1. Department of Earth Sciences, University of Cambridge , Downing Street, Cambridge CB2 3EQ, UK

The northeast Atlantic Ocean is of key importance for modulating climate on glacial–interglacial timescales because of heat loss to the atmosphere from the North Atlantic Current and its continuation through the Norwegian Sea, and the convective formation of deep water in the Nordic/Arctic seas, which together provide a temperate climate to northwestern Europe. The poleward flux of warm saline Atlantic waters is partly counterbalanced and maintained by the deep, dense, return flow of ISOW which crosses the Iceland–Scotland ridge mainly through the Faeroe Bank channel toenter the Iceland basin. Together with a similar dense overflow through the Denmark Strait, this process constitutes the initial step of the global ocean conveyor-belt model⁷. The thermohaline conveyor is unstable, and is sensitive to the fresh water/salt balance of the region⁸. Marine isotope stage 3 (60–30 kyr before present, BP) was characterized by extreme millennial-scale climate instability, resulting in the Dansgaard–Oeschger temperature fluctuations seen in ice-core and marine records of temperature, and also resulting in ice rafting⁹. Our results show that even during the generally stable Holocene there is an underlying fluctuation in the strength of ISOW flow south of Iceland with a similar periodicity, which may be linked to climate changes.

The data we present have been obtained from kasten core NEAP-15K (56° 21.92' N, 27° 48.68' W) recovered from 2,848 m depth near the crest of Gardar drift in the south Iceland basin (see map in Fig. 1inset). In this area Holocene sedimentation rates are greatly enhanced (Fig.1b) owing to the interaction of bottom currents with the sea-bed topography, and the core contains a 450-cm-long Holocene sediment record. We use a sedimentological near-bottom palaeocurrent speed proxy, the "sortable silt" mean size¹⁰ (

is the mean grain size of the 10–63-m terrigenous silt fraction, a parameter that varies independently of sediment supply in current-sorted and deposited muds and for which higher values represent relatively greater near-bottom flow speeds). The mean Holocene deposition rate for the northeast Atlantic is 4 cm kyr^-1 ( ref. 12). The core site is far from any sources of downslope mass transport, but is under the deep western boundary current of the Iceland basin which carries material from south Iceland and its continental slope¹³ to create the 1,100-km-long Gardar drift¹⁴. The very high sedimentation rates and particle sizes are thus primarily current-controlled, although fine sediment supply by ice rafting during the Holocene cannot be ruled out altogether. However, its effects would be negligible in core NEAP-15K as its Holocene accumulation ratesfor the terrigenous fine fraction (<63 m) average 20 g cm^-1 kyr ^-1 with peaks of 80 g cm^-2 kyr ^-1, whereas in full glacial conditions within the main ice-rafting belt just south of our site (at 51° 07.0' N, 21° 52.0' W) the flux from the sea surface for the same sediment component was only 1.5 g cm ^-2 kyr^-1 (ref. 15). The accelerator mass spectrometry (AMS) ¹⁴C dates used for the age model in Fig. 1chave been converted to calendar years BP¹⁶ (where 'present' is AD1950) after applying a 400-year reservoir correction. High-resolution dating was also successfully employed to splice the topmost part of NEAP-15K, which overpenetrated by 35 cm and scrambled the upper sediment section, with box core NEAP-16B recovered from the same site, thereby extending the record to the present day. Sedimentation rates (Fig. 1b) are fairly constant from 9.4 kyr ago to the present, but are significantly higher between 10.4 and 9.4 kyr BP(70–160 cm kyr ^-1) due to the mobilization of sediments accumulated in glacial times by renewed ISOW flow¹⁷. The

varies over the range 10.5 to 16 m (Fig. 2b), implying thatsediment-sorting currents of significantly varying magnitude characterized the Holocene history of ISOW flow. We also note thata number of the fluctuations are small (2 m peak–trough but,nevertheless, significantly bigger than the precision of the method¹⁸), suggesting that some of the changes were subtle, probably reflecting the overall stable nature of sedimentation over the past 10,000 years.

