Southern Ocean: THC, Ocean-Ice Interactions and Global Climate

Lecturer: Doug Martinson, LDEO, Columbia University (

Summary by: Fiamma Straneo, WHOI (


This talk begins with a relatively brief overview of the thermohaline circulation (THC) of the Southern Ocean, of the mechanisms and sites of dense water formation, as well as some of the remaining unresolved issues. Fundamental to Southern Ocean deep water formation is the growth and decay of sea ice. Sea ice, sitting at the sensitive interface between the ocean and atmosphere has the advantage of being highly visible (from space) and easily monitored. Hence, it represents a natural variable to observe and study as a potential indicator of changes in deep water formation, though sea ice plays an important role in global climate independent of its role in deep water formation. After a review of the physics of the sea-ice interaction, including feedbacks and sensitivities, we will discuss some recent work which documents how the Southern Ocean's sea-ice fields covary with extrapolar climate. A number of mechanisms responsible for this covariability have been proposed and will be discussed in the final part of this lecture.

I. The Thermohaline Circulation and the Southern Ocean

The Southern Ocean plays a fundamental role in the global Thermohaline Circulation (THC) both in terms of local dense water production as well as through an active dynamical role. In the earlier schematics of the THC, however, such as that of Stommel and Aarons or in Broecker's Conveyor Belt diagram, the Southern Ocean's role is typically under-represented. A more recent (and more accurate) cartoon by Schmitz (1996), based on extensive observations, finally gives the Southern Ocean its due (the Antarctic is, in fact, the center of the world in this representation). Two striking features are apparent from this schematic (fig 1). First, that the Southern Ocean is responsible for the production of a large number of dense water components, including bottom and intermediate waters. The influence of these waters on the global ocean can be quantified by considering the fact that 57% of the global ocean is colder than 4C and, since North Atlantic Deep Water (NADW) is warmer than this, it follows that these cold waters must all originate around Antarctica. This leads to the second striking feature: that the dense global bottom water becomes subsurface water in the Southern Ocean because of its relative "lightness". Through the production of these water masses, the Southern Ocean can impact global climate in a number of ways. Amongst these are the propagation of ocean property anomalies throughout the global ocean and the influence on the surface ocean's properties of both the Southern Atlantic's and Pacific's subtropical gyres through the interaction with the underlying intermediate waters formed around Antarctica.

II. Dense Water Formation Mechanisms and Sites

Consider the processes of dense water formation and the role of sea ice in these mechanisms. As early as 1921, Brennecke hypothesized that the source of the deep waters originating in the Southern Ocean must be the intense cooling and salt rejection during sea ice formation occurring along the continental shelves of Antarctica. Wst, in 1928, argued that this mechanism could not provide enough deep water and that convection must also be occurring in the open ocean in the Weddell region. Current observations attribute the majority of dense water production in the Southern Ocean to the shelf mechanism (see Baines and Condie, 1998) though open-ocean convection is also known to occur (see Martinson et al., 1981, and references within). As we briefly review both mechanisms, it will become evident how sea-ice plays an important role in both cases. The net result of these dense water production mechanisms and subsequent entrainment processes are an estimated 20 Sv (or so) of AABW entering the world ocean.

For open-ocean convection to occur, the surface layer must become dense enough to sink (and mix) with the deep waters. In the case of deep convection, this requires elimination of the pycnocline (separating the light surface layer from the dense deep water) through the removal of buoyancy. This is not easily achieved by cooling alone both because the water is already close to the freezing point (the temperature is ~2C in the peak of summer) and, in general, because at these temperatures the density of sea-water is primarily controlled by salinity. Once the water is cooled to the freezing point, any further cooling will result in ice formation and the associated brine rejection. The resulting salt flux is an efficient mechanism in making the surface water denser, driving convection; if enough ice forms, the convection can penetrate the permanent pycnocline, causing overturn and deep water formation. When this occurs, mixing with the deep warm waters tends to warm the surface layer and potentially melting the overlying sea-ice. Any further surface cooling of the water column will cause convection to occur again and no new sea-ice can form until the whole water column has cooled to the freezing point (an unlikely event). Thus regions of open ocean convection at these latitudes are characterized by the absence of overlying sea ice; a feature that has allowed them to be identified from satellite images such as the Weddell Polynya of the mid 1970s (see Martinson et al. 1981). The breakdown of the permanent pycnocline, required for deep convection, is not necessary for intermediate water formation. The latter can instead be formed either by mid-depth convection due to a surface buoyancy loss (as for SubAntarctic Mode Water; SAMW, formation described by McCartney, 1977) or via subduction of cold fresh Antarctic surface waters, flowing north, as they encounter southward flowing subtropical waters.

