North Pacific seasonality and theglaciation of North America 2.7millionyears agoGerald H. Haug1, Andrey Ganopolski2, Daniel M. Sigman3, Antoni Rosell-Mele4, George E. A. Swann5, Ralf Tiedemann6, Samuel L. Jaccard7,Jorg Bollmann7, Mark A. Maslin5, Melanie J. Leng8 & Geoffrey Eglinton9
1Geoforschungszentrum Potsdam (GFZ), and 2Potsdam Institute for Climate Impact Research (PIK), 14473 Potsdam, Germany3Department of Geosciences, Princeton University, Princeton, New Jersey 08544, USA4ICREA and ICTA, Autonomous University of Barcelona, 08193 Bellaterra, Catalonia, Spain5Environmental Change Research Centre, Department of Geography, University College London, London, WC1H 0AP, UK6IFM-Geomar, 24148 Kiel, Germany7Department of Earth Sciences, ETH Zurich, 8092 Zurich, Switzerland8NERC Isotope Geosciences Laboratory, British Geological Survey, Keyworth, Nottingham NG12 5GG, UK9Biogeochemistry Centre, University of Bristol, Bristol BS8 1TS, UK
In the context of gradual Cenozoic cooling, the timing of the onset of significant Northern Hemisphere glaciation 2.7 million yearsago is consistent with Milankovitch’s orbital theory, which posited that ice sheets grow when polar summertime insolation andtemperature are low. However, the role of moisture supply in the initiation of large Northern Hemisphere ice sheets has remainedunclear. The subarctic Pacific Ocean represents a significant source of water vapour to boreal North America, but it has beenlargely overlooked in efforts to explain Northern Hemisphere glaciation. Here we present alkenone unsaturation ratios and diatomoxygen isotope ratios from a sediment core in the western subarctic Pacific Ocean, indicating that 2.7 million years ago late-summer sea surface temperatures in this ocean region rose in response to an increase in stratification. At the same time, wintersea surface temperatures cooled, winter floating ice became more abundant and global climate descended into glacial conditions.We suggest that the observed summer warming extended into the autumn, providing water vapour to northern North America,where it precipitated and accumulated as snow, and thus allowed the initiation of Northern Hemisphere glaciation.
To initiate and sustain the large Northern Hemisphere ice sheets ofthe Plio-Pleistocene ice ages, two requirements are broadly recog-nized. First, the more polar continental areas must be sufficientlycold for precipitation to fall as snow rather than rain and for snowand ice to survive the warm summer melting season. Second,adequate moisture must be introduced to high northern latitudesto promote the accumulation of glacial ice. In attempts to explainthe initiation of the major Northern Hemisphere glaciation2.7million years (Myr) ago, much attention has been given to thetemperature requirements of continental glaciation. The timeinterval between 4.5 and 3.1Myr was dominated by a pronouncedlong-term minimum in the amplitude of the 41-kyr cycle in theobliquity of the Earth’s rotation1, which would have failed toproduce particularly cold Northern Hemisphere summers—thekey requirement posited by Milankovitch for the onset of NorthernHemisphere glaciation. During this time interval, there may havebeen several aborted shifts toward glaciation, for example, between4.1–3.9Myr and 3.5–3.3Myr (ref. 2; Fig. 1). During the late Plioceneand early Pleistocene, a high amplitude in the obliquity cycleresulted in periods of low tilt angle, which, in turn, would haveyielded periods with cold summers in the Northern Hemisphere.Thus, it has been suggested that the progressive increase in theamplitude of the obliquity cycle tipped the scale between 3.1–2.5Myr, allowing for long-term expansion of Northern Hemisphereice1. In short, our long-held view of the temperature requirement ofglaciation is largely consistent with the timing of the onset ofNorthern Hemisphere glaciation.
However, the onset of Northern Hemisphere glaciation hasproved to be inconsistent with ideas regarding the water vapourrequirement3,4. It has been suggested that glaciation began inresponse to increased North Atlantic Deep Water formation and
the flow of warm Gulf Stream waters into the high-latitude NorthAtlantic, associated with the closure of the Panama seaway5,6.However, recent studies show that this closure and associatedchanges in North Atlantic circulation occurred 4.6Myr ago, wellbefore the onset of intense Northern Hemisphere glaciation4,5.Thus, it is unknown whether and how a change in water vapoursupply encouraged the initiation of intense Northern Hemisphereglaciation.
Seasonality of the modern subarctic North PacificThe modern subarctic Pacific surface is dominated by a permanent‘halocline’, or salinity-driven density gradient in the upper 300m,that reduces exchange between the surface layer and the oceaninterior7. Seasonal changes in the temperature of the surface mixedlayer are thus only minimally buffered by the heat capacity of theocean subsurface, resulting in a sea surface temperature (SST) thathas one of the largest annual ranges of any open ocean region, withwinter (February) surface ocean temperatures in the subarcticNorthwest Pacific of about þ1 8C, late-summer (September) tem-peratures of þ12 8C, and a seasonal thermocline during summerand autumn8.This seasonal variation in the physical conditions of the subarctic
Pacific leads to strong seasonality in the biological productivity ofthe region. Winter mixing transports nutrients from the subsurfaceinto the euphotic zone. During spring, as the euphotic zone deepensand the mixed layer shoals, a diatom-dominated bloom begins,lasting until early summer, when most of the nutrients are con-sumed, silicate in particular9,10. However, during late summer andautumn, when the water column is most stable, a secondarybiogenic bloom typically occurs, this time dominated by cocco-lithophores11,12. Alkenones accumulating in the sediments below
this region indicate a modern temperature of 10.1 8C, consistentwith the late-summer and autumn growth of the coccolithophorids,which are among the prymnesiophytes that produce these com-pounds11.
