Andrew J. Weaver

One of the central questions in understanding past and future climates concerns the North Atlantic conveyor and how it might respond to increasing anthropogenic greenhouse gases and changes in the net flux of freshwater into its basin. Rahmstorf1, on page xxxx of this issue has for the first time undertaken a detailed study of the sensitivity of the conveyor to changes in the freshwater input into the North Atlantic using a global ocean model coupled to a very simple atmospheric model. His results suggest that the present-day conveyor may be less stable than many researchers have thought. He further suggests that 0.06 x 106 m3s-1 of additional freshwater continually maintained into the North Atlantic could irreversibly shut down the conveyor with a pronounced cooling effect on the surrounding areas.

In the North Atlantic, the ocean acts as a large-scale conveyor that transports heat from low to high latitudes. Most of the oceanic heat transport in the North Atlantic is thought to be associated with the thermohaline circulation (that part of the ocean's circulation which is driven by fluxes of heat and freshwater through the ocean's surface). In the present climate, high latitude cooling together with low latitude heating accelerates the thermohaline circulation with poleward flow at the surface. Through analogy between the ocean conveyor and an automobile we can consider the thermal forcing to act as the accelerator pedal. On the other hand, net high latitude precipitation, runoff and ice melt and low latitude evaporation tend to oppose the thermally-driven thermohaline circulation (much like the brake pedal in our automobile). In today's North Atlantic the thermal forcing dominates over the haline (freshwater) forcing and the conveyor belt moves forward with deep water forming in the Greenland, Iceland and Norwegian (GIN) Seas through intense heat loss to the overlying atmosphere. In addition, the Labrador Sea is a source of deep water (LSW) formed through intense winter cooling which overrides and mixes with the denser GIN Sea waters as it flows equatorward as upper North Atlantic Deep Water.

No such deep sinking exists in the Pacific. If one compares the climates of Bodö, Norway (67deg.17'N, 14deg.25'E), with average January temperature of -2deg.C and average July temperature of 14deg.C, to that of Nome, Alaska (64deg.30'N, 165deg.26'W), with average January temperature of -15deg.C and average July temperature of 10deg.C (both of which are at similar latitudes and on the western flanks of continental land masses), one directly sees the impact of the poleward heat transport of the thermohaline circulation. Clearly then any interruption of the conveyor would have significant effect on the climate of Europe and the North Atlantic region in general.

Rahmstorf began by integrating his global model under present-day forcing to an equilibrium climate with about 20 Sv (1 Sv [[equivalence]] 106 m3s-1) of North Atlantic Deep Water formation (see Fig. 1). He then applied an external perturbation which slowly increased the amount of freshwater into the North Atlantic at a rate of 0.05 Sv/year. As expected from our analogy with the automobile, as the brake was depressed the conveyor slowed down. At a critical value of about 0.03 Sv convection in the Labrador Sea shut off permanently. If, at this stage the perturbation was slowly reversed at the same rate then the point 2 on Fig. 1 is reached which has the same forcing as point 1 but with a different strength of the conveyor. If instead the brake is continually applied then eventually, once the perturbation has reached about 0.1 Sv, the conveyor completely halted. Rahmstorf also reversed the perturbation from this stage and reached point 3 when present day forcing was recovered. Thus, under present day forcing he found three possible configurations of the North Atlantic conveyor: on, off and on with no LSW. In other terms, the automobile has at least three gears and so operates at three speeds under the same pressure on the accelerator and brake pedals. The conveyor, like the automobile, rapidly jumps and jolts between speeds as the gears are changed. The existence of three equilibria of the North Atlantic conveyor is not in itself a new result (see refs., 2-3) but the beautiful illustration of the transition between these equilibria associated with small changes in the hydrological cycle is a novel result.

States very close to the present day are recovered as one traces around the "hysteresis" curve of Fig. 1. Of additional importance is the experiment in which Rahmstorf determined the "point of no return" for the North Atlantic conveyor (X on Fig. 1). He showed that if the forcing were kept steady at 0.05 Sv of extra freshwater into the North Atlantic, then the weakened conveyor (with no LSW formation) remained stable. This was not the case when the extra freshwater flux forcing was retained at only 0.06 Sv (and the conveyor had a strength of 12 Sv). In this case the conveyor eventually ground to a halt. A further remarkable result from the Rahmstorf study is the illustration of a region in Fig. 1 (just before the point where LSW ceases) where no steady circulation is possible and the system oscillates with interdecadal timescale.

What do these model results have to say with respect to the real world and are they reliable? Let us first focus on the prediction that LSW formation shuts down with an additional 0.03 Sv of freshwater into the North Atlantic (about twice the discharge of the St. Lawrence River). The North Atlantic actually has a historical example known as the Great Salinity Anomaly4 which caused LSW to cease from about 1968-19725. From ref. 6 we can estimate the magnitude of the Great Salinity Anomaly perturbation to be about 0.032 Sv for a period of 2 years (consistent and strikingly similar to the number found by Rahmstorf). Of course, when the source of freshwater (Arctic ice export6) was removed LSW formation recovered.