Figure 1: Data from kasten core NEAP-15K with an inset map showing its location in the south Iceland basin and the simplified regional flow of ISOW.

a, Sortable silt mean size with error bars after ref. 18. b, Sedimentation rates between calendar years coinciding with ¹⁴C AMS data used in age model. c, Filled circles represent ¹⁴C AMS dates used to construct the age model, and filled diamonds are dates not included in the age model which were predominantly used to splice the uppermost part of NEAP-15K with the bottom of box core NEAP-16B. All dates arefrom monospecific assemblages of the planktonic foraminifera species Globigerina bulloides in the >150-m fraction. The error range for the calendar dates is represented by horizontal bars above and below the solid circles and is defined as 2 standard deviations.

High resolution image and legend (52K)

Figure 2: North Atlantic Holocene palaeoenvironmental proxy records on a calendar years BP(and AD/BC) basis.

a, ¹⁸O data from central Greenland GISP2 ice core with gaussian interpolation using a 300-yr window. Solid and dashed arrows between 10 and 7.5 kyr BPrepresent periods of general warming or cooling which match relative decreases or increases in the flow intensity of ISOW vigour (b), respectively. b, Sortable silt mean size record for NEAP-15K with gaussian interpolation using a 300-yr window. c, Planktonic foraminiferal ¹⁸O data from theSargasso Sea (ref. 26), which mainly reflects changes in sea surface temperature.

High resolution image and legend (27K)

The documented history of climate change in northern Europe over the past few millennia is marked by the alternation of cooler and warmer periods. At present we are still recovering from a time of colder climate known as the Little Ice Age¹⁹, centred at 400 yr BP, which, in our record, coincides with reduced ISOW flow intensity ( Fig. 2b). The

shows that since then, deep-water flow vigour has been increasing. Modern values are comparable to the last warm interval in European history, known as the Mediaeval Warm Period, which peaked at different times in various regions surrounding theNorth Atlantic basin between 750 and 1,050 yr BP(AD900 to1250)¹,²⁰. The climatic history in the few millennia before the Mediaeval Warm Period is less clear, but Europe appears to have enjoyed a warmer spell at 2,000 yr BP, also referred to as the Roman Warm Period, followed by cooling and glacier advance in the Dark Ages (AD500 to 1000)¹,². In our record, a peak in deep-current speed centred at 1,850 yr BP(AD100) coincides with the Roman Warm Period. From these observations and the flow speed fluctuations revealed by Fig. 2bwe infer that periods comparable to the Little Ice Age and the Mediaeval Warm Period were a recurrent feature of earlier parts of Holocene climatic history, with the warm intervals coinciding with faster near-bottom water flow in the south Iceland basin.

The Holocene ¹⁸O record in the GISP2 Greenland ice core, which is thought to reflect mainly palaeotemperature changes²¹, reaches typical Holocene values shortly after 10 kyr BP(Fig. 2a). From that time until 7.5 kyr BP the broad temperature trends of the ice coreare similar to our

results (Fig. 2b), with intervals of lower temperatures over Greenland coinciding with times of more vigorous ISOW flow (we note the cycle between 8.9 and 7.5 kyr BP). After 7.5 kyr BP, no clear or consistent relationship exists between the two (note that the faster flow during the cold "8.2 kyr event"²² is well constrained within the 2 range of dated points in Fig. 1a, c). We suggest that, because sea level was rising at a sustained rate until and just beyond 7.5 kyr BP²³, the flux of melt water to the northern North Atlantic must have been greater during warmer times; this would have had the effect of reducing the density and hence slowing down the flow of ISOW in the Iceland basin. The system flipped into its present state after 7.5 kyr BP , when most of the remaining glacial ice had been melted and changes in temperature did not cause significant variations in the freshwater budget of the North Atlantic region.

