12.3. PALEOCLIMATE 273

eastward (through eastward propagationof a wave of depression on the thermo-cline); this deepens the thermocline in theeast Pacific some two months later. This inturn raises the SST in the east. The basicpostulate—that the ocean responds to theatmosphere—has been confirmed in sophis-ticated ocean models forced by ‘‘observed’’wind stresses during an El Nino event.

Third, the El Nino-Southern Oscillationphenomenon arises spontaneously as an oscil-lation of the coupled ocean-atmosphere sys-tem.Bjerknes first suggested that what wenow call ENSO is a single phenomenon anda manifestation of ocean-atmosphere cou-pling. The results noted previously appearto confirm that the phenomenon dependscrucially on feedback between ocean andatmosphere. This is demonstrated in cou-pled ocean-atmosphere models of varyingdegrees of complexity, in which ENSO-like fluctuations may arise spontaneously. Itappears that stochastic forcing of the systemby middle latitude weather systems, whichcan reach down into the tropics to induce‘‘westerly wind bursts,’’ can also play a rolein triggering ENSO events.

Once the El Nino event is fully devel-oped, negative feedbacks begin to dominatethe Bjerknes positive feedback, loweringthe SST and bringing the event to its endafter several months. The details of thesenegative feedbacks involve some very inter-esting ocean dynamics. In essence, whenthe easterlies above the central Pacific startweakening at the beginning of the event, itleads to the formation of an off-equatorialshallower-than-normal thermocline signal,which propagates westward, reflects offthe western boundary of the Pacific, andthen travels eastwards. After a few monthsdelay the thermocline undulation arrivesat the eastern boundary, causing the ther-mocline to shoal there, so terminating thewarm event.

12.2.4. Other modes of variability

The ENSO phenomenon discussed pre-viously is a direct manifestation of strong

coupling between the tropical atmosphereand tropical ocean and it gives rise to coher-ent variability in the coupled climate. Thereare other modes of variability that ariseinternally to the atmosphere (i.e., would bepresent even in the absence of coupling tothe ocean below). Perhaps the most impor-tant of these is the annular mode, a meridionalwobble of the subtropical jet stream. The cli-matological position of the zonal-average,zonal wind, u, is plotted in Fig. 5.20. Butin fact the position and strength of the jetstream maximum varies on all timescales;when it is poleward of its climatologicalposition, u is a few m s−1 stronger thanwhen it is equatorward. These variations inu extend through the depth of the tropo-sphere and indeed right up into the strato-sphere. Importantly for the ocean below,the surface winds and air-sea fluxes alsovary in synchrony with the annular mode,driving variations in SST and circulation.The manifestation of the annular mode inthe northern hemisphere, is known as theNorth Atlantic Oscillation, or NAO for short;the annular mode in the southern hemi-sphere is known as SAM, for southern annularmode. Both introduce stochastic noise intothe climate system that can be reddened byinteraction with the ocean as discussed inSection 12.1.1.

12.3. PALEOCLIMATE

Here we briefly review something ofwhat is known about the evolution of cli-mate over Earth history. Fig. 12.12 listsstandard terminology for key periods ofgeologic time. Study of paleoclimate is anextremely exciting area of research, a fas-cinating detective story in which scientistsstudy evidence of past climates recordedin ocean and lake sediments, glaciers andicesheets, and continental deposits. Prox-ies of past climates are myriad, and to theuninitiated at least, can be bizarre (packratmiddens, midges...), including such mea-surements as the isotopic ratios of shellsburied in ocean sediments, thickness and

274 12. CLIMATE AND CLIMATE VARIABILITY

FIGURE 12.12. The names and dates of the key periods of geologic time. The unit of time is millions of yearsbefore present (M y), except during the Holocene, the last 10,000 years (10k y).

density of tree rings, chemical compositionof ice, and the radioactivity of corals. More-over new proxies continue to be developed.What is undoubtedly clear is that climatehas been in continual change over Earth his-tory and has often been in states that arequite different from that of today. However,it is important to remember that infer-ences about paleoclimate are often basedon sparse evidence,7 and detailed descrip-tions of past climates will never be available.Musing about paleoclimate is neverthe-less intellectually stimulating (and greatfun), because we can let our imaginations

wander, speculating about ancient worldsand what they might tell us about howEarth might evolve in the future. More-over, the historical record challenges andtests our understanding of the underlyingmechanisms of climate and climate change.With such a short instrumental record,paleoclimate observations are essential forevaluating climate variations on timescalesof decades and longer. One can be sure thatthe laws of physics and chemistry (if notbiology!) have not changed over time, andso they place strong constraints on whatmay or may not have happened.

