Repeat hydrographic sections occupied in 1995 and 2005 reveal a rapid decline in the salinity and density of Antarctic Bottom Water throughout the Australian Antarctic Basin. The basin-wide shift of the deep potential temperature-salinity (θ − S) relationship reflects freshening of both the Indian and Pacific sources of Antarctic Bottom Water. The θ − S curves diverge for waters cooler than −0.1°C, corresponding to a layer up to 1000 m thick over the Antarctic continental slope and rise. Changes over the last decade are in the same direction but more rapid than those observed between the late 1960s and the 1990s. When combined with recent observations of similar freshening of North Atlantic Deep Water, these results demonstrate that dense water formed in both hemispheres is freshening in response to changes in the high latitude freshwater balance and rapidly transmitting the signature of changes in surface climate into the deep ocean.
 The thermohaline (or overturning) circulation is the dominant mechanism responsible for ocean heat transport and therefore has a strong influence on global and regional climate patterns. Climate models [Cubasch et al., 2001] and studies of past climate [Dansgaard et al., 1993; Broecker, 1997] suggest the climate system is sensitive to changes in the strength of the thermohaline circulation. The possibility that increased freshwater input to the high latitude ocean could cause a slowing of the thermohaline circulation, driving an abrupt change in climate, has attracted considerable interest [Alley et al., 2003]. Most attention has focused on the North Atlantic, where a significant decrease in the salinity of North Atlantic Deep Water has been observed during the past four decades [e.g., Dickson et al., 2002]. The Southern Ocean is also an integral part of the thermohaline circulation: the circumpolar channel of the Southern Ocean allows a global-scale circulation to exist; southern and northern sources of dense water make similar contributions to the ventilation of the deep ocean [Orsi et al., 2002]; and water mass transformations in the Southern Ocean connect the deep and shallow limbs of the overturning circulation [Speer et al., 2000; Rintoul et al., 2001]. Evidence from repeat hydrographic sections is presented here to show that the Antarctic Bottom Water (AABW) that forms the southern arm of the deep overturning circulation is also freshening rapidly.
 The Weddell Sea is the largest source of AABW, supplying 60% of the total [Orsi et al., 2002]. The remainder is split between the Ross Sea and the Mertz Polynya region of the Adélie Land coast near 145°E [Orsi et al., 2002; Rintoul, 1998]. Both of the latter sources drain into the Australian Antarctic Basin (Figure 1), making it the best ventilated of the deep basins surrounding Antarctica, as is evident from the large inventory of chlorofluorocarbons observed in the bottom water there [Orsi et al., 1999, 2002]. Measurements of bottom water properties in this basin can therefore be used to monitor changes in the Indian and Pacific sources of AABW.
 Earlier studies have used bottle data to document changes in the AABW of the Australian Antarctic Basin. Whitworth  compared deep T-S properties throughout the basin and concluded there was a shift from a salty mode to a fresh mode of bottom water around 1994. Jacobs [2004, 2006] examined the region between 140°E and 150°E and south of 65°S and concluded that salinity had declined between 1950 and 2004. However, these previous studies aggregated data from broad regions and from all seasons and are potentially sensitive to aliasing of spatial and seasonal variability. For example, most of the recent stations used by Whitworth  are closer to the continent, and nearer the bottom water sources, than the earlier data. In the region considered by Jacobs [2004, 2006], bottom water properties vary strongly with season [Fukamachi et al., 2000] and with location, reflecting the narrow bottom currents and rough bathymetry over the continental rise and slope [e.g., Chase et al., 1987]. Aoki et al.  avoided the aliasing problem by only examining differences between co-located stations from the same season, but they considered only a single longitude. In the present study the same approach is taken with a new set of repeat hydrographic sections to provide the first basin-wide view of bottom water changes, free of the ambiguity caused by spatial and temporal aliasing.
 Changes in deep ocean properties are assessed using reoccupations of hydrographic stations made a decade earlier during the World Ocean Circulation Experiment (WOCE, Figure 1). (The WOCE lines used include I9S, I8S, SR3, P11A, and S4; these data and older historical observations used here are available from Orsi and Whitworth .) The repeat sections were carried out in the same season and at the same locations as the original stations. The sections cut across the outflow of new AABW at five locations along a 5000 km long path: two of the sections are located immediately downstream of the two source regions, allowing changes in the individual sources to be assessed, and the remaining three are located at increasing distance along the system of deep boundary currents carrying new AABW away from the formation region. In some cases, relatively high quality measurements of temperature and salinity were also made at nearby locations in the late 1960s and early 1970s [Orsi and Whitworth, 2005], allowing comparison to an earlier snapshot of bottom water properties.
