For an updated and more detailed version, see the following paper (pdf, 3MB):
Rahmstorf, S., 2006: Thermohaline Ocean Circulation. In: Encyclopedia of Quaternary Sciences, Edited by S. A. Elias. Elsevier, Amsterdam.
Books: (click on image)
since 5 Jan 2009
What is the thermohaline circulation (THC)?As opposed to wind-driven currents and tides (which are due to the gravity of moon and sun), the thermohaline circulation (Fig. 1) is that part of the ocean circulation which is driven by density differences. Sea water density depends on temperature and salinity, hence the name thermo-haline. The salinity and temperature differences arise from heating/cooling at the sea surface and from the surface freshwater fluxes (evaporation and sea ice formation enhance salinity; precipitation, runoff and ice-melt decrease salinity). Heat sources at the ocean bottom play a minor role.
Figure 1. Schematic representation of the global thermohaline circulation.
Surface currents are shown in red, deep waters in light blue and bottom waters in dark blue. The main deep water formation sites are shown in orange. (After , modified by S.R.)
- Deep water formation: the sinking of water masses, closely associated with (but not to be confused with) convection, which is a vertical mixing process, ). Deep water formation takes place in a few localised areas: the Greenland-Norwegian Sea, the Labrador Sea, the Mediteranean Sea, the Wedell Sea, the Ross Sea.
- Spreading of deep waters (e.g., North Atlantic Deep Water, NADW, and Antarctic Bottom Water, AABW), mainly as deep western boundary currents (DWBC).
- Upwelling of deep waters: this is not as localised and difficult to observe. It is thought to take place mainly in the Antarctic Circumpolar Current region, possibly aided by the wind (Ekman divergence).
- Near-surface currents: these are required to close the flow. In the Atlantic, the surface currents compensating the outflow of NADW range from the Benguela Current off South Africa via Gulf Stream and North Atlantic Current into the Nordic Seas off Scandinavia (Fig. 2). (Note that the Gulf Stream is primarily a wind-driven current, as part of the subtropical gyre circulation. The thermohaline circulation contributes only roughly 20% to the Gulf Stream flow.)
|Figure 2. Thermohaline circulation of the Atlantic. |
This highly simplified cartoon of Atlantic currents shows warmer surface currents (red) and cold north Atlantic Deep Water (NADW, blue). The thermohaline circulation heats the North Atlantic and Northern Europe. It extends right up to the Greenland and Norwegian Seas, pushing back the winter sea ice margin. (From .)
Some observational dataThe volume transport of the overturning circulation at 24 N has been estimated from hydrographic section data () as 17 Sv (1 Sv = 106 m3/s), its heat transport as 1.2 PW (1 PW = 1015 W). More recently, an inverse model by  yielded 15+/-2 Sv NADW overturning in the high latitudes. (Note: when comparing these numbers with models care needs to be taken what exactly is compared - in models, the most common measure of NADW overturning is the maximum of the zonally integrated transport stream function in the North Atlantic, sometimes also the outflow value at 30 S.)
The short answer would be: high-latitude cooling. In cold regions the highest surface water densities are reached, this causes convective mixing and sinking of deep water, which drives the circulation.
What drives the THC?
Reality is more complex. Pressure gradients at depth, resulting from density gradients in the overlying waters, are the driving force in the equations of motion. As the density forcing occurs at the surface (see above), a subtle question is why the density differences and the circulation affect the whole ocean depth and are not confined to a near-surface layer.  showed that a deep circulation only arises when heating (buoyancy source) is at depth and cooling at the surface. The reason that there is a deep circulation after all is turbulent mixing, which brings down the heat on a time scale of ~1,000 years. It has been shown that in the long-term equilibrium the strength of the thermohaline circulation in models depends on the turbulent mixing coefficient , and that the energy required for this turbulent mixing comes to a large extent from the moon via tidal currents ().
This discussion can be labelled: is the THC pushed or pulled ()? I.e., pushed by formation of cold deep water, or pulled by downward diffusion of heat through the thermocline? The answer is a question of time scale: ultimately, in the long run, it is pulled. But on shorter time scales, up to centuries, it can be considered pushed in the sense that it is density changes in the deep water formation regions which affect the circulation strength. If this density drops too much so that deep water formation is not possible, the circulation stops. Ultimately, on the long time scale of turbulent mixing, the deep ocean density will drop as well until new deep water formation can start.
As mentioned above, highest surface densities in the world ocean are reached where water is very cold, while lower densities are found in the saltier but warmer tropical and subtropical areas. In this sense the THC is thermally driven. Nevertheless, the influence of salinity is important and is what causes the non-linearity of the system. This was first described in a classic paper by  with the help of a simple box model. Salinity is involved in a positive feedback: higher salinity in the deep water formation area enhances the circulation, and the circulation in turn transports higher salinity waters into the deep water formation regions (which tend to be regions of net precipitation, i.e., freshwater would accumulate and surface salinity would drop if the circulation stopped). Put simply, in Stommel's model the high-latitude salinity increases linearly with the flow, and the flow increases linearly with high-latitude salinity, which combined gives a quadratic (i.e., non-linear) equation. This leads to two possible equilibrium states, the system is bistable in a certain parameter range. This becomes more than an academic point as complex circulation models behave in the same way, and as the present North Atlantic in many models is in the bistable regime (). The first coupled climate model to show these two equilibria (discovered quite by accident) is the one by .