7One should qualify this statement by recognizing the key role of observation in paleoclimate. There are some ‘‘hard’’paleo observations. For example the glacial terminus (moraines) of North America are incontrovertible evidence of theextent of past glaciations. As our colleague Prof. Ed Boyle reminds us, ‘‘Ice Ages would be polite tea-party chit-chat wereit not for geologists climbing mountains in muddy boots.’’

12.3. PALEOCLIMATE 275

Theory and modeling of paleoclimate andclimate change is still rudimentary. This is inpart because we must deal not only with thephysical aspects of the climate system (diffi-cult enough in themselves) but also with bio-geochemical transformations, and on verylong timescales, geology and geophysics(cf. Fig. 12.1). Understanding biogeochem-istry is particularly important because itis often required to appreciate and quan-tify the proxy climate record itself. More-over, because greenhouse gases, such asH2O, CO2, and CH4, are involved in life,we are presented with a much more chal-lenging problem than the mere applicationof Newton’s laws of mechanics and the lawsof thermodynamics to the Earth. There aremany ideas on mechanisms driving climatechange on paleoclimate timescales, only afew on which there is consensus, and evenwhen a consensus forms, there is often littlesupporting evidence.

Here we have chosen to focus on thoseaspects of the paleoclimate record for which,it seems to us, there is a broad consensus andare less likely to be challenged as new evi-dence comes to the fore. In Section 12.3.1 wereview what is known about the evolutionof climate on the billion year timescale, andthen in Section 12.3.2, focus in on the last70M y or so. Warm and cold climates are dis-cussed in Sections 12.3.3 and 12.3.4, respec-tively. We finish by briefly reviewing theevidence for glacial-interglacial cycles andabrupt climate change (Section 12.3.5) and,very briefly, global warming (Section 12.3.6).

12.3.1. Climate over Earth history

Earth has supported life of one form oranother for billions of years, suggesting thatits climate, although constantly changing,has remained within somewhat narrow lim-its over that time. For example, ancient rocksshow markings that are clear evidence oferosion due to running water, and primitive

life forms may go back at least 3.5B y.One might suppose that there is a natu-ral ‘‘thermostat’’ that ensures that the Earthnever gets too warm or too cold. One mightalso infer that life finds a way to eke out anexistence.

It is clear that some kind of thermostatmust be in operation because astrophysi-cists have concluded that 4B y ago the Sunwas burning perhaps 25–30% less stronglythan today. Simple one-dimensional climatemodels of the kind discussed in Chapter 2suggest that if greenhouse gas concentra-tions in the distant past were at the samelevel as today, the Earth would have frozenover for the first two thirds or so of itsexistence.8 This is known as the faint earlySun paradox (see Problem 4 at the end ofthe chapter). A solution to the conundrumdemands the operation of a thermostat,warming the Earth in the distant past andcompensating for the increasing strengthof the Sun over time. If the thermostatinvolved carbon, an assumption that per-haps needs to be critically challenged but iscommonly supposed, then we must explainhow CO2 levels in the atmosphere mighthave diminished over time.

On very long timescales one must con-sider the exchange of atmospheric CO2with the underlying solid Earth in chemicalweathering. Carbon is transferred from theEarth’s interior to the atmosphere as CO2gas produced during volcanic eruptions.This is balanced by removal of atmosphericCO2 in the chemical weathering of conti-nental rocks, which ultimately deposits thecarbon in sediments on the sea floor; see theschematic Fig. 12.13. It is remarkable that therate of input of CO2 by volcanic activity andthe rate of removal by chemical weatheringhas remained so closely in balance, eventhough the input and output themselves areeach subject to considerable change.

Volcanic activity is unlikely to be part ofa thermostat, because it is driven by heat

8Indeed there are hints in the paleoclimate record that Earth may have come close to freezing over during severalperiods of its history (most likely between 500M and 800M y ago), to form what has been called the ‘‘snowball Earth.’’