Figure 2a shows changes in the potential temperature-salinity (θ − S) relationship at WOCE section I9S (nominally 115°E). Over the continental rise (61°S – 63.3°S), the densest water was fresher (by 0.018) and less dense (by 0.005 kg m−3) in January 2005 than in January 1995. The θ − S curves in 2005 were offset to the fresh side of the 1995 curves for all waters cooler than −0.1°C (neutral density anomaly greater than 28.30 kg m−3), corresponding to a wedge of water up to 1000 m thick over the continental slope and rise. For water warmer than −0.1°C and for stations north of the mid-ocean ridge, the θ − S profiles for the two cruises closely agree, confirming that the observed differences are not due to a shift in calibration. Water of the same density was significantly cooler and fresher in 2005 (eg by 0.09°C and 0.016 at γn = 28.34 kg m−3). A similar signal is observed in the deep basin further offshore (Figure 2b), where the densest water was 0.008 fresher and 0.012 kg m−3 lighter in 2005 than in 1995. The decrease in salinity and density of the AABW over the last decade continues a shift toward fresher and lighter bottom water observed between 1970 and 1995 (Figure 2b) [Whitworth, 2002]. The rate of change in salinity and density increased in the last decade: the rate of freshening between 1995 and 2005 is about 30% greater than that observed between 1970 and 1995.
 Freshening of AABW is also observed further downstream from the source regions. In the Princess Elizabeth Trough at 80°E, θ − S curves from 1971 were characterised by a salinity maximum at the bottom and maximum γn of 28.35 kg m−3 (Figure 2c). In the mid-1990s, the deep part of the θ − S profile (γn > 28.30 kg m−3) was nearly isohaline at a salinity between 34.670 and 34.675 and the near-bottom waters were warmer and lighter than observed in 1971. In 2005, water near the sea floor was warmer, fresher and less dense than observed in the mid-1990s. A similar pattern of changes is observed in the deep western boundary current carrying dense AABW northward east of the Kerguelen Plateau (Figure 2d). A consistent change in shape of the θ − S profile (from a salinity maximum at the bottom in 1970–71 to a salinity minimum at the bottom in 2005) is found in each region and provides a diagnostic of water mass change which is insensitive to any shift in salinity calibration between the recent and historical data.
 The changes in bottom water properties reflect changes in the sources of dense water in the Ross Sea and Mertz Polynya. Ross Sea shelf waters freshened by more than 0.1 between 1963–1978 and 2000 [Jacobs et al., 2002]. Fresher shelf waters would be expected to result in fresher Ross Sea Bottom Water (RSBW), which is made up of about 35% shelf water [Orsi et al., 1999]. Comparison of bottom water properties just downstream of the outflow from the Ross Sea shows that the salinity of new RSBW decreased between the late 1960s and the early 1990s (Figure 3a). Further downstream at 150°E, the salinity of the RSBW has been reduced by mixing with surrounding fresher waters, although a clear salinity maximum (∼34.71) is still present near the sea floor. This maximum is less prominent in the 1990s and the densest water is significantly lighter than observed in the 1960s (Figure 3b). Between 128°E and 140°E, immediately downstream of the Mertz Polynya source, a weak salinity maximum (∼34.70) at the bottom was still observed in the 1960s [Gordon and Tchernia, 1972] (Figure 3c). In the 1990s, the shape of the deep θ − S profiles was very different: a distinct salinity minimum at the bottom was observed at every station and the density of water near the sea floor was lower than observed in the earlier period. Figure 3 suggests that both the Ross Sea and Mertz sources produced fresher and less dense bottom water between 1992 and 1996 compared to the period between 1967 and 1971. The freshening observed since 1995 in the western basin (Figure 2) and in a series of repeat sections between 1994 and 2003 at 140°E [Aoki et al., 2005] suggests the freshening trend is continuing to the present time.
 Comparison of repeat sections shows a rapid and basin-wide shift in bottom water properties has occurred in the Australian Antarctic Basin. Bottom water was fresher and less dense in 2005 than in the mid-1990s; changes of similar magnitude and sign were observed between the early 1970s and mid-1990s. The largest changes were observed near the two bottom water formation regions, but changes were detected throughout the basin. The decrease in bottom salinity of 0.018 in ten years observed in the densest AABW at 115°E is comparable to the freshening found in the deep North Atlantic at similar distances downstream of the dense outflows [Dickson et al., 2002; Jacobs, 2006]. Freshening of both the Ross Sea and Adélie Land sources of dense water have contributed to the observed basin-wide shift in the θ − S curve. While the AABW is fresher and lighter in the most recent observations, there is no evidence of a change in rate of bottom water production. Oxygen concentrations in the bottom water at 115°E were marginally (∼5 μmol kg−1) higher in 2005 than in 1995, suggesting perhaps a slight increase in the supply of oxygen-rich bottom water to the deep ocean or a small change in the proportion of the high oxygen end member in the mixture forming AABW.