North Pacific changes 2.7Myr agoBefore about 2.7Myr ago, the accumulation of diatomaceoussediments in the subarctic Pacific was roughly four to five timesgreater than it is today (Fig. 1; ODP Site 882: 508 21 0 N, 1678 35 0 E,water depth of 3,244m). Records from both the western and easternbasins of the subarctic Pacific indicate an abrupt drop in opalaccumulation rate at isotope stage G6, synchronous with themassive onset of ice-rafted debris (IRD, Fig. 1). The nearly completeconsumption of silicate in the modern subarctic Pacific summer-time surface and a sedimentary N isotope change across the 2.7-Myrtransition conspire to indicate that the drop in opal accumulationwas associated with a drop in the supply of major nutrients from theocean interior to the surface ocean13. Thus, the biogeochemicaldata point to the development of the subarctic Pacific haloclineat 2.7Myr, closely associated with the onset of major NorthernHemisphere glaciation13 (Fig. 1).
The close association of subarctic Pacific halocline formationwith major Northern Hemisphere glaciation as well as the abruptand dramatic nature of both changes suggest a positive feedbackbetween the two. We have previously focused on how climatecooling increased the vertical stability of the North Pacific13,14.This work raised atmospheric CO2 as the possible mechanism bywhich polar stratification could, in turn, cause global cooling andthus participate in a positive feedback. The sediment core data andclimate model output reported here provide a more direct mecha-nism by which the development of the subarctic Pacific halocline setthe scene for ice-sheet growth in the Northern Hemisphere.
The d18O of microfossil calcite from the planktonic foraminiferG. bulloides increases at 2.7Myr by approximately 2‰ (ref. 14;Fig. 1). This has been taken to indicate a drop in SSTof about 5 8C,taking ice-volume variations into account15. The d18O increaseoccurs shortly (,3 kyr) before the drop in opal accumulation andcoincides with the first step in IRD increase. Poor preservationmakes such a foraminifera-derived reconstruction difficult in thesenearly carbonate-free sediments, especially after the 2.7-Myr tran-sition. Nevertheless, the sharp increase in IRD across the transitionand the evidence from other regions16 confirm that the overall senseof change at 2.7Myr was a dramatic cooling.
We have measured alkenone unsaturation ratios17–22 (UK37 and
UK37
0) across the 2.7-Myr transition at ODP Site 882, to provide an
additional constraint on surface temperature changes in the sub-arctic North Pacific. Prymnesiophytes, including the coccolitho-phorids, produce the long-chain alkenones that are used in thistemperature reconstruction. As phytoplankton, these organisms areconcentrated in the upper euphotic zone of polar waters, whereasG. bulloides can live at a variety of depths and also forms agametogenic crust in the subsurface. Moreover, coccolithophoridstend to bloom in the middle to late summer in the western subarcticPacific, after the diatom bloom9–12, whereas foraminiferal pro-duction tends to follow the productivity of the entire phytoplanktonpool and thus is at a maximum in the spring9,10. For these reasons,significant differences should be expected between alkenone- andforaminiferal-based temperature reconstructions. Even with thisexpectation, the alkenone-derived temperature change is in surpris-ing contrast to the indicators of cooling at the 2.7-Myr transition:the alkenone data indicate a warming of$7 8C across the transition(Fig. 1).
Given this unexpected result, possible sources of artefact mustbe considered. The alkenone content of these old, high-latitudesediments is quite low, requiring the use of a high-sensitivitygas chromatography-chemical ionization mass spectrometry(GC-CIMS) method for measurement of the degree of alkenone
unsaturation23 (see Supplementary Information). Diagenesis andchanges in light and nutrient conditions are not expected to bias theunsaturation ratios to the degree that would be required to explainthe transition at 2.7Myr (refs 24, 25). A recent concern isthat alkenones can be transported laterally, associated with finesediments, and can perhaps be remobilized from ancient sedi-ments26. This is an unlikely concern for the sediments of ODPSite 882. Sediment transport in the region is, if anything, from theNorth, and there is no evidence for a radical change in lateraltransport at the 2.7-Myr transition. For remobilization and sub-sequent incorporation of alkenones from older sediments to havecaused the apparent decrease in UK
37 at the 2.7-Myr transition,extremely old sediments would need to be eroded to produce such a‘warm’ UK
37 signature. This assumption is not supported by thecomposition of the coccolith assemblage, which is dominated bycoccoliths typical/indicative of the time period analysed (mainlygephyrocapsids and reticulofenestrids).The alkenone evidence for post-2.7-Myr summertime warming is
corroborated by the d18O of the silica frustules produced by large(75–150 mm) autumn-living subarctic North Pacific diatom species(Coscinodiscus marginatus and C. radiatus, see SupplementaryInformation). Their d18O decreases by approximately 5‰ acrossthe 2.7-Myr transition (Fig. 1), indicating some combination ofwarming and freshening in the late-summertime/autumn surface.Development of the full modern halocline in the subarctic Pacific at2.7Myr can explain only,1‰ of this d18O decrease8, whereas onewould expect ice volume to have caused a global ocean d18O increaseof,0.5‰ (ref. 2). This leaves a,4.5‰ decrease to be explained bylate-summertime/autumn sea surface warming. Published coeffi-cients for the dependence of diatom silica d18O on temperaturerange from 0.2‰ to 0.5‰ per 8C (refs 27–29). Thus, the diatomd18O data appear to require a warming of$9 8C at 2.7Myr, which issimilar to the warming estimate from the alkenones. Althoughsignificant uncertainties remain in the use of diatom d18O , they arecompletely different from those that apply to the alkenones. Inparticular, lateral transport or exhumation fromolder sediments arenot plausible concerns for these large and extremely well-preserveddiatoms (see Supplementary Information).