The

record was spectrally analysed (Fig. 3) following gaussian interpolation in the time domain using a 300-yr window and a 90-yr sampling interval (Fig. 2b). Only one broad but pronounced spectral density peak centred at 1,500 yr was obtained. This frequency of fluctuations in the intensity of ISOW flow is very similar to that of ice-rafting events recently reported for the past 30,000 yr in the northeast Atlantic centred at 1,470 yr (ref. 4), to that of precipitation in western Canada⁵, and to that of monsoon-related aridity/humidity cycles in Arabian dust (which have a 1,450–1,470-yr period⁶); it is also similar to one of the periodicities recorded in the sea surface and deep-water geochemical and faunal data from the Feni drift between 500 and 340 kyr BP²⁴; and, finally, it is comparable with the 1,450-yr period present in the GISP2 ice-core chemical data produced by variation in wind strength and storminess due to changing atmospheric circulation patterns over the past 110 kyr ( ref. 25). Our results therefore lend strong support to the idea that, as in glacial times⁷, deep-water masses originating inthe high-latitude North Atlantic play an important role in modulating climate in the present interglacial.

Figure 3: Spectral analysis by the Blackman-Tukey technique³⁴ of the sortable silt mean size record from NEAP-15K using data as shown in Fig. 2b.

The 1,500-yr peak accounts for 26% of the total signal in the range above the Nyquist frequency (1/180 yr^-1) analysed (including red noise).

High resolution image and legend (11K)

We also observe several similarities in the fluctuations between the

record and a high-resolution history of surface-water foraminiferal ¹⁸O in the Sargasso Sea²⁶ for the past 3.5 kyr (Fig.2b, c). The correlation, which shows ¹⁸O minima (warm) coinciding with periods of faster ISOW flow and vice versa, is r = 0.47 for the past 1,500 yr and 0.76 for the past 1,200 yr after introducing a lead of 90 yr to the

data (this is acceptable as it is still within the errors of the dating techniques employed). Modern oceanographic studies show that, on decadal timescales under the influence of the North Atlantic Oscillation (NAO), convective overturning in the Sargasso and Greenland seas are linked²⁷. However, we cannot explain in the same terms the apparent connection between these two regions on much longer, millennial timescales because the available evidence suggests that varying deep-water convection in the modern Greenland Sea does not affect the intensity of the overflows south of the Greenland–Scotland ridge²⁸. The present 1990s extreme of the NAO has only slight convection in theGreenland–Norwegian seas area, but no diminution in the overflows²⁷,²⁹. This may be because a large component of ISOW volume flux is thought to originate from the North Atlantic Current entering the Arctic via cooling in the Barents Sea, and returning through Fram Strait, without any significant interaction with the Greenland gyre³⁰. Larger-scale atmospheric feedbacks and reorganizations must be invoked to explain the connections observed between the surface northwest Atlantic and the flow intensity of ISOW, particularly in view of the pervasive 1,500-yr periodicity we observe in our data and which is found in a variety of other proxies from the North Atlantic region⁴,⁵,²⁵ (in glacial as well as interglacial times) and from as far away as the Arabian Sea⁶.

There is now no doubt that the Earth's climate is highly unstable on millennial timescales. However, our data do not provide an explanation for the ultimate forcing mechanism(s) of these events, and more than one process could be responsible for the changes in ISOW flow vigour recorded in core NEAP-15K. The speed of the overflow south of Iceland could be controlled by the influence of changes in either the salt- or freshwater flux on convection in the Nordic seas; alternatively, a shift in the density of shallow and intermediate water masses entrained by ISOW after overflowing the Iceland–Scotland ridge could also be responsible for the fluctuations of Fig. 2b³¹. Despite the problems in identifying the exact origin of the ISOW flow signal, our data shed light on both the sensitivity and the various operational modes of the thermohaline current system. In an initial early Holocene mode, freshwater flux from the land-based ice melted by warmer conditions is argued to be associated with reduced deep-water activity³², resulting in a less dense and more sluggish ISOW flow over Gardar drift. Then, for the past 7.5 kyr, higher temperatures are in phase with more vigorous deep-water flow, under conditions similar to those at present with little melt water and an oscillation of the supply (by the North Atlantic Current) of saline water and heat to the Nordic seas. So far,no clear 1,500-yr periodicity in either ¹⁴ C or ¹⁰Be has been reported from ice cores, suggesting that solar variation is an unlikely forcing mechanism and that an oceanic internal oscillation in 'conveyor' strength is more probable.

