In 1751, the captain of an English slave-tradingship made a historic discovery. While sailing at latitude25°N in the subtropical North Atlantic Ocean, CaptainHenry Ellis lowered a “bucket sea-gauge”, devisedand provided for him by a British clergyman, the Rev-erend Stephen Hales, down through the warm surfacewaters into the deep. By means of a long rope and asystem of valves, water from various depths could bebrought up to the deck, where its temperature wasread from a built-in thermometer. To his surprise Cap-tain Ellis found that the deep water was icy cold.

He reported his findings to Reverend Hales in aletter: “The cold increased regularly, in proportion tothe depths, till it descended to 3900 feet: from whencethe mercury in the thermometer came up at 53 de-grees (Fahrenheit); and tho’ I afterwards sunk it to thedepth of 5346 feet, that is a mile and 66 feet, it cameup no lower.”

These were the first ever recorded temperaturemeasurements of the deep ocean. And they revealedwhat is now known to be a fundamental and strikingphysical feature of all the world oceans: deep water isalways cold1. The warm waters of the tropics and sub-tropics are confined to a thin layer at the surface; theheat of the sun does not slowly warm up the depths asmight be expected.

Ellis’ letter to Hales suggests he had no inkling ofthe far-reaching significance of his discovery. Hewrote: “This experiment, which seem’d at first butmere food for curiosity, became in the interim veryuseful to us. By its means we supplied our cold bath,and cooled our wines or water at pleasure; which isvastly agreeable to us in this burning climate.”2

In fact, Ellis had struck upon the first indication ofthe ‘thermohaline circulation’, the system of deepocean currents that circulates cold waters of polar ori-gin around the planet, often referred to as the ‘GreatOcean Conveyor Belt’.

But it was not until several decades later, in 1797,that another Englishman, Count Rumford, published acorrect explanation for Ellis’s “useful” discovery: “It ap-pears to be extremely difficult, if not quite impossible,to account for this degree of cold at the bottom of thesea in the torrid zone, on any other supposition thanthat of cold currents from the poles; and the utility ofthese currents in tempering the excessive heats ofthese climates is too evident to require any illustra-tion.”3

Now, two hundred years later, using the most ad-vanced super-computers our century can provide, we

are beginning to understand the intricate dynamicsruling the complex system of deep ocean circulationand, what Rumford found so evident, the role it playsin climate. It is a subject which may be of fundamentalimportance to our future.

An Ocean in the Computer

My work is ‘Climate Modelling’; I simulate the cur-rents of the world’s ocean in a computer and investi-gate their transport of heat across the globe. Themodel I most frequently work with was developed atthe Geophysical Fluid Dynamics Laboratory in Prince-ton and is used by many oceanographers around theworld; I have adjusted it to best suit my experiments.The surface of the planet is divided into grid cells. Mypresent model has 194 cells in longitude, 96 in latitudeand 24 vertical levels: altogether almost half a milliongrid points.

Fig. 1. Current velocities in cm/s (colour scale) in the oceanmodel. Top panel, at 100 m depth. Bottom panel, at 2,000 mdepth.

At each point where there is ocean, the tempera-ture and salinity of the water and the velocity of thecurrents are computed using basic hydrodynamic andthermodynamic equations for each time step whichhas been programmed. If the model is run for, say, 100simulated years, roughly 100,000 time steps would berequired. Multiply this by the number of grid points,

Gulf Stream

Antarctic Circumpolar Current

Benguela

100 m

NADW

2000 m

Antarctic Circumpolar Current

Currents of Change

Investigating the Ocean’s Role in Climate

Essay for the McDonnell Foundation Centennial Fellowship 1999

by Stefan Rahmstorf

Currents of Change

2

Stefan Rahmstorf

and it becomes clear why even the fastest super-computers available will take quite some time to per-form the huge calculations. In fact, to reach a steadystate (or equilibrium) in the ocean circulation takesseveral thousand simulated years; so many calcula-tions are necessary that a super-computer takesseveral weeks to perform them.