A link between stratification and seasonalityDespite the initially counterintuitive nature of these results, warm-ing is completely consistent with the development of the subarcticPacific halocline at 2.7Myr. Because the halocline acts to reduceexchange between the surface and ocean interior, the developmentof the halocline at 2.7Myr should have caused the seasonal variationof surface ocean temperature to reflect more fully the seasonal cyclein insolation and air temperature. That is, upon stratification, theseasonal variation in surface temperature should have increasedtoward the ,11 8C range that characterizes the modern subarcticPacific. The UK
37-index reflects the SST maximum that coincideswith the late-summer/autumn coccolith bloom, and the diatomsanalysed here also grow at the surface during the late summer10. Bycontrast, the foraminiferal calcite is biased to record the springdiatom bloom and is also strongly influenced by the temperature ofthe shallow subsurface through growth below the mixed layerand the formation of gametogenic crust. Consequently, theapparent paradox between the cooling at 2.7Myr as indicated bythe planktonic foraminifera d18O and the warming as indicatedby the UK
37-index and diatom d18O is best interpreted as anexpression of the amplified seasonal contrast that should havebeen expected from the development of the permanent haloclineat that time.The high heat capacity of sea water causes surface waters to
remain warm into the autumn and cool into the spring. Stratifica-tion of the subarctic Pacific will decrease the thermal inertia of theupper ocean and thus reduce the phase lag between land and oceantemperature. However, the amplification of the seasonal cycle
should overwhelm this effect, so that stratification will cause thesubarctic Pacific surface to be significantly warmer than the landfurther into the autumn.Such an enhanced temperature contrast with continental climate,
which responds rapidly to seasonal insolation changes, is well suitedfor driving glaciation in North America. The subarctic Pacific is adominant source of water vapour to boreal North America30, andwarmer SSTs in the autumn would cause a larger fraction of thewater vapour delivery to occur when continental climate is coldenough for snow to accumulate. In this way, the stratification of thesubarctic Pacific would allow for adequate water vapour supply tofeed glaciers when global climate cooling would otherwise drive adecrease in water vapour transport and limit ice-sheet growth.Thus, the obliquity minimum within isotope stage G6 at 2.7Myrmay have succeeded in beginning the age of intense NorthernHemisphere glaciations specifically because it triggered the devel-opment of the subarctic Pacific halocline, which then continued toprovide water vapour to boreal North America even as the globallyaveraged atmosphere became colder and drier.
North Pacific seasonality and glaciationTo test the links among subarctic Pacific stratification, SST and ice-sheet growth, we carry out a suite of experiments with CLIMBER-2,an Earth system model of intermediate complexity31,32. To controlstratification, we vary the freshwater input into the subarctic NorthPacific region. If this input is reduced frommodern forcing by 0.2 Sv(,20% of the total precipitation over the North Pacific), the modelcomes to equilibriumwith a ‘destratified’ subarctic Pacific that lacksits modern halocline. An increase in freshwater input is unlikely tohave been the specific cause of subarctic Pacific stratification at2.7Myr (ref. 13); it merely represents a simple model strategyfor changing polar stratification31 (see also Supplementary Infor-mation).The ‘destratified’ and ‘stratified’ equilibria differ in ways that are
consistent with their representation of pre- and post-2.7Myrconditions, respectively. Relative to the ‘destratified’ equilibrium,the modern equilibrium maintains much colder winter and springSSTs in the subarctic Pacific and has significant seasonal sea-icecover, which the ‘destratified’ equilibrium lacks. Despite the overallcooling associated with the modern equilibrium, late-summer andautumn SST is actually warmer in this modern equilibrium, whichwe explain above as the result of reduced thermal inertia associated
with stratification. Relative to the destratified state, the coldspring in the stratified state reduces spring and summer snowmelt(Fig. 2a, b). At the same time, the warm autumn SSTof the stratifiedstate maintains the moisture supply to North America (Fig. 2c, d)despite annually averaged cooling.
To assess the significance of the differences between the stratifiedand destratified states for the build-up of ice sheets in the NorthernHemisphere, we performed additional experiments using anextreme orbital configuration called ‘cold orbit’. A high-resolutionsnow pack model coupled to CLIMBER-2 was used to diagnose thearea of permanent snow cover, which can be considered as aminimum footprint for the ice sheets. For the modern climatestate (stratified North Pacific) and the ‘cold orbit’, a large area ofNorth America is perennially covered by snow (Fig. 3b). In contrast,with a destratified North Pacific, the area of perennial snow cover isrestricted to the Arctic archipelago and small mountainous areas(Fig. 3a). Growth of the ice sheet provides an additional positivefeedback, which explains part of the large temperature differencebetween the stratified and destratified North Pacific climate states(Fig. 3b, see also Supplementary Information).
