAZ with a core depth of 1100 1200 m. Also the CDW in the SAZ is saltier and warmer than in the Tasman outflow, indicating its more recent Southern Ocean origin. As a result, the total RMS errors in the c 2500 versus T h relationship are higher at the core level of AAIW and CDW, because the water mass variability is largest at these depths. The errors are lower at the water mass interfaces, where the T-S properties of both flows are indistinguishable. [22] This example of a two-spline approximation south and north of the STF indicates that in regions where the water properties change (e.g., the transition from subant- 5of14

Figure 4. Comparison of cumulative transport (integrated from south to north) above 2500 dbar derived from the empirical relationship between c 2500 and T 1250 shown in Figure 2 (right) (thin line) and transports calculated from the full CTD data (thick line). Differences in transport at each station pair are shown along the x axis. arctic to subtropical regimes), the relationship between c 2500 and T h may also change. Additional scatter is introduced by temporal variability of inflows characterized by distinctly different T-S properties even at the deeper levels of intermediate and deep waters. Similar conclusions have also been drawn by Sun and Watts [2001], who found that the highest scatter in GEM fields in the Southern Ocean is in the Agulhas Retroflection region caused by energetic temporal variations in the volume of subtropical waters transported into the region by the Agulhas Current. 3.3. Spatial Variability of the C T h Relationship in Drake Passage [23] To explore how the relationship between upper ocean temperature and baroclinic transport function varies across the whole Drake Passage region, we used climatological data from Olbers et al. [1992]. We investigate the spatial scales over which the c T h relationship varies, how the spatial variability compares to temporal variability along a single line, and identify the processes contributing to the spatial variability in the relationship. Varying the depth h used to define T h indicates that all deviations (median, maximum and in the last temperature bin) from the mean curve are somewhat smaller than those obtained for individual hydrographic sections (not shown). The median RMS error in transport is near or less than 2 Sv at all depths below 400 m. The maximum error is also reduced, especially for the temperature in the upper 1000 m. For h between 600 and 1200 m depth the median and total errors in baroclinic transport both are about 1 Sv. 6of14

Figure 5. Dependence of the RMS error in c 2500 versus T h relationship on depth h, as in Figure 3, here estimated from all the sections in the Drake Passage region shown in Figure 1. [24] If we take the climatological temperature field at some depth h where the regional deviations in c from the mean curve are small, we can calculate the difference between the observed climatological stream function (c clim ) and that estimated from the c T h relationship (c Th ). The spatial pattern of these anomalies, c 0 2500 = c clim c Th, will correspond to the areas where the local T-S properties diverge from the average T-S curves along the baroclinic transport streamlines passing through the Drake Passage region. For example, Figure 9 shows that anomalies from the mean c T 750 relationship are generally small throughout the Drake Passage region. The expected median error in volume transport within the Drake Passage region due to regional deviations of the T-S relationship along the streamlines is also small (less than 1 Sv, not shown). The maximum error in the volume transport does not exceed 2 Sv. However, a large positive anomaly in c 2500 of more than 0.1 10 7 observed outside the Drake Passage in the southern Argentine Basin (north of 48 S, east of 53 W, Figure 9) indicates that T-S properties in that region are distinctly different from the Drake Passage, and the c T h relationship obtained for the Drake Passage would fail there. [25] Peterson and Whitworth [1989] were probably the first who traced in detail the full path of the SAF east of the Drake Passage. Using observations made in 1979 1980 from R/V Atlantis II they showed that the Falkland Current is the northward extension of the SAF and it can be traced to the Brazil-Falkland confluence zone and then southward to the southern Argentine Basin. They found that the CDW advected from the south over the Falkland Plateau into the Argentine Basin is joined by the North Atlantic Deep Water (NADW) along the seaward side of the Falkland Current return. The NADW, the parent water mass of the Lower Circumpolar Deep Water (LCDW), has higher maximum salinity values (greater that 34.8) comparing to LCDW. The core salinity values of the LCDW in the SAF in the Drake Passage do not exceed 34.72 [see, e.g., Garsía et al., 2002]. The AAIW along this route also gets progressively saltier because of steady erosion of the AAIW minimum [Meredith et al., 1999]. [26] In Figure 10 we plotted T-S properties in the upper 1000 m along the c clim =1.7 10 7 contour (the contour is shown in Figure 9d) passing through the region. Within the Drake Passage the contour follows the Polar Front Zone (PFZ) between the SAF and Polar Front (PF), and runs through the strongest positive and negative anomalies in the Drake Passage. Positions of the SAF and PF derived using the temperature contours of 4 and 2 C at 200 m depth are shown in Figure 9a, and follow very closely the representation of the fronts given by Peterson and Whitworth [1989] and Orsi et al. [1995]. Outside the Drake Passage region in the southern Argentine Basin the T-S properties along the contour change dramatically. There the contour coincides with the SAF which is 7of14

