lable at ScienceDirect

Quaternary Science Reviews 107 (2015) 11e24

Contents lists avai

Quaternary Science Reviews

journal homepage: www.elsevier .com/locate/quascirev

Invited review

Quaternary glaciations: from observations to theories

Didier PaillardLaboratoire des Sciences du Climat et de l'Environnement, IPSL, CEA-CNRS-UVSQ, Centre d'Etudes de Saclay, 91191 Gif-sur-Yvette, France

a r t i c l e i n f o

Article history:Received 2 July 2014Received in revised form3 October 2014Accepted 4 October 2014Available online 4 November 2014

Keywords:QuaternaryPaleoclimatologyIce agesModelsCarbon cycle

E-mail address: didier.paillard@lsce.ipsl.fr.

http://dx.doi.org/10.1016/j.quascirev.2014.10.0020277-3791/© 2014 Elsevier Ltd. All rights reserved.

a b s t r a c t

Ice ages are known since the mid-nineteenth century. From the beginning, they have been at the centerof theories of climate and climate change. Still, the mechanisms behind these large amplitude oscillationsremain poorly understood. In order to position our current knowledge of glacialeinterglacial cycles, it isuseful to present how the notion of climate change appeared in the XIXth century with the discovery ofglacial periods, and how the two main theories, the astronomical one and the geochemical one, emergedprogressively both from sound physical principles but also from extravagant ideas. Major progresses ingeochemistry in the XXth century led first to the firm evidence of an astronomical pacemaker of thesecycles thanks to the accumulation of paleoceanographic data. Still, the Milankovitch's theory predicts anice age cyclicity of about 41,000 years, while the major periodicity found in the records is 100,000 yr.Besides, ice cores from Antarctica proved unambiguously that the atmospheric carbon dioxide was lowerduring glacial periods. Even more importantly, during the last termination, the atmospheric pCO2 in-creases significantly by about 50 ppm, several millenia before any important change in continental icevolume. This fact, together with many other pieces of information, strongly suggests an active role ofgreenhouse gases in the ice age problem, at least during deglaciations. Since terminations are precisely atthe heart of the 100-ka problem, we need to formulate a new synthesis of the astronomical andgeochemical theories in order to unravel this almost two-century-old question of ice ages. The foun-dations of such a theory have already been put forward, and its predictions appear in surprisingly goodagreement with many recent observations.

© 2014 Elsevier Ltd. All rights reserved.

1. Introduction

It is too often stated in textbooks but also in many scientific pa-pers, that Quaternary glaciations are “caused” by the astronomicalforcing, according to theMilankovitch theory. Consequently, there issometimes a diffuse feeling, even within scientific communitiesworking on climatic questions, that this problem was fully settledseveral decades ago. But if indeed ice ages are “paced” by the as-tronomy, the key physical mechanisms are far from being under-stood. On the other extreme, quite too often, some recent papers arediscussing aspects of glacial cycles, like the role of the differentorbital parameters on Earth's climate, without providing any cluetowards the 150-year long historical context of theories and dis-coveries ofQuaternary Sciences. Butexplaining thatweare “standingon the shoulders of giants” is not only a question of acknowledg-ment: it is also a necessity to articulate our understanding of a sci-entific question. As a result, there is often some confusion amongstudents and non-specialists, on what is the current knowledge onglacial cycles and what are the key questions. In the following, I will

therefore try to providea shorthistorical accountof thedevelopmentof ideas on ice ages and how they relate to the building of climatesciences in general. In particular, I will present the birth and thedevelopment of the twomajor theories of ice ages, the astronomicalone and the geochemical one, as well as the development of themajor paleoclimatic reconstruction techniques, and how these newdata have shifted the balance of evidence on one or the other side.Then, I will show why we currently need a synthesis of these twotheories and finally I will provide a track towards this goal.

2. Historical background

Though it is probably difficult to pinpoint the emergence of thenotion of Earth's climatic change in the history of Sciences, it isinteresting to note that the evidence of a changing environment wasdescribed since the Antiquity. For instance, in the context of the as-tronomical theory of ice ages, Aristotle's writings (Meteorologica,Book 1, Chapter 14, translation by Lee, 1951) might seem quiteprophetical:

“The same parts of the earth are not always moist or dry, butchange their character according to the appearance or failure of

D. Paillard / Quaternary Science Reviews 107 (2015) 11e2412

rivers. ...sea replaces what was once dry land, and where there isnow sea, is at another time land. This process must however besupposed to take place in an orderly cycle.”

Then he speculates on the causes of such changes:

“we should suppose that the cause of all these changes is that,just as there is a winter among the yearly seasons, so at fixedintervals in some great period of time, there is a great winterand excess rain”.

Quite clearly Aristotlewas not talking of ice ages, whichwere notknown at this time. He describes hydrological changes that wereeither observedhistorically, or inferredmore indirectly, inEgyptor inGreece, and insists that the changes aremostly local or regional ones,with different regions experiencing sometimes opposite changes.This would be now called evidences of local or regional climaticchanges. Still, Aristotle's writings were certainly very influencial inthe construction ofmodern science in general, including geology. Forinstance, the above short sentence is cited by C. Lyell (1830) as a“theory of periodical revolutions of the inorganic world”. The sameword “revolution” is used by Cuvier in his book “Discours sur lesr�evolutions de la surface du globe” (1825) (A Discourse on the Rev-olutions of the Surface of theGlobe). The use of theword “revolution”is quite meaningful, and the same word is also used by Adh�emar(1842) for the title of his book on the first astronomical theory ofice ages (“R�evolutions de la mer, d�eluges p�eriodiques”). It probablyaims at reminding, in someway, Copernicus' book “De revolutionibusorbium coelestium”which is oftenpresented as the foundation stone,onwhich Kepler and Newton developed modern physics. There wasobviously among natural scientists a strong desire to build Geologyon a similarly footing, ie. the occurrence of cyclic changes, whoseunknown ultimate causes might possibly be related to some “cos-mic” phenomena, either astronomical or divine depending on au-thors. Cuvier's book presents his theory of cataclysmic changes thatpaced the succession of fossils, which together with the work ofWilliam Smith, led to the foundation of stratigraphy. Since the An-tiquity, geologists had foundmarine fossils over lands, even on top ofmountains. This led to centuries of discussions on how these mighthave formed, and the dominant view at the beginning of the XIXthcentury was that recurrent floods or catastrophes occurred in thepast, the last one being the “Great Flood” from the Bible. Looking foran explanation of these “revolutions” was a scientific challenge, asnoted by the astronomer John Herschel (1830):

“Impressed with the magnificence of that view of geologicalrevolutions which regards them rather as regular and necessaryeffects of great and general causes, than as resulting from aseries of convulsions and catastrophes regulated by no laws andreducible to no fixed principles, the mind naturally turns tothose immense periods with whose existence in the planetarysystem the astronomer is familiar”.

Still, the notion that climate might change through time, inparticular on a rather large or even a global scale, was controversialat the beginning of the XIXth century. Many fossils from formertropical environments were found in high northern latitude, buttheir interpretation was not necessarily straightforward. Forinstance, if mammoths, ie. “siberian elephants”, were often cited asa proof of a former warm “african type” climate in arctic regions,such analogies were also strongly (and rightly) criticized by otherscientists (eg. J. Fleming, 1829). The dominant view was, never-theless, that the Earth probably experienced a gradual cooling, froma hot Paleozoic time period, with giant ferns growing even in arctic

regions as evidenced in the coal mines, to a warm Eocene and thento the present-day temperate climate. This idea was also somewhatin accordance with the “plutonist” view that rocks formed in fireand that the Earth was initially molten: For instance, in Les �epoquesde la nature (1778) Buffon computed an age of the Earth based onthe cooling rate of iron, which led to an age of 75,000 years (thoughwhen accounting for the slow diffusion of heat, Lord Kelvin in 1897obtained an age between 20 and 40 millions of years).

In this general context, the discovery of ice ages was a crucialstep forward. Indeed, for the first time the evidences of climaticchange were not based on paleontological interpretations, but onmuch less ambiguous, physically based observations, like moraines,erratic boulders, or glacial striations. It furthermore strongly sug-gested that climate evolution was not a simple long term coolingtrend, as usually believed.

3. Discovery of ice ages and early theories of climate

3.1. The evidence of past glaciations

The erratic blocks found in lots of places in the Alps and innorthern Europe had been subject to many speculations for a longtime, often involving giants or trolls, while the usual scientificexplanation involved again catastrophic floods or diluvium. But anew suggestion slowly emerged in the first half of the XIXth cen-tury: glaciers should be the cause. The morphology and geology ofthe Alps was indeed investigated by swiss scientists, in particular byVenetz (1833), de Charpentier (1836) and Agassiz (1840), who havecarefully described and analyzed many morphological features inthe Swiss Alps, with the unambiguous conclusion that the only validexplanation for moraines, erratic boulders, or glacial striations,involved episodes of significant glacial advances in the past. Similarobservations and suggestions were also made in Scandinavia byEsmark (1827) (Andersen, 1992) or in Scotland by Jameson, whounfortunately did not publish his findings. A very nice and detailedaccount of the history of this discovery is given in Berger (1988,2012) or in Bard (2004) or more recently in Krüger (2013) orWoodward (2014).

As noted by Lyell, geological theories were still too often associ-ated with cosmogonies and philosophical preconceptions. In thisrespect, in contrast to the plutonists, the neptunists favored the ideathat all rocks are sedimentary in origin and were formed in water.Theywere alsomore inclined to follow the ideaof a cold origin for theEarth, like the famous poet, but also minister of mining, J. W vonGoethe, a proponent of ice ages, who presents his doctor Faust infavorof neptunism,whileMephistopheleswas, of course, a plutonist.Thoughnecessarily very speculative, apossible scientific explanationfor such glacial periods was put forward by J. Esmark. He notes thataccording to William Whiston's theory, Isaac Newton's successor atCambridge University, the Earth was initially a “comet” on a veryeccentric orbit. It was therefore in a frozen statewhen far away fromthe Sun. Accordingly, ice ages would occur when the Earth was veryyoung, something which soon will be proved wrong. This wasnevertheless probably the first suggestion of a possible link betweenthe eccentricity of the Earth and ice ages. But it stoods somewhat incontradiction with astronomical knowledge at this time. J. Herschel(1830) states that “Geometers have demonstrated the absolute invari-ability” of the Earth'smajor axis, and therefore the length of the year:Increasing the eccentricity would only warm up the mean climate,since “the total quantity of heat received by the earth from the sun inone revolution is inversely proportional to the minor axis of theorbit”. He further adds that for any significant mean annual change,the seasons would be so extreme that they would “produce a climateperfectly intolerable”. He also considered that neither obliquity norprecessional changes (see below) could account for significant

D. Paillard / Quaternary Science Reviews 107 (2015) 11e24 13

climatic changes (see eg. Paillard, 2010; for a brief presentation of theclimatically relevant astronomical parameters).