276 12. CLIMATE AND CLIMATE VARIABILITY

FIGURE 12.13. Carbon from the Earth’s interior isinjected in to the atmosphere as CO2 gas in volcaniceruptions. Removal of atmospheric CO2 on geologicaltimescales is thought to occur in the chemical weather-ing of continental rocks, being ultimately washed intothe ocean and buried in the sediments. The two pro-cesses must have been in close, but not exact balance,on geological time scales.

sources deep within the Earth that cannotreact to climate change. Chemical weath-ering of rocks, on the other hand, may besensitive to climate and atmospheric CO2concentrations, because it is mediated bytemperature, precipitation, vegetation, andorographic elevation and slope, which areclosely tied together (remember the discus-sion in Section 1.3.2). So, the argumentgoes, if volcanic activity increased for aperiod of time, elevating CO2 levels in theatmosphere, the resulting warmer, moisterclimate might be expected to enhance chem-ical weathering, increasing the rate of CO2removal and reducing greenhouse warmingenough to keep climate roughly constant.Conversely, in a cold climate, arid con-ditions would reduce weathering rates,leading to a build-up of atmospheric CO2and a warming tendency. Scientists vigor-ously debate whether such a mechanismcan regulate atmospheric CO2; it is cur-rently very difficult to test the idea withobservation or models.

Whatever the regulatory mechanism,when the fragmentary paleoclimate recordof atmospheric CO2 levels and temperaturesis pieced together over geologic time, a con-nection emerges. Figure 12.14 shows a syn-thesis of evidence for continental glaciationplotted along with estimates of atmosphericCO2 (inferred from the geological record andgeochemical models) over the past 600M y.Such reconstructions are highly problem-atical and subject to great uncertainty. Wesee that CO2 levels in the atmosphere werethought to have been generally much grea-ter in the distant past than at present, per-haps as much as 10 to 20 times presentlevels 400–500M y ago. Moreover, glaciationappears to occur during periods of low CO2and warm periods in Earth history seem tobe associated with elevated levels of CO2.That the temperature and CO2 concentra-tions appear to co-vary, however, should notbe taken as implying cause or effect.

Factors other than variations in thesolar constant and greenhouse gas forc-ing must surely also have been at workin driving the changes seen in Fig. 12.14.These include changes in the land-sea dis-tribution and orography (driven by platetectonics), the albedo of the underlyingsurface, and global biogeochemical cycles.One fascinating idea—known as the Gaiahypothesis—is that life itself plays a rolein regulating the climate of the planet,optimizing the environment for continuedevolution. Another idea is that ocean basinshave evolved on geological timescalesthrough continental drift, placing chang-ing constraints on ocean circulation andits ability to transport heat meridionally.For example, Fig. 12.15 shows paleogeo-graphic reconstructions from the Jurassic(170M y ago), the Cretaceous (100M y ago),and the Eocene (50M y ago). We have seenin Chapters 10 and 11 how the circula-tion of the ocean is profoundly affected bythe geometry of the land-sea distributionand so we can be sure that the pattern ofocean circulation in the past, and perhapsits role in climate, must have been very

12.3. PALEOCLIMATE 277

FIGURE 12.14. (a) Comparison of CO2 concentrations from a geochemical model (continuous line) with acompilation (Berner, 1997) of proxy CO2 observations (horizontal bars). RCO2 is the ratio of past atmosphericCO2 concentrations to present day levels. Thus RCO2 = 10 means that concentrations were thought to be 10 timespresent levels. (b) CO2 radiative forcing effects expressed in W m−2. (c) Combined CO2 and solar radiance forcingeffects in W m−2. (d) Glaciological evidence for continental-scale glaciation deduced from a compilation of manysources. Modified from Crowley (2000).

different from that of today. It has beenhypothesized that the opening and clos-ing of critical oceanic gateways—narrowpassages linking major ocean basins—havebeen drivers of climate variability by regu-lating the amount of water, heat, and saltexchanged between ocean basins. This, forexample, can alter meridional transport ofheat by the ocean and hence play a rolein glaciation and deglaciation. There are anumber of important gateways.

Drake Passage, separating South Americafrom Antarctica, opened up 25–20M y ago,leaving Antarctica isolated by what wenow call the Antarctic Circumpolar Cur-rent (Fig. 9.13). This may have made it moredifficult for the ocean to deliver heat tothe south pole, helping Antarctica to freezeover. However this hypothesis has timingproblems. Ice first appeared on Antarctica35M y ago, before the opening of Drake Pas-sage, and the most intense glaciation overAntarctica occurred 13M y ago, significantly

after it opened. Uplift of Central Americaover the past 10M y closed a deep ocean pas-sage between North and South America toform the Isthmus of Panama about 4 millionyears ago. Before then the Isthmus wasopen, allowing the trade winds to blowwarm and possibly salty water between theAtlantic and the Pacific. Its closing couldhave supported a Gulf Stream carrying trop-ical waters polewards, as in today’s climate,possibly enhancing the meridional over-turning circulation of the Atlantic basin andhelping to warm northern latitudes in theAtlantic sector, as discussed in Chapter 11.Finally, it has been suggested that closing ofthe Indonesian seaway, 3–4M y ago, was aprecursor to East African aridification.