Figure 6. A comparison of the c 2500 T h relationship at the SR3 section south of Australia (thick line) to the SR1 section in Drake Passage (dashed line). At SR3, different water mass properties north and south of the Subtropical Front require two splines to be fitted, one for profiles on either side of the front (see text for further explanation). The average temperature between (left) 600 and 700 m and (right) 1200 and 1300 m is used to define T h. Figure 7. The potential temperature ( C) versus salinity relationship for the Drake Passage region (blue curves) and WOCE SR3 section south of Australia (red curves). Voids in q S space correspond to fronts across which water mass properties change. 8of14

Figure 8. Distribution of RMS error in c 2500 versus T h relationship as in Figure 3. Estimates were obtained for the WOCE section SR3 south of Australia. Data come from six occupations of the SR3 section. manifested in strong increase in temperature near the sea surface in comparison to the Drake Passage. The subsurface salinity minimum corresponding to the core of the AAIW is located at a depth of 300 m there, with salinities much greater than those of Antarctic Surface Water (AASW) found along the contour in the PFZ of the Drake Passage. At the core level of the deep salinity maximum, the LCDW along the contour in the Drake Passage region is significantly fresher (S = 34.73) and the core is deeper (z core = 2800 m) than of the modified LCDW found in the SAF in the southern Argentine Basin (S = 34.78, z core = 2400 m; not shown). [27] Within the Drake Passage the c 2500 anomalies are small, and T-S properties are similar along the streamlines. For example, T-S anomalies on neutral density surfaces along the c clim =1.7 10 7 contour in the Drake Passage do not exceed 0.1 C in temperature and 0.01 in salinity in the water column below the AASW (not shown). A sequence of negative (positive) anomalies in c 2500 along this contour follows closely the depth changes of the interface between the AASW and Upper Circumpolar Deep Water (compare the depth of 2.75 C contour and c 0 2500 shown in Figure 10). Since we used the temperature at 750 m to determine c Th, higher (lower) than average temperatures at the depth of 750 m result in higher (lower) c Th. Note that by definition the depth-integrated density above 2500 m is constant along the c clim contour: small density anomalies at the depth h must be compensated by density anomalies somewhere else in the water column. 3.4. Deriving Baroclinic Transport From XBT Data [28] We have shown that using the c T h relationship derived in Drake Passage from individual hydrographic sections, the baroclinic transport relative to and above 2500 db could be estimated from the XBT data with an error of less than 2 Sv. Spatial deviations from the mean relationship across the region are also small, with median and maximum errors in transport of about 1 Sv if the temperature is taken at depths between 600 and 1200 m. However, this technique for estimating baroclinic transport from subsurface temperature is applicable only for the part of the section where the ocean depth is greater than 2500 m, and some transport would be missed inshore of the 2500 m isobath. [29] Taking the WOCE SR1 section as an example, consider the uncertainty in total transport attributed to the transport variability inshore of the 2500 m isobath. The meridional distribution of mean baroclinic transport across the SR1 section derived from four repeats between 1993 and 1997 is presented in Figure 11. The transport distribution clearly shows a broad peak of strong eastward flow between 55 S and 57 S associated with the main jet of the ACC (the Subantarctic Front, SAF), the Polar Front jet between 57 S and 58 S, and three weaker transport maxima 9of14

Figure 9. (a) Temperature at 750 m in the Drake Passage region; (b) observed climatological c clim field; (c) c T750 derived from the c 2500 T 750 relationship and the temperature field in Figure 9a in the box limited by 53 64 S, 54 70 W; (d) the difference between the observed and estimated c 2500 fields c 0 2500 = c clim c T750. Depths less than 3000 m depth are shaded. Climatological temperature and c fields are derived from the Olbers et al. [1992] climatology. The c clim =1.7 10 7 contour is shown in Figure 9d. FC is Falkland Current; FP is Falkland Plateau; and AB is Argentine Basin. south of 58 S. The prominent maxima at the southern end of the section at 60.7 S corresponds to the southern ACC front. [30] The meridional distribution of variability in the cumulative transport integrated from the south of the section (Figure 11) indicates that the largest standard deviations in baroclinic transport are found in the Polar Front and in the SAF. Variability south of 59 S is weak (standard deviations less than 3 Sv). Variability in integrated transport in the SAF and further north is almost constant (of about 10 Sv) indicating strong variability at the northern end of the section. (Note that these standard deviations in integrated transport calculated in latitude bins reflect both shifts of front location and changes in current strength). While during the 1994 and 1997 occupations of SR1 the northern extent of the SAF was located at 55.8 S and at 55.2 S respectively, in the 1993 section the SAF was observed much further north, attached to the continental slope. In the 1996 section an even more energetic SAF jet was also found north of the mean location, with a large portion of the flow found inshore of the 2500 m isobath (Figure 12). 10 of 14