A time-evolving experiment simulates the development of stra-tification at 2.7Myr (Fig. 1i). The experiment begins at 3.1millionyears ago (Fig. 1i) from the destratified state and gradually increasesthe freshwater flux to the subarctic Pacific at a constant rate of 0.2 Svper million years. The simulation also includes orbital parametervariation1. At 2.75Myr, stratification sets in, winter/springtime(March/April) subarctic Pacific SST drops by ,5 8C, and summer-time/autumn (August/September) SST increases by ,3 8C. Thetiming of the gradual freshwater increase has been optimized toyield stratification at 2.7Myr. However, the abrupt development ofstratification from the gradual change in freshwater input was anatural response of the model, indicating that a 2.7-Myr stratifica-tion event may have been a threshold response to a gradual changein forcing.
Previous work on the connection between water vapour supplyand Northern Hemisphere glaciation has focused on the NorthAtlantic. The connection between deep convection and meridionalheat transport in the North Atlantic has represented a centralmotivation for this focus33. It is obvious that the North Pacific, alarge oceanic region that is upstream of North America in
Figure 2 Output for two equilibrium states of the CLIMBER-2 Earth system model.
a, Seasonal variation in North Pacific SST at 558 N (blue) and 458 N (red). The solid lines
correspond to the ‘modern’ (stratified) equilibrium, the dashed lines to the ‘Pliocene’
(unstratified) equilibrium state. b–d, ‘Modern’ minus ‘Pliocene’ differences in seasonal
variation of temperature (b) and precipitation (as snow and rain) zonally averaged over
Northern America at 608 N (d), and evaporation from the North Pacific at 558 N (c).
Figure 3 Simulated area of permanent snow cover (shaded) for the ‘cold orbit’
configuration in the destratified (a) and stratified (b) equilibria. b, Isolines (green) show
differences in annual surface air temperature between climate states corresponding to
stratified and destratified equilibria under the ‘cold orbit’ configuration.
atmospheric circulation, could play a critical role in the develop-ment of Northern Hemisphere glaciation. Ironically, it was theisolation of the subarctic Pacific surface from the ocean interior thatset the stage for major Northern Hemisphere glaciation at2.7Myr. A
Received 18 October; accepted 30 December 2004; doi:10.1038/nature03332.
1. Berger, A. & Loutre, M. F. Insolation values for the climate of the last 10 million years. Quat. Sci. Rev.
10, 297–317 (1991).
2. Shackleton, N. J., Hall, M. & Pate, D. Pliocene stable isotope stratigraphy of Site 846. Proc. ODP Sci.
Res. 138, 337–357 (1995).
3. Cane, M. A. & Molnar, P. Closing the Indonesian seaway as a precursor to East African aridification
around 3–4 million years ago. Nature 411, 157–162 (2001).
4. Ravelo, A. C., Andreasen, D. H., Lyle, M., Olivarez Lyle, A. & Wara, M. W. Regional climate shifts
caused by gradual cooling in the Pliocene epoch. Nature 429, 263–267 (2004).
5. Haug, G. H. & Tiedemann, R. Effect of the formation of the Isthmus of Panama on Atlantic Ocean
Over the past 50 million years, theEarth’s climate has been cooling(Fig. 1). Although Antarctica has
been glaciated for at least the past 35 millionyears1, large ice sheets did not appear in theNorthern Hemisphere until about 2.7 mil-lion years ago. Earth scientists largely agreethat overall climate cooling is associatedwith decreasing levels of carbon dioxide inthe atmosphere2,3, and that ice sheets canonly grow if sufficient moisture is availableand winter snow survives the summer heat4.But what triggered the onset of the ice ages 2.7 million years ago? Explanations havefocused on continental temperatures4,5, withidentification of potential moisture sourcesfrom the Atlantic5,6, but there remain manyopen questions7.
Haug et al.8 (page 821 of this issue) con-tribute an important piece to the ice-agepuzzle. Geochemical evidence suggests that,2.7 million years ago, the seasonal tempera-ture contrast of the subarctic Pacific Oceansea surface became larger as summerswarmed and winters cooled.Warmer summersea-surface temperatures result in a warmeratmosphere that can hold more moisture.Like a snow gun blasting away at ski slopes,westerly winds blow the moisture onto thecold North American continent where it falls as snow and accumulates as ice (Fig. 1,inset; Fig.2,overleaf).
Haug et al. have combined geochemicalexpertise with numerical modelling to pre-sent an integrated approach to the origin ofthe ice ages.Evidence comes from the floor ofthe subarctic Pacific Ocean, on which theremains of certain species of marine plank-ton (diatoms, coccolithophores and foram-inifera) have accumulated over time. Theprimary evidence for summertime warming2.7 million years ago stems from the bio-chemistry of coccolithophores, which variesaccording to temperature9. Augmenting this well-established index are the 18O/16Oratios in the siliceous tests of diatoms, acomparatively more complex measure ofpalaeotemperatures10.
At first glance, the results from these tworecorders contrast with other climate indica-tors in this region. Foraminiferal 18O/16Oratios — a classical indicator11 — from thesame deep-sea sediments suggest sea-surfacecooling 2.7 million years ago. This particularevidence is corroborated by perhaps themost intuitive indicator of climatic cooling,
an increase in the amount of debris of conti-nental origin delivered to the site by icebergs.
