Royal Society Publishing

Observations of change in the Southern Ocean

Stan Jacobs


The Southern Ocean has been in a state of disequilibrium with its atmosphere and cryosphere during recent decades. Ocean station and drifting float observations have revealed rising temperatures in the upper 3000 m. Salinity has declined in intermediate waters and more rapidly in the sparsely sampled high latitudes. Dissolved oxygen levels may also have decreased, but measurement accuracy is inconsistent. Sea ice area increased from 1979 to 1998, particularly in the Ross Sea, while a decline in ice extent since the early 1970s has been led by the Amundsen–Bellingshausen sector. Fresher waters with lower oxygen isotope content on the Pacific–Antarctic continental shelf are consistent with increased melting of continental ice. Newly forming bottom water has become colder and less salty downstream from that region, but generally warmer in the Weddell Sea. Many ice shelves have retreated or thinned, but others have grown and no trend is apparent in the large iceberg calving rate. Warming and isotherm shoaling within the polar gyres may result in part from changes in the Southern Annular Mode, which could facilitate deep-water access to the continental shelves. Sea-level rise over the past half century has a strong eustatic component and has recently accelerated. Observations over longer periods and with better spatial coverage are needed to better understand the processes causing these changes and their links to the Antarctic ice sheet.


1. Introduction

Change in the Southern Ocean can take many forms, and over many spatial and temporal scales. This brief review focuses mainly on thermohaline changes over the last half century, variability longer than a decade and observations at high latitudes. Temperature is the most easily and accurately measured property, followed by salinity, and both have experienced substantial trends in different regions. Dissolved oxygen, carbon dioxide and transient tracer concentrations are less well documented, and biological variability will not be described here. Changes in atmospheric forcing are considered in relation to potential impacts on sea ice and ocean circulation. Ice shelf and iceberg attrition influence seawater thermohaline properties and recent analyses of those characteristics have altered some perceptions of the ongoing sea-level rise (SLR) enigma.

2. Temperature

Using annually gridded and objectively analysed historical ocean temperatures, Levitus et al. (2000) were the first to fully quantify recent warming of the world ocean. They reported changes in heat content, documenting an increase of ca 20×1022 J between the surface and 3000 m from the mid-1950s to the mid-1990s. Decadal variability in the southern oceans (figure 1) correlated positively with, but explained a lower percentage of the variance than in the corresponding northern oceans. A global volume mean temperature increase of 0.31 °C in the upper 300 m accounted for about half the total increase in heat content. Levitus et al. indicated that the observed warming was likely a combination of natural variability and anthropogenic effects and noted the sparse data coverage in the Southern Ocean.

Figure 1

Heat content anomalies in the upper 3000 m of each of the southern oceans, combined for the Southern Hemisphere, and based on 5 year running composites. Vertical lines represent ±1 standard error for the 5 year estimates, and linear trends apply to the 1955–1996 period. Modified from fig. 4 in Levitus et al. (2000).

A larger warming was reported at mid-depth in the Southern Ocean by Gille (2002, 2003), based on comparisons between observations by autonomous floats drifting between 700 and 1100 m and earlier ocean station data (figure 2). Considerable variability was evident, but many of the more positive temperature changes were recorded at latitudes associated with the Antarctic Circumpolar Current (ACC). Little data were available south of the ACC in regions with seasonal sea ice. Some of the floats would also have been drifting in Antarctic Intermediate Water that had originated near the sea surface between the Polar and Subantarctic Fronts, where air temperature increases occurred over the same period (Jacka & Budd 1998). Gille noted that some portion of the apparent warming of ca 0.2 °C since the 1950s could have resulted from a southward shift of the ACC, consistent with an earlier inference from deep measurements in the Southeast Pacific (Swift 1995). Because isotherms and isopycnals rise southward across the ACC, ocean variables compared at specific locations can represent dynamic as well as property changes (Bindoff & McDougall 2000). Aoki et al. (2003) provided another example at depths of 200–900 m in a 400×1600 km zonal band around 70 °E, 62 °S, where warming over three decades suggested that the Southern ACC Front had shifted southward.

Figure 2

Differences between 900 m temperatures recorded by drifting floats during the 1990s and prior ocean atlas data. The latter were at least 10 years earlier, but after 1930 and less than 220 km from the float data. Gaps indicate regions without comparable observations or with formal errors greater than one standard deviation. This is fig. 3a in Gille (2003), reproduced by permission of the American Meteorological Society.