Like open-ocean convection in the Antarctic region, shelf-formation also relies on sea ice formation. Gill first noted that as one moved westward along the continental shelves of the Weddell-Enderby Basin, the water became more saline, eventually reaching values that should promote deep water formation. He attributed the long-shore salt gradient to an excess of ice formation followed by the removal of ice due to wind. Later, Killworth proposed a mechanism for the export of the dense water formed which. He found that contrary to what would be expected on the basis of geostrophy alone, frictional effects allow the dense water to flow down to great depths even under rotational constraints, while entrainment along the way can greatly increase the volume of dense water produced (see Foster and Carmack, 1976). One more mechanism involving sea ice is responsible for Ice Shelf Water formation in the Antarctic region. It involves the melting of glacial ice by the shelf waters, which, in the process, become cooler and fresher. This water then sinks because it is colder than the surrounding waters, and the sinking is accelerated by the thermobaric effect driving a strong plume along the shelf. The water formed in this manner has a very distinct temperature and chemical tracer signature, which has been identified in deep waters and has been captured during its descent down the shelf by current meter moorings (Foldvik, 1990). The main physical processes associated with convection and mixing in polar oceans are represented in fig 2.

III. Different Aspects of the Role of Sea-Ice in Climate

The role of sea-ice in dense water formation, and, in general, in the THC is perhaps not as striking as some of the other ways in which sea-ice impacts our climate. First the high reflectivity of sea-ice (the high albedo) cause the reflection of 80-90% of the sun's energy away from the white polar regions. Second, sea-ice has a strong insulating effect on the underlying ocean drastically reducing the exchange of heat and moisture between the atmosphere and the ocean. Finally, sea-ice is deeply tied to the freshwater cycle with obvious implications for our climate. Because the albedo and insulating effects are so strong, even small changes in the amount of sea ice cover might be expected to drive large changes in the regional, and ultimately, global climate. A qualitative estimate of the importance of sea ice for global climate is provided by the doubled carbon dioxide scenario modeled by Rind et al. (1995). Their results show that approximately 38% of the global warming which follows a doubling of the atmospheric CO2 concentration in their GCM is a consequence of changes in the sea ice (70% of the change occurred in the Southern Ocean sea ice).

From a practical standpoint, sea ice is a climatically important and highly visible (from space) variable, residing at the sensitive ocean-atmosphere interface. In principle, then, if we knew the relationship between sea ice variability and that of the THC and global climate, sea ice would provide a powerful early warning diagnostic or predictor of change elsewhere. Further motivation for monitoring the arctic regions comes from the amplification of the global temperature change is these regions as shown by Wallace's comparison of the annual average temperature change over the Arctic and the non-polar regions over last 50 years (Fig 3).

IV. Physics of Air-Sea-Ice (ASI) Interaction

Now consider the nature of the ASI interaction in a little more detail. Fig 4 shows a typical Antarctic winter upper ocean profile. A well mixed surface layer 100m thick is maintained by surface mechanical stirring. At its base rests a sharp pycnocline (10-40 m thick) characterized by a thermal contrast of 2-3C. The strong diffusive heat flux (around 20-25 W/m2) across this interface satisfies a large fraction of the approximately 35 W/m2 of heat loss to the atmosphere. The remaining heat loss must be provided by latent heat of fusion: ice growth. Let us assume that the system is in steady state. Any further cooling will drive new ice growth, which in turn implies a salinization of the mixed layer. This, therefore, becomes prone to static instability which will, in turn, cause the erosion of the pycnocline, a warming of the mixed layer, an increase in the sensible heat flux and a decrease in ice-growth. It is an example of a negative feedback mechanism which inhibits further ice-growth and maintains the system in a balance typical of the winter season.