The present-day distribution of temperatures was also not wellexplained. At the beginning of the XIXth century, a big difficultywas to understand why the Southern hemisphere appears colderthan the Northern hemisphere, and in particular to account for therecently discovered Antarctica, a new continent that was still un-explored, but appeared covered by a huge amount of ice. A first ideawas to look for asymmetries in the astronomical forcing. C. Lyellwrites in his book (Princ of Geol Vol 1 p174):

“Before the amount of difference between the temperature of thetwo hemispheres was ascertained, it was referred by many as-tronomers to theprecessionof theequinoxes,or theaccelerationofthe earth's motion in its perihelium; in consequence of which thespring and summer of the southern hemisphere are now shorter,by nearly eight days, than those seasons north of the equator”.

But immediately continues, citing Herschel (1830):

“But Sir J. Herschel reminds us that … it is demonstrable thatwhatever be the ellipticity of the earth's orbit, the two hemi-spheres must receive equal absolute quantities of light and heatper annum, the proximity of the sun in perigee exactlycompensating the effect of its swifter motion”.

In other words, the astronomical forcing, or at least the pre-cessional one, cannot easily explain the current southern hemi-sphere climate, and Lyell was convinced that the present-daydistribution of temperatures was mostly explained in terms of thegeographical distribution of lands and seas, with less land area inthe southern hemisphere. Consequently, he speculates that climatemay have changed through time, due to the vertical movements ofthe Earth crust, as evidenced by marine fossils at high elevations:

“let the Himalaya mountains, with the whole Hindostan, sinkdown, and their place be occupied by the Indian ocean, while anequal extent of territory andmountains, of the same vast height,rise up between North Greenland and the Orkney islands. Itseems difficult to exaggerate the amount towhich the climate ofthe northern hemisphere would then be cooled”.

3.2. Astronomical theories

Though Adh�emar (1842) was aware of Herschel's objection tothe role of precession, he suggested that, if temperature depends onthe input of solar radiation, it might also depend on other factors.He therefore built the first ice age theory, based on precession asexposed above by Lyell, inwhich longer winters are associated withglacial periods, as “demonstrated” by the Southern hemispheretoday. But he went as far as suggesting that the southern oceanwascurrently much deeper than the arctic ocean, due to the gravita-tional attraction of Antarctic ice that he assumed to be almost100 km thick, therefore the small amount of emerged lands in thesouth since they are currently flooded or in a “diluvium” period.

Croll (1867) also discussed the link between glaciations and sealevel, but with more realistic assumptions concerning Antarctica:

“A mile of ice removed from off the antarctic continent (… )would raise the general level of the sea 206 feet. And to thismust be added the rise resulting from the displacement of theearth's centre of gravity, which in the latitude of Scotland mouldamount to about 74 feet”.

Though the periodicity of ice ages was not proven, it becameclear with the work of Archibald Geikie (1863) and his brother

James Geikie (1874) in Scotland, that a succession of colder andwarmer periods occurred. As explained by Croll (1864):

“The recurrence of colder and warmer periods evidently pointsto some great, fixed, and continuously operating cosmical law.(… ) The true cosmical cause must be sought for in the relationof our earth to the sun”.

Croll's theory of ice ages is a more elaborate version ofAdh�emar's theory. He particularly insists on the role of eccen-tricity, as a modulation of the precessional forcing, and intenseglaciations would therefore occur only when eccentricity is largeenough. Still, the mechanism by which precession could act wasnot well understood, though Croll suggests that “the great accu-mulation of ice and snow in the northern regions, arising from theseverity of the seasons, would tend to keep the air in the northernregions of the globe much colder”. Croll (1867) also recognizedthat the obliquity could play a major role. Though “the increase inobliquity would not sensibly affect the polar winter”, it would helpmelt the ice in the summer hemisphere. This latter idea appearsextremely close to Milankovitch's theory. Interestingly, Joseph J.Murphy (1869, 1876) suggested that cool summers, not coldwinters, were more important to grow glaciers, as evidenced bynumerous field data. He therefore suggested that Croll's theoryshould be reversed, with glaciations associated with summer atthe aphelion, in accordance with Milankovitch's theory and alsowith our current understanding of ice sheet mass balance. Murphy(1876) still notes that:

“There is one fact of physical geography, which seems at firstsight to support Mr. Croll's theory (… ). In the Antarctic regionsthere is a glacial climate now”

but concludes correctly that this was mostly the result of therepartition of land and sea, not the consequence of the astronom-ical forcing. In other words, the theoretical foundations of the as-tronomical theory of climatewere readily available in the late XIXthcentury. Still, these ideas were not properly quantified. The greatachievement of Milankovitch was to provide a rigorous computa-tion of all the necessary elements, from the astronomical planetaryparameters, down to regional seasonal insolation values. For thefirst time, it was therefore possible to discuss the respective role ofeach astronomical parameter e eccentricity, obliquity and preces-sion e on the incoming radiation. This paved the road for con-structing a chronology of ice ages, and finally to the success of histheory. He writes (1941):

“I have dealt with Croll's theory (… ) and I have found that itsinadequacy lies in the fact that the influence of the variability ofthe obliquity upon insolation is not sufficiently taken intoaccount”.

Indeed, the obliquity forcing is dominant at high latitudes, thusits influence on the timing of glaciations. This is even more truewhen integrating the forcing over a fraction of the year, as done byMilankovitch in his “caloric seasons” since, as explained above, theprecessional forcing fully disappears in annual mean. Interestingly,these questions are still subject to discussions now (eg. Huybersand Wunsch, 2005; Raymo et al., 2006).

3.3. Geochemical theories

A fundamental advantage of the astronomical theories of iceages was their ability to predict the timing of glaciations. This wasalso their weakness, as noted by Arrhenius (1896):

D. Paillard / Quaternary Science Reviews 107 (2015) 11e2414

“It seems that the great advantage which Croll's hypothesispromised to geologists, viz. of giving them a natural chronology,predisposed them in favour of its acceptance. But this circum-stance, which at first appeared advantageous, seems with theadvance of investigation rather to militate against the theory”

Croll's theory was indeed rejected mostly on the grounds thatgeological data on the last Glacial period indicated an age of about10e30 ka BP (before present), thus invalidating the expectationsfrom Croll's theory that the Last Glacial time was coïncident of theeccentricity maxima between 80 and 140 ka BP. Consequently, therole of past changes in atmospheric CO2 was considered an inter-esting and credible alternative theory by geologists.

Indeed, if astronomers or mathematicians were inclined to-wards astronomical theories of climatic change, chemists often hada different opinion. Since the work of J. Fourier (1824), it was sus-pected that greenhouse gases have a significant impact on Earthclimate. The first suggestion that the Earth carbon cycle might notbe constant through time and could possibly affect climate, wasprobably made by J. Ebelmen. In 1845, he described the majorgeochemical processes controlling the atmospheric composition ofcarbon dioxide and oxygen on geologic time scales. He then con-cludes that:

“many circumstances nonetheless tend to prove that in ancientgeologic epochs the atmosphere was denser and richer in car-bonic acid and perhaps oxygen, than at present. To a greaterweight of the gaseous envelope should correspond a strongercondensation of solar heat and some atmospheric phenomenaof a greater intensity. ”

Though his results were later mostly forgotten (see Berner andMaasch, 1996), this conclusion was further reinforced by the workof J. Tyndall. Tyndall was the first to measure the infrared absorp-tion and emission spectra of atmospheric gases and to discoverthat, while the major atmospheric constituents, nitrogen and oxy-gen, weremostly transparent to infrared radiations, the greenhouseeffect was due mostly to water vapor and carbon dioxide. Tyndallwas also an accomplished mountaineer: he was the first to climbthe Weisshorn swiss peak and wrote a book on the dynamics ofglaciers (1860). He was well aware of the climatic and geologicimplications of his spectrometric measurements (1861):

“Now if, as the above experiments indicate, the chief influencebe exercised by the aqueous vapor, every variation of this con-stituent must produce a change of climate. Similar remarkswould apply to the carbonic acid diffused through the air (… )Such changes in fact may have produced all the mutations ofclimate which the researches of geologists reveal. ”

This led the Swedish chemist Svante Arrhenius to perform thefirst computation of the role of CO2 on climate. As explained by S.Arrhenius (1896), his motivation was not theoretical, but wasaiming at answering the ice age problem:

“I should certainlynothaveundertaken these tedious calculationsif an extraordinary interest had not been connectedwith them. Inthe Physical Society of Stockholm there have been occasionallyvery lively discussions on the probable causes of Ice Age … ”.

He computed that a one-third reduction in atmospheric pCO2would result in a global 3 �C cooling, something which wasconsistent with the available data. He also explained the warmEocene climates by multiplying pCO2 levels by two or three, thus a

warming of 5 �Ce10 �C. Svante Arrhenius pointed out that, theatmosphere being a small carbon reservoir in contrast to the oceanor rocks, it is rather easy to change its concentration in the longterm by accumulating small imbalances, for instance betweenvolcanic sources, and chemical weathering, the major carbon sinkon geological time scales. It is worth emphasizing that, on all thesepoints, Arrhenius was basically correct, even if his climate modelwas erroneous (eg. Dufresne, 2009).

The american geologist Thomas Chamberlin also explained thatthere is no reason for the geological sources and sinks of carbon tocompensate, and even described (1897) what might be the firstcoupled carbon e ice-sheet self-sustained oscillatory model, withcolder oceans favoring organic preservation and thus pCO2lowering in a positive feedback, while in a later stage the ice-sheetadvance reduces silicate weathering, thus providing a mechanismto increase pCO2 again. A simple mathematical implementation ofthis idea is shown on Fig. 1 for illustration.

