Changes in the Atlantic Meridional Overturning Circulation

In particular, it damps the temperature rise due to the emission of enhanced greenhouse gases. The MOC is associated with the thermohaline circulation (THC), which is the large-scale ocean circulation driven by fluxes of heat and freshwater at the surface. It is difficult to define the THC exactly. In the real world we cannot disentangle the various driving forces of the ocean circulation, and exclude, for instance, the wind forcing. 

Even in idealized models in which the wind stress is neglected, a THC which only responds to the ‘push’ of surface waters, becoming convectively unstable and sinking, cannot be sustained. Also the ‘pull’ by small-scale mixing, that gradually lightens the deep waters, is necessary. And the main energy source for the small-scale mixing is mechanical (winds and tides). Clearly, the THC is only a theoretical concept. For practical purposes, oceanographers prefer to speak of the MOC: The zonally-integrated mass transport in the ocean. In the Atlantic Ocean it consists of northward flowing water above and southward flowing water below (Figure 1). This typical ‘estuarine circulation’ captures the essential part of the THC. In the North Atlantic, surface water cools and sinks, forming North Atlantic Deep Water which spreads southward into the deep ocean. As a result, the MOC is associated with a large northward heat flux: North of 30°N, the MOC dominates global ocean heat transport1). The loss of heat from the ocean to the atmosphere in the North Atlantic keeps North-Western Europe relatively mild.

The thermohaline part of the MOC is essentially nonlinear: Small changes in the thermohaline forcing can have large effects. As a result, the THC can be subject to sudden transitions. In the popular literature this is often phrased in terms of whether the ‘Gulf Stream’ will halt or not. In reality, the Gulf Stream is the western boundary current of a large wind-driven subtropical gyre, that will continue to flow as long as the earth rotates and the winds keep blowing. But the wind-driven gyre is not part of the MOC. The part that branches off between 30°N and 40°N and flows further northward along the West European and Scandinavian coastline constitutes the MOC, and this branch is vulnerable to anthropogenic climate change. In the media this branch is often identified with the Gulf Stream; in reality it is one of its many extensions. Here, we report on a number of studies performed at KNMI on the driving mechanisms of the MOC, its potential change and climate impact, and the detection of a slow down induced by anthropogenic climate change.

Forcing of the MOC and abrupt transitions

The Atlantic MOC is dominated by one large, basin-wide circulation of northward flowing waters between 0 and 1500 m depth, and southward flowing waters between 1500 and 3000 m depth. The strength of the MOC is surprisingly well related to the thermohaline forcing, although the MOC itself is partly wind-driven. The sinking branch occurs where water at the surface becomes sufficiently dense. In the present climate this happens in the Labrador Sea and Greenland-Iceland-Norwegian basin, because of cooling. The circulation intensifies when the cooling becomes stronger. It turns out that the circulation depends on the (air) temperature-difference between the tropics and the subpolar regions, just south of the sea-ice margin. However, not only the temperature, but also the salinity affects the density of water. At the surface, the salinity is changed by evaporation, precipitation, and the growth or melt of sea ice. In the tropics evaporation makes the water dense, in the subpolar regions precipitation lightens water. The combination of these salinity-related forcings tends to drive an opposite circulation. Without temperature-effects the MOC would reverse, with sinking in the tropics and rising in the subpolar regions.

So, when the thermal forcing dominates, the result is a strong MOC characterized by northern sinking. When the haline forcing dominates, a much weaker MOC with sinking in the tropics occurs. Most interesting is the regime where both forcings are of comparable strength. In this regime both states are possible, that is, the MOC is bi-stable. The actual state depends on the history of the MOC. When it was already dominated by thermal forcing, it prefers to stay so; the same holds for the haline state. Such behaviour is called hysteresis2). Then, a thermohaline perturbation can induce an abrupt transition from one state to the other within a few decades. The most conceivable perturbation is a freshwater pulse in the North Atlantic (e.g., by glacier or ice-cap melting), which will always tend to bring the MOC from the thermally forced state to the haline state at which the MOC is strongly reduced in strength or even reversed.

Abrupt MOC reduction in future climate

Anthropogenic forcing due to rising emissions of greenhouse gases tends to weaken the thermal forcing: Polar and subpolar temperatures will rise much faster than tropical temperatures in the twenty-first century3). At the same time the haline forcing counteracting the MOC is expected to increase. In a warmer world the hydrological cycle is intensified and sea and land-ice will melt. As a result, most climate models project a decrease of the MOC between 0 and 50% for the next century, although none of them simulates a collapse3). But the concern remains that the MOC not only weakens; also its stability could decrease in the future because it approaches, or further penetrates into the bi-stable regime.