Updating the world ocean warming situation, Levitus et al. (2005) extended the record into 2003 and made use of an additional 1.7 million temperature profiles from earlier years. That resulted in a significant downward adjustment of ocean heat absorption, to 14.5×1022 J (figure 3), but clearly demonstrated that the ocean is carrying by far the largest share of the observed warming, as anticipated (e.g. Hansen et al. 1997). More than 80% of the Earth's thermal imbalance has been absorbed by the ocean, even after other heat sinks are maximized. For example, a large circumpolar reduction in Antarctic sea ice extent during this period seems untenable (Ackley et al. 2003). Also, the heat allocated to melt continental ice, 0.8×1022 J, is the value that would apply if the full SLR of 1.8 mm yr−1 had resulted only from melting of the Antarctic and Greenland ice sheets. By most accounts the role of those ice sheets is considerably smaller, if not zero (Church et al. 2001; Zwally et al. 2005).

Figure 3

Estimates of Earth's heat balance components (1022 J) for the 1955–1998 period, modified from Levitus et al. (2005). This world ocean heat absorption applies for pentads from 1955–1959 to 1994–1998 and from 0–3000 m. About three quarters of the ocean heat content increase was in the upper 700 m and about half in the Southern Hemisphere.

Levitus et al. (2005) also reported changes in ocean heat content vs. time for 0–300, 0–700 and 0–3000 m integrations, finding more than half of the total in the Atlantic sector. Vertical sections of the upper 1500 m, for which the southern hemisphere values are on the left-hand side of figure 4, show a strongly positive trend centred near the sea surface around 40 °S. This area lies above the ca 900 m depth of the observations in figure 2, and may depict surface warming of Subantarctic Mode Waters. The authors noted that the strongly negative trend near 150 m in the tropics may be associated with a late 1970's polarity reversal of the Pacific Decadal Oscillation. Of particular interest at high-southern latitudes is the cooling above ca 600 m and below ca 1100 m, the formation regions of shelf and bottom waters and the warming between those depths, where circumpolar deep water reaches the Antarctic continental margin.

Figure 4

Linear trends of zonally integrated heat content of the upper ocean, in 1018 J yr−1, for the 1955–2003 period. Positive trends in dark shading, negative in light shading. Modified from fig. 2 of Levitus et al. (2005).

Robertson et al. (2002) reported recent warming of ca 0.3 °C at the depth of the Warm Deep Water (WDW) temperature maximum in the southeast and northwest Weddell Gyre (figure 5a). That feature is usually located between 300 and 700 m where the outer Gyre interacts with the continental margin and more toward the deeper end of that range in the northwest sector. The increase occurred from the mid-1970s to 2000 in the northwest Gyre outflow area around 64 °S, 46 °W, parallel to a warming in the southeast Gyre inflow region around 69 °S, 15 °W. This was observed to be coherent with a 1989–1995 trend in the temperature of Weddell Sea Bottom Water (Fahrbach et al. 2004), a 1970–1998 increase in sea ice surface temperatures (Comiso 2000), and a thermal recovery from the Weddell Polynya events of the mid-1970s (McPhee 2003). The similar trends in the southeast and northwest study areas also supported the idea that more or warmer circumpolar deep water was being entrained into the Weddell Gyre over time, perhaps associated with a southward shift of the southern boundary of the ACC. An update of the northwest sector of the figure suggests a slight cooling in 2002, preceded by a decline after 1995 in the inflow region and consistent with recent cooling in a broad WDW band across the Weddell Gyre (Fahrbach et al. 2004).

Figure 5

(a) Potential temperature at the Warm Deep Water (WDW) temperature maximum (Tmax) for the inflow (triangles) and outflow (dots) regions in the southeast and northwest Weddell Gyre. The black dots/triangles show annual averages for the grey individual measurements, vertical bars 1 standard deviation and dashed lines regressions of the average values. (b) Depths associated with those WDW Tmax values. Observations derived mostly from summer ocean profile data in 3000–4500 m water depths. The inflow area extends ca 250 km seaward of the continental shelf, from ca 0–20 °W; the outflow area extends ca 250 km north of 65 °S, from ca 40–53 °W. Modified from fig. 8 in Robertson et al. (2002).

3. Salinity

Boyer et al. (2005) have documented ocean salinity changes during the last half of the last century, showing more regions with negative than positive trends (figure 6). The areas of strongest increases in the Southern Hemisphere have been near surface in the subtropics and the largest decreases have been south of 70 °S, i.e. in the Pacific sector and Weddell Sea. Their study was based on edited historical ocean stations and derived climatologies in the World Ocean Database 2001, supplemented by World Ocean Circulation Experiment (WOCE) and profiling float measurements. Boyer et al. cautioned that uneven spatial and temporal data distribution is the largest source of uncertainty, a particular problem in the Southern Ocean. An earlier comparison between WOCE and historical observations suggested that systematic errors in the data were of similar magnitude as characteristic changes resulting from spatial property gradients, with pre-WOCE positive salinity biases averaging as high as 0.006 in the 1955–1965 period (Gouretski & Jancke 2001). To the extent that analysis reflected real bias and not actual change, it could shift the figure 6 contours by ca 1×10−4 yr in a positive direction. Another shift worth noting is the increasing realization that salinity change is not forced entirely at the sea surface. Meltwater and brine resulting from Southern Ocean intrusions beneath ca 1.5×106 km2 of floating ice shelves is added to seawater at depths more than 250 m. Where these products and iceberg meltwater are concentrated along the Antarctic continental margin (Silva et al. 2006), they will influence the density field and ocean circulation.