3.4. Some remarks

The available paleoclimatic data was very sparse in the XIXthand early XXth century. It was mostly qualitative and sometimesdifficult to interpret. Besides, the timing of the geological past waslargely an open question, continental drift was unknown, and thephysical mechanisms governing meteorology or oceanographywere not fully understood. It was therefore very difficult for sci-entists to sort out the different phenomena and to build theoriesfocused specifically towards only one of them. In retrospect, whilemany of these ideas are to some extent currently still valid ones, thetheories exposed by the scientists were too ambitious and there-fore, in the end, incorrect. For instance, Adh�emar was right inbuilding an astronomical theory of climate, but wrong to apply it to“periodic diluvia”, which led him to an extravagant hypothesis.Croll was right to include the effect of eccentricity in this astro-nomical theory, but wrong to explain the present day Antarcticclimate with it, which probably led him to reject Murphy's sug-gestion on the role of summer temperature on the ice sheets.

Similarly, proponents of the role of CO2 where eager to explainEocene climates as well as glacial climates in the same generalframework of small imbalances in the carbon fluxes. Opinions wereoften rather clear cut, probably to compensate for the sparsity ofgeological information. Still, a few scientists tried to accommodatein some ways for what each theory had best to offer on each point.For instance, Ekholm (1901) rejected Croll's theory based on pre-cession, being convinced that CO2 was responsible for ice ages, asexplained by Arrhenius. Still, he explained that smaller climaticvariations, such as the cooling observed in Scandinavia over theHolocene, was best accounted for by the decline in obliquity, aconclusion that certainly holds today (eg. Marchal et al., 2002). Heconcluded that:

“the present burning of pit-coal is so great that … it will un-doubtedly cause a very obvious rise of the mean temperature ofthe Earth… By such means also the deterioration of the climateof the northern and Arctic regions, depending on the decrease ofthe obliquity, may be counteracted. ”

4. From paleoceanographic discoveries to modernastronomical theories

4.1. Marine geology and isotope geochemistry

During the XXth century, major advances on the ice age problemconcerned the accumulation of many different paleoclimatological

Fig. 1. A “Chamberlin carbon-Ice sheet oscillator”. Solution of the coupled system { V' ¼ (C0eC)/tV; C' ¼ a C e (V0e V)/tc }, ie. the ice volume V(t) decreases when it is warm, i.e. whenatmospheric carbon dioxide concentration C(t) is high, while C(t) increases when the ice volume is large due to reduced silicate weathering (here C(0) ¼ 300 ppm; V(0) ¼ 120 m sealev equi., C0 ¼ 300, V0 ¼ 100, tV ¼ 10, tc ¼ 5, a ¼ 0). The organic matter feedback a is destabilizing: if a > 0, the model would spiral up in an unbounded fashion (if a < 0, theoscillations are damped). A more realistic model should include non-linear terms.

D. Paillard / Quaternary Science Reviews 107 (2015) 11e24 15

records, togetherwith theirmore andmoredetailed interpretations.Key ingredients were the development of marine geology, isotopicgeochemistry, geochronology and also plate tectonics, whichtogether provided the necessary general informations on ice ages. Inthis respect, the Swedish expedition on the Albatross (1947e1948)was a landmark event. Indeed, the slow accumulation of sedimentsin the bottom of the oceanwas long suspected to provide the time-continuous records needed for a better understanding of past cli-mates. The techniques used before, on the Challenger expedition(1872e1876), or on the german shipMeteor (1925e1927), could notrecover more than about 1 m of sediment. B. Kullenberg devised anew piston-corer that could easily provide sediment cores up to tenor 20 m long (eg. Pettersson, 1953). Equiped with this new device,theAlbatrosswas able to recover in herworld-roundexpedition over200 deep ocean sediment cores, and some of them were used asrosetta stones for the ice age problem. They were studied by GustafArrhenius, Svante Arrhenius' grandson, who was on board. Inparticular, he observed the evidence of very clear cycles in carbonatedeposition (Arrhenius, 1952).

At about the same time, Harold Urey, who received the nobelprice for his discovery of deuterium, discussed the thermody-namics of isotopic exchanges. He writes (1947):

“These calculations suggest investigations of particular interestto geology … Accurate determinations of the 18O content ofcarbonate rocks could be used to determine the temperature atwhich they were formed”.

Consequently, using some cores from the Swedish expedition,Emiliani (1955) performed the isotopic analysis of fossil forami-nifera, thus providing for the first time an unequivocal and quan-tified record of the succession of ice ages over the last few hundredthousands of years.

It is also at this time that the radiocarbon method was devel-oped, as well as methods based on thorium isotopes. This gaveEmiliani some chronological constraints, but only in the youngestpart of the records, with the Last Glacial maximum (LGM) at about20 ka BP, and a previous cold period at about 65 ka BP, which henamed stage 4. The ages of the older parts were based on simpleextrapolation, which finally led to results shown on Fig. 2.

Today it is clear that stage 2 and 4 are only parts of the sameglacial period within the last climatic cycle. Even if Emiliani (1955)notes:

“some inconsistencies. For example the lower value of theinsolation peak at �45,000 years corresponds to the low-temperature maximum of stage 3”

it was extremely tempting to associate the climatic cycles to theinsolation cycles from the Milankovitch theory. These are largelydominated by obliquity, therefore the 40 ka cyclicity of Emiliani'sice ages in 1955. A decade later, in a compilation of isotopic data,Emiliani (1966) suggested significantly older ages for the earlierstages, but the correlation with insolation, or more precisely“caloric seasons”, was still holding quite well, with ice age cyclescorresponding to either one or two obliquity cycle.

The association of light carbonate d18O isotopic values withwarmer temperatures was largely confirmed by many comple-mentary studies, in particular on micropaleontological faunal as-semblages (eg. Ericson and Wollin, 1956; Imbrie and Kipp, 1971).But its precise quantitative interpretation became more clear, withthe work of Shackleton (1967) who showed that, in contrast toEmiliani's assumptions, the isotopic content of the ocean was asignificant part of the signal. In addition to being paleother-mometers, marine isotopic curves were also rather direct records ofthe ice-sheets volume, ie. a global signal that could, in particular, beused for stratigraphy. This was further reinforced by the dating offossil coral reefs (Chappell, 1974; Chappell and Shackleton, 1986)which clearly confirmed the link between the marine isotopic re-cords and sea level changes.

4.2. Absolute chronology and the 100 ka problem

Dating has always been a central difficulty in geology. If manyradiometric techniques were developed in the 1940's and 1950's,reliable dates could only be obtained with technological advancesin mass-spectrometers and better controls on the geochemicalcontext of the samples. Consequently, a reliable chronology of theice ages became available only in the 1970's. In contrast to expec-tations, it became progressively clear that ice ages were not simplypaced by obliquity. Broecker and van Donk (1970) identified a clear

Fig. 2. a/summer insolation according to Milankovitch theory and b/the definition of marine isotopic stages with odd numbers for warm periods (redrawn from Emiliani, 1955); c/adecade later, though earlier stages are significantly older, they still correlate with the obliquity-driven insolation forcing (redrawn from Emiliani, 1966); d/this is no more the casewith a chronology comparable to present-day knowledge (redrawn from Emiliani and Shackleton, 1974) with two alternative time scales for termination II.

D. Paillard / Quaternary Science Reviews 107 (2015) 11e2416

asymmetric “sawtoothed” characteristic shape, with long periodsof glacial growth, followed by rapid deglaciations, or “terminations”.In addition to these typical cycles, there are secondary oscillationssuperimposed that closely parallel precessional changes. They alsonote that:

“The two preceding terminations appear to correlate with timesof maximum eccentricity necessary to the generation of pre-cession maxima. ”

In other words, the link between ice ages and the insolationforcing appeared at the same time more robust, but also morecomplex than initially thought. This is finally synthesized in thelandmark paper of Hays et al. (1976) which is often presented as the“proof” of the astronomical theory, since it is demonstrated thatastronomical periodicities are indeed present in the geological re-cords. As explained in the title, the astronomical forcing is, withoutdoubt, the “pacemaker of the ice ages”. The main 100-ka cycle ap-pears clearly associated to eccentricity, and more precisely thepaper demonstrates the “association of glacial times with intervals oflow eccentricity”. But this left entirely open the question of themechanisms involved, and the authors conclude that:

“the 100,000-year climate cycle is driven in some way bychanges in orbital eccentricity. As before, we avoid the obliga-tion of identifying the physical mechanism of this response, andinstead characterize the behavior of the system only in generalterms. Specifically, we abandon the assumption of linearity”.

This is the precise statement of the so-called “100-ka problem”:there is an obvious link between eccentricity and ice ages, but noobvious mechanism. In particular, once again, the expectationsfrom Croll are entirely reversed: if ice ages are indeed linked toeccentricity changes, they are associated with low values, not highones.

4.3. Ice-only theories

Climate is often presented as a “complex system” with manydifferent interacting components. Physicists, chemists or

astronomers in the XIXth and early XXth century were eager todiscuss theories on climate, but progresses in meteorology and,later, the appearance of numerical models of the atmosphere,introduced new severe constraints that were difficult to ignore.This might be the reasonwhy, in the 1970's, no scientist would darebuilding a numerical model of ice ages, even though the newlyavailable paleoceanographic data had provided some revolutionaryinsights into the question. The task was therefore taken up by ajournalist, or science writer, Nigel Calder, whowas thenworking ona BBC film on climate. He cautions that (1974):

“Meteorological processes are so notoriously nonlinear that myassumptions are almost frivolous”

But in retrospect, though the shape of the predicted cycles is notvery good, Calder's model (1974) actually did correctly predict thetiming of almost all terminations (see Fig. 3), something that manymuch more recent models are unable to reproduce. This timingseemed initially quite wrong, but the discrepancy was linked to anincorrect estimation in the data of the absolute age of the Brun-heseMatuyama magnetic reversal. With the revised ages obtainedabout 20 years later (Bassinot et al., 1994), the timing predicted bythe Calder model appears now very good. This probably stands asthe best proof, that the 100 ka cycles of the ice ages are stronglylinked to the eccentricity forcing, as noted in the Hays et al. (1976)paper, as shown on Fig. 3.