12.3.2. Paleotemperatures over the past70 million years: the δ18O record

Let us zoom into the last 70M y periodof Fig. 12.14. The paleorecord suggests that

278 12. CLIMATE AND CLIMATE VARIABILITY

FIGURE 12.15. Paleogeographic reconstructions for (top) the Jurassic (170M y ago), (middle) the Cretaceous(100M y ago), and (bottom) the Eocene (50M y ago). Panthalassa was the huge ocean that in the paleo worlddominated one hemisphere. Pangea was the supercontinent in the other hemisphere. The Tethys Sea was the bodyof water enclosed on three sides (and at times, almost four sides) by the generally ‘‘C-shaped’’ Pangea.

12.3. PALEOCLIMATE 279

over the last 55M y there has been a broadprogression from generally warmer to gen-erally colder conditions, with significantshorter-term oscillations superimposed.How can one figure this out? Some keysupporting evidence is shown in Fig. 12.16based on isotopic measurements of oxygen.Sediments at the bottom of the ocean pro-vide a proxy record of climate conditions inthe water column. One key proxy is δ18O—ameasure of the ratio of two isotopes ofoxygen, 18O and 16O—which is recorded insea-bed sediments by the fossilized calciteshells of foraminifera (organisms that livenear the surface or the bottom of the ocean).It turns out that the δ18O in the shells is afunction of the δ18O of the ocean and thetemperature of the ocean (see Appendix A.3for a more detailed discussion of δ18O). Therecord of δ18O over the last 55M y indicates acooling of the deep ocean by a massive 14◦C.In other words, deep ocean temperatureswere perhaps close to 16◦C (!!) compared to2◦C as observed today (cf. Fig. 9.5). If over

this period of time the abyssal ocean wereventilated by convection from the poles asin today’s climate (note how temperaturesurfaces in the deep ocean thread back tothe pole in Fig. 9.5), then one can concludethat surface conditions at the poles mustalso have been very much warmer. Indeed,this is consistent with other sources ofevidence, such as the presence of fossilizedremains of palm trees and the ancestors ofmodern crocodiles north of the Arctic Circle60M y ago.

To explain such a large cooling trend,sustained over many millions of years, oneneeds to invoke a mechanism that persistsover this enormous span of time. Follow-ing on from the discussion in Section 12.3.1,at least two important ideas have beenput forward as a possible cause. Firstly,it has been suggested that the balance inFig. 12.13 might have changed to reduceCO2 forcing of atmospheric temperaturesover this period due to (a) decreased inputof CO2 from the Earth’s interior to the

Millionsof years

FIGURE 12.16. A compilation of δ18O measurements from the fossilized shells of benthic foraminifera analyzedfrom many sediment cores in the North Atlantic over 70M y. Modified from Miller et al (1987).

280 12. CLIMATE AND CLIMATE VARIABILITY

climate system, as the rate of sea-floorspreading decreases over time, reducingvolcanic activity, and (b) increased removalof CO2 from the atmosphere due to en-hanced physical and chemical weatheringof unusually high-elevation terrain drivenby tectonic uplift. Secondly, it has been sug-gested that poleward ocean heat transportprogressively decreased because of changesin the distribution of land and sea, andthe opening up and closing of gateways, asbriefly discussed in Section 12.3.1.

Whatever the mechanisms at work, asshown in Fig. 12.14, the paleorecord sug-gests that, over the past 100M y or so, theEarth has experienced great warmth andperiods of great cold. We now briefly reviewwhat ‘‘warm’’ climates and ‘‘cold’’ climatesmight have been like.