So what happens when such a simulation isrun? First one has to specify a way to calculate theexchange of heat, freshwater (through evaporation,precipitation and river run-off) and momentum (fromthe wind) at the ocean surface. This is called “forc-ing”- these forces drive the ocean circulation. Themodel then computes what kind of currents developin the virtual ocean. The results of such a simulation,with a forcing based on the present-day climate, areillustrated in Fig. 1. The computer model has repro-duced all the major ocean currents known from ship-board measurements. The Gulf Stream and its ex-tension towards Britain and Scandinavia (illustratedin the top panel) and the southward flow of North At-lantic Deep Water out of the Atlantic (labelled NADWin the bottom panel) are of particular importance.These currents work together as a kind of conveyorbelt bringing warmth to Europe.

The Conveyor Belt: A Controversial Concept

The concept of an ocean ‘Conveyor Belt’ wasfirst formed by Wallace Broecker4, 5 of Columbia Uni-versity in 1987 to illustrate the idea that all theoceans in the world were connected through one co-herent circulation system, which transported heatand salt between them. Fig. 2 shows a version ofBroecker’s famous sketch of 1987, depicting the sys-tem as a conveyor belt transporting warm wateralong the surface and cold water back through thedepths.

Fig. 2. The Global Conveyor Belt after Broecker6.

Broecker believed that the global conveyor beltwas driven by the atmosphere’s transport of watervapour from the Atlantic basin to the Pacific - waterevaporating from the Atlantic and raining down in thePacific catchment area (see the section “Its driving

force” below). His theory was that the strength of theconveyor belt flow was proportional to this vapourtransport. If the vapour transport were reduced (byless evaporation, for example) the entire systemwould slow down, just as a real conveyor belt doeswhen its power is reduced.

It was a brilliant and provocative idea and ‘TheGlobal Ocean Conveyor Belt’ became a standardterm for describing world ocean circulation in popularpublications and many scientific ones.

Some oceanographers, however, questionedthe metaphor and the theory behind it, and in subse-quent years a number of research papers has beenpublished which give a more accurate picture of theway the ocean circulation functions. In the work I’vebeen doing since 1991, I’ve found that the circulationloops in the Atlantic and Pacific are only weakly con-nected and that the oceans don’t respond as onesystem. Within the Atlantic Ocean, however, the cir-culation loop functions very much like a ‘conveyorbelt’ - warm water is transported north by a system ofcurrents including the Gulf Stream to near Green-land, where it drops down to become the cold NorthAtlantic Deep Water flowing south (Fig. 3).

And if extra freshwater is added in the computermodel to the region near Greenland (as could hap-pen with extra rainfall or the melting of ice in the realworld), the whole system slows down as one, fromthe North Atlantic Drift right down to the BenguelaCurrent of the South Atlantic.

The Role of the Atlantic Conveyor in Climate

The cold water discovered in the subtropical At-lantic by Ellis in 1751 was, as Rumford theorised,brought there by a current which had originated inthe polar region; temperature measurements in thereal ocean and computer models show there is asouthward outflow of cold deep water from the Arcticthroughout the Atlantic. This cold water is replacedby warm surface waters, which gradually give offtheir heat to the atmosphere as they flow northwardtowards Europe. This acts as a massive “centralheating system” for all the land downwind.

The heat released by this system is enormous:it measures around 1015 W, equivalent to the outputof a million large power stations. If we compare plac-es in Europe with locations at similar latitudes on theNorth American continent, its effect becomes obvi-ous. Bodö in Norway has average temperatures of-2°C in January and 14°C in July; Nome, on the Pa-cific Coast of Alaska at the same latitude, has a muchcolder -15°C in January and only 10°C in July7. Andsatellite images show how the warm current keepsmuch of the Greenland-Norwegian Sea free of iceeven in winter, despite the rest of the Arctic Ocean,even much further south, being frozen.

If the Atlantic ‘Conveyor Belt’ circulation is

Currents of Change

3

Stefan Rahmstorf

switched off in a computer model, a different climateforms in the virtual world. There is little change inocean temperatures near the Equator, but the NorthAtlantic region becomes much colder than it is in re-ality, and the South Atlantic and other parts of theSouthern Hemisphere become warmer. This experi-ment reveals that the Atlantic circulation moves heatfrom the South Atlantic below the Equator across thetropics to the North Atlantic - the heat is not comingdirectly out of the tropical region.