Calder was using “caloric seasons” as defined byMilankovitch asa forcing. Later on, it became more usual to use daily insolation atsome specific orbital positions, like the summer solstice or mid-july, since it can be much more easily computed and it better“fits” the last termination (eg. Broecker and van Donk, 1970; Berger,1978), and since the precessional imprint, and therefore the ec-centricity modulation, is much more pronounced in the dailyinsolation than in the « caloric seasons », as shown on Fig. 4. Still,with a simple “rectifier model”, ie a model that extract power fromthe envelop of the precessional forcing, like in Imbrie and Imbrie(1980), we are faced with the “400-ka problem”: the model ex-tracts not only the 100-ka periodicity, but also (and inevitably) themuch larger 400-ka one that characterizes eccentricity variations.The lack of this 400-ka signature in the ice volume record is a

Fig. 3. a/The data used as a reference in order to evaluate b/Calder's model (redrawn from Calder, 1974); c/comparison with present-day knowledge of the chronology of ice ages(stack LR04, Lisiecki and Raymo, 2005). The timing of Calder's model appears a posteriori extremely good. The reason can be found in the link, in this model, between glacialmaxima and eccentricity minima. This synchronization is also noted by Hays et al., 1976.

D. Paillard / Quaternary Science Reviews 107 (2015) 11e24 17

strong indication that the influence of eccentricity on ice ages isprobably not straightforward.

In spite of the Hays et al. (1976) paper, or of the predictions ofthe Calder model, the conclusion that ice ages are predictable andpaced “in some way” by eccentricity, is still today challenged (eg.Crucifix, 2013) and many authors insist on introducing other forc-ings or other mechanisms to account for the main (about 100-ka)periodicity, like explaining it as multiples of obliquity (eg. Huybersand Wunsch, 2005), or induced by orbital inclination (Muller and

Fig. 4. Caloric seasons in red (ie. here the mean power in W.m�2 received during thehalf-year centered around the june solstice) vs. mean insolation in blue (the meanpower in W.m�2 received between the march equinox and the september equinox),both in the time domain (top) and frequency domain (bottom). The difference resultsfrom the varying length of seasons (from 169 to 196 days between equinoxes). Bothseries have the same 41-ka content, but the precessional imprint on caloric seasons issignificantly smaller. The eccentricity modulation appears through the doubling ofprecession lines. The usual daily summer solstice insolation is very similar to the bluecurve.

MacDonald, 1997). Clearly, in order to account for observed largeglacial cycles that are phase-locked to eccentricity, some non-linearmechanism is needed. Suggested concepts are stochastic resonance(Benzi et al., 1982), internal oscillations (eg. K€all�en et al., 1979;Saltzman and Moritz, 1980; Saltzman et al., 1981; Gildor andTziperman, 2000), combination tones (eg. Ghil and Le Treut,1981; Le Treut and Ghil, 1983) or chaotic systems (eg. Saltzmanand Maasch, 1990; for some parameter settings, see; Mitsui andAihara, 2014). A key question for all these suggestions remainsthe synchronization of glacial maxima with eccentricity minima, asobserved in the records. It is worth mentioning that for many ofthese models, the phase-locking to the astronomical forcing hasinteresting non-trivial properties that are currently under investi-gation (eg. De Saedeleer et al., 2013; Mitsui and Aihara, 2014).

4.4. Relaxation oscillations

Among the many possibilities that have been suggested, prob-ably the simplest and most efficient mechanisms are based on arelaxation oscillation. Indeed, this concept explains quite well themost important qualitative features of the glacial cycles over thelast million years: the fast deglaciations and slower glaciations, andthe more or less constant amplitude of cycles, ie. the general saw-tooth shape of the records. Furthermore, this generic mechanismcan easily be locked to the precessional forcing. A number ofmodels can be classified into this category, based on simple climaticbox models (eg. Gildor and Tziperman, 2000), on simplified ice-sheet physics (eg. Pollard, 1983; Saltzman and Sutera, 1984), onempirical rules (Paillard, 1998; Imbrie et al., 2011) or on simplerformulations like the van der Pol oscillator (Saltzman and Moritz,1980; Crucifix, 2012). Though generally not required in allmodels, it is useful to have some insolation forcing threshold belowwhich the slowglaciation phasemay start. This glaciation thresholdwill determine rather directly the duration of interglacials. Theinevitable conclusion is that, in case of small insolation changeswhen the eccentricity is small, like today, the interglacials will beexceptionally long (eg. Paillard, 2001; Berger and Loutre, 2002) asshown on Fig. 5. But the most critical aspect of these relaxationoscillators is the triggering of terminations: indeed, a commonfeature of all these models is a slow glaciation influenced by theprecessional and obliquity forcings, up to a point where the ice-sheet is so large, that it triggers some new phenomenon, like alack of moisture due to sea ice extent (Gildor and Tziperman, 2000),

Fig. 5. Results from simple “ice-only” conceptual models. From top to bottom: bluecurve is the insolation forcing (at 65�N, at the summer solstice, Laskar et al., 2004);purple curve (Calder, 1974); red curve (Imbrie and Imbrie, 1980), thick black curve(Paillard, 1998) and thin black curve (idem, with a lower threshold for the triggering ofglaciations).

D. Paillard / Quaternary Science Reviews 107 (2015) 11e2418

enhanced catastrophic calving (Pollard, 1983), isostatic rebound(Oerlemans, 1983; Birchfield and Grumbine, 1985) or some non-specified mechanism in rule-based or mathematical models. Thisgeneral picture appears also to hold when using a more complexthree-dimensional ice-sheet model coupled asynchronously to anatmospheric general circulation model (Abe-Ouchi et al., 2013).Interestingly, a closer inspection of the deglacial triggering in rule-based models suggests that terminations should be triggered notonly by a large ice-sheet as explained above, but also by a favorableastronomical forcing, in which case it is possible to account ratheraccurately for the varying phase and durations of individual tran-sitions (Parrenin and Paillard, 2003, 2012; Imbrie et al., 2011). Thisfurther suggests that other climatic mechanismsmay have a criticalrole in the triggering of terminations. As we will see below, a verylikely candidate is the observed rise in atmospheric CO2concentration.

4.5. Physically based models

Since the first attempts to provide simple physically based icesheets models (Weertman, 1976; Pollard, 1978) for the ice ageproblem, it is clear that a precise estimation of accumulation andablation over ice sheets is a critical and difficult question. Forinstance, a rather small 1 cm/year error in snow accumulation willrapidly grow into several hundred meters of ice sheet height errorwithin a glacial cycle. Besides, the ablation zone on the margins ofthe ice sheet has a geographically very restricted area of typically afew tens of kilometers. The dynamics of ice sheets is also largelycontrolled by small-scale processes involving kilometer-sized gla-ciers and ice-streams, that are usually not captured by ice-sheetmodels. Last but not least, the data sets available on the

meteorological and glaciological states of present day ice-sheets arerather sparse due to the logistic difficulties involved in obtaining insitu measurements, and covers mostly only the last one or twodecades. Building a realistic coupled ice-sheet e climate model istherefore still a difficult challenge. This may have some seriousimplications on the estimations of the future evolution of ice-sheets. Interestingly, the simulation of their past evolution mightbe a useful benchmark in this context. To address this question, it isnecessary to integrate the climatic evolution over hundreds ofthousands of years, something fully out reach of general circulationmodels. This is mostly why “intermediate complexity”models weredeveloped (Claussen et al., 2002). If the first coupled ice-sheetclimate models were two-dimensional (Deblonde and Peltier,1991; Gall�ee et al., 1991; Loutre and Berger, 2000; Crucifix et al.,2001), more recent models have a coarse but realistic geographicrepresentation (Calov and Ganopolski, 2005; Ganopolski and Calov,2011). More recently, Abe-Ouchi et al. (2013) have used manysnapshot simulations from an atmospheric general circulationmodel to cover various astronomical, CO2 and ice sheet extent.Witha three dimensional ice-sheet model, they have computed acoupled climate e ice sheet evolution by interpolating between theclimatic results. In these different experiments, it is generally moredifficult to obtain a strong 100-ka oscillation, when the atmo-spheric 100-ka CO2 forcing is absent.

5. From glaciological discoveries to integrated theories ofglacial cycles

5.1. A more complete picture of ice ages: continents, atmosphericCO2 and ocean circulation

If paleoceanography did provide some critical advances on thequestion of ice ages, contributions from continental records werealso very important. In particular, it was possible to correlate quiteprecisely many observations from loess records (eg. Kukla, 1977) orfrom pollen records (eg. Woillard, 1978). Soon, it became clear thatthe marine isotopic stratigraphy should be used as the reference,because of its continuity and its global significance as a record ofglobal ice volume, as explained by Kukla (1977):

“In summary, the isotopic record of deep-sea cores provides themost accurate hitherto known information on Pleistocene cli-mates, unique in its continuity and global extent. Validity of anyother climatostratigraphic system of Pleistocene must be testedby comparison with the deep-sea record. ”

It became progressively accepted to refer to regional climaticchanges, like temperature records, monsoons, or any environ-mental or biogeochemical changes, in a unique common strati-graphic framework linked to the Earth global ice volume: themarine isotopic stages as defined by Emiliani (1955). This led to thefamous reference stacked isotopic record SPECMAP, with anastronomically tuned timescale (Imbrie et al., 1984), that as beenused for about 20 years as the main stratigraphic framework of theQuaternary. A better and longer stack record has been build morerecently (Lisiecki and Raymo, 2005).