12.3.3. Greenhouse climates

In the Cretaceous period Earth was a‘‘greenhouse world.’’ There were no ice capsand sea level was up to 100–200 m higherthan present, largely due to the melting ofall ice caps and the thermal expansion of theoceans that were much warmer than today.The great super continent of Pangea hadbegun to break apart, and by 100M y agoone can already recognize present-day con-tinents (Fig. 12.15). High sea level meantthat much of the continental areas wereflooded and there were many inland lakesand seas. Indeed, the meaning of the wordCretaceous is ‘‘abundance of chalk,’’ reflect-ing the widespread occurrence of limestonefrom creatures living in the many inlandseas and lakes of the period. Broad-leavedplants, dinosaurs, turtles, and crocodiles allexisted north of the Arctic Circle.

It is thought that the Cretaceous wasa period of elevated CO2 levels—perhapsas much as five times preindustrial con-centrations (see Fig. 12.14)—accounting inpart for its great warmth. CO2 forcingalone is unlikely to account for such warmpoles where temperatures were perhaps25◦C warmer than today. One proposed

explanation is that the oceans carried muchmore heat poleward than today, renderingthe poles warmer and the tropics colder.There is speculation that the deep oceanmay have been much warmer and saltierthan at present, possibly due to convectionin the tropics and/or subtropics triggeredby high values of salinity, much as observedtoday in the Eastern Mediterranean. Indeed,the configuration of the continents may havebeen conducive to such a process; the pres-ence of the Tethys Sea (see Fig. 12.15 middle)and a large tropical seaway extending upin to subtropical latitudes, underneath thesinking branch of the Hadley cell bringingdry air down to the surface, could haveincreased evaporation and hence salinity tothe extent that ocean convection was trig-gered, mixing warm, salty water to depth.But this is just speculation. It is very difficultto plausibly quantify and model this processand efforts to do so often meet with failure.

Another major challenge in understand-ing the paleorecord in the Cretaceous isthe evidence that palm trees and reptileswere present in the interior of the conti-nents. Crocodiles and (young) palm treesare not frost resistant, indicating that tem-peratures did not go below freezing evenduring the peak of winter at some latitudesnorth of 60◦ N and in the middle of conti-nents, away from the moderating effects ofthe ocean. Models, however, simulate freez-ing conditions in the continental interiors inwinter even when CO2 is increased very dra-matically. Note the large seasonal changein temperature in the interior of continentsobserved in the present climate (Fig. 12.2).Perhaps lakes and small inland seas helpedto keep the interior warm.

12.3.4. Cold climates

Most of the time in the last 1M y, Earthhas been much colder than at present, andice has encroached much further equator-ward (Fig. 12.14). The glacial climate that weknow the most about is that at the height ofthe most recent glacial cycle—the last glacial

12.3. PALEOCLIMATE 281

maximum (LGM) between 18 and 23k y ago,during which ice sheets reached their great-est extent 21k y ago. Reconstruction ofclimate at the LGM was carried out in theCLIMAP project9 employing, in the main,proxy data from ocean sediments. Thick icecovered Canada, the northern United States(as far south as the Great Lakes), northernEurope (including all of Scandinavia, thenorthern half of the British Isles, and Wales)and parts of Eurasia. The effect on sur-face elevation is shown in Fig. 12.17, whichshould be compared to modern conditionsshown in Fig. 9.1. Where Chicago, Glasgow,and Stockholm now stand, ice was over1 km thick. It is thought that the LaurentideIce sheet covering N. America had roughlythe volume of ice locked up in present-dayAntarctica. Sea level was about 120–130 mlower than today. Note that the coastlineof the LGM shown in Fig. 12.17 revealsthat, for example, the British Isles wereconnected to Europe, and many islandsthat exist today were joined to Asia andAustralia. Most of the population lived inthese fertile lowlands, many of which arenow under water. Ice sheets on Antarcticaand Greenland extended across landexposed by the fall of sea level. More-over, sea ice was also considerably moreextensive, covering much of the Greenlandand Norwegian Seas, and persisted throughthe summer. In the southern hemisphere,Argentina, Chile, and New Zealand wereunder ice, as were parts of Australia andSouth America.