How can these apparently contradictoryobservations be reconciled? Haug et al.8
point to seasonal changes in the biologicalcommunities of the subarctic Pacific Oceanwhere, in modern times, different planktoncommunities populate the various seasons.Coccolithophores and those species ofdiatoms used for the geochemical analysesprefer the warm ocean surface of latesummer and autumn. The particular fora-miniferan species used for analysis, on theother hand, are more prolific during latewinter and spring when the sea surface is fertile due to mixing with deeper, colder,nutrient-rich water.
Thus the two seemingly opposing tem-perature trends 2.7 million years ago simplyreflect an increase in seasonality in the sub-arctic Pacific Ocean, which is consistent withother reconstructions of events in the NorthPacific Ocean7. This then provides the con-figuration on which to build an ice age:
late winter cooling reflects climate cooling,allowing snow to accumulate; late summerwarming increases the atmosphere’s poten-tial to hold moisture and to load the snowgun (Fig.1, inset).
What,then,caused the sudden increase inlate summer temperatures? To answer thisquestion, Haug et al.8 refer to the physicalproperties of sea water itself.Water has a highheat capacity, which means that the surfaceocean remains warm long after overlying airand adjacent land masses have cooled. Ifthere is mixing of the surface ocean withdeeper and cooler water, however, surfacewaters cannot warm up. This was the situa-tion before 2.7 million years ago, evidencefor which comes from the high accumula-tion rates of diatom skeletal remains at thestudy site, implying vigorous diatom pro-ductivity in the overlying sea surface andtherefore a continuous supply of nutrientsfrom deeper waters.
At 2.7 million years ago, the abundance of diatom remains plummets, suggesting a
Snow maker for the ice agesKatharina Billups
In the Northern Hemisphere, large-scale glaciation was initiatedcomparatively recently. Paradoxically, it seems that the trigger was aseasonal warming of the sea surface in an upwind oceanic region.
Figure 1 Global climate change over the past 60 million years. This record, showing a mainly coolingtrend, is inferred from foraminiferal oxygen-isotope records from all major ocean basins1 with18O/16O ratios plotted as per mil deviation from a standard (�18O). The horizontal grey bars indicatethe relative extent of polar ice sheets — light grey, ice volumes less than half of the maximum extent;dark grey, ice volumes close to the maximum extent (after ref. 1). Haug et al.8 provide evidence thatthe initiation of the Northern Hemisphere ice ages, 2–3 million years ago (arrow), was linked to thedevelopment of a stratified sea surface in the subarctic Pacific, which resulted in warmer sea-surfacetemperatures in late summer. Inset, the warmer sea surface was a source of moisture for the overlyingatmosphere, with westerly winds loading the snow gun that produced large ice sheets on the NorthAmerican continent. The star indicates the authors’ study site8.
decrease in the nutrient availability, like thatbrought about by the sea surface being cut offfrom the deeper ocean, at least on a seasonalbasis. At the same time, the reduction in ver-tical mixing allows the sea surface to warm.Thus the development of a seasonally lay-ered, or stratified, surface ocean 2.7 millionyears ago, which was probably a regionalresponse to the large-scale climatic changesat this time12, allowed late summer/autumnwarming of the sea surface and provided amoisture source for ice growth.
Haug et al.8 test the interpretations of thegeochemical records with a suite of numeri-cal computer-model experiments. The sim-ulated ocean is ‘stratified’ and ‘destratified’to determine whether this mechanism canaccount for the geochemically derivedchanges in temperature. And it can. Thestratified model state produces more extremeseasons and a larger North American icesheet than does the destratified model.
This is an exemplary study. The individ-ual climate indicators may not have with-stood the uncertainties and assumptionsthat limit each of them, but put together byHaug et al. they tell a cogent story of theorigin of the ice ages. ■
Katharina Billups is at the College of MarineStudies, University of Delaware, 700 PilottownRoad, Lewes, Delaware 19958, USA.e-mail: [email protected]. Zachos, J. et al. Science 292, 686–693 (2001).
2. Raymo, M. E. & Ruddiman, W. F. Nature 359, 117–122 (1992).
3. Pearson, P. N. & Palmer, M. R. Nature 406, 695–699 (2000).
4. Milankovitch, M. Serb. Akad. Beogr. Spec. Publ. 132 (1941).
5. Haug, G. H. & Tiedemann, R. Nature 393, 673–676 (1998).
6. Driscoll, N. W. & Haug, G. H. Science 282, 436–438 (1998).
7. Ravelo, A. C. et al. Nature 429, 263–267 (2004).
8. Haug, G. H. et al. Nature 433, 821–825 (2005).
9. Wefer, G., Berger, W. H., Bijma, J. & Fischer, G. in Use of Proxies
in Paleoceanography (eds Fischer, G. & Wefer G.) 1–68
(Springer, Berlin, 1999).
10.Brandiss, M. E., O’Neil, J. R., Edlund, M. B. & Stoermer, E. F.
Geochim. Cosmochim. Acta 62, 1119–1125 (1998).
11.Emiliani, C. J. Geol. 63, 538–578 (1955).
12.Haug, G. H. et al. Nature 401, 779–782 (1999).
Hearing
Aid from hair forceCorné Kros
Mammals hear with exquisite sensitivity and precision over a hugerange of frequencies; tiny amplifiers in the inner ear make this possible.New results challenge current thinking on how these amplifiers work.