Fig. 3. The Atlantic Conveyor Belt. Orange circles show theregions of convection in the Greenland/Norwegian and Lab-rador Seas. The outflow of North Atlantic Deep Water(NADW) is shown in blue.

So Rumford did not get it quite right: the oceancurrents do not seem to do much to cool the “exces-sive heats” of the tropics, although they certainly playan important role in preventing excessive cold in Brit-ain, Scandinavia and the rest of Northern Europe.

Some of my scientific colleagues have com-pared climates with and without the Atlantic ‘Convey-or Belt’ in computer simulations using coupledocean-atmosphere models8. These show that with-out the ocean heating, air temperatures would coolby up to 10°C averaged over a year. The chill isgreatest near Scandinavia, but extends to a lesserextent right across Europe and much of the NorthernHemisphere.

Driving the Conveyor

What drives this remarkable circulation? Whydoes it occur only in the Atlantic, why don’t the Pacificand Indian Oceans have similar heating systems?

In general, ocean currents are driven either bywinds or by density differences. Density in the oceandepends on temperature and salinity, and the Atlan-tic Conveyor is a thermohaline circulation (from the

Greek words for ‘heat’ and ‘salt’). When surface wa-ters become dense enough - through cooling or be-coming saltier (or a combination of these twofactors), a mixing process takes place in which theysink and form deep water.

Count Rumford understood the basics of thismechanism and reported them in his 1797 publica-tion. The deep mixing takes place because of con-vection; sea water from near the surface sinks down2km or more and then spreads at this depth. This iscalled ‘deep water formation’.

Because this convection only takes place atspecific sites, the image of a plug-hole in a bathtub issuggestive. At certain locations on the broad ex-panse of ocean, the surface waters descend into thedeep. They don’t exactly gurgle down the plug-hole,rather they subside so slowly that it is hard to meas-ure directly. In the present climate, deep water formsin the Greenland-Norwegian Seas and in the Labra-dor Sea (marked by orange points in Fig. 3). Thereare no deep convection sites in the North Pacific.The only other place where the global ocean’s deepwater is formed is near the Antarctic continent in theSouthern Hemisphere.

At these convection sites, the water has becomedense enough to push away the underlying deep wa-ter and sink because it has cooled - cold water isdenser than warm. In the Indian Ocean, the water istoo warm to sink; the ocean’s basin does not extendvery far north of the Equator.

But cold is not the only factor. The waters of theNorth Pacific are cold, but even as they approachfreezing point they still lack the density which wouldenable them to sink down and thus drive a conveyorflow. This is because the North Pacific is less saltythan the North Atlantic (around 32‰, compared to35‰ in the northern North Atlantic), and salt is thesecond crucial factor in the density of ocean water.

In fact, the high salinity of the North Atlantic isthe key to understanding how the thermohaline circu-lation works. For decades, two theories were widelyaccepted as explanations of what drives the flow ofNorth Atlantic Deep Water. Both were frequently cit-ed. At first glance they seem complimentary - but mywork led me to examine them more closely.

Wallace Broecker of Columbia University rea-soned that the global conveyor (see Fig. 2.) was driv-en by evaporation. He explained the salinitydifference between the Atlantic and the Pacific as aresult of excess freshwater evaporating from the At-lantic basin and being blown across into the Pacificcatchment by Easterly winds - what is known as va-pour transport. In other words, water was evaporat-ing from the Atlantic leaving its salt behind, andraining back down as freshwater which diluted thePacific.

So in Broecker’s theory, the relatively high salin-

Currents of Change

4

Stefan Rahmstorf

ity of the North Atlantic water, coupled with the coldtemperatures, created the convection ‘pumps’ whichdrove the global conveyor. The salinity budget of theoceans was balanced by the conveyor transportingfresher Pacific water back to the North Atlantic.

If, the theory continues, the conveyor were togrind to a halt for some reason, then the salinity ofthe North Atlantic would start to rise, as evaporationwould continue to leave salty water behind, althoughthe inflow of fresher water would have stopped.