Figure 12.17b shows the difference bet-ween average August SST centered on theLGM and August SST for the modern era.Many details of this reconstruction havebeen challenged, but the broad featuresare probably correct. The average SST was4◦C colder than present and North AtlanticSSTs were perhaps colder by more than 8◦C.It appears that low latitude temperatures

were perhaps 2◦C lower than today. Windsat the LGM were drier, stronger, and dustierthan in the present climate. Ice sheets, bygrinding away the underlying bedrock, arevery efficient producers of debris of all sizes,which gets pushed out to the ice margin.At the LGM, windy, cold, arid condi-tions existed equatorward of the ice. Windsscooped up the finer-grained debris, result-ing in great dust storms blowing across theEarth’s surface with more exposed shelfareas. Indeed, glacial layers in ice coresdrilled in both Greenland and Antarcticacarry more dust than interglacial layers.Forests shrank and deserts expanded. Todaythe N. African and Arabian deserts arekey sources of dust; at the LGM desertsexpanded into Asia. One very significantfeature of glacial climates evident in thepaleorecord is that they exhibited consider-ably more variability than warm climates.For example, in an event known as theYounger Dryas, which occurred about 12k yago, the climate warmed only to suddenlyreturn to close to LGM conditions for sev-eral hundred years; see Fig. 12.23 and thediscussion in Section 12.3.5.

Key factors that may explain the dra-matically different climate of the LGMare the presence of the ice sheets them-selves, with their high albedo reflectingsolar radiation back out to space, and(see below) lower levels of greenhousegases. It is thought that the pronouncedclimate variability of glacial periods sug-gested by the paleorecord may have beenassociated with melting ice producinglarge inland lakes that were perhapscut off from the oceans for hundredsof years, but which then intermittentlyand perhaps suddenly discharged intothe oceans. It has been argued that suchsudden discharges of buoyant fluid overthe surface of the northern N. Atlanticcould have had a significant impact on

9CLIMAP (Climate: Long-range Investigation, Mapping and Prediction), was a major research project of the 1970s and1980s, which resulted in a map of climate conditions during the last glacial maximum based on proxy data from oceansediments.

282 12. CLIMATE AND CLIMATE VARIABILITY

FIGURE 12.17. (a) CLIMAP reconstruction of elevation at the Last Glacial Maximum (LGM). The white (black)areas represent terrain with a height in excess of (less than) 1.5 km and are indicative of ice-covered areas. Thedepth of the ocean is represented with a grey scale (dark is deep). The white contour marks the 4 km deep isobath.This figure should be compared with Fig. 9.1. Note the modification of the coast line relative to the modern, dueto the 120 m or so drop in sea level. (b) August SST at LGM (from CLIMAP) minus August SST for the modernclimate (◦C). The brown areas represent negative values, the green areas positive values.

the strength of the ocean’s meridionaloverturning circulation and its ability totransport heat polewards.

12.3.5. Glacial-interglacial cycles

The left frame of Fig. 12.18 shows theδ18O record over the past 2.5 million years,recorded in the calcite of foraminifera in

sediments of the subpolar North Atlantic.Before about 800k y ago, one observes rem-arkable oscillations spanning 2M y or so,with a period of about 40k y. After 800k yago the nature of the record changes andfluctuations with longer periods are super-imposed. These are the signals of greatglacial-interglacial shifts on a roughly 100k ytimescale. There have been about 7 such

12.3. PALEOCLIMATE 283

FIGURE 12.18. Left: δ18 O over the last 2.5 million years recorded in the calcite shells of bottom dwellingforaminifera in the subpolar North Atlantic. Shown is the average of tens of δ18 O records sampled from variousmarine sediment cores (Huybers, 2006). Values are reported as the anomaly from the average δ18 O over the pastmillion years. More negative values (rightward) indicate warmer temperatures and less ice volume. Right: δ18 O ofice over the last 50 k y measured in the GISP2 ice-core (Grootes and Stuiver, 1997). In contrast to the δ18 O of marineshells, less negative values in the δ18 O of ice indicate warmer atmospheric temperatures, in this case in the vicinityof Greenland.

cycles, during which temperate forests inEurope and North America have repeat-edly given way to tundra and ice. Icehas periodically accumulated in the NorthAmerican and Scandinavian areas until itcovered hills and mountains to heights of2–3 km, as was last observed at the LGM (seeFig. 12.17) and today only in Greenland andAntarctica.

Such glacial-interglacial signals arenot limited to the North Atlantic sector.Qualitatively similar signals are evidentin different kinds of paleorecords takenfrom around the world, including deepsea sediments, continental deposits ofplants, and ice cores. These reveal amarked range of climate on Earth, cyclingbetween glacial and interglacial conditions.