Figure 6

Linear trends in salinity (10−4 yr−1) of zonally averaged pentadal salinity anomalies, 1955–1959 to 1994–1998 for the world ocean. Positive salinity trends are in dark shading, negative in light shading. Modified from fig. 1d of Boyer et al. (2005).

Most areas near Antarctica in the Pacific sector of the Southern Ocean reveal decreasing salinity over recent decades (figure 7). The lines labelled GVC and GVR near the top of the figure show changes that have occurred between 140 and 150 °E on the George V continental shelf and rise, updated from Jacobs (2004). At greater depths in the Australian-Antarctic Basin (AAB), salinity also declined in a comparison between 1994 and 1996 and pre-1994 measurements (Whitworth 2002). Upstream from (east of) that region, the solid lines near the centre of the diagram depict the interannual variability that can arise with more frequent observations (Jacobs et al. 2002), here at 500 m on the continental shelf near Ross Island (SWR) and near the depth of the surface water temperature minimum in the Ross Gyre (RSG). The SWR rate of −0.003 yr−1 is essentially the same as observed near the eastern end of the Ross Ice Shelf and in historical measurements between 450 and 550 m on the western Ross Sea continental shelf. The shallower, stronger and more variable RSG salinity decreases could result in part from short term fluctuations in sea ice cover and precipitation, along with changes in mixed layer depth.

Figure 7

Salinity changes during recent decades in several Southern Ocean regions, compared with North Atlantic trends over similar periods. GVC and GVR, George V continental shelf and rise; AAB, Australian Antarctic Basin; 32S and 17S, zonal WOCE transects at 32 and 17 °S in the Indian and Pacific Oceans versus earlier measurements, from Wong et al. (1999); ASC, Amundsen Sea continental shelf; SWR, southwest Ross Sea; RSG, Ross Sea Gyre; DNA, Dickson et al. (2002) North Atlantic, showing maximum, mean and minimum trends. Vertical scales are the same for each set of measurements.

The short ASC line in figure 7 shows the salinity change between sixteen 1994 stations in the Amundsen Sea and their reoccupations (less than 20 km separation) in 2000 at the same time of year. Those are the only repeat salinity observations available from a broad sector of the Amundsen shelf and upper slope, identified as a likely source of strong freshening downstream in the Ross Sea (Jacobs et al. 2002). The salinity decrease there was accompanied by cooling throughout most of the water column, in an area where Shepherd et al. (2004) have inferred that a warming ocean is eroding the west Antarctic ice sheet. Near Antarctica, most salinity data have been acquired in the austral summer and even then suffer from large spatial and temporal gaps. At the bottom of figure 7, dashed lines illustrate the widespread and uniform freshening of the deep North Atlantic (Dickson et al. 2002). While more measurements were available to calculate those trends, the high southern latitude examples shown here are comparable in magnitude over the same time frame. They are also congruous with Antarctic Intermediate Water salinity decreases shown for 17 and 32 S in the Indian and Pacific Oceans and with the larger-scale trends in figure 6.

It is instructive to look more closely at one of those Southern Ocean freshening regions, the continental rise off George V Land. In figure 8, potential temperature is plotted against salinity for the deepest measurement on each of 77 stations in 1900–3000 m of water between 140 and 150 °E. Averages within discrete observational periods reveal a persistent freshening and cooling (Jacobs 2004) even if the single value prior to 1969 is ignored. Of course the station distribution is far from ideal, but the finding has since been confirmed by repeat measurements along 140 °E (Aoki et al. 2005). As the freshening had a larger impact than the cooling on density, this newly forming Antarctic bottom water has become lighter during recent decades. Where it is too light to reach the abyssal sea floor, it may intrude into the lower deep water, as previously observed directly east of this sector (Carmack & Killworth 1978). Within each time interval the measurements tend to parallel isopycnals, represented here by density anomaly surfaces relative to ca 500 m. From one period to the next, however, the averaged TS values have moved roughly along the heavy dashed line (M) that depicts how those properties will be altered as continental ice melts into seawater (Jenkins 1999). One implication of that evolution in bottom water properties is that their shelf and surface water components, whether derived locally or from upstream, are increasingly influenced by meltwater from continental ice. A similar inference has been drawn from changing oxygen isotope concentrations near the coastline in the upstream Ross Sea (Jacobs et al. 2002).