Figure 10. (a) c 2500 anomaly along the c clim =1.710 7 contour shown in Figure 9d. Vertical sections of water properties along the contour: (b) temperature (contour interval 0.25 (1.0) for temperatures below (above) 4 C) and (c) salinity (contour interval 0.05). 11 of 14

Figure 11. (a) Mean volume transport per unit width (so area under the curve is equal to transport, area of scale bar = 10 Sv) along the WOCE SR1 line. (b) The standard deviation of cumulative transport integrated from south to north along the SR1 line. The SR1 mean transports are estimated from four repeats of SR1. [31] The repeat occupations of SR1, with their very tight station spacing over the steep continental slopes, allow estimation of the likely transport errors if flow inshore of the 2500 m isobath is not resolved. These transport components and their variability estimated from four repeats of SR1 line are presented in Table 1. As expected from Figure 11, the mean transport and its variability inshore of the 2500 m isobath at the southern end of the section is small (1.1 ± 1.0 Sv). At the northern end, the inshore transport of 5.6 Sv is significant, and its standard deviation of 3.6 Sv is large compared to the total transport variability of 4.6 Sv across the whole section. Furthermore, because the main ACC jet was located to the north of its mean position at SR1 in the 1993 and 1996 sections, and close to the continental slope, part of the SAF flow was inshore of the 2500 m isobath (Figure 12). As a consequence, the transport between the 2500 m isobaths is biased low relative to the true net baroclinic transport, and the apparent variability (9.1 Sv) is twice as large as the observed variability in net transport. This fact must be taken into account when designing observational programs in the region, as well as when interpreting historical transport estimates based on hydrographic sections with poor spatial resolution. [32] Transport over the northern continental slope could be estimated using a number of c z T h approximations where c z is integrated to a set of depths z less than 2500 m. Then the transport observed inshore of the 2500 m isobath could be calculated in the traditional way, relative to the deepest common depth at each station pair. However, the larger uncertainty in the spline fit at the limits of the observed temperature range makes it difficult to apply this approach. Larger transport errors also will arise for the shallow regions because T h would be taken at shallower depths, where the relationship is less tight and approximation errors in transport become comparable with transport variability (Figures 2, 3, and 5). Accurate baroclinic transport estimates over the continental slope probably require direct measurements of temperature and salinity profiles from ships or moorings. 4. Summary [33] We have explored a proxy technique for estimating baroclinic transport from subsurface temperature and its applicability in two regions of the Southern Ocean where high-quality repeat hydrographic sections are available. The upper ocean temperature versus baroclinic stream function relationship produced reliable transport estimates along the SR1 line with an error of less than 2 Sv, as evaluated by comparing to transport estimates calculated using the complete CTD data. The method performs well along the entire latitude band spanned by the SR1 section, with median errors of about 1 Sv at individual station pairs and with maximum errors less than 4 Sv. The mean relationship also works well across the whole Drake Passage region. Nevertheless, while errors associated with the spline approximation of the c T h relationship are small, significant transport observed inshore of the 2500 m isobath at the northern side of Drake Passage adds additional uncertainty to estimates of the net transport and its variability if flow over the continental slope is not accounted for. [34] We also found that in regions where the water properties change (e.g., from subantarctic to subtropical characteristics), the relationship between c 2500 and T h also changes. Additional scatter can be introduced by inflows of water with different T-S properties, even at the level of intermediate and deep waters, and thus a different relationship may be needed in such regions. However, as indicated by the analysis of the spatial variability of the c T h relationship in the Drake Passage region, this method can be used as a valuable tool to reveal anomalous T-S properties at various depths with respect to the baroclinic stream function field. [35] The proxy method explored here has much in common with the GEM parametrization technique developed by 12 of 14