Glaciologists also used the isotopic composition of the ice fromGreenland and Antarctica to provide estimations of temperaturechanges over current ice sheets, Greenland and Antarctica. But anew critical piece information appeared in the 1980's. Based on airbubbles recovered from Antarctic ice cores, it was possible tomeasure rather directly the air composition of the Last Glacialmaximum (Delmas et al., 1980; Neftel et al., 1982). As assumed by S.Arrhenius almost a century before, pCO2 levels appeared indeedsignificantly lower during glacial periods by almost 100 ppm, and

D. Paillard / Quaternary Science Reviews 107 (2015) 11e24 19

the evolution of atmospheric CO2 levels was shown to match quiteclosely the evolution of temperature over Antarctica for each glacialcycles (Barnola et al.,1987, Petit et al., 1999; Lüthi et al., 2008). SuchCO2 changes certainly had some role on the Earth climate, andcurrent climate modeling results suggest that it would account forabout half the cooling during glacial times. Still a critical questionconcerned the relative timing between ice ages and CO2 variations,in order to untangle the causes and consequences. Indeed, if carbonentered the scene as a potential important player in the ice ageproblem, there is no doubt that the periodicity is controlled by theastronomical forcing. In some ways, both Arrhenius and Milanko-vitch were right after all, and the astronomical theory certainlyneeds a geochemical component to account for observed paleocli-matic changes. But this adds complexity to the problem, since atheory of ice ages now needs to account not only for ice sheetchanges, but also for atmospheric CO2 variations. Geochemistswere surprised by the size of the glacialeinterglacial change inpCO2, and even now it is still quite difficult to find an explanationfor such a large difference. If many theories have been suggested,most scientists agree that the carbonwas stored more efficiently inthe ocean during glacial periods. But the precise physical orbiogeochemical mechanisms responsible for this oceanic glacialcarbon storage are still currently under investigation.

Progresses in paleoceanography did also provide some impor-tant informations on past ocean circulation. By using the carbonisotopic composition of benthic foraminifera, it was shown that theAtlantic circulation underwent significant changes between theLGM and today (Duplessy et al., 1988), with a significant reductionduring the LGM of the depth of the North Atlantic Deep Waters(NADW) that remained above about 2000m deep, while they reachtoday down to 4000 m. In contrast the bottom of the Atlantic wasfilled with Antarctic bottom waters. More generally, a deep strati-fication at about 2000 m appears to be present not only in theAtlantic, but also in the Indian ocean (Kallel et al., 1988). Morerecently, it has been suggested that very salty and dense waterswere filling the deep ocean from the Southern Ocean (Adkins et al.,2002). As explained below, it is very likely that such a differentocean state had a strong impact on the carbon cycle.

The discovery of rapid climatic variability in ice cores or Dans-gaardeOeschger events (Dansgaard et al., 1982; Johnsen et al.,1992) as well as the evidence of abrupt iceberg discharges in theNorth Atlantic, the so-called Heinrich events (Heinrich, 1988; Bondet al., 1992) also demonstrated that Quaternary climates could besignificantly influenced by internal variability on millenial scales,probably evenwithout any external forcing. This was a considerablesurprise and prompted the awareness that glacial cycles, paced bythe astronomy, were obviously not the “final story” of Quaternaryclimates. This paradigm shift suggested that ice ages might belargely influenced by other changes, in the carbon cycle or in thedeep ocean, and that looking for an ice-sheet-only theorymight notbe the best strategy for explaining Quaternary climates.

5.2. The return of geochemical theories

Concerning Milankovitch's theory, it is worth insisting on thepoint that the central object of interest is not “climate”, ie. thetemperature over different regions of the Earth, but the ice-sheets,and more precisely ice-sheets in high northern latitudes for theQuaternary period. It can be said that Milankovitch's theory isclearly not a theory of “climate”, but a theory of ice-sheets. Incontrast to the geochemical theorywhich states that CO2 and globalclimate are driving ice sheet changes, the hypothesis of Milanko-vitch is just the opposite: ice-sheets are driving global climaticchanges over the Quaternary, in particular through their large al-bedo effect. This mechanism is largely confirmed by LGM

simulations with state-of-the-art models, since the ice sheets areresponsible for a half of the LGM cooling. But the Milankovitchtheory does not account for other possible climatic features, likeabrupt changes, CO2 variations or any other mechanisms that arenow recognized climatically relevant. It is therefore not entirelysurprising that its predictions of a quasi-linear, mainly obliquitydriven oscillation, is not corresponding exactly to observations.Indeed, climatic changes might, in turn, also have an impact on ice-sheets. For instance, it is now a classical result from general circu-lation models that CO2 accounts for almost the other half of theLGM cooling (cf Fig. 6).

But the most important piece of information is the relativetiming between the last deglaciation, ie. the melting of the icesheets, and the rise in CO2 as shown on Fig. 7.

As already noted for many years (eg. Pichon et al., 1992), tem-peratures in the Southern ocean and Antarctica increased severalmillenia before the melting of northern hemisphere ice sheets andthe warming in the North. This observation was since confirmedmany times by more systematic studies (eg. Shakun et al., 2012).Obviously, this cannot be explained within the Milankovitch the-ory. Therefore, if Milankovitch theory is clearly necessary to explainthe rhythm of ice ages, it is probably not sufficient to account for iceages dynamics. It now seems indeed very unreasonable to believethat the evolution of Quaternary climate, and a fortiori of even oldertime periods, will only be explained by the astronomy. In contrastto scientists in the early XXth century that were eager to find asimple unique mechanism for ice ages, we now have many paleo-climatic records to build up more complex scenarios, that arefurthermore consistent with results obtained with current com-puter simulations. The most natural and promising direction is totry to account for CO2 changes by including the geochemical part ofthe story. This was quite well understood already in the 1980's withthe first attempts to build a coupled ice sheetecarbon cycle model(Saltzman and Sutera, 1987; Saltzman and Maasch, 1988). Thesemodels were build on a set of three non-linear (polynomial) ordi-nary differential equations, in the spirit of dynamical system the-ory, representing ice volume, some deep ocean property(temperature or overturning) and atmospheric CO2. Throughproviding many interesting results, a weak point of these models isthe difficulty to identify and quantify the key physical processesinvolved, in particular in the carbon cycle part of the equations.

5.3. Explaining lower CO2 during glacial times

The lack of a well accepted theory of the glacial carbon cycle iscurrently the main missing piece in the puzzle of Quaternary cli-mates. Many possible physical and biogeochemical processes havebeen suggested (eg. Archer et al., 2000; Sigman andBoyle, 2000) butthere is currently no consensus on the most important ones. Mostlikely, the deep ocean contained more carbon during glacial timessince this is the only large carbon reservoir involvedwithin the timescales of interest.Most likelyalso, the Southernocean is the keyarea,since the evolution of atmospheric CO2 closely parallels the evolu-tion of Southern records, as examplified by the strong correlationbetween pCO2 and temperature reconstructions over Antarctica(Petit et al., 1999; Lüthi et al., 2008). Most studies are trying tosimulate the carbon cycle during the Last Glacialmaximumwith thehope of obtaining atmospheric levels about 100 ppm lower thanduring pre-industrial times, as recorded in ice cores. In other words,most studies are trying to explain lower pCO2 as a consequence ofcold, ice-sheet induced, climate. But as outlined above, we knowquite well the succession of events, at least during the deglaciation,and the causal link between ice-sheets and carbon appears to be justthe opposite. Furthermore, we also need to identify a “switchmechanism” in order to explain terminations and the 100-ka cycles

Fig. 6. (left) Last Glacial maximum regional cooling in gray shades compared simulations in PMIP-2 (circles): current ocean-atmosphere general circulation models simulatereasonably well temperatures during the Last Glacial maximum, the major forcings (right) being the prescribed ice-sheet and greenhouse gases concentrations, each one accountingfor about half of the cooling (redrawn from IPCC Fourth Assessment Report, 2007).

D. Paillard / Quaternary Science Reviews 107 (2015) 11e2420

as explained in the preceding sections, and Fig. 7 suggests that pCO2is a very plausible candidate. In other words, the key questionmightnot be “why is pCO2 lower during glacial times”, but “whydoes pCO2rise just after glacial maxima”. This dynamical perspective providesa very stringent constraint on the possible carbon cyclemechanismsinvolved. Indeed, most of the suggested processes for lowering CO2are most efficient during glacial maxima. According to the datashown on Fig. 7, amechanism for lowering pCO2 should break down

Fig. 7. From top to bottom: insolation at 65�N at the june solstice (Laskar et al., 2004);sea level records (Fairbanks, 1989; Bard et al., 1996) Greenland d18O record as a proxyof Northern hemisphere climate (Dansgaard et al., 1993); atmospheric CO2 (Monninet al., 2001) and antarctic dD as a proxy of Southern hemisphere climate (Stenniet al., 2001). At the onset of the BollingeAllerod, about 14.6 ka BP, the sea level iswithin glacial values, while atmospheric CO2 has increased by about 50 ppm. Moregenerally, the southern hemisphere is warming several thousands of years before thenorthern hemisphere one.

Fig. 8. Scheme of the mechanism (Paillard and Parrenin, 2004). Top: current situation,where salty waters from brine rejection are mostly mixed with open ocean waters.Middle: with lower sea-level, more intense sea ice formation, less ice shelf melting, itis likely that the salty waters will overspill down to the bottom of the Ocean. Bottom:shortly after the glacial maximum in the north, Antarctica will reach its maximum areaand will cover a substantial part of the continental shelves (schemes from Boutteset al., 2012). The critical parameter is the fraction (“frac”) of salty “brine” watersreaching the bottom of the Ocean, which relates to the green curve of Fig. 10.

Fig. 9. Top: pCO2 and antarctic temperatures according to ice cores (Monnin et al.,2001) to be compared with two model scenarios of salty bottom water formation,either an abrupt stop at 18 ka BP, or a more progressive one (Bouttes et al., 2012).

D. Paillard / Quaternary Science Reviews 107 (2015) 11e24 21

within a few millenia of the glacial maximum. According to therelaxationoscillatorhypothesis for explaining the100-ka cycles, thismechanism should in fact break down mainly as a direct conse-quence of the glacial maximum. Following these lines, a conceptualmodel for the coupled ice sheetecarbon cycle system has been

Fig. 10. Conceptual model of coupled ice sheet e CO2 evolution over the last 5 million yhemisphere ice volume and Antarctica extent; black and red: isotopic records of ice ages;oceanic storage of carbon (from Paillard and Parrenin, 2004).

constructed (Paillard and Parrenin, 2004), in which the key processfor storing carbon in the deep ocean is linked to the formation ofvery dense and salty bottom waters on the continental margins ofAntarctica during cold climates, due to sea ice formation and brinerejection. In particular, these dense waters can reach the bottom ofthe Ocean when they overspill the continental shelf and flow asgravity currents down the topography (eg. Foldvik et al., 2004). Thismechanismstops naturallywhen theAntarctic ice sheet coversmostof the continental shelves, since there is almost no margin left forbottom water formation, as shown on Fig. 8. This expansion ofAntarctica results largely because of the reduced sea level at theglacial maximum (Ritz et al., 2001; Anderson et al., 2002).