Figure 8

Potential temperature versus salinity, mostly within 20 m of the sea floor, in water depths of 1900–3000 m on the continental rise north of George V Land. Individual values during the time intervals indicated (with number of observations) are shown by the open symbols and averages over those periods by the larger solid symbols. Dotted lines are density anomaly surfaces relative to 2000 dbar and the dashed line (M) illustrates the T/S relationship for glacial ice melting into seawater. Modified and updated from fig. 13 in Jacobs (2004).

4. Dissolved oxygen

A more commonly measured ocean parameter, even in the remote AAB, is dissolved oxygen (O2). Keeling & Garcia (2002) noted that ocean circulation models predict outgassing and a decrease in the oceanic O2 inventory due to warming, consistent with the ocean's biogeochemical response and with observations of declining O2 concentrations. As an example they cited Matear et al. (2000), who indicated that an O2 change of −5 to −15 μmol kg−1 occurred between 1968 and 1995–1996 at depths more than 400 m in a zonal band from 50–60 °S and 110–170 °E (figure 9). Their model results compared favourably with the difference between meridional WOCE transects south of Australia and area averaged measurements on scattered oceanographic stations that were occupied three decades earlier from the research vessel Eltanin.

Figure 9

(a) Region south of Australia over which temporal oxygen change was analysed. (b) Modelled (dashed lines) and observed oxygen content versus 1500 dbar density anomaly, 1995–96 minus 1968. Error bars on the observations denote the 95% confidence interval. Modified from fig. 5 in Matear et al. (2000).

Concerned that those old O2 measurements might not be sufficiently accurate for such a purpose led to figure 10, where the Eltanin data and three repeats of a WOCE section southwest of Tasmania data are plotted for a smaller region. While the earlier data are still noisier, probably due in part to greater spatial and temporal variability, this comparison appears to confirm the notion that O2 levels have declined since the late 1960s. However, the range of the deep O2 concentrations reported from 1991 to 1996 is significant, suggesting that analytical problems could account for a considerable portion of the reported temporal change. This interpretation may be supported by the large corrections that have been necessary to include some O2 data in ocean atlases (lower bar in figure 10). In addition, Garcia et al. (2005) have documented more variability than trend in the O2 content of the top 100 m of the world ocean, where any multidecadal signal might be expected to be strongest.

Figure 10

Dissolved oxygen measurements, 50–60 °S and 130–150 °E, from June–November 1968 Eltanin cruises and from July–October reoccupations of WOCE section SR3 in the 1990s. Oxygen is plotted against density referenced to 1500 dbar, with 2nd order polynomials fit to each dataset. The horizontal bar at bottom illustrates adjustments that were made to dissolved oxygen data from early Discovery (D) and other cruises prior to inclusion in a Southern Ocean Atlas (Gordon et al. 1982).

Min & Keller (2005) have also evaluated O2 measurements in this region, beginning with the postulate that changes in the historical observation network might introduce errors, depending on the method of data analysis. Using O2 measurements during the 1970s and 1990s within the rectangle outlined in figure 9, and including the 1996 WOCE SR3 line, they found that crossover and objective analyses were less prone to spatial sampling bias than area averaging. The latter technique, also used above for figure 8, yielded errors of ca 2.7±1 μmol kg−1 along an isopycnal near the O2 minimum. That is useful information, but is less than the offset shown by Matear et al. (2000) at the same level, or the spread of curves in figure 10, and much less than anomalies in the historical records. Along with comparable offsets in preliminary data from recent cruises this suggests the need for international quality controls, perhaps similar to the standard seawater used for salinity, if dissolved oxygen measurements are to be used in modelling climate change.

Many other ocean constituents have also changed during recent decades in response to altered source functions. These include components of the carbon system and associated combustion water, with increasing impacts on salinity and sea level (Brewer 1997). Some introduced chemical tracer concentrations have declined, such as tritium since the cessation of nuclear testing in the atmosphere, while others like the chlorofluorocarbons (CFCs) have continued to increase. Repeated sections along the front of the Ross Ice Shelf, e.g. showed rising CFC-11 levels in surface and shelf waters from 1984 through 2000, although open water equilibrium saturation peaked in the mid-1990s (Smethie & Jacobs 2005).

5. Sea ice

Detailed analyses of satellite passive microwave records in the Southern Ocean have typically shown an increase in circumpolar ice extent since late 1978 (Zwally et al. 2002). This correlated positively with average winter temperatures over the Antarctic pack and at several Antarctic coastal stations (Comiso 2000), and would be consistent with the cooler shallow ocean temperatures at the southern end of figure 4. A two-decade cooling trend implied an increase in ice extent of 0.70% K−1, corresponding to an ice edge advance of ca 0.1° latitude (Zwally et al. 2002). Much larger retreats predicated on biological data, from 1.0 to 2.8° latitude between pre-1950 and post-1973 northern ice edge positions, may be more regional than circumpolar (Ackley et al. 2003; Wolff 2003), as observed during the satellite era (figure 11).