Systematic computer simulations by myself andother oceanographers found flaws in this theory. Wediscovered that the present thermohaline circulationcould be maintained without any airborne transportof water vapour from the Atlantic to the Pacific, andeven with a weak reverse vapour transport. And, sur-prisingly, within such a scenario the North Atlanticcontinued to be saltier than the North Pacific.

Finding the explanation for this paradox wassimple: the conveyor does not need evaporation inorder to operate, it transports salty water into theNorth Atlantic by itself, thus maintaining the high sa-linity there. It is a classic chicken-and-egg situation -the Atlantic conveyor functions because salinity ishigh in the North Atlantic, and salinity is high in theNorth Atlantic because the conveyor is functioning. Itis a positive feedback which makes the conveyor aself-sustaining system.

This peculiar property had already been exam-ined much earlier, in 1961, by the famous Americanoceanographer Henry Stommel in a simple but pow-erful conceptual model. One consequence of thepositive feedback was that if it was somehow inter-rupted, the conveyor belt would grind to a permanenthalt. In the absence of ongoing circulation, the NorthAtlantic’s salinity would drop so much that no moredeep water could form.

After I had been working on the thermohalinecirculation for some time, I realised there was a con-tradiction between these two theories - Broecker’sevaporation-driven conveyor and Stommel’s self-maintaining conveyor - which meant both could notbe true. Each was widely cited in oceanographic lit-erature, both were accepted as valid and yet on clos-er examination I found they were mutually exclusive.

In Broecker’s theory, the direction of the fresh-water transport by the Atlantic conveyor was north-ward; in Stommel’s theory it was southward. This iswhy, if the circulation halted, the salinity of the NorthAtlantic would increase according to Broecker, butdecrease according to Stommel. Obviously this re-sult was directly dependent on whether the upper,northward flowing branch of the Atlantic conveyor(see Fig. 3) had a higher or lower salinity than theoutflow of North Atlantic Deep Water.

Existing hydrographic measurements from thereal ocean were unable to resolve this issue because

of the complicated salinity layering of the Atlantic wa-ters. Above the layer of North Atlantic Deep Water,northward flowing layers exist with both higher andlower salinities. From the data, it is not possible toidentify which of the near-surface currents belong tothe thermohaline conveyor and which are simplydriven by winds. So I set the computer model to thetask.

Through a series of model experiments, I foundthat in the model world the conveyor belt transport offreshwater is indeed southward, and that salinity de-creases throughout most of the Atlantic when theconveyor is shut down. There was net evaporationfrom the Atlantic in my model, as in the real world, butthis did not affect the functioning of the conveyormuch. Also the freshwater loss to the atmospherewas not balanced by the conveyor, but rather bywind-driven, near-surface currents.

In this sense, Stommel’s theory was the moreaccurate of the two. But his work, done in 1961, hadbeen based on a very simple, ‘box’ model of a theo-retical thermohaline circulation. It was limited to onehemisphere and driven by the density difference be-tween the water of the tropics and that of the high lat-itudes. Model experiments and observational datademonstrate that the real Atlantic does not work likethis, although some of my colleagues have tried toargue for a theory of two more or less disconnectedcells, one in each hemisphere.

My model results showed the existence of onecross-hemispheric conveyor belt in the Atlantic,transporting heat from the southern to the northernhemisphere, which is driven by the density differenc-es between North and South Atlantic water. If thenorthern part of the Atlantic thermohaline circulationis slowed down by adding virtual freshwater to themodel ocean, the whole system slows down, includ-ing the Benguela current off South Africa in theSouthern Hemisphere.

To overturn established beliefs is not alwayseasy. In 1995, I wrote a paper pointing out the inher-ent contradiction in the two major thermohaline circu-lation theories and arguing for a southward directionof freshwater transport. I submitted it to a highly re-spected scientific journal. My work was given thethumbs down: one of the two reviewers criticised myresults for not being new, saying everyone in the fieldhad known all this for a long time; the other wrote thatmy results were wrong and flying in the face of all ev-idence. This outcome all too accurately emphasisedthe problem - fortunately I was able to publishelsewhere9.