284 12. CLIMATE AND CLIMATE VARIABILITY

In particular, ice cores taken from glaciersyield local air temperature,10 precipitationrate, dust, and direct records of past trace gasconcentrations of CO2 and CH4. The deepestcore yet drilled (� 3 km), from Antarctica,records a remarkable 700k y history ofclimate variability shown in Fig. 12.19.The core reveals oscillations of Antarcticair temperature, greenhouse gas concen-trations which more-or-less covary with aperiod of about 100k y. Note, however, thatthe oscillations do not have exactly the sameperiod. Of the six or seven cycles seen in theAntarctic record, the two most recent have asomewhat longer period than the previouscycles.

The 100k y signals evident in Figs. 12.18and 12.19 are thought to be representative ofclimate variability over broad geographicalregions. Scientists vigorously debate whe-ther, for example, changes over Antarcticaled or lagged those over Greenland, orwhether CO2 changes led or lagged tem-perature changes. This is very difficult totie down because of uncertainty in theprecise setting of the ‘‘clock’’ within andbetween records. Here we simply state thatat zero order the low frequency signalsseem to covary over broad areas of theglobe, strongly suggestive of global-scalechange.

The oscillations seen in Fig. 12.19 havea characteristic ‘‘saw-tooth’’ pattern, typ-ical of many records spanning glacial-interglacial cycles, with a long period ofcooling into the glacial state followed by

rapid warming to the following interglacial.Abrupt increases in CO2 occur during theperiod of rapid ice melting. Superimposedon the sawtooth are irregular higher fre-quency oscillations (to be discussed below).Typically, the coolest part of each glacialperiod and the lowest CO2 concentrationsoccur just before the glacial termination.Temperature fluctuations (representativeof surface conditions) have a magnitudeof about 12◦C and CO2 levels fluctuatebetween 180 and 300 ppm. The Antarcticdust record also confirms continental arid-ity. Dust transport was more prevalentduring glacial than interglacial times, asmentioned in Section 12.3.4. Finally, it isworthy of note that present levels of CO2(around 370 ppm in the year 2000; cf.Fig. 1.3) are unprecedented during thepast 700k ys. By the end of this centurylevels will almost certainly have reached600 ppm.

Milankovitch cycles

It seems that climate on timescales of10k y–100k ys is strongly influenced byvariations in Earth’s position and ori-entation relative to the Sun. Indeed, aswe shall see, some of the expected peri-ods are visible in the paleorecord, butdirect association (phasing and amplitude)is much more problematical. Variation inthe Earth’s orbit over time—known asMilankovitch cycles11—cause changes in theamount and distribution of solar radiation

10Note that 18O/16O ratios in ice cores have the opposite relationship to temperature than that of 18O/16O ratios inCaCO3 shells (see Appendix A.3). Snow produced in colder air tends to have a lower δ18O value than snow producedin warmer air. Consequently, the δ18O value of glacial ice can be used as a proxy for air temperature, with low valuesindicating colder temperatures than higher values (see Fig. 12.18).

11 Milutin Milankovitch (1879–1958), the Serbian mathematician, dedicated his career to for-mulating a mathematical theory of climate based on the seasonal and latitudinal variationsof solar radiation received by the Earth. In the 1920s he developed improved methods ofcalculating variations in Earth’s eccentricity, precession, and tilt through time.

12.3. PALEOCLIMATE 285

FIGURE 12.19. Ice-core records of atmospheric carbon dioxide (left) and methane (middle) concentrationsobtained from bubbles trapped in Antarctic ice. Values to 400 k y ago are from Vostok (Petit et al, 1999), whereasearlier values are from EPICA Dome C (Siegenthaler et al, 2005; Spahni et al, 2005). (right) δ D concentrationsfrom EPICA Dome C (EPICA community members, 2004) measured in the ice, as opposed to the bubbles, areindicative of local air temperature variations, similar to δ18 O of ice measurements. A rightward shift correspondsto warming.

286 12. CLIMATE AND CLIMATE VARIABILITY

reaching the Earth on orbital timescales.Before discussing variations of the Earth’sorbit over time, let us return to ideas intro-duced in Chapter 5 and review some simplefacts about Earth’s orbit around the Sunand the cause of the seasons.