Figure 11

Monthly anomalies of sea ice area and linear trends in the Ross Sea and the Bellingshausen–Amundsen Seas, from late 1978 through mid-2004. Note the different vertical scales visually exaggerate short-term variability in the Bel-Adm sector. Figure courtesy of Comiso (2005).

The full (1973 onward) satellite era record of sea ice extent has long been negative (Folland et al. 2001), and the trend is now close to zero since 1978 (J. Comiso 2005, personal communication). In addition, there are remarkable regional differences, particularly between the adjacent Amundsen–Bellingshausen and Ross sectors (figure 11). Increased ice area in the Ross Sea has more than balanced a decrease in the Amundsen–Bellingshausen, even exceeding the circumpolar increase in area over the same time frame. Large short-term changes have also occurred within these subregions, with the recent Ross Sea decline from a peak in late 1998 masking a coincident increase in summer sea ice cover over the continental shelf. For most of the post-1978 period, the increasing Ross Sea ice area in figure 11 seems at odds with the freshening in figure 7. This situation could arise from a thinner ice cover or the import of more ice or lower salinity water from the east (Jacobs et al. 2002). But sea ice anomalies are driven primarily by the atmosphere, which necessitates a slight diversion to consider its variability, along with potential effects on the underlying ocean.

6. Atmospheric forcing

Most of the observations discussed here have been obtained over the last few decades, during which the index of the Southern Annular Mode (SAM) has become increasingly positive (figure 12). Also called the Antarctic Oscillation, SAM refers to the zonally symmetric fluctuations of mid-latitude westerly winds that characterize the primary mode of atmospheric variability in the Southern Hemisphere (e.g. Hall & Visbeck 2002). In its positive phase, surface westerlies are stronger and shifted southward over the circumpolar ocean near 60 °S, and are weaker farther north. This enhances northward Ekman drift south of the Polar Front, a divergent flow consistent with increased sea ice extent. A positive SAM also favours cooling over much of Antarctica, but anomalously strong westerlies upwind of the Antarctic Peninsula may decrease the incidence of cold air outbreaks and increase warm air advection from the Southern Ocean (Kwok & Comiso 2002a; Thompson & Solomon 2002). Interacting with the SAM are shorter term oscillatory phenomena like ENSO (El Nino–Southern Oscillation), with extrapolar links that contribute to large anomalies or dipoles in the Pacific and Weddell sectors (e.g. Kwok & Comiso 2002b; Yuan 2004).

Figure 12

Seasonal (grey bars) and annual (black line) values of the Southern Annular Mode (SAM), a dimensionless index based on observations of mean sea-level pressure at stations near 40 °S (increasing), minus equivalent measurements from stations near 65 °S (decreasing). This is fig. 1 of Marshall et al. (2004), reproduced by permission of the American Geophysical Union.

The SAM is also likely to have a variety of impacts on the Southern Ocean, as diagrammed in figure 13 for its positive phase. Stronger westerlies will slightly increase the strength of the ACC (Meredith et al. 2004), steepening its isopycnals and potentially decreasing meridional heat transport (Hall & Visbeck 2002). However, continuity resulting from increased northward flow of surface water would enhance upwelling near the continental margin, which in turn may facilitate the access of WDW to the continental shelf. The effectiveness of heat transport across the shelf break will depend in part on related forcing over the continental shelf, and on the role of the Antarctic Slope Front (Jacobs 2004). Greater heat transport onto the shelf at depth is likely to increase the melting of continental ice, while enhanced vertical heat flux and upwelling would reduce the volume of sea ice produced, both freshening the shelf waters. The system also contains potential negative feedbacks, such as the cooling from increased deep melting, which could in turn damp shallow melting and enhance sea ice production. It is not yet clear how these processes are balanced or influenced by the SAM, which may not continue to evolve in the direction shown in figure 12.

Figure 13

Diagram of possible ocean circulation and related changes accompanying a more positive SAM index. The ‘greater than’ symbol means ‘stronger’ and arrows (including heads and tails) indicate mean flow directions. The deep water (light shading) temperature maximum over the upper continental slope is often located near the depth of the continental shelf break. Currents are typically stronger along than across the intervening Antarctic Slope Front (ASF). Adapted from fig. 12 in Hall & Visbeck (2002).