Unstable Currents

There is another crucial difference between thetwo views of the conveyor. A circulation driven byevaporation and high-latitude cooling would be very

Currents of Change

5

Stefan Rahmstorf

stable. But Stommel’s self-maintaining conveyor de-pends on precariously balanced forces: cooling pullsin one direction, while the input of freshwater fromrain, snow, melting ice and rivers pulls in the other.This freshwater threatens to reduce the salinity, andtherefore the density, of the surface waters; only bya constant flushing away of the freshwater and re-plenishing with salty water from the south does theconveyor survive. If the flow slows down too much,there comes a point where it can no longer keep upand the conveyor breaks down.

Fig. 4. A schematic stability diagram shows how the flow ofthe conveyor depends on the amount of freshwater entering

the North Atlantic. Units are 106 m3/s on both axes.

A look at a simple stability diagram shows howthis works (Fig. 4). The key feature is that there is adefinite threshold (labelled S and called a ‘bifurca-tion’) for how much freshwater input the conveyorcan cope with. Such thresholds are typical for com-plex, non-linear systems. This diagram is based onStommel’s theory, adapted for the Atlantic conveyor,and on global circulation model experiments10.

Different models locate the present climate (P)at different positions on the stability curve - for exam-ple, models with a rather strong conveyor are locatedfurther left in the graph, and require a larger increasein precipitation to push the conveyor ‘over the edge’(transition a in Fig. 4). The stability diagram is thus aunifying framework that allows us to understand dif-ferent computer models and experiments.

My experiments also revealed another kind ofthreshold where the conveyor can suddenly change.While the vulnerability in Stommel’s model arisesfrom the large-scale transport of salt by the convey-or, this new type of threshold depends on the verticalmixing in the convection areas (e.g. Greenland Sea,Labrador Sea). If the mixing is interrupted, then theconveyor may break down completely in a matter ofyears, or the locations of the convection sites mayshift. Such a shift in convection sites is indicated astransition b in Fig. 4.

Although the effect on climate of a cessation ofAtlantic circulation had come to be generally under-stood, until 1994 no one had considered the possibil-ity of the shifting of convection sites (or ‘convectiveinstability’ as it is now known), let alone the conse-quences. That year, quite by chance, I discoveredthat such an occurence would have a major effect:regional climates would change radically in less thana decade.

The paper I published on the subject11 struck achord with paleo-climatologists, scientists who hadbeen seeking explanations for the evidence they hadfound of abrupt climatic changes taking place thou-sands of years ago (see section ‘The Past..’).

The speed at which these convective shifts takeplace is frightening. With Stommel’s theory, once thefreshwater threshold has been exceeded, the con-veyor circulation slowly grinds to a halt over a centu-ry or more. But if a convective instability is triggered,within some few years the conveyor stops transport-ing heat to the far north.

We do not yet know where these critical limits ofconvection are, nor what it would take to set off suchan event. Current climate models are not powerfulenough to resolve such regional processes clearly.

Icy Times

By looking back at past climates we can under-stand a little more about the effect of the North Atlan-tic circulation. The ancient ice caps of Greenland andAntarctica have preserved a unique and detailedrecord of the history of climate, layed down in year af-ter year of snow that never melted, going back atleast 100,000 years. Several cores have been drilledright through these mountain-high ice caps in recentyears (Fig. 5), and from the exact composition of theice and enclosed air bubles in different layers muchinformation about the past climate can be recovered.

Fig. 5. Scientists drilling an ice core in Greenland.

Other valuable records of the past are containedin sea corals, tree rings and in ancient pollen - polleneven reveals information about vegetation cover atdifferent times in the planet’s history. And cores tak-en from the sediments at the ocean bottom give awealth of clues about past ocean circulation and cli-mate. From these data, it has been possible to re-construct conditions at the height of the last Ice Age

−0.1 0 0.1

0

20

Freshwater Forcing [Sv]

NA

DW

Flo

w [S

v]

oS

30

S

P

a

b

Currents of Change

6

Stefan Rahmstorf

(the so-called ‘Last Glacial Maximum’), around21,000 years ago. Huge ice sheets, several kilom-eters thick, covered the northern parts of NorthAmerica and Eurasia. In Europe, ice covered all ofScandinavia and reached down as far as Berlin. Asfar south as France, cold, dry steppe extendedacross the continent, stalked by mammoths.