 Studies of southern hemisphere climate variability and change are severely hampered by the lack of sustained observations. The three hydrographic snapshots used here are clearly insufficient to distinguish between a trend and poorly-sampled high frequency variability. However, the monotonic freshening observed in the time series of sections at 140°E [Aoki et al., 2005] (eight sections, sampling five out of ten summers between 1994 and 2003), and the fact that a consistent signal is observed around one quarter of the circumpolar belt, supports the argument that the changes described here reflect a sustained freshening of bottom water, rather than aliased interannual variability.
 The most likely cause of the observed freshening is an increase in the influence of meltwater from continental ice [Jacobs, 2006]. For example, Jacobs et al.  showed that the magnitude and oxygen isotopic signature of the freshening of Ross Sea shelf water requires an inflow of glacial melt from the Amundsen Sea, where rapid basal melt rates have been inferred [Rignot and Jacobs, 2002; Shepherd et al., 2004; Zwally et al., 2005]. While the small number of repeated stations in the Australian Antarctic Basin makes it difficult to determine how much freshwater input is required to explain the freshening, a rough estimate can be made to check whether it is plausible that glacial melt can supply the freshwater required. Using conservation of salt, ρ1S1H = ρ2S2(H + F), where ρ is density, S is the average salinity in a layer of thickness H, F is the height of freshwater added or removed per unit area, and the subscripts refer to observations at two time periods. The repeat sections at 80°E, 115°E and 140°E were used to calculate the change in the average salinity in a 500 m thick layer above the bottom between 1995 and 2005, within 500 km of the 3000 m isobath. (Most of the freshening has occurred in waters denser than γn = 28.32 kg m−3 (e.g., Figure 2), which lies about 500 m above the sea floor.) A freshwater input of 13 ± 5 Gt a−1, where the error bar reflects uncertainty in the interpolation between sparse observations, is required to explain the freshening observed between 1995 and 2005 in the Australian Antarctic Basin (80°E to 140°E). This value reflects additional freshwater supplied by both the Mertz and Ross Sea sources. The freshwater input is roughly compatible with estimates of a mass loss of 11.94 ± 2.7 Gt a−1 for floating ice in the Adelie Land sector between 1992 and 2002 [Zwally et al., 2005] and freshening of Ross Sea shelf waters of 46 Gt a−1 as the result of inflow of melt from glaciers to the east [Jacobs et al., 2002; Shepherd et al., 2004].
 Widespread warming of the upper Southern Ocean has been observed in recent decades [Gille, 2002; Levitus et al., 2005] and enhanced melt of floating glacial ice has been attributed to this ocean warming [Rignot and Jacobs, 2002; Shepherd et al., 2004]. The temperature changes, in turn, are likely linked to changes in high latitude winds in the southern hemisphere [Hall and Visbeck, 2002]. The Southern Annular Mode (SAM) is the dominant mode of variability of the southern hemisphere atmosphere [Thompson et al., 2000]. Over the past thirty years, a positive trend in the SAM has resulted in a strengthening and poleward shift of the westerly winds at high southern latitudes [Marshall et al., 2004; Thompson and Solomon, 2002]. The trend in the winds is expected to drive stronger Ekman transport and upwelling of relatively warm deep water, a southward shift and strengthening of the Antarctic Circumpolar Current, and a poleward expansion of warm and moist atmospheric conditions [Hall and Visbeck, 2002]. Enhanced upwelling of Circumpolar Deep Water and a southward shift of the ACC would both act to bring warmer water into contact with the base of floating glacial ice and drive enhanced melt [Jacobs, 2006]. The trend in the SAM has been attributed to both decreases in stratospheric ozone [Thompson and Solomon, 2002] and to enhanced greenhouse warming [Fyfe et al., 1999; Kushner et al., 2001]. Natural variability of the coupled ocean-atmosphere-ice system may also have contributed. While the cause of the observed trend in high latitude climate is not yet known with certainty, the freshening and decrease in density of AABW in recent decades during a period of increasingly positive SAM may serve as an analogue for the response to future climate change, given that climate models suggest the atmospheric response to increasing greenhouse gases projects strongly onto the SAM [Kushner et al., 2001].
 Repeated oceanographic observations suggest that dense water formed in the Indian and Pacific sectors of the Southern Ocean is significantly fresher and lighter in 2005 than observed in 1995 at the same locations and season. The rate of change is similar in sign but larger in magnitude than observed between the late 1960s and the 1990s. The decline in salinity is comparable in magnitude and rate to changes observed in the North Atlantic during this time period. These observations demonstrate that changes in the high latitude freshwater balance are affecting the properties of dense water sinking in both the northern and southern limbs of the global overturning circulation.
 This study is a contribution to the CSIRO Climate Change Science Program. The work was supported by the Australian Government's Cooperative Research Centres Programme through the Antarctic Climate and Ecosystems Cooperative Research Centre and by the CSIRO Wealth from Oceans National Research Flagship. I thank Louise Bell for assistance with Figure 1, Mark Rosenberg and the captain and crew of RSV Aurora Australis for their efforts at sea, and two anonymous reviewers for their comments.