While a full description of the ASI interactions requires a complex system of coupled dynamic and thermodynamic equations, it is possible to express the climatically-meaningful ASI characteristics through a number of simple scaling laws relating them to the external parameters of the system (Martinson, 1990). These enable us to tie the local response to the regional forcing which in turn is tied to the global response, thus providing a link between the three. Furthermore, they are particularly useful in the analysis of observations. As an example of how this can be done, let us consider two parameters which dominate the bulk properties of the system, obtained via vertical integration and seasonal averaging (see Martinson and Iannuzzi, 1998; MI98, for details). One is the thermal barrier (TB), which is the enthalpy (relative to the freezing point) residing within the permanent thermocline (see fig 5). The second is the salt deficit (SD), which is the freshwater surplus in the surface layer relative to the deep water (fig 5). SD stabilizes the water column and is removed predominantly by salt rejection during ice growth; for convenience it is expressed as the amount of in situ ice growth required to eliminate it. TB adds stability to the water column via the negative feedback mechanism described above. It too is expressed in terms of ice growth, in this case, the amount of ice that can be melted by venting the thermal barrier. These two parameters can be combined in a variety of ways to provide physically meaningful bulk quantities that encapsulate some of the most important mechanisms of ASI interactions. Here we will consider just two of the most climatically-relevant ones: bulk stability (S) and the total seasonally averaged ocean heat flux. S (S=TB+SD) is the total amount of in situ ice growth, or equivalent latent heat loss, before destabilizing the water column. The heat flux (FT) instead is given by the sum of the entrainment and diffusive fluxes and it dictates the ice growth and the deep water ventilation rates.

As an example of how bulk properties may be used to produce physically relevant parameters from observations, to be then used in correlation analyses or model diagnostics, consider two stations from the eastern Weddell gyre shown in fig 6. One represents an "entrainment" profile and the other a "diffusive" profile, both are characterized by a strong SD but the former has a broad thermocline while the second a sharp one. As shown in fig 6, even though the individual diffusive and latent heat fluxes are very different for the two stations, the total sensible heat fluxes are nearly identical.

V. Climatologies of the bulk properties and parameters

These quantities effectively capture the principal characteristics of the Weddell gyre region, as seen in an analysis of a 25-year climatology of the bulk properties. First consider the Thermal Barrier (fig 7a): it is larger in areas of minimal upwelling where the pycnocline is not compressed near the surface (diminishing the net amount of available enthalpy). On the other hand, near the core of the cyclonic gyre enhanced upwelling pushes the pycnocline closer to the surface and by compressing it, diminishes the net amount of enthalpy it contains. Thus the regional distribution is consistent with our understanding of gyre-scale dynamics. Moreover, the bulk property formulations derived in MI98 enable us to quantify the sensitivity of various local processes to changes in the gyre-dynamics (themselves reflecting changes in the larger scale climatic state).

By combining TB and SD one can obtain the net climatological bulk stability of the gyre as a function of space (fig 7b). It shows that the gyre center is the region with the weakest overall bulk stability. From this, it seems surprising that the system does not produce a polynya here more often, but an examination of the sensitivity of the various scalings shows that the feedbacks introduce a remarkable degree of stability and self-regulation. Additional analysis (see MI98) shows that the bulk stability is overwhelmingly dominated by TB implying that it is the deep water heat more than the surface buoyancy that controls the vast majority of the system's evolution.

The diffusive heat flux (fig 7c, expressed as the amount of ice that is melted or prevented from growing over a 5 month winter period) shows considerable spatial variability across the gyre. It is highest in the eastern and central portions of the gyre, reflecting the gyre-scale dynamics and/or persistent storm tracks, which compress the pycnocline in the gyre center. Comparable spatial variability is exhibited by the entrainment heat flux (again expressed as the amount of ice that is melted or prevented from growing, over a 5 month winter period) though the two are strongly anti-correlated. Physically, this can be explained by considering that in regions where the pycnocline is compressed by stronger upwelling (such as the gyre's center) the diffusive flux will be large while entrainment will be prevented and thus the associated heat flux small. The opposite is true of regions of low upwelling.