Elements of this scenario can be illustrated by changes that have occurred in the western Weddell Sea. The zonal section in figure 14 shows the WDW temperature maximum at about the latitude of the northern Larsen-C Ice Shelf, about 500 km southwest of the outflow region values in figure 5. The subsurface Antarctic slope front rises above the 500 m shelf break, and the modified WDW appears to leak across that front onto the continental shelf. Superimposed on that section are lines representing approximate depths of the temperature maximum in the northwest Weddell Sea in the mid-1970s and in 2000 (figure 5b). Shoaling of that feature over this period is consistent with an increasingly positive SAM index, and with a related process sometimes referred to as ‘gyre spinup’, one aspect of which is elevating the deep water temperature maximum relative to the shelf break. Whether more and/or warmer water then flows onto the continental shelf will be mediated by currents associated with the slope front, local bottom topography and the sinking of shelf and deep water mixtures.

Figure 14

Zonal temperature section in the upper ocean near 67°40′ S in the western Weddell Sea, from stations (numbers at top) occupied in May 1992. Pressure (dbar, ca depth in metres) and potential temperature (ca in situ temperature). Bold dashed lines illustrate shoaling of the WDW temperature maximum from ca 1975–2000 (figure 5). Modified from fig. 8a in Gordon (1998).

7. Ice Shelves

Roughly 150 km northwest of the figure 14 temperature section, several ocean stations were occupied within ca 25 km of the northern end of Larsen-C Ice Shelf in February 2002 (Nicholls et al. 2004). Their T/S diagrams of the vertical profile data (figure 15) include an off-shelf station within the Weddell Gyre outflow region. Stations 99 and 100 best illustrate the Modified Warm Deep Water (MWDW) that had reached the ice shelf vicinity at depths of 300–400 m, about 2° colder and 0.1 fresher than at the off-shelf edge of the Gyre. It is now well known, and illustrated by the bold dashed lines in figure 15b, how the thermohaline properties of seawater change as that water melts meteoric ice (see also figure 8). Nicholls et al. were thus able to demonstrate that the Ice Shelf Water on stations 96–98, i.e. water colder than the sea surface freezing point, could not have resulted from the MCDW on stations 99 and 100. Alternatively, they proposed that it must have been derived from colder shelf waters produced in winter, at which time the temperature maximum could have been eroded and water column properties would lie along the dotted line at the right ends of the dashed lines. This in turn raised questions about the hypothesis of Shepherd et al. (2003) that the relatively warm temperatures at depth on stations 99 and 100 could account for their calculated thinning of Larsen-C.

Figure 15

(a) Potential temperature/salinity diagram of February 2002 ocean stations occupied from 66°–66°36′ S, 59°50′–60°30′ W, near the northern edge of Larsen-C Ice Shelf, and one off-shelf profile at 64°03′ S, 48°34′ W. (b) Enlargement of the boxed area in (a), with dashed lines representing the direction seawater moves in T/S space as meltwater is added. The near-horizontal lines indicate the sea surface freezing temperature and the heavy dashed lines in (a) the WDW warming from figure 5. ASW, Antarctic Surface Water; MWDW and WDW, Modified and Warm Deep Water and ISW, Ice Shelf Water; HSSW, High Salinity Shelf Water. Modified from fig. 4 in Nicholls et al. (2004).

We do not know how the temperature of the MWDW changed as the WDW in the Gyre warmed since the mid-1970s (figures 5 and 15). Similar warm intrusions are known to extend to the front of the Ross Ice Shelf for much of the year, so by analogy the −1.4 to −1.6° MWDW in May and February in this area (figures 14 and 15) could persist near the Larsen-C in winter. It contains sufficient heat to melt metres of ice/year off the base of Larsen-C, but that would require a relatively high annual mean velocity into the cavity, and efficient heat transfer at the ice/seawater interface. Where seawater does not release all of its sensible heat in basal warming and melting, it can emerge from ice shelf cavities at temperatures well above the surface freezing point, as observed in the Ross and Amundsen Seas, and noted in the local context by Nicholls et al. (2004).

Northern sections of the Larsen Ice Shelf have retreated significantly during recent decades, changes attributed largely to atmospheric warming and melt pond penetration of cracks in the ice (Scambos et al. 2003). On the longer term that shelf ice will have thinned in part by basal melting, but it is not known whether its melt rate was increasing, or at what point a critical thickness might allow rapid collapse. That scenario may be more plausible west of the Antarctic Peninsula, where ice shelf retreat has also occurred and the shelf seas are warmer at depth. Around the continent there are few direct measurements of basal melting and freezing in ice shelf cavities, while indirect and model rate estimates vary widely. In some locations, ocean properties near the ice fronts have varied on longer than seasonal time scales (Smethie & Jacobs 2005), and inferences of decadal thinning and thickening are becoming available from analyses of satellite measurements (Shepherd et al. 2004; Zwally et al. 2005). Net basal melting has been estimated to vary by 1–2 orders of magnitude from the cold Amery and Filchner–Ronne Ice Shelf cavities to the warm Amundsen Sea regime (Jacobs et al. 1996; Shepherd et al. 2004). Basal attrition depends on source water properties, cavity configuration, grounding line depth and ocean circulation, including tidal energy. Ocean temperature appears to be a primary factor (Potter & Paren 1985; Rignot & Jacobs 2002), but temporal increases in basal melting have yet to be conclusively demonstrated or linked to ocean warming.