Figure 12. Geostrophic velocity (relative to the deepest common depth, in m/s) across the northern end of the SR1 section in (a) 1993, (b) 1994, (c) 1996, and (d) 1997. Positive velocities are to the east. Sun and Watts [2001]. The GEM parametrization allows estimates of T-S profiles to be made from integral properties like dynamic height, while the c T h relationship is used to estimate baroclinic stream function field (or c) from temperature profiles. Both methods rely on the relative stability of the T-S relationship along streamlines, a property that appears to hold for the circumpolar extent of the ACC. References Cunningham, S. A., S. G. Alderson, B. A. King, and M. A. Brandon (2003), Transport and variability of the Antarctic Circumpolar Cur- 13 of 14

rent in Drake Passage, J. Geophys. Res., 108(C5), 8084, doi:10.1029/ 2001JC001147. Garsía, M. A., I. Bladé, A. Cruzado, Z. Velásquez, H. Garsía, J. Puigdefàbregas, and J. Sospedra (2002), Observed variability of water properties and transports on the World Ocean Circulation Experiment SR1b section across the Antarctic Circumpolar Current, J. Geophys. Res., 107(C10), 3162, doi:10.1029/2000jc000277. Killworth, P. D. (1992), An equivalent-barotropic mode in the Fine Resolution Antarctic Model, J. Phys. Oceanogr., 22, 1379 1387. Le Traon, P. Y., F. Nadal, and N. Ducet (1998), An improved mapping method of multisatellite altimeter data, J. Atmos. Oceanic Technol., 15, 522 534. Meredith, M. P., K. E. Grose, E. L. McDonagh, K. J. Heywood, R. D. Frew, and P. F. Dennis (1999), Distribution of oxygen isotopes in the water masses of Drake Passage and South Atlantic, J. Geophys. Res., 104, 20,949 20,962. Olbers, D., V. Gouretski, G. Seiß, and J. Schröter (1992), Hydrographic Atlas of the Southern Ocean, 17 pp. + 82 plates, Alfred Wegener Inst. for Polar and Mar. Res., Bremerhaven, Germany. Orsi, A. H., T. Whitworth III, and W. D. Nowlin Jr. (1995), On the meridional extent and fronts of the Antarctic Circumpolar Current, Deep Sea Res., Part I, 42, 641 673. Peterson, R. G., and T. Whitworth III (1989), The Subantarctic and Polar Fronts in relation to deep water masses through the southwestern Atlantic, J. Geophys. Res., 94, 10,817 10,838. Rintoul, S. R., and J. L. Bullister (1999), A late winter hydrographic section from Tasmania to Antarctica, Deep Sea Res., Part I, 46, 1417 1454. Rintoul, S. R., and S. Sokolov (2001), Baroclinic transport variability of the Antarctic Circumpolar Current south of Australia (WOCE repeat section SR3), J. Geophys. Res., 106, 2815 2832. Rintoul, S. R., S. Sokolov, and J. A. Church (2002), A 6 year record of baroclinic transport variability of the Antarctic Circumpolar Current at 140 E derived from XBT and altimeter measurements, J. Geophys. Res., 107(C10), 3155, doi:10.1029/2001jc000787. Sokolov, S., and S. R. Rintoul (2000), Circulation and water masses along WOCE section P11: Papua New Guinea to Tasmania, J. Mar. Res., 58, 223 268. Sokolov, S., and S. R. Rintoul (2002), Structure of Southern Ocean fronts at 140 E, J. Mar. Syst., 37, 151 184. Stommel, H. (1947), Note on the use of the T-S correlation for dynamic height anomaly computations, J. Mar. Res., 6, 85 92. Sun, C., and D. R. Watts (2001), A circumpolar Gravest Empirical Mode for the Southern Ocean hydrography, J. Geophys. Res., 106, 2833 2856. Watts, D. R., and H. T. Rossby (1977), Measuring dynamic heights with inverted echo sounders: Results from MODE, J. Phys. Oceanogr., 7, 345 358. Watts, D. R., C. Sun, and S. R. Rintoul (2001), A two-dimensional Gravest Empirical Mode determined from hydrographic observations in the Subantarctic Front, J. Phys. Oceanogr., 31, 2186 2209. Whitworth, T., III (1983), Monitoring the net transport of the Antarctic Circumpolar Current at Drake Passage, J. Phys. Oceanogr., 13, 2045 2057. Whitworth, T., III, and R. G. Peterson (1985), The volume transport of the Antarctic Circumpolar Current from three-year bottom pressure measurements, J. Phys. Oceanogr., 15, 810 816. B. King, Southampton Oceanography Centre, Southampton SO14 3ZH, UK. (brian.a.king@soc.soton.ac.uk) S. R. Rintoul and S. Sokolov, CSIRO Marine Research and Antarctic Climate and Ecosystems CRC, GPO Box 1538, Hobart, Tasmania 7001, Australia. (steve.rintoul@csiro.au; serguei.sokolov@csiro.au) R. L. Rojas, Centro Nacional de Datos Oceanogrficos de Chile (CENDOC), Errazuriz 232, Playa Ancha, Cholesteremia Hidrografico y Oceanografico de la Armada (SHOA), Casilla 324, Valparaiso, Chile. (rrojas@shoa.cl) 14 of 14