This conceptual idea has since been tested in an intermediatecomplexity model of the climate and carbon cycle (Bouttes et al.,2010). Interestingly, the corresponding simulations of the carbonisotopic composition in the deep ocean is in good agreement withthe geographic patterns observed in paleoceanographic data(Bouttes et al., 2011), thus explaining quite well the mystery of thestrong deep ocean stratification during glacial times (Kallel et al.,1988; Curry and Oppo, 2005), as well as the observed bottom saltywaters (Adkins et al., 2002). The corresponding simulation ofradiocarbon (Mariotti et al., 2013) is also in good agreement withD14Cmeasurements in the atmosphere and ocean. Furthermore, thetime evolution during the last deglaciation of pCO2 and carbonisotopes at different ocean depths is quitewell represented (Boutteset al., 2012) with a simultaneous increase in pCO2 and southerntemperatures (see Fig. 9), both being linked to the de-stratificationof the southern ocean. This synchronicity was recently confirmedthrough a careful analysis of the ice core chronology (Parrenin et al.,2013).

Recent observations of dense salty water formation on thecontinental shelves of Antarctica associated with sea ice brinerejection (Ohshima et al., 2013) gives undoubtedly some support tothe hypothesis outlined above, since this processescontribute today to some non-negligible part of AABW (Meredith,2013) and is likely to be much more important in a colderclimate. As mentioned above, at the glacial maximum the Antarcticice-sheet covers a significant part of the continental shelf and thismechanism breaks down. A key open question is the dynamics ofthe gravity currents that transport these salty shelf waters down tothe bottom of the Southern ocean (Foldvik et al., 2004; Ohshimaet al., 2013). Unfortunately, these processes are not well

ears. From top to bottom, black: simulated pCO2; blue and red, simulated northerngreen: the critical parameter that controls the formation of salty deep waters and the

D. Paillard / Quaternary Science Reviews 107 (2015) 11e2422

understood, and not well represented by current state-of-the-artclimate models. Another important mechanism is the basalmelting under the ice shelves, that provides today a significantamount of fresh water to the continental shelf waters, thusreducing their density and their ability to contribute to bottomwater formation. More generally, the coupling between the oceanand basal melting appears today to be widely distributed aroundAntarctica and more important than previously acknowledged(Rignot et al., 2013). According to regional ocean model studies ofthe caverns beneath Antarctic ice shelves, reducing the ice basalmelting would greatly enhance the formation of bottom waters(Hellmer, 2004) which would again becomemore dense and saltier.This is likely to occur during glacial periods thus favoring saltyAABW formation and lower pCO2. If the mechanisms outlinedabove clearly need further confirmations, many indications frompaleoclimatic data, from present day observations, from regionalmodeling and from intermediate complexity models, altogetherseem to confirm the initial hypothesis.

6. Conclusion

Modeling the coupled climate e ice sheetecarbon cycle evolu-tion over the Quaternary is clearly a challenge. The short history ofice age theories outlined in this paper outlines the close connec-tions between paleoclimatic observations and the understanding ofthe dynamics of climate changes. Since the XIXth century, a centralquestion of climate sciences has been to quantify the respectiveroles of the astronomical forcing and of greenhouse gases changes,in causing past observed major climatic evolutions. A century and ahalf later, this question is far from being settled, in spite of thetremendous progresses achieved in observing and simulating thepresent day climate. In the current context of global warming, manyefforts are aiming towards better predictions of the impacts ofelevated greenhouse gases concentrations on our environment,mostly on the century scale. But obviously, given the time scalesinvolved, we are currently just at the start of a “super-deglaciation”,which will certainly strongly interfere with future ice ages (eg.Archer and Ganopolski, 2005; Paillard, 2006). On much shortertime scales, it is now clear that extremely abrupt reorganizationsoccurred during deglaciations in terms of climate (eg. Dansgaardet al., 1989) or in terms sea-level (Deschamps et al., 2012). Ittherefore appears quite paradoxical that comparatively littlemodeling efforts have been put into a better understanding of pastterminations. Still, thanks to available paleoclimatic data, we nowhave a rather consistent picture of the succession of events, inparticular during the last deglaciation. As explained above, thisallows for the building of a plausible, physically relevant scenariothat accounts for most of the observations: the evolution of sealevel, of northern and southern hemisphere climate, of the carboncycle, and also of isotopic proxy data in the deep ocean. This sce-nario relies heavily on mechanisms involving the Antarctic icesheet and the Southern ocean, in particular in their role in bottomwater formation and carbon oceanic storage. Unfortunately, thesemechanisms are still not well documented and not represented incurrent models. Nevertheless, it seems that we are getting veryclose to a more complete understanding of the dynamics of glacialcycles. By building the synthesis between the astronomical theoryand the geochemical theory of ice ages, we will soon close a 150-year long dispute on the mechanisms behind the largest, well-documented, recent climatic changes, which are Quaternary cycles.

Acknowledgments

Comments from M. Crucifix and from an anonymous reviewersignificantly improved the manuscript.

References

Abe-Ouchi, A., Saito, F., Kawamura, K., Raymo, M.E., Okuno, J., Takahashi, K.,Blatter, H., 2013. Insolation-driven 100,000-year glacial cycles and hysteresis ofice-sheet volume. Nature 500, 190e194.

Adh�emar, J., 1842. R�evolutions de la mer, d�eluges p�eriodiques. Carilian-Goeury & V.Dalmont, Paris, France.

Adkins, J.F., McIntyre, K., Schrag, D., 2002. The salinity, temperature and d18O of theglacial deep ocean. Science 298, 1769e1773.

Agassiz, L., 1840. �Etudes sur les glaciers. Jent & Gassmann, Neuchatel, Switzerland.Andersen, B.G., 1992. Jens Esmark e a pioneer in glacial geology. Boreas Pioneers

Ser. 21, 97e102.Anderson, J., Shippe, S., Lowe, A., Wellner, J., Mosola, A., 2002. The Antarctic Ice

Sheet during the Last Glacial Maximum and its subsequent retreat history: areview. Quat. Sci. Rev. 21, 49e70.

Archer, D., Ganopolski, A., 2005. A movable trigger: fossil fuel CO2 and the onset ofthe next glaciation. Geochem. Geophys. Geosyst. 6 http://dx.doi.org/10.1029/2004GC000891.

Archer, D., Winguth, A.M.E., Lea, D.W., Mahowald, N.M., 2000. What caused theglacial/interglacial atmospheric pCO2 cycles? Rev. Geophys. 38, 159e189.

Aristotle, translation by H.D.P. Lee, 1951. Meteorologica. Book 1, (Chapter 14). Har-vard University Press.

Arrhenius, S., 1896. On the influence of carbonic acid in the air upon the temper-ature of the ground. Philos. Mag. 41, 237e276.

Arrhenius, G., 1952. Sediment Cores from the East Pacific. In: Report Swedish Deep-sea Expedition 1947-1948, vol. 5, pp. 1e228.

Bard, E., 2004. Greenhouse effect and ice ages: historical perspective. C. R. Geosci.336, 603e638.

Bard, E., Hamelin, B., Arnold, M., Montaggioni, L., Cabioch, G., Faure, G., Rougerie, F.,1996. Deglacial sea-level record from Tahiti corals and the timing of globalmeltwater discharge. Nature 382, 241e244.

Barnola, J.-M., Raynaud, D., Korotkevitch, Y., Lorius, C., 1987. Vostok ice core pro-vides 160,000-year record of atmospheric CO2. Nature 329, 408e414.

Bassinot, F., Labeyrie, L., Vincent, E., Quidelleur, X., Shackleton, N., Lancelot, Y., 1994.The astronomical theory of climate and the age of the BrunheseMatuyamamagnetic reversal. Earth Planet. Sci. Lett. 126, 91e108.

Benzi, R., Parisi, G., Sutera, A., Vulpiani, A., 1982. Stochastic resonance in climaticchange. Tellus 34, 10e16.

Berger, A., 1978. Long-term variations of daily insolation and Quaternary climaticchange. J. Atmos. Sci. 35, 2362e2367.

Berger, A., 1988. Milankovitch theory and climate. Rev. Geophys 26, 624e657.Berger, A., 2012. A brief history of the astronomical theories of paleoclimates. In:

Berger, Andr�e, Mesinger, Fedor, Sijacki, Djordje (Eds.), Climate Change: In-ferences from Paleoclimate and Regional Aspects.

Berger, A., Loutre, M.-F., 2002. An exceptionally long interglacial ahead? Science297, 1287e1288.

Berner, R.A., Maasch, K.A., 1996. Chemical weathering and controls on atmosphericO2 and CO2: fundamental principles were enunciated by J. J. Ebelmen in 1845.Geochim. Cosmochim. Acta 60, 1633e1637.

Birchfield, E., Grumbine, R., 1985. Slow physics of large continental ice sheets andunderlying bedrock and its relation to the pleistocene ice ages. J.G.R 90,11,294e11,302.

Bond, G., Heinrich, H., Broecker, W., Labeyrie, L., McManus, J., Andrews, J., Huon, S.,Jantschik, R., Clasen, S., Simet, C., Tedesco, K., Klas, M., Bonani, G., Ivy, S., 1992.Evidence of massive discharges of icebergs into the North Atlantic ocean duringthe Last Glacial period. Nature 360, 245e249.

Bouttes, N., Paillard, D., Roche, D.M., 2010. Impact of brine-induced stratification onthe glacial carbon cycle. Clim. Past. 6, 575e589.

Bouttes, N., Paillard, D., Roche, D.M., Brovkin, V., Bopp, L., 2011. Last GlacialMaximum CO2 and d13C successfully reconciled. Geophys. Res. Lett. 38, 1e5.