Consistent with the fact that the two fluxes appear to be anti-correlated, their sum (fig 7d) shows remarkably little spatial variability across the gyre, ranging from 25-35 W/m2 over most of the area and varying by less than 30% over the gyre core where the component fluxes were varying by a factor of 5. One of the key conclusions which can be drawn from this is that the ASI system, ignoring feedbacks with the atmosphere, seems capable of self-regulating itself so as to deliver a fairly consistent sensible heat flux regardless of the mechanism employed to obtain it. Furthermore, observations suggest that the system is in a quasi-balanced state whereby the ocean provides the atmosphere with the exact amount of heat it requires over the winter, in the form of sensible heat originating from the deep water. This stability is also found in a number of model experiments where seasonal perturbations give rise to interseasonal feedbacks which bring the system back to this balanced state.

Finally, in terms of the dense water formation processes discussed above, the bulk properties provide an indication of the regions that are most susceptible to open-ocean convection. It seems that the system is currently in a state where most of the dense water production is occurring on the shelf. However, one can imagine that some perturbation might lead to a stabilization of the shelf region (e.g. through a reduction in the offshore transport of sea-ice) which would in turn lead to a destabilization of the central region. This seems to suggest, conceptually, a balance by which dense water forms either along the coast or in the interior, but not simultaneously in both places.

VI. Covariation of the bulk parameters with the broader-scale climate indices

The next step is to examine how the bulk parameters, which represent the essence of the physical system, have varied over the last 25 years, and determine whether they covary with broader-scale polar or extra-polar climate indices. Since a number of studies (e.g. White and Peterson, 1996; Simmonds and Jacka, 1995) have shown that the position of the Antarctic ice edge, relative to its climatological average, varies with the El Nio-Southern Oscillation (ENSO) index, as well as with coastal polynyas (Drinkwater et al., 1998), we analyze the covariance of the ENSO index versus stability and versus the total heat flux from data collected in the proximity of the Greenwich Meridian (the region with the greatest number of available data). We found that stability does indeed show a strong relationship with the ENSO index with a correlation of 0.88 and a lag of approximately 9 months.

Because data in the Antarctic region are still too sparse and sporadic to show a high signal to noise ratio, Martinson and Iannuzzi (2000) use the optimal analysis of Kaplan et al. (1998) to construct a gridded set of parameter values from which to compute the covariance matrix of various upper ocean bulk properties. They then compute the EOFs from the matrix, and reduce the space by eliminating those EOFs which contain uncorrelated noise and account for little variance. From this analysis one can examine the correlation of the Principal Component of the low-order EOFs with the extra-polar climate and the spatial variability of these correlations throughout the Weddell gyre region. This method tends to emphasize the gyre-scale coherent structure, as the more locally confined variations are lost in the reduction of the space. Subpolar Upper Ocean and sea-ice fields show considerable covariability with extrapolar climate and the dominant response is to ENSO. The controlling physical changes in the Weddell gyre's upper ocean bulk property distributions seem to suggest the following regional response to ENSO variability. During El Nio years, an enhanced low pressure system in the Weddell Sea, propagated from the Pacific and triggered by changes in the tropical atmosphere or ocean, leads to an intensified cyclonic gyre with increased northward diverging winds. These tend to force more ice to the northern rim of the gyre, where it drives a decrease in the surface salinity and increase in the stability. The enhanced cyclonic forcing causes the pycnocline base to move up, thus decreasing the Thermal Barrier and leading to an overall decrease in stability and increase in heat flux. Though these changes tend to compensate, the change in stability is considerably larger than that in the heat flux so that, 9-12 months after an ENSO, there is a net tendency for destabilization. These changes will, in turn, influence the dense water formation mechanisms and the surface ice cover, thus introducing additional feedbacks on a variety of time scales. On a global scale these results are consistent with the GCM simulations of Rind et al. (2000) which show that equatorial warm anomalies drive an equatorward shift of the subtropical jet (through a decrease in the meridional temperature gradient) thus decreasing the cyclone activity and forcing of the Pacific subpolar gyres. In the Atlantic, the warm anomaly drives an opposite response resulting in an increase in the cyclonic forcing over the subpolar gyre. This hemispheric mechanism also explains the anti-phasing of the Antarctic Dipole, also found in Yuan and Martinson (2000), and is consistent with our broader understanding of Southern Hemispheric circulation and teleconnections, as summarized in the excellent review of Carleton (2000).


Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7a

Figure 7b

Figure 7c

Figure 7d