Our ability to hear relies on cells in theinner ear called hair cells — namedafter the bundle of 100 or so hair-
like projections that protrudes from theirupper surfaces. Sound bends the hair bun-dles, causing small electrical (‘transducer’)currents to flow, which in turn makes thehair cells signal the reception of sound tothe brain. In mammals, the silent majorityof the hair cells (the outer hair cells) do nottalk to the brain, instead helping the innerhair cells — the true sensory receptors — todo so with more clarity than they could
achieve by themselves. But how is this done?For two decades scientists have sought
the answer in the extraordinary ability of theouter hair cells to change their length rapidlywhen stimulated. Now, however, Kennedy,Crawford and Fettiplace1 (page 880 of thisissue) and Chan and Hudspeth2 (in NatureNeuroscience) present provocative evidencethat the main component of the elusive‘cochlear amplifier’ may instead reside in thehair bundles of the outer hair cells.
Sound waves that reach the ear lead tovibrations inside the cochlea — a fluid-filled,
coiled tube forming the auditory part of theinner ear (Fig. 1a). The sensory hair cellsreside in a thin strip of tissue, the organ ofCorti, that is wedged between two mem-branes of the cochlea. The vibrations cause ashearing motion between these two mem-branes, which bends the hair bundles. Like arolled-up piano, one end of the organ ofCorti vibrates best at low frequencies and theother at high frequencies. In normal ears, thevibration is boosted and sharpened for softsounds by what has become known as thecochlear amplifier3.
Twenty years ago, a remarkable discoveryby Brownell and colleagues4 seemed to showwhat the cochlear amplifier is made of: theyfound that electrical stimulation of the outerhair cells made them lengthen and shortentheir cell bodies. The idea is that, in vivo, theelectricity produced by bending the hairbundles would drive this lengthening andshortening, or electromotility, as fast assound could vibrate the bundles. The strate-gic position of the outer hair cells wouldlocally boost the vibration of the organ ofCorti, and in this way stimulate, by fluidcoupling, the bundles on inner hair cells.
At a molecular level, this mechanism isthought to rely on a motor protein calledprestin, named from the musical term for avery fast tempo. The basolateral membranesof the outer hair cells are packed with thisprotein5, which changes shape as fast as youcan change the voltage across the membrane,over a range of frequencies up to at least 100kHz (ref. 6). But there is a snag: althoughprestin is quick, it is not clear whether thetransmembrane voltage in vivo changesmuch over the period of the sound wave, atsound frequencies greater than a few kilo-hertz. This is because the receptor potentialdue to the transducer currents is severelyattenuated at higher frequencies by theelectrical impedance of the cell7.
An alternative source of force that is notvoltage-dependent may thus be needed topower the cochlear amplifier. Kennedy andcolleagues1 report large forces generated bythe hair bundles of rat outer hair cells in vitro,when stimulated by a flexible glass fibre.Youwould expect the tip of the fibre to move lesswhen attached to the bundle than when it isfreely moving in the fluid.So there must havebeen disbelief in the lab when, in some cases,the fibre moved further when coupled to thehair bundles, implying that a force in thebundle drags the fibre along, instead of theother way round.
This force — an order of magnitude largerthan the force that is a necessary by-productof opening the ion channels through whichthe transducer current enters hair cells8 — isnot there at the moment the hair bundle is moved by the fibre, but develops within a fraction of a millisecond. Its time course is closely coupled to that over which thetransducer current adapts to a steady
Figure 2 Icy evidence in the Northern Hemisphere: a present-day ice-sheet on the Svalbard Islands.
37’ was measured by analyzing the organic extracts from sediment samples by gaschromatography coupled to chemical ionization mass spectrometry (GC-CIMS). This techniquehas the advantage over conventional procedures using gas chromatography with flame ionizationdetector (GC-FID) that it is more sensitive and selective to determine the relative abundance ofthe alkenones (Rosell-Melé et al., 1995). The technique of GC-CIMS also yields UK
37’ resultswhich are independent of the concentration of the alkenones in the analyzed aliquots by GC overa wider range of concentrations than by GC-FID (Rosell-Melé et al., 1995). Moreover, bothtechniques yield results which are intercomparable over a range of UK
37’ values (Rosell-Melé etal., 2001).
Rosell-Mele, A., Carter, J., Parry, A. & Eglinton, G. Novel procedure for the determination of theUK
37 index in sediment samples. Analytical Chemistry 67, 1283-1289 (1995).Rosell-Mele, A., Bard, E., Emeis, K.C., Grimalt, J.O., Müller, P., Schneider, R., Bouloubassi, I.,
Epstein, B., Fahl, K., Flügge, A., Freeman, K., Goni, M., Guntner, U., Hartz, D., Hellebust,S., Herbert, T., Ikehara, M., Ishiwatari, R., Kawamura, K., Kenig, F., de Leeuw, J., Lehman,S., Mejanelle, L., Ohkouchi, N., Pancost, R.D., Pelejero, C., Prahl, F., Quinn, J., Rontani,J.F., Rostek, F., Rullkötter, J., Sachs, J., Blanz, T., Sawada, K., Schultz-Bull, D., Sikes, E.,Songzoni, C., Ternois, Y., Versteegh, G., Volkman, J.K. & Wakeham, S. Precision of thecurrent methods to measure the alkenone proxy UK
37’ and absolute alkenone abundance insediments: Results of an interlaboratory comparison study. Geochemistry GeophysicsGeosystems 2, U1-U28 (2001).