Bouttes, N., Paillard, D., Roche, D.M., Waelbroeck, C., Kageyama, M., Lourantou, A.,Michel, E., Bopp, L., 2012. Impact of oceanic processes on the carbon cycleduring the last termination. Clim. Past. 8, 149e170.

Broecker, W., van Donk, J., 1970. Insolation changes, ice volumes and the O18 recordin deep-sea cores. Rev. Geophys. Space Phys. 8, 169e197.

Buffon, G.-L.L., 1778. Les �epoques de la nature (Paris).Calder, N., 1974. Arithmetic of ice ages. Nature 252, 216e218.Calov, R., Ganopolski, A., 2005. Multistability and hysteresis in the climate-

cryosphere system under orbital forcing. Geophys. Res. Lett. 32, L21717.Chamberlin, T.C., 1897. A group of hypotheses bearing on climatic changes. J. Geol. 5,

653e683.Chappell, J., 1974. Relationships between sealevels, 18O variations and orbital per-

turbations, during the past 250,000 years. Nature 252, 199e202.Chappell, J., Shackleton, N.,1986. Oxygen isotopes and sea level. Nature 324,137e140.Claussen, M., Mysak, L., Weaver, A.J., Crucifix, M., Fichefet, T., Loutre, M.-F., Weber, s,

Alcamo, J., Alexeev, V.A., Berger, A., Calov, R., Ganopolski, A., Goosse, H.,Lohmann, G., Lunkeit, F., Mokhov, I.I., Petoukhov, V., Stone, P., Wang, Z., 2002.Earth system models of intermediate complexity: closing the gap in the spec-trum of climate system models. Clim. Dyn. 18, 579e586.

Croll, J., 1864. On the physical cause of the change of climate during geologicalepochs. Lond. Edinb. Dublin Philos. Mag. J. Sci. 28, 121e137.

Croll, J., 1867. On the change in the obliquity of the ecliptic, its influence on theclimate of the polar regions, and level of the sea. Trans. Geol. Soc. Glasg177e198.

D. Paillard / Quaternary Science Reviews 107 (2015) 11e24 23

Crucifix, M., 2012. Oscillators and relaxation phenomena in Pleistocene climatetheory. Philos. Trans. R. Soc. A: Math. Phys. Eng. Sci. 370, 1140e1165.

Crucifix, M., 2013. Why could ice ages be unpredictable? Clim. Past 9, 2253e2267.Crucifix, M., Loutre, M.-F., Lambeck, K., Berger, A., 2001. Effect of isostatic rebound

on modelled ice volume variations during the last 200 kyr. Earth Planet. Sci.Lett. 184, 623e633.

Curry, W., Oppo, D., 2005. Glacial water mass geometry and the distribution of deltaC-13 of sigma CO2 in the western Atlantic Ocean. Paleoceanography 20, PA1017.

Dansgaard, W., Clausen, H., Gundestrup, N., Hammer, C., Johnsen, S.,Kristinsdottir, P., Reeh, N., 1982. A new Greenland deep ice core. Science 218,1273e1277.

Dansgaard, W., White, J.W.C., Johnsen, S., 1989. The abrupt termination of theYounger Dryas climate event. Nature 339, 532e534.

Dansgaard, W., Johnsen, S., Clausen, H.B., Dahl-Jensen, D., Gundestrup, N.,Hammer, C., Hvidberg, C., Steffensen, J., Sveinbj€ornsdottir, A.E., Jouzel, J.,Bond, G., 1993. Evidence for general instability of past climate from a 250-kyrice-core record. Nature 364, 218e220.

de Charpentier, J., 1836. Account of one of the most important results of the in-vestigations of M. Venetz, regarding the present and earlier condition of theglaciers of the Canton du Valais. Edinb. New. Philos. J. 21, 210e220.

De Saedeleer, B., Crucifix, M., Wieczorek, S., 2013. Is the astronomical forcing areliable and unique pacemaker for climate? A conceptual model study. Clim.Dyn. 40, 273e294.

Deblonde, G., Peltier, W.R., 1991. Simulations of continental ice sheet growth overthe Last Glacial-interglacial cycle: experiments with a one-level seasonal en-ergy balance model including realistic geography. J. Geophys. Res. Atmos. 96(D5), 9189e9215.

Delmas, R.J., Ascencio, J.M., Legrand, M., 1980. Polar ice evidence that atmosphericCO2 20,000 BP was 50% of present. Nature 284, 155e157.

Deschamps, P., Durand, N., Bard, E., Hamelin, B., Camoin, G., Thomas, A.L.,Henderson, G.M., Okuno, J., Yokoyama, Y., 2012. Ice-sheet collapse and sea-levelrise at the Bølling warming 14,600years ago. Nature 483, 559e564.

Dufresne, J.-L., 2009. L'effet de serre : sa d�ecouverte, son analyse par la m�ethode despuissances nettes �echang�ees et les effets de ses variations r�ecentes et futuressur le climat terrestre. Habilitation's thesis. Universit�e Pierre et Marie Curie.

Duplessy, J.-C., Shackleton, N., Fairbanks, R., Labeyrie, L., Oppo, D., Kallel, N., 1988.Deepwater source variations during the last climatic cycle and their impact onthe global deepwater circulation. Paleoceanography 3, 343e360.

Ekholm, N., 1901. On the variations of the climate of the geological and historicalpast and their causes. Q. J. Roy. Meteor. Soc. 27, 1e61.

Emiliani, C., 1955. Pleistocene temperatures. J. Geol. 63, 538e578.Emiliani, C., 1966. Isotopic paleotemperatures. Science 154, 851e857.Emiliani, C., Shackleton, N., 1974. The brunhes Epoch: isotopic paleotemperatures

and geochronology. Science 183, 511e514.Ericson, D., Wollin, G., 1956. Correlation of six cores from the equatorial Atlantic and

the Caribbean. Deep-Sea Res. 3, 104e125.Esmark, J., 1827. Remarks tending to explain the geological history of the earth.

Edinb. New Philos. 107e121.Fairbanks, R., 1989. A 17,000 year glacio-eustatic sea level record : influence of

glacial melting rates on the Younger Dryas event and deep ocean circulation.Nature 342, 637e642.

Fleming, J., 1829. On the value of the evidence from the animal kingdom tending toprove that the arctic regions formerly enjoyed a milder climate than at present.Edinb. New. Philos. 277e286.

Foldvik, A., Gammelsrod, T., Osterhus, S., Fahrbach, E., Rohardt, G., Schroder, M.,Nicholls, K.W., Padman, L., Woodgate, R.A., 2004. Ice shelf water overflow andbottom water formation in the southern Weddell Sea. J. Geophys. Res-Oceans109, C02015.

Fourier, J., 1824. Remarques g�en�erales sur les temp�eratures du globe terrestre et desespaces plan�etaires. Ann. Chim. Phys. 27, 136e167.

Gall�ee, H., van Ypersele, J.-P., Fichefet, T., Tricot, C., Berger, A., 1991. Simulation of theLast Glacial cycle by a coupled, sectorially averaged climate-ice sheet model. 1.The climate model. J.G.R 96 (D7), 13139e13161.

Ganopolski, A., Calov, R., 2011. The role of orbital forcing, carbon dioxide andregolith in 100 kyr glacial cycles. Clim. Past. 7, 1415e1425.

Geikie, A., 1863. On the phenomena of the glacial drift of Scotland. Geol. Soc. Glasg.Trans. 1e2, 1e190.

Geikie, J., 1874. The Great Ice Age and its Relation to the Antiquity of Man, 1re �ed., W.Isbister, London, 575 pp. (2e �ed., E. Stanford, London, 1877, 624 pp..

Ghil, M., Le Treut, H., 1981. A climate model with cryodynamics and geodynamics.J.G.R 86 (C6), 5262e5270.

Gildor, H., Tziperman, E., 2000. Sea ice as the glacial cycles' climate switch: role ofseasonal and orbital forcing. Paleoceanography 15, 605e615.

Hays, J., Imbrie, J., Shackleton, N., 1976. Variations in the earth's orbit: pacemakersof the ice ages. Science 194, 1121e1132.

Heinrich, H., 1988. Origin and consequences of cyclic ice rafting in the NortheastAtlantic Ocean during the past 130,000 years. Quat. Res. 29, 142e152.

Hellmer, H., 2004. Impact of Antarctic ice shelf basal melting on sea ice and deepocean properties. Geophys. Res. Lett. 31, 1e4.

Herschel, J., 1830. On the astronomical causes which may influence geologicalphænomena. Trans. Geol. Soc. Lond. 293e299 (published 1835).

Huybers, P., Wunsch, C., 2005. Obliquity pacing of the late Pleistocene glacial ter-minations. Nature 434, 491e494.

Imbrie, J., Imbrie, J.Z., 1980. Modelling the climatic response to orbital variations.Science 207, 943e953.

Imbrie, J., Kipp, N., 1971. A new micropaleontological method for paleoclimatology:application to a late Pleistocene Caribbean core. In: Turekian, K. (Ed.), The LateCenozoic Glacial Ages, pp. 71e181.

Imbrie, J., Hays, J., Martinson, D., McIntyre, A., Mix, A.C., Morley, J.J., Pisias, N.,Prell, W., Shackleton, N., 1984. The orbital theory of Pleistocene climate: supportfrom a revised chronology of the marine d18O record. Milankovitch Clim.269e305.

Imbrie, J.Z., Imbrie-Moore, A., Lisiecki, L.E., 2011. A phase-space model for Pleisto-cene ice volume. Earth Planet. Sci. Lett. 307, 94e102.

IPCC, 2007. Climate Change 2007: The Physical Science Basis. Contribution ofWorking Group I to the Fourth Assessment Report of the IPCC. CambridgeUniversity Press, Cambridge, United Kingdom and New York, NY, USA.

Johnsen, S., Clausen, H., Dansgaard, W., Fuhrer, K., Gundestrup, N., Hammer, C.,Iversen, P., Jouzel, J., Stauffer, B., Steffensen, J., 1992. Irregular glacial in-terstadials recorded in a new Greenland ice core. Nature 359, 311e312.