Evidence points to the ocean’s thermohaline cir-culation being quite different from today. Reconstruc-tions show that the Atlantic conveyor did not nearlyextend as far north. The convection regions weresouth of Iceland, and the water sank only to interme-diate depths. The bottom of the North Atlantic was in-stead filled by waters of Antarctic origin which werepushing in from the south12.

In 1997, my colleagues and I at the Potsdam In-stitute for Climate Impact Research in Germany per-formed a computer simulation of the Ice Ageclimate13 (Fig. 6) including both atmospheric andoceanic circulations - a world first. Our model pro-duced exactly the ocean circulation changes de-scribed above, and we could establish that thechanges in the conveyor had a significant effect onIce Age temperatures. We found that with the chang-es in the ocean circulation, the Northern Hemispherewas on average 9°C colder in the Ice Age than today,but if we experimented by deliberately preventing theocean circulation changes, the temperatures wereonly 6°C colder.

Fig. 6. Cooling of surface temperatures at the height of thelast Ice Age during Northern Hemisphere summer, as simu-lated by the Potsdam coupled climate model13.

It is a combination of such model simulationsand further detective work on ice cores and similardata which will ultimately lead to a detailed under-standing of the forces shaping the climate of ourplanet.

The Past: A Roller-Coaster Ride

The coldness of the last Ice Age was punctuatedby many sudden and erratic swings in the climate.The Greenland ice cores show sudden temperatureshifts of about 5°C that happened over about a dec-ade, but last for centuries (Fig. 7). Many of theseevents were not local to Greenland, but had reper-cussions that have been detected as far afield as

South America and New Zealand. When we lookback over the history of climate, the past 10,000years (the Holocene) appear as an unusually stableperiod. It is probably no coincidence that this is thetime in which agriculture was invented and humancivilisation developed.

Fig. 7. Temperature record of the past 100,000 years, de-rived from a Greenland ice core.

The cause of these rapid fluctuations puzzlespaleo-climatologists. Subtle and gradual changes inthe energy the Earth receives from the sun, due towobbles in our orbit (the so-called Milankovich cy-cles) are the major reason for Ice Ages and other cli-mate changes of the past. But why does the climatenot respond in a smooth, gradual way? This is one ofthe greatest riddles in climatology.

Sediment cores from the sea bottom reveal thatocean currents were changing in sync with theweather over Greenland and other land areas. Theshifts in ocean circulation (through the convective in-stability mechanism) together with instabilities in thelarge ice sheets may well be the culprits responsiblefor these erratic climate changes.

Even the relatively stable climate of theHolocene has not been an entirely smooth ride forhumanity. An as yet unexplained cold snap occurredaround 8,000 years ago - perhaps also caused by achange in ocean circulation. The warm ‘HoloceneOptimum’ followed - around 6,000 years ago - whenthe Sahara was green and dotted with lakes, like theone at the mouth of the ‘Cave of Swimmers’ made fa-mous by the book and film “The English Patient”.

Our own Millennium started with the ‘MedievalOptimum’, in which Vikings settled now icy Green-land and grapevines grew in Yorkshire, England.Then, from about 1550 to 1850, the so-called ‘LittleIce Age’ took hold of Europe: temperatures werearound 1°C below those of the present century. Thewinter landscapes of Pieter Brueghel were paintedduring this period. Lake Constance, Western Eu-rope’s largest lake, (incidentally, where I grew up)regularly froze over completely - something whichhas happened only once this century, in 1963. InEngland, festivals took place on the ice of the RiverThames. The Great Frost of 1608 was later de-

100 80 60 40 20 0

0

-20

∆T

AGRICULTURE BEGINS

HO

LOC

EN

E

THOUSANDS OF YEARS BEFORE PRESENT

LAST ICE AGE

(°c)

Currents of Change

7

Stefan Rahmstorf

scribed in a novel by Virginia Woolf*: “Birds froze inmid-air and fell like stones to the ground. At Norwicha young countrywoman started to cross the road inher usual robust health and was seen by the onlook-ers to turn visibly to powder and be blown in a puff ofdust over the roofs as the icy blast struck her at thestreet corner. The mortality among sheep and cattlewas enormous. It was no uncommon sight to comeupon a whole herd of swine frozen immovable uponthe road.”