The most worrying prediction is the "point of no return" of only 0.06 Sv for the sustained perturbation of the freshwater forcing. Recent coupled atmosphere-ocean modelling results7-10 aimed at understanding the climatic response to increasing greenhouse gases have all shown that as the climate warms, the northward transport of water vapour in the atmosphere is enhanced, thereby increasing the supply of freshwater to the North Atlantic Ocean. In addition, one model8 reveals that the thermohaline circulation initially weakens but eventually recovers if the forcing is held fixed once the atmospheric concentration of CO2 reaches a level of twice the present day. This is not the case in the response under 4xCO2 forcing8, although it appears that the deep ocean may not have fully-equilibrated in this latter experiment. These results, together with Rahmstorf's analysis, suggest that greenhouse gas induced warming over the North Atlantic and Europe would be smaller (or perhaps even with net cooling) than it would otherwise be if the thermohaline circulation were to remain active. The effects on the ocean environment are probably far-reaching but not quantifiable at this stage.

On page xxxx of this same issue Manabe and Stouffer11 undertake a study to try and quantify the response of the coupled atmosphere-ocean system to large perturbations of freshwater associated with the melting of continental ice sheets during the transition from the last glacial period to the present interglacial period. By adding a large perturbation of 1 Sv of freshwater uniformly released over the 50deg. - 70deg. latitude band of the North Atlantic for a 10 year period, they show that the thermohaline circulation does indeed significantly weaken but appears to have recovered (after a few oscillations) by year 300. How can we relate this result to the work of Rahmstorf as they initially appear to be in conflict? Closer inspection reveals subtle differences between the independent experiments. Rahmstorf maintains his much smaller freshwater flux perturbation for a long period of time while Manabe and Stouffer apply their perturbation over a rapid 10 year period which is too short a time to, in the words of Rahmstorf, invoke an advective spindown of the circulation. This spindown timescale is of the order of a few centuries and is linked to the timescale of the overturning of the North Atlantic conveyor. An extremely interesting experiment would be to rerun the glacial runoff experiment of Manabe and Stouffer by reducing the perturbation by a factor of 10 but increasing the time over which it is applied by a factor of 10. The net freshwater perturbation would have the same size but it would persist over a longer period perhaps causing a shutdown of the North Atlantic Conveyor as suggested by Rahmstorf's results.

The work of Rahmstorf has put many pieces of the climate puzzle together. He has shown (Fig. 1) the existence of multiple modes of operation of the conveyor under present-day forcing; the existence of rapid transitions between these modes; the existence of unstable, variable regimes. Most importantly he has provided a dramatic illustration of where our present climate sits on the stability curve. Combined with the results from previous coupled atmosphere-ocean modelling studies, it sends out a warning as to possible responses of the North Atlantic conveyor to increased high latitude precipitation associated with increasing atmospheric greenhouse gases. Nevertheless, a cornerpiece of the puzzle is missing: That is validation of these results using fully coupled atmosphere-ocean models which resolve the annual cycle, do not drift away from the present-day climate and do not need to employ flux corrections to prevent this drift. Such models are perhaps a few years away as is the computer technology which will allow them to be integrated many times for thousands of years.

Andrew J. Weaver is at the School of Earth and Ocean Sciences, University of Victoria, PO Box 1700, Victoria, B.C., Canada, V8W 2Y2.

1. Rahmstorf, S. Nature 377, xxxx-xxxx (1995)

2. Weaver, A.J. & Hughes, T.M.C. Nature, 367, 447-450 (1994)

3. Mikolajewicz, U., Maier-Reimer, E., Crowley, T.J. & Kim, K.-Y. Paleoceanogr., 8, 409-426 (1993)

4. Dickson, R.R., Meincke, J., Malmberg, S.A & Lee, A.J. Prog. Oceanogr., 20, 103-151 (1988)

5. Lazier, J.R.N., Atmos.-Ocean, 18, 227-238 (1980)

6. Mysak, L.A., Manak, D.K. & Marsden, R.F. Clim. Dyn., 5, 111-133 (1990)

7. Stouffer, R.J., Manabe, S., Bryan, K. Nature, 342, 660-662 (1989)

8. Manabe, S. & Stouffer, R.J. J. Climate, 7, 5-23 (1994)

9. Cubasch, U. et al.Climate Dyn. 8, 55-69 (1992)

10. Murphy J.M., Mitchell J.F.B. J. Climate, 8, 57-80 (1995)

11. Manabe, S. & Stouffer, R.J. Nature 377, xxxx-xxxx (1995)

Figure 1: Schematic diagram showing the behaviour of North Atlantic conveyor as a freshwater perturbation is slowly varied. The abscissa shows the size of the perturbation applied while the strength of the conveyor is represented in the ordinate. Point 1 indicates the present day conveyor strength under present day forcing. Points 2 and 3 indicate two other possible configurations of the conveyor under the same present day forcing. Point 4 indicates where the formation of Labrador Sea water ceases while the X marks the location of the "point of no return", beyond which the conveyor inevitably shuts down. Interdecadal oscillations of the conveyor occur just before the point where Labrador Sea Water formation shuts down.