Kallel, N., Labeyrie, L., Juillet-Leclerc, A., Duplessy, J.-C., 1988. A deep hydrologicalfront between intermediate and deep-water masses in the glacial Indian Ocean.Nature 333, 651e655.

K€all�en, E., Crafoord, C., Ghil, M., 1979. Free oscillations in a climate model with ice-sheet dynamics. J. Atmos. Sci. 36, 2292e2303.

Krüger, T., 2013. Discovering the Ice Ages: International Reception and Conse-quences for a Historical Understanding of Climate. Brill Ed.

Kukla, G., 1977. Pleistocene landdsea correlations I. Europe. Earth-Sci. Rev. 13,307e374.

Laskar, J., Robutel, P., Joutel, F., Gastineau, M., Correia, A.C.M., Levrard, B., 2004.A long-term numerical solution for the insolation quantities of the Earth.Astron. Astrophys. 428, 261e285.

Le Treut, H., Ghil, M., 1983. Orbital forcing, climatic interactions and glaciation cy-cles. J.G.R 88, 5167e5190.

Lisiecki, L., Raymo, M., 2005. A Pliocene-Pleistocene stack of 57 globally distributedbenthic d18O records. Paleoceanography 20, PA1003. http://dx.doi.org/10.1029/2004PA001071.

Loutre, M.-F., Berger, A., 2000. No glacial-interglacial cycle in the ice volumesimulated under a constant astronomical forcing and a variable CO2. Geophys.Res. Lett. 27, 783e786.

Lüthi, D., Le Floch, M., Bereiter, B., Blunier, T., Barnola, J.-M., Siegenthaler, U.,Raynaud, D., Jouzel, J., Fischer, H., Kawamura, K., Stocker, T.F., 2008. High-res-olution carbon dioxide concentration record 650,000e800,000 years beforepresent. Nature 453, 379e382.

Lyell, C., 1830. Principles of Geology. John Murray, London.Marchal, O., Cacho, I., Stocker, T.F., Grimalt, J., Calvo, E., Martrat, B., Shackleton, N.,

Vautravers, M.J., Cortijo, E., Van Kreveld, S., Andersson, C., Koç, N., Chapman, M.,Sbaffi, L., Duplessy, J.-C., Sarnthein, M., Turon, J.-L., Duprat, J., Jansen, E., 2002.Apparent long-term cooling of the sea surface in the northeast Atlantic andMediterranean during the Holocene. Quat. Sci. Rev. 21, 455e483.

Mariotti, V., Paillard, D., Roche, D., Bouttes, N., Bopp, L., 2013. Simulated Last Gglacialmaximum D14Catm and the deep glacial ocean carbon reservoir. Radiocarbon 55,1595e1602.

Meredith, M., 2013. Oceanography: replenishing the abyss. Nat. Geosci. 6, 166e167.Mitsui, T., Aihara, K., 2014. Dynamics between order and chaos in conceptual

models of glacial cycles. Clim. Dyn. 42, 3087e3099.Monnin, E., Indermühle, A., D€allenbach, A., Flückiger, J., Stauffer, B., Stocker, T.F.,

Raynaud, D., Barnola, J.-M., 2001. Atmospheric CO2 concentrations over the LastGlacial termination. Science 291, 112e114.

Muller, R., MacDonald, G., 1997. Glacial cycles and astronomical forcing. Science 277,215e218.

Murphy, J.J., 1869. On the nature and cause of the glacial climate Quarterly. J. Geol.Soc. 25, 350e356.

Murphy, J.J., 1876. The glacial climate and the polar ice-cap. Q. J. Geol. Soc. Lond 32,400e406. http://dx.doi.org/10.1144/GSL.JGS.1876.032.01-04.45.

Neftel, A., Oeschger, H., Schwander, J., Stauffer, B., Zumbrunn, R., 1982. Ice coresample measurements give atmospheric CO2 content during past 40,000 years.Nature 295, 220e223.

Oerlemans, J., 1983. A numerical study on cyclic behaviour of polar ice sheets. TellusA 35, 81e87.

Ohshima, K.I., Fukamachi, Y., Williams, G.D., Nihashi, S., Roquet, F., Kitade, Y.,Tamura, T., Hirano, D., Herraiz-Borreguero, L., Field, I., Hindell, M., Aoki, S.,Wakatsuchi, M., 2013. Antarctic bottom water production by intense sea-iceformation in the Cape Darnley polynya. Nat. Geosci. 6, 235e240.

Paillard, D., 1998. The timing of Pleistocene glaciations from a simple multiple-stateclimate model. Nature 391, 378e381.

Paillard, D., 2001. Glacial cycles: toward a new paradigm. Rev. Geophys. 39,325e346.

Paillard, D., 2006. What drives the ice age cycle ? Science 313, 455e456.Paillard, D., 2010. Climate and the orbital parameters of the Earth. C. R. Geosci. 342,

273e285.Paillard, D., Parrenin, F., 2004. The Antarctic ice-sheet and the triggering of de-

glaciations. Earth Planet. Sci. Lett. 227, 263e271.Parrenin, F., Paillard, D., 2003. Amplitude and phase of glacial cycles from a con-

ceptual model. Earth Planet. Sci. Lett. 214, 243e250.Parrenin, F., Paillard, D., 2012. Terminations VI and VIII (~530 and ~720 kyr BP) tell

us the importance of obliquity and precession in the triggering of deglaciations.Clim. Past. 8, 2031e2037.

Parrenin, F., Masson-Delmotte, V., Kohler, P., Raynaud, D., Paillard, D., Schwander, J.,Barbante, C., Landais, A., Wegner, A., Jouzel, J., 2013. Synchronous change of

D. Paillard / Quaternary Science Reviews 107 (2015) 11e2424

atmospheric CO2 and antarctic temperature during the last deglacial warming.Science 339, 1060e1063.

Petit, J.-R., Jouzel, J., Raynaud, D., Barkov, N., Barnola, J.-M., Basile, I., Bender, M.,Chappellaz, J., Davis, M., Delaygue, G., Delmotte, M., Kotlyakov, V.M.,Legrand, M., Lipenkov, V., Lorius, C., P�epin, L., Ritz, C., Saltzman, E.S.,Stievenard, M., 1999. Climate and atmospheric history of the past 420,000 yearsfrom the Vostok ice core, Antarctica. Nature 399, 429e436.

Pettersson, H., 1953. Westward Ho with the Albatross (New York).Pichon, J.-J., Labeyrie, L., Bareille, G., Labracherie, M., Duprat, J., Jouzel, J., 1992.

Surface water temperature changes in the high latitudes of the SouthernHemisphere over the Last Glacial-interglacial cycle. Paleoceanography 7,289e318.

Pollard, D., 1978. An investigation of the astronomical theory of the ice ages using asimple climate-ice sheet model. Nature 272, 233e235.

Pollard, D., 1983. A coupled climate-ice sheet model applied to the Quaternary iceages. J.G.R 88, 7705e7718.

Raymo, M.E., Lisiecki, L., Nisancioglu, K., 2006. Plio-Pleistocene ice volume, antarcticclimate, and the global d18O record. Science 313, 492e495.

Rignot, E., Jacobs, S., Mouginot, J., Scheuchl, B., 2013. Ice-shelf melting aroundAntarctica. Science 341, 266e270.

Ritz, C., Rommelaere, V., Dumas, C., 2001. Modeling the evolution of Antarctic icesheet over the last 420,000 years: implications for altitude changes in theVostok region. J.G.R 106 (D23), 31,943e31,964.

Saltzman, B., Maasch, K., 1988. Carbon cycle instability as a cause of the latepleistocene ice age oscillations: modeling the asymmetric response. Glob.Biogeochem. Cycles 2, 177e185.

Saltzman, B., Maasch, K., 1990. A first-order global model of late Cenozoic climaticchange. Trans. R. Soc. Edinb Earth Sci. 81, 315e325.

Saltzman, B., Moritz, R., 1980. A time-dependent climatic feedback system involvingsea-ice extent, ocean temperature, and CO2. Tellus 32, 93e118.

Saltzman, B., Sutera, A., 1984. A model of the internal feedback system involved inlate Quaternary climatic variations. J. Atmos. Sci. 41, 736e745.

Saltzman, B., Sutera, A., 1987. The mid-Quaternary climatic transition as a freeresponse of a three-variable dynamical model. J. Atmos. Sci. 44, 236e241.

Saltzman, B., Sutera, A., Evenson, A., 1981. Structural stochastic stability of a simpleauto-oscillatory climatic feedback system. J. Atmos. Sci. 38, 494e503.

Shackleton, N., 1967. Oxygen isotope analyses and Pleistocene temperatures re-assessed. Nature 215, 15e17.

Shakun, J.D., Clark, P.U., He, F., Marcott, S.A., Mix, A., Liu, Z., Otto-Bliesner, B.,Schmittner, A., Bard, E., 2012. Global warming preceded by increasing carbondioxide concentrations during the last deglaciation. Nature 484, 49e54.

Sigman, D., Boyle, E., 2000. Glacial/interglacial variations in atmospheric carbondioxide. Nature 407, 859e869.

Stenni, B., Masson-Delmotte, V., Johnsen, S., Jouzel, J., Longinelli, A., Monnin, E.,R€othlisberger, R., Selmo, E., 2001. An oceanic cold reversal during the lastdeglaciation. Science 293, 2074e2077.

Tyndall, J., 1860. The Glaciers of the Alps (London).Tyndall, J., 1861. On the absorption and radiation of heat by gases and vapours, and

on the physical connexion of radiation, absorption and condbuction. J. Philos.Trans. R. Soc. Lond. 151, 1e36.

Urey, H., 1947. The thermodynamic properties of isotopic substances. J. Chem. Soc.562e581.

Venetz, I., 1833. M�emoire sur les variations de la temp�erature des Alpes de la Suisse.M�em. Soc. Helv. Sci. Nat. 1, 1e38.

Weertman, J., 1976. Milankovitch solar radiation variations and ice age ice sheetsizes. Nature 261, 17e20.

Woillard, G., 1978. Grande pile peat bog: a continuous pollen record for the last140,000 years. Quat. Res. 9, 1e21.

Woodward, J., 2014. The Ice Age: a Very Short Introduction. Oxford Univ. Press,p. 184.