Non-linear behaviour of the THC
The situation can be described with a simple stability diagram showing strength of the THC as a function of the freshwater input into the North Atlantic. This shows the bistable regime and a saddle-node bifurcation point where the circulation breaks down. It is discussed in more detail (but for the non-specialist) in .
An important point is that the salt transport feedback is not the only feedback rendering the system non-linear. The convective mixing process is itself a highly non-linear, self-sustaining process. In models this can lead to multiple stable convection patterns ([14, 15]), which on one hand can cause artefacts related to the coarse model grid. On the other hand this may be part of a real mechanism for shifts in convection location, as have apparently occured during glacial times.
The bottom line is: salinity leads to non-linearity which causes the existence of multiple equilibria and thresholds in the THC.
A related question is: why is no deep water formed in the North Pacific? Salinity there is too low, but why? A body of literature exists on this topic; it is discussed, e.g., in . My opinion is: for geographical reasons so much freshwater enters the North Pacific that it is far in the monostable regime where no deep water formation is possible.
The climatic effect of the THC is still to some extent under discussion, and is due to the heat transport of ~1 PW of this circulation. Back-of-the-envelope calculations suggest that this amount of heat transported into the northern North Atlantic (north of 24 N) should warm this region by ~5K. This is indeed roughly the difference between sea surface temperature (SST) in the North Atlantic as compared to the North Pacific at similar latitudes. A look at sea ice margins suggest that they are pushed back by the warm surface currents in the Atlantic sector as compared to the North Pacific (Fig. 1), this in turn leads to reduced reflection of sunlight and thus warming (albedo feedback). A look at global surface air temperatures is also quite suggestive: over the three main deep water formation regions of the world ocean, air temperatures are warmer by up to ~10K compared to the latitudinal mean.
The effect on climate
These observations are, however, no quantitative proof of the climatic effect of the THC, and other explanations can be invoked, such as planetary waves in the atmosphere, locked in place by the geography (Rocky mountains).
One way to estimate the effect of the THC is to switch it off in coupled climate models (by adding a lot of freshwater to the northern Atlantic), and compare the surface climate before and after switching it off. Roughly, this leads to a cooling with a maximum of ~10K over the Nordic Seas (e.g., [12, 17]). The maximum tends to occur near the sea ice margin due to the ice albedo effect. Unfortunately, the details of this cooling are model-dependent: one model shows cooling up to 22K in annual mean and 33K in winter (). Models also differ in how widespread the cooling is: most tend to affect temperatures over land in northwestern Europe (Scandinavia, Britain) by several degrees, others show strong cooling further west affecting Canada ().
|Figure 3. Deviation of surface air temperature from zonal mean.|
Deviations are shown in degree C. Based on NCAR surface air temperature climatology, reproduced from .
Sediment data document that the THC has undergone major changes in the history of climate (e.g., [21, 22]). Three major circulation modes were indentified: a warm mode similar to the present-day Atlantic, a cold mode with NADW forming south of Iceland in the Irminger Sea, and a switched-off mode (). The latter appears to have occurred after major input of freshwater, either from surging glacial ice sheets (Heinrich events) or in form of meltwater floods (e.g., Younger Dryas event). The most dramatic climate events recorded in Greenland, the Dansgaard-Oeschger (D/O) events, were probably associated with north-south shifts in convection location, i.e. transitions between warm and cold modes of the Atlantic THC. Recent simulations of such shifts show encouraging agreement with paleoclimatic data ().
History of the THC
Global warming can affect the THC in two ways: surface warming and surface freshening, both reducing the density of high-latitude surface waters and thus inhibiting deep water formation.  was the first to warn that this could lead to a breakdown of the THC and to abrupt climate change. Subsequently, [26, 27] showed that this could indeed occur for strong global warming (i.e., for a quadrupling, but not for a doubling of CO2). In these scenarios there was no surface cooling, as the high CO2 levels more than compensated for the reduced ocean heat transport. The possibility of a real cooling (both a relative cooling, i.e. a drop back to roughly pre-industrial temperatures after an initial warming phase, and in the longer run an absolutecooling below preindustrial values) as a result of anthropogenic warming was first demonstrated in a sensitivity study by . Significant absolute cooling can arise after CO2 levels decline, but the THC remains switched off after its collapse is triggered in a rapid warming phase.
The THC in anthropogenic global warming
A THC collapse is now widely discussed as one of a number of "low probability - high impact" risks associated with global warming. More likely than a breakdown of the THC, which only occurs in very pessimistic scenarios, is a weakening of the THC by 20-50%, as simulated by many coupled climate models (). Key open questions include:
- What changes in freshwater input to the North Atlantic will result from global warming? (Uncertainty e.g. due to uncertain estimates of Greenland meltwater runoff, ignored so far in most models, and due to possible changes in ENSO ().)
- What is the risk of exceeding a threshold for THC collapse for a given warming?
- What other thresholds exist? (e.g., a local shutdown of convection in the Labrador Sea as simulated by , rather than a full THC collapse.)
- What consequences would result for marine ecosystems?
- How would temperatures over land be affected by a collapse scenario? (Just a reduced warming, or a warming followed by abrupt cooling?)