8. Icebergs

The retreat of ice shelves on both sides of the Antarctic Peninsula over the last three decades has amounted to an areal reduction of 12 840 km2 (Scambos et al. 2003). That would have generated many icebergs of many sizes, and increasing numbers of large icebergs have been tracked by the National Ice Center (NIC) over the same period. However, jumps in the number of icebergs being tracked have coincided with the introduction of Operational Linescan System images in 1986 and scatterometer data in later years (Long et al. 2002). Those technological advances, coupled with expected low frequency calving events from the large ice shelves, did not suggest any trend in the records. Of course improved observational tools might parallel a greater calving rate, which may also have been anomalous throughout the short satellite era. In addition, numbers may not represent volume, and most icebergs less than 10 km in length are not monitored. If more ice streams have been accelerating than slowing recently, that could lead to a higher iceberg production, or vice-versa, with or without net changes in ice shelf areas or ice front positions.

Silva et al. (2006) have edited the NIC database and calculated an annual large iceberg calving rate of 1089 Gt a−1 during 1979–2003. While that rate is strongly dependent on the infrequent breakout of giant icebergs, it is only slightly larger than the 1008 Gt a−1 estimate of Jacobs et al. (1992) for the first half of that period. Since the large icebergs calve many smaller ones, Silva et al. used the observed attrition to revise downward a small iceberg calving rate derived from a database maintained at the Norsk Polarinstitutt. By combining their results with recent estimates of in situ ice shelf melting, apparent losses can be compared with reported accumulation. This mass budget approach has large uncertainties, but suggests a negative balance for the west Antarctic ice sheet, consistent with other recent estimates (Bentley 2004; Zwally et al. 2005).

Gladstone et al. (2001) and Silva et al. (2006) also investigated the likely meltwater distribution resulting from the modelled and observed drifts of small and large Antarctic icebergs. A large fraction of the iceberg tracks and the modelled meltwater release (figure 16), was concentrated near the continental margin, in accord with prior reports of icebergs following the westward coastal currents (e.g. Tchernia & Jeannin 1983). Since their melting will both cool and freshen the seawater, the observations in figures 4–8, along with the oxygen isotope reductions in the Ross Sea, could imply iceberg increases in the coastal waters.

Figure 16

Distribution of iceberg meltwater released into the Southern Ocean, calculated from observed large iceberg trajectories, 1987–2003, and from modelled small iceberg trajectories seeded with a climatological calving flux. Grey scale rates vary from <0.1 to >1.0 m m−2 yr−1. This is fig. 4c of Silva et al. (2006), reproduced by permission of the American Geophysical Union.

9. Sea level

SLR, an important ocean change with potential connections to the cryosphere, has continued at a relatively slow pace (1.7–1.8 mm yr−1) over the past half century, with evidence for a more rapid rate over the last decade (Cazenave & Nerem 2004; Holgate & Woodworth 2004). It has long been thought that the rise was largely being driven by thermal expansion and receding temperate glaciers, if not Greenland, with a zero to negative contribution from the Antarctic Ice Sheet (Church et al. 2001). That left a substantial fraction of the observed SLR unaccounted for, and re-evaluation of a possible overestimate did not find that ocean tide gauges were located in regions of abnormally high warming (Cabanes et al. 2001a; Miller & Douglas 2004). Meanwhile, analyses of large ocean databases (Levitus et al. 2000, 2005) showed that the associated thermosteric (volume change from heating) component accounts for a smaller percentage of SLR than models had estimated (Antonov et al. 2002). In addition, declining world ocean salinity from a smaller global dataset implied a significant eustatic (mass change) increase from continental or other sources (Antonov et al. 2002; Munk 2003; Miller & Douglas 2004).

The linear trends from several SLR calculations (figure 17) show a combined thermosteric and eustatic rate of ca 1.8 mm yr−1 from the mid-1950s through the mid-1990s (Munk 2003). That rate includes a correction for freshwater input from sources that alter salinity much more than sea level, mainly retreating and thinning Arctic sea ice. If that correction were larger, then the ‘explained’ rate would be lower, such as the 1.0 mm yr−1 line from Wadhams & Munk (2004), who did not consider ice shelf changes. A satellite altimetry rate of 3.1 mm yr−1 for 1993–2003 includes a glacial isostatic adjustment (Cazenave & Nerem 2004), and is similar to results from coastal tide gauge records north of 40 °S (Holgate & Woodworth 2004). The latter study also showed high values around 1980 and considered the issue of coastal change preceding the global average. Antonov et al. (2005) estimated that the thermosteric trend from 1993 to 2003 left a larger residual to be accounted for by the addition of mass than over the 55 year record, while Church et al. (2005) showed that about 0.5 mm yr−1 of increased SLR over the last decade could have been a recovery from effects of the Mt Pinatubo volcanic eruption in 1991.