Fig. 8. Pieter Brueghel the elder: Winter Landscape with aBird Trap (1565)

What role the ocean circulation played in theseclimate changes is as yet unclear. New data fromsediment cores14 strongly suggest that these eventswere part of a more or less regular, 1,500-year longcycle, involving major shifts in the North Atlanticocean currents. With recent advances in computermodelling and the increase of data available on pastclimates, we may be on the verge of developing anunderstanding of such cycles.

The Future: Risk of a Sea Change?

The climate of the next century will be defined byan ongoing increase in the concentration of carbondioxide and other greenhouse gases in the atmos-phere. All climate models are predicting that this willlead to a substantial temperature increase (around2°C by the year 2100)15. The hydrological cycle ofevaporation and precipitation is also expected to in-crease, as in a warmer world the atmosphere canhold more moisture.

How will the Atlantic ocean circulation respondto these changes? Given its past instability, this is avery real concern. A warmer climate will mean lesscooling and more precipitation, possibly also extrafreshwater from a melting of the Greenland Icesheet. The delicate balance in which the presentconveyor operates may cease to exist. Model sce-narios for the twenty-first century consistently predicta weakening of the conveyor by between 15% and

50% of its present strength16.What has not yet been determined is whether

sudden temperature shifts like those seen in theGreenland ice core could take place. Due to their lim-ited resolution, the current generation of climatemodels cannot properly represent the processeswhich might lead to these sudden changes (e.g. theconvective instability). We do not yet know if or whenwe would cross a threshold in the climate systemwhich could dramatically change our future17. It is arisk we cannot afford to ignore.

Stefan Rahmstorf

References

1.Warren, B.A. & Wunsch, C. Deep Circulation of the WorldOcean, in Evolution of Physical Oceanography (ed. Warren, B.A.)6-40 (MIT Press, Cambridge, 1981).

2.Ellis, H. Philosoph. Transact. Roy. Soc. London 47, 211-214 (1751).

3.Thompson, B. in The complete works of Count Rumford(published in 1870 by the American Academy of Sciences, Bos-ton) (1797).

4.Broecker, W. The Biggest Chill. Nat. Hist. Mag. 97, 74-82(1987).

5.Broecker, W.S. The great ocean conveyor. Oceanogr. 4,79-89 (1991).

6.Broecker, W. Chaotic Climate. Scientific American, 44-50(1995).

7.Weaver, A. Driving the ocean conveyor. Nature 378, 135-136 (1995).

8.Manabe, S. & Stouffer, R.J. Two stable equilibria of a cou-pled ocean-atmosphere model. J. Clim. 1, 841-866 (1988).

9.Rahmstorf, S. On the freshwater forcing and transport ofthe Atlantic thermohaline circulation. Clim. Dyn. 12, 799-811(1996).

10.Rahmstorf, S. Bifurcations of the Atlantic thermohalinecirculation in response to changes in the hydrological cycle. Na-ture 378, 145-149 (1995).

11.Rahmstorf, S. Rapid climate transitions in a coupledocean-atmosphere model. Nature 372, 82-85 (1994).

12.Labeyrie, L.D., et al. Changes in the vertical structure ofthe North Atlantic ocean between glacial and modern times. Qua-ternary Science Review 11, 401-413 (1992).

13.Ganopolski, A., Rahmstorf, S., Petoukhov, V. & Claus-sen, M. Simulation of modern and glacial climates with a coupledglobal climate model. Nature, accepted (1997).

14.Bond, G., et al. A pervasive millennial-scale cycle inNorth Atlantic Holocene and glacial climates. Science 278, 1257-1266 (1997).

15.Houghton, J.T., et al. Climate Change 1995 The IPCCReport. (Cambridge University Press, Cambridge, 1995).

16.Rahmstorf, S. Shifting seas in the greenhouse? Nature399, 523-524 (1999).

17.Rahmstorf, S. Risk of sea-change in the Atlantic. Nature388, 825-826 (1997).

*Woolfe’s description in “Orlando” is based on a con-temporary report by Thomas Dekker