The main concern for future climate must be that a possible increase in melting of the Greenland ice sheet resulting from anthropogenically induced atmospheric warming may reach a critical level where the 'conveyor belt' will flip to its early Holocene operational mode³³. The resulting perturbations could conceivably result in climate extremes exceeding those of the Little Ice Age for northern Europe. Without such perturbations, the climate looks likely to be warm for several hundred years ( ref. 6).

References

1. Lamb, H. H. Climatic History and the Future (Princeton Univ. Press, ( 1985)).
2. Röthlisberger, F. 10,000 Jahre Gletschergeschichte der Erde(Sauerländer, Aarau, (1986)).
3. O'Brien, S. R. et al. Complexity of Holocene climate as reconstructed from a Greenland ice core. Science 270, 1962–1964 (1995). | ChemPort |
4. Bond, G. et al. Apervasive millennial-scale cycle in North Atlantic Holocene and glacial climates. Science 278, 1257–1266 (1997). | Article | ISI | ChemPort |
5. Campbell, I. D. , Campbell, C. , Apps, M. J. , Rutter, M. W. & Bush, A. B. G. Late Holocene 1500 year climatic periodicities and their implications. Geology 26, 471–473 (1998). | Article | ISI | ChemPort |
6. Sirocko, F. , Garbe-Schonberg, D. , McIntyre A. & Molfino, B. Teleconnections between the subtropical monsoons and high-latitude climates during the last deglaciation. Science 272, 526– 529 (1996). | ISI | ChemPort |
7. Broecker, W. S. & Denton, G. H. The role of the ocean-atmosphere reorganizations in glacial cycles. Geochim. Cosmochim. Acta 53, 2465–2501 ( 1989).
8. Broecker, W. S. , Bond, G. & Klas, M. Asalt oscillator in the glacial Atlantic? 1. The concept. Paleoceanography 5, 469–477 ( 1990).
9. Rasmussen, T. L. , Thomsen, E. , van Weering, T. C. E. & Labeyrie, L. Rapid changes in surface and deep water conditions at the Faeroe Margin during the last 58,000 years. Paleoceanography 11, 757–771 (1996). | ISI |
10. McCave, I. N. , Manighetti, B. & Robinson, S. G. Sortable silt and fine sediment size/composition slicing: parameters for palaeocurrent speed and palaeoceanography. Paleoceanography 10, 593–610 ( 1995). | Article | ISI |
11. McCave, I. N. , Manighetti, B. & Beveridge, N. A. S. Circulation in the glacial North Atlantic inferred from grain-size measurements. Nature 374, 149–152 (1995).  | Article | ISI | ChemPort |
12. Balsam, W. L. & McCoy, F. W. Atlantis sediments: glacial/interglacial comparisons. Paleoceanography 2, 531– 542 (1987).
13. Shor, A. N. Bottom currents and abyssal sedimentation processes south of Iceland.Thesis, Woods Hole Oceanographic Inst.((1980)).
14. McCave, I. N. & Tucholke, B. E. in The Western North Atlantic RegionVol. M (eds Vogt, P. R. & Tucholke, B. E.) (451– 468) (Geol. Soc. Am., Boulder, (1986)).
15. Manighetti, B. & McCave, I. N. Depositional fluxes, palaeoproductivity, and ice rafting in the NE Atlantic over the past 30 ka. Paleoceanography 10, 579– 592 (1995). | ISI |
16. Stuiver, M. & Reimer, P. J. Extended ¹⁴Ca data-base and revised calib 3.0 C-14 age calibration program. Radiocarbon 35, 215–230 ( 1993). | ISI |
17. Keigwin, L. D. & Jones, G. A. Glacial-Holocene stratigraphy, chronology, and paleoceanographic observations on some North Atlantic sediment drifts. Deep-Sea Res. 36, 845–867 (1989). | Article | ISI | ChemPort |