Figure 17

Linear trends of sea-level rise (SLR), its components and distribution from (a) Antonov et al. (2005), (b) Munk (2003), (c) Wadhams & Munk (2004), (d) Cazenave & Nerem (2004) and (e) Cabanes et al. (2001b) where A=ocean surface area. Important information on temporal variability, spatial coverage and confidence levels may be found in those original sources. The Ross Sea vertical line shows dynamic height change relative to 500 m due to freshening from 1960 to 2000. Thermosteric refers to volume expansion due to warming, usually in the upper 3000 m, and eustatic to mass additions from continental sources.

The trends displayed in figure 17 are sometimes interpreted as global, although coverage is poor in the polar regions. The figure inset, from Cabanes et al. (2001b), suggests a disproportionately large fraction of recent SLR coming from the Southern Ocean, relative to its area, but based on measurements north of 60 °S. It is worth noting that the ocean volume change from freshening can be large where freshening is strong. That is illustrated in figure 17 for the southern Ross Sea, where dynamic height rose 46 mm in four decades from measured freshening of 0.003 yr−1. Some fraction of that change can probably be attributed to the west Antarctic ice sheet (Jacobs et al. 2002; Thomas et al. 2004; Zwally et al. 2005).

10. Summary

Measurements made from research vessels and drifting submersible floats have revealed significant warming of the ocean during the last half century. While overall data coverage is biased toward summer months south of the polar front and subsurface data are in short supply, the Southern Ocean appears to have absorbed about half of the planetary heat increase over that period. The warming decreases with depth, has not been steady in time or spatially homogeneous, and extends to the Antarctic continental margin within the Circumpolar Deep Water. Areas of cooling have been documented where shelf and bottom waters form in that region. Salinity and other oceanic chemical parameters are also changing, although uneven data distribution and accuracy add uncertainty to some of the apparent trends. Freshening at high latitudes exceeds salinity increases in the subtropics, and in some areas is comparable in magnitude to the ‘Great Salinity Anomaly’ of the North Atlantic. The freshening is strongest near Antarctica, where the overall shift in thermohaline properties and oxygen isotope content suggest that the primary agent is glacial meltwater. Circumpolar sea ice extent is little changed since 1978, with the largest positive and negative anomalies in the adjacent Ross and Bellingshausen–Amundsen sectors.

Both ocean and sea ice parameters are influenced by interannual and longer-term cycles in atmospheric forcing. This variability is dominated by the SAM and complicates interpretations of short-term observations. The SAM has potential influences on ocean dynamics that include modulating sea ice extent and the upwelling of warming deep water onto the Antarctic continental shelf. The strength of SAM impacts on the ocean remain to be documented, but could also involve sea ice thickness, and the melting of glacial ice and icebergs that tend to drift along the Antarctic Slope Front and accumulate in the shallower shelf regions. Where increased sensible oceanic heat melts more ice, the seawater will be cooled and freshened, enhancing upwelling and altering the properties if not the formation rates of surface, shelf and bottom waters.

It is not yet known how much Antarctica has contributed to ocean freshening and SLR over the last several decades. Small negative mass changes have been identified from portions of the marine West Antarctic Ice Sheet, but may be at least partially balanced by interior growth (Jacobs et al. 2002; Joughin & Tulaczyk 2002; Thomas et al. 2004; Davis et al. 2005; Zwally et al. 2005). High-southern latitude ocean freshening exceeds increases that might be attributable to an enhanced hydrological cycle, but could result in part from perturbations related to the SAM, ENSO and sea ice thickness. How effectively ocean warming is transmitted to and utilized in melting of the ice shelves, and the role of shelf ice in the flow of ice streams and mass balance of the grounded ice sheet are important problems for continued research. Answers to such questions should improve our understanding of the likely trajectory of SLR and other Southern Ocean changes over the next half century.


This work has been supported by the National Science Foundation (OPP-01-25172 and 02-33303), the National Aeronautics and Space Administration (JPLCIT-1260804 and OCEANS/04-0272-0142), and by the Lamont-Doherty Earth Observatory (LDEO) of Columbia University. Assistance with figures by P. Mele, T. Boyer, J. Comiso, S. Gille, R. Robertson, T. Silva and A. Payne was most appreciated, as were comments by anonymous reviewers and participants in the October 2005 Royal Society discussion meeting on ‘Evolution of the Antarctic Ice Sheet: new understanding and challenges’. LDEO contribution no. 6855.


  • One contribution of 14 to a Discussion Meeting Issue ‘Evolution of the Antarctic Ice Sheet: new understanding and challenges’.


View Abstract