Oxygen isotopes of diatom opalSamples from the 75-150 µm size fraction at ODP site 882 were analysed for δ18O(diatom) havingbeen cleaned using 30% H2O2 and 5% HCl to produce material over 97% free of non-diatomcontamination. c.98% of the total volume of biogenic silica in all samples was comprised ofCoscinodiscus marginatus (Ehrenb.) and Coscinodiscus radiatus (Ehrenb.). Annual peak fluxesof both C. marginatus and C. radiatus in the modern Pacific occur in the autumn/early winter(Takahashi, 1986; Takahashi et al., 1996) resulting in the δ18O(diatom), two diatom species, signalhere being representative of these conditions. Changes in the relative abundance of C. marginatusand C. radiatus are not coeval with changes in δ18O(diatom) indicating inter-species vital affects tobe negligible.
Diatom hydrous layers were stripped during a pre-fluorination outgassing stage in nickel reactiontubes using a BrF5 reagent at 250ºC for six minutes before full reaction at 450ºC. After 14 hours aplateau in sample δ18O(diatom) values was obtained, indicating full removal of oxygen from thehydroxyl and loosely bonded water in diatoms. Oxygen was converted to CO2 following themethodology of Clayton and Mayeda (1963) with δ18O(diatom) measured on a optima dual inletmass spectrometer. δ18O(diatom) values were converted to the SMOW scale using a within-runlaboratory standard (BFCmod) calibrated against NBS-28 and corrected according to Craig (1957).The standard error on the analysis of diatom oxygen is +/- 0.3per mil (2 sigma).
The hand-picked diatoms Coscinodiscus marginatus (Ehrenb.) and Coscinodiscus radiatus(Ehrenb.). are large and pristine (see supplementary methods Figure 1). These intricate andfragile diatom microfossils, had they been scraped off the shelves and transported by icebergs to
the sediment of Site 882, would not be nearly so well-preserved. In all likelihood, they would nothave survived as unbroken frustules. This large diatom is also extremely unlikely to undergoextensive seafloor transport, especially without being broken.
supplementary information Fig. 1 (top): Diatom Cosinodiscus marginatus picked for isotopeanalyses under a light microscope from ODP Site 882 (size = c.90 um in diameter). (base) SEMphoto from a hand-picked ODP 882 Cosinodiscus marginatus with a 10 µm scale bar (black line).
Juillet-Leclerc, A. & Labeyrie, L. Temperature dependence of the oxygen isotope fractionationbetween diatom silica and water. Earth and Planet. Sci. Lett. 84, 69-74 (1987).
Shemesh, A., Charles, C. D. & Fairbanks, R. G. Oxygen isotopes in biogenic silica: globalchanges in ocean temperature and isotopic composition. Science 256, 1434-1436 (1992).
Brandriss, M. E., O'Neil, J. R., Edlund, M. B. & Stoermer, E. F. Oxygen isotope fractionationbetween diatomaceous silica and water, Geochimica et Cosmochimica Acta 62, 1119-1125(1998).
Clayton R. N., O’Neil, J. R. & Mayeda T. K. Oxygen isotope exchange between quartz andwater. Journal of Geophysical Research 77, 3057–3067 (1972).
Craig H. Isotopic standards for carbon and oxygen and correction factors for mass spectromaticanalysis of carbon dioxide. Geochimica et Cosmochimica Acta 12, 133–149 (1957).
Takahashi, K., Seasonal fluxes of pelagic diatoms in the subarctic Pacific, 1982-1983. Deep-SeaResearch 33, 1225-1251 (1986).
Takahashi, K., Hisamichi, K., Yanada, M. & Maita, Y. Seasonal changes of marinephytoplankton productivity: a sediment trap study. Kaiyo Monthly 10, 109-115 (1996). [InJapanese]
Climate Modeling
Model: Earth system model CLIMBER-2 includes processes associated with the atmosphere,ocean, sea ice, vegetation and land surface (Petukhov et al., 2000). Atmosphere and land surfacehave 10o latitudinal and 51o longitudinal resolution. The World Ocean is represented by threemeridional sectors, Atlantic, Pacific, and Indian, with meridional resolution of 2.5o and 22 unevenvertical levels. The model also includes an ice sheet component with a resolution 0.75o in latitudeand 1.5o in longitude (Calov et al. 2004). In this study, we did not use the full model of ice sheetdynamics, instead simulating only the area of perennial snow cover in high resolution, throughthe snow pack model imbedded into the ice sheet model. An important addition to the oceaniccomponent of the model is a dependence of the summer mixed layer depth on verticalstratification. It is prescribed that, in the presence of strong vertical stratification, summer mixedlayer depth is 25 m, while it is set to 50 m otherwise. During the winter period, the depth ofmixed layer is computed using the standard convective adjustment procedure.
The seasonality of the mixed layer is a higher-resolution process than paloceanographicallyfocused climate models such as CLIMBER-2 previously been expected to included, given thelong run times involved. The seasonal cycle of the mixed layer does depend strongly on thelarger-scale, year-round vertical density gradient of the upper ocean, so we are absolutelyjustified in including this physical dependence into the model. Nevertheless, two importantquestions arise.