18. Bianchi, G. G. , Hall, I. N. , McCave, I. N. & Joseph, L. Measurement of the sortable silt current speed proxy using the Sedigraph 5100 and Coulter Counter IIe: Precision and accuracy. Sedimentology(in the press).
19. Grove, J. M. The Little Ice Age(Methuen, London, (1988)).
20. Grove, J. M. & Switsur, V. R. Glacial geological evidence for the Medieval Warm Period. Clim. Change 26, 143–169 (1994). | ISI |
21. Jouzel, J. et al. Validity of the temperature reconstruction from water isotopes in ice cores. J. Geophys. Res. 102, 26471– 26488 (1997). | Article | ISI | ChemPort |
22. Alley, R. B. et al. Holocene climatic instability: A prominent, widespread event 8,200 yr ago. Geology 25, 483–486 (1997). | Article | ISI |
23. Bard, E. et al. Deglacial sea-level record from Tahiti corals and the timing of global meltwater discharge. Nature 382, 241– 244 (1996).  | Article | ISI | ChemPort |
24. Oppo, D. W. , McManus, J. F. & Cullen, J. L. Abrupt climate events 500,000 to 340,000 years ago: evidence from subpolar North Atlantic sediments. Science 279, 1335–1338 (1998).  | Article | PubMed | ISI | ChemPort |
25. Mayewski, P. A. et al. Major features and forcing of high-latitude northern hemisphere atmospheric circulation using a 110,000-year-long glaciochemical series. J. Geophys. Res. 102, 26345–26366 (1997). | Article | ISI | ChemPort |
26. Keigwin, L. D. The Little Ice Age and Medieval Warm Period in the Sargasso Sea. Science 274, 1504–1508 ( 1996). | Article | PubMed | ISI | ChemPort |
27. Dickson, R. R. , Lazier, J. , Meincke, J. , Rhines, P. & Swift, J. Long-term coordinated changes in the convective activity of the North Atlantic. Prog. Oceanogr. 38, 241–295 (1996). | Article |
28. Dickson, R. R. & Brown, J. The production of North Atlantic Deep Water: Sources, rates and pathways. J. Geophys. Res. 99, 12319–12341 ( 1994).
29. van Aken, H. M. & Becker, G. Hydrography and through-flow in the north-eastern North Atlantic Ocean: the NANSEN project. Prog. Oceanogr. 38, 297– 346 (1996). | Article |
30. Mauritzen, C. Production of dense overflow waters feeding the North Atlantic across the Greenland-Scotland Ridge. Part 1: Evidence for a revised circulation scheme. Deep-Sea Res. I 43, 769– 806 (1996). | Article | ISI |
31. Price, J. F. & Baringer, M. O'N. Outflows and deep water production by marginal seas. Prog. Oceanogr. 33, 161 –200 (1994). | Article | ISI |
32. Sakai, K. & Peltier, W. R. Amultibasin model of the global thermohaline circulation: Paleoceanographic analyses of the origins of ice-age climate variability. J. Geophys. Res. 101, 22535–22562 (1996). | ISI |
33. Broecker, W. S. Thermohaline circulation, the Achilles heel of our climate system: Will man-made CO₂upset the current balance? Science 278 , 1582–1588 (1997).  | Article | PubMed | ISI | ChemPort |
34. OS-3 ARAND SYSTEM: Documentation and ExamplesVol. 1 (Computer Center, Oregon State Univ., (1973)).

Acknowledgements

We thank S. Crowhurst for help in the spectral analysis of the data presented here, N.Shackleton for comments on an earlier draft, and R. Dickson for hydrographic observations. This work was supported by UK NERC for the North East Atlantic Palaeoceanography and Climate Change project (NEAPACC).