The transition point where the mono-stable, thermally forced state changes to the bi-stable state, at which an abrupt transition can take place, is marked by the sign of the net freshwater transport into the Atlantic basin by the MOC4). The Atlantic is a net evaporative basin; easterly trade winds carry water vapour from the Atlantic to the Pacific across the narrow isthmus of Panama. This water vapour loss is compensated for by the ocean circulation. We discriminate two mechanisms by which the ocean circulation can import freshwater into the Atlantic. When at the southern boundary the upper layers are fresher than the lower layers, the MOC imports freshwater. When at 30°S the water at the eastern boundary has lower salinity than the water at the western boundary, the wind-driven, azonal circulation imports freshwater into the Atlantic. When the MOC imports freshwater into the Atlantic it is mono-stable. This can be understood as follows: Let u assume a positive freshwater anomaly in the north, for instance as a result of increased sea-ice melt. This anomaly decreases the surface density. As a result, the sinking of deep water reduces and the MOC is weakened. If the MOC imports freshwater, a weaker MOC will make the Atlantic more saline; the (positive) freshwater anomaly is damped. Also a negative freshwater anomaly will be damped. But if the MOC exports freshwater, the opposite occurs and freshwater anomalies in the north are amplified by the response of the MOC. In this case the MOC is bi-stable. It is possible that a positive freshwater anomaly induces a collapse of the MOC. A freshwater export by the MOC is possible when the wind-driven circulation imports enough freshwater to allow for freshwater loss by both the MOC and through net evaporation. Although the error bars are large, a best estimate from the observations indicates that the latter is the case in present climate.

A recent analysis of a suite of coupled models used for paleoclimate research has revealed that nearly all models are biased towards the mono-stable regime4). In these models the evaporation over the Atlantic is too large, and freshwater import by the azonal circulation is too small. In a follow-up study we have repeated this analysis for a series of IPCC-class climate models that simulate both the present as well future climate. The same bias appears again, but more interesting, the MOC tends to shift in the future towards the bi-stable regime in (nearly) all models, despite the future increase of net evaporation over the Atlantic basin which all models predict (Figure 2). This result suggests two important facts: In present-day climate models the MOC is too stable for perturbations; the risk of abrupt climate change induced by a strong reduction of the MOC increases due to the enhanced greenhouse effect is underestimated.

KNMI and the University of Utrecht, in collaboration with SARA, have performed a series of climate change scenario runs with an IPCC-class model in which a MOC collapse was artificially forced, to study its impact on climate (Figure 3). We recover a 3°C cooling over Western Europe which is a well known result. Together with anthropogenic warming this would result in a cooling of only 0.5°C in 2050. Of specific interest is the pattern of the temperature change: Cooling over the North Atlantic, heating over the South Atlantic. The associated Sea Surface Temperature (SST)-anomaly induces a shift in the Inter Tropical Convergence Zone, the net precipitation, and the Hadley circulation. Climate models tend to underestimate the atmospheric response (and possible feedbacks) to such a collapse, as simulated oceanic mixed layers in the tropical Atlantic are too deep and the SST changes are too weak there. Also, such heating patterns might induce a shift in the Atlantic storm tracks inducing additional cooling of North-Western Europe5), but this response is very sensitive to the amplitude and pattern of the SST-anomaly 

Detecting changes in the MOC

Recently a slowing of the MOC over the past fifty years was reported6) that was much stronger than simulated by any climate model. The reported decrease, however, may have been severely biased by under-sampling of the natural variability of the MOC. The analysis was based on five ‘snapshot’ measurements over a period from 1950 to 2005 (in situ hydrographic sections across the Atlantic by research vessels). In general, before a long-term change can be detected, one must be able to estimate whether the change is associated with a real trend, or whether it represents a fluctuation that results from the high-frequency internal variability. For a more complete continuous observational assessment of the MOC, a mooring-array was deployed by the UK and USA in 2004.

We simulated the detection problem by analyzing the anthropogenic signal and noise associated with the natural variability of the MOC in a large, 62-member ensemble of climate model simulations, all forced with increasing levels of greenhouse gas concentrations in the atmosphere. We also estimated the noise associated with the measurement error7). The signal-to-noise ratio was separately estimated for snapshot measurements and continuous monitoring. The signal-to-noise ratio peaks near 30°N and 30°S at 2000 m depth, confirming the rationale for the mooring-array at 26°N. Also, detection occurs much faster when the MOC is continuously monitored, due to the decrease in noise (Figure 4). The apparent slow-down of the MOC that was recently suggested6) must have been due to unresolved internal variability.