First, is our parameterization appropriate? Fig. 3a shows that mixed layer depth derived fromdensity criteria (Levitus World Ocean Atlas) is approximately twice as large in the northern partof North Atlantic (latitudinal belt 50 to 60oN) as compared to the North Pacific during the ‘warmseason’ (May to October, similar relationship is also held for the individual months). Thisdifference cannot be attributed solely to the wind speed, because it differs only a little betweenthe Atlantic and Pacific oceans (Fig. 2c). Instead, both within each basin and between the twobasins, the mixed layer depth is strongly (inversely) correlated with the vertical stratification,reflected here as the annually averaged density gradient between surface and 100 m depth. Thevertical stratification in the Pacific is higher by a factor of 3 and more than that in the Atlantic,and the summer mixed layer is correspondingly about half as deep. This justifies theparameterization which we employ in the model for this study. .
Second, how critical is the mixed layer parameterization to the model results? In supplementaryinformation Fig. 3, results of two different experiments with modern (stratified Pacific) arecompared with Pliocene (destratified Pacific) case. The blue curve represents the experiment withstratification-dependent mixed layer depth, while the green curve represents simulation ofstratified North Pacific but without the added summer mixed layer depth sensitivityparameterization (constant 50 m mixed layer depth). In the experiment with constant 50 m mixedlayer, the SST in the North Pacific is overall colder compared to the stratification-dependentmixed layer depth. Nevertheless, it is clear that stratification, with or without the mixed layerparameterization, increases the amplitude of the seasonal SST cycle relative to the destratifiedcase (solid lines versus dashed line in panel a). As a result, both evaporation from the North
Pacific and precipitation over North America are affected by stratification, in the way describedin the manuscript. However, to actually observe summer time warming in the face of wintercooling (rather than just less summer cooling than winter cooling), the mixed layerparameterization is needed. That is, the summer mixed layer parameterization improves themodel fit to the data. This shows that shallowing of the summer mixed layer in the North Pacificplays an important role in the seasonal SST cycle and in its effect on water vapor transport to thecontinent. From our perspective, these results indicate that (a) the link between stratification andthe amplitude of seasonal SST cycle is a robust model result but that (b) mixed layer depthconsiderations are required in the model if it is to fit the data quantitatively. Given thatourparameterization is a sensible one based on modern ocean data (see above), we include it in themodel runs that are shown in the manuscript.
supplementary information Fig. 2 Longitudinal distribution of the averaged between 50oN and60oN (a) ‘warm season’ (May-October) mixed layer depth (Levitus World Ocean Atlas); (b)annually averaged vertical density gradient between surface and 100 m depth (same stronganticorrelation noted with (a) if densities at the surface and 300 m are compared); and (c) surfacewind speed in warm season
supplementary information Fig. 3 The same as Fig. 2 in the main manuscript, with the exceptionof the green curve, which shows (a.) modern conditions with a constant 50 m mixed layer depth,and (b-d.) the difference between modern and ‘Pliocene’ conditions for the standard version ofthe model (constant 50 m mixed layer depth).
Experiments. The CLIMBER-2 model simulates the modern hydrological regime of the NorthPacific, which is characterized by a strong halocline and very shallow convection in winterseason. To produce a “destratified” North Pacific, we applied anomalous negative freshwaterforcing of –0.2 Sv uniformly distributed over the North Pacific. A corresponding amount offreshwater was added to the rest of the World Ocean to conserve global salinity. Under suchanomalous freshwater flux, the halocline disappears in the North Pacific and intermediate watermasses are formed via mid-depth convection north of 50oN. A gradual decrease in the magnitudeof anomalous freshwater flux results in small changes in the North Pacific until a critical value ofabout -0.13 Sv is crossed, at which point permanent stratification develops, associated withstronger winter cooling and enhanced seasonality of SST variations in the northern part of theNorth Pacific.
To analyse the role of permanent stratification in the North Pacific for the seasonal variation oftemperature and the atmospheric hydrological cycle (Fig. 2 a-d), we compare the experiment withto anomalous freshwater flux (-0.2 Sv) to the simulation of the modern climate state. In bothcases, we use the same preindustrial CO2 concentration (280 ppm) and the same orbitalconfiguration (present day).
To assess the significance of permanent stratification in the subpolar North Pacific for the build-up of ice sheets in the Northern Hemisphere, we performed additional experiments using anextreme orbital configuration called “cold orbit” (Fig. 3). In this configuration, the orbital
eccentricity is 0.06 (close to its maximum value), and the Earth is in aphelion at the northernsummer solstice. As in all experiments, we use preindustrial CO2 concentration.
In the transient experiment shown in Fig. 1a, we start from the equilibrium climate statecorresponding to the anomalous freshwater flux –0.2 Sv (“destratified” Pacific) and orbitalparameters corresponding to 3.1 million years BP. We run the model for 600 Kyr changing onlyorbital configuration following Berger and Loutre (1991) and gradually reducing anomalousfreshwater flux to the North Pacific with a uniform rate of about 0.2 Sv per million years.
Berger, A. & Loutre M.F. Insolation values for the climate of the last 10 million years.Quaternary Science Reviews 10, 297-317 (1991).
Calov, R., Ganopolski, A., Claussen, M., Petoukhov, V. & Greve, R. Transient simulation of thelast glacial inception. Part I: Glacial inception as a bifurcation in the climate system.Submitted to Climate Dynamics (2004).