Imagine for a moment that the Earth trav-elled around the Sun in a circular orbit,as in Fig. 12.20a (left). If the Earth’s spinaxis were perpendicular to the orbital plane(i.e., did not tilt), we would experienceno seasons and the length of daytime andnighttime would never change throughoutthe year and be equal to one another. But

FIGURE 12.20. (a) The eccentricity of the Earth’sorbit varies on 100k y & 400k y timescales from (almost)zero, a circle, to 0.07, a very slight ellipse. Theellipse shown on the right has an eccentricity of 0.5,vastly greater than that of Earth’s path around theSun. (b) The change in the tilt of the Earth’s spinaxis—the obliquity—varies between 22.1◦and 24.5◦

on a timescale of 41k y. The tilt of the Earth is cur-rently 23.5◦. (c) The direction of the Earth’s spin vectorprecesses with a period of 23k y.

now suppose that the spin axis is tilted as aconstant angle, as sketched in Fig. 5.3, and,moreover, that the direction of tilt in space isconstant relative to the fixed stars. Now, asdiscussed in Section 5.1.1, we would expe-rience seasons and the length of daytimewould vary throughout the year. When thenorthern hemisphere (NH) is tilted towardthe Sun, the Sun rises high in the sky, day-time is long, and the NH receives intenseradiation and experiences summer condi-tions. When the NH tilts away from theSun, the Sun stays low in the sky, thedaytime is short and the NH receives dimin-ished levels of radiation and experienceswinter. These seasonal differences culmi-nate at the summer and winter solstices. Inmodern times, the longest day of the yearoccurs on June 21st (the summer solstice)and shortest day of the year on Decem-ber 21st (the winter solstice) (see Fig. 5.4).The length of the day and night becomeequal at the equinoxes. Thus we see thatseasonality and length of day variations arefundamentally controlled by the tilt of theEarth’s axis away from the orbital plane.This tilt of the Earth’s axis away from theorbital plane is known as the obliquity (seeFig. 12.20b). It varies between 21.1◦ and 24.5◦

on about 41k y timescales; at the presenttime it is 23.5◦. Obliquity affects the annualinsolation in both hemispheres simultane-ously. When the tilt is large, seasonality athigh latitudes becomes more extreme butwith little effect at the equator.

The Earth’s orbit is not exactly circular,however. As shown in Figs. 5.4 and 12.20,Earth moves around the Sun following anelliptical path; the distance from the Sunvaries between 153 million km at perihelion(closest distance of the Earth to the Sun) and158 million km at aphelion (farthest dis-tance between the Earth and Sun). As can beseen in Fig. 5.4, in modern times the Earth isslightly closer to the Sun at the NH wintersolstice. Winter radiation is slightly higherthan it would be if the Earth followed a per-fectly circular orbit. Conversely, at the NHsummer solstice the Earth is slightly farther

12.3. PALEOCLIMATE 287

away from the Sun, and so NH summerradiation is slightly lower than it would beif the Earth followed a perfectly circularorbit. This is a rather small effect, how-ever, because the Earth-Sun distance onlyvaries by 3% of the mean. Neverthelessthe eccentricity of the Earth’s orbit aroundthe Sun (see Fig. 12.20a), a measure of itsdegree of circularity, enhances or reducesthe seasonal variation of the intensity ofradiation received by the Earth. The eccen-tricity varies with periods of about 100k yand 400k y. It modulates seasonal differ-ences and precession, the third importantorbital parameter.

Precession measures the direction of theEarth’s axis of rotation, which affects themagnitude of the seasonal cycle and is ofopposite phase in the two hemispheres.Earth’s spin axis precesses at a period of27k y with respect to the fixed stars. How-ever, this is not the climatically relevantperiod because the direction of the majoraxis of Earth’s eccentric orbit also moves.Thus climatologists define the climatic pre-cession as the direction of Earth’s spin axiswith respect to Earth’s eccentric orbit. Thishas a period of about 23k y. Today the rota-tion axis points toward the North Star, sosetting the dates during the year at which theEarth reaches aphelion and perihelion on itsorbit around the Sun (see Fig. 5.4). At thepresent time, perihelion falls on January 3rd,only a few weeks after the winter solstice,and so the northern hemisphere winter andsouthern hemisphere summer are slightlywarmer than the corresponding seasons inthe opposite hemispheres.

We discussed in Chapters 5 and 8 thosefactors that control the annual-mean tem-perature as a function of latitude and inparticular the importance of the latitudi-nal dependence of incoming solar radiation.This latitudinal dependence is criticallymodulated by orbital parameters. Becauseof their different periodicity (see Fig. 12.21),the composite variations in solar radiationare very complex. They are functions ofboth latitude and season, as well as time.

FIGURE 12.21. Variations in eccentricity, preces-sion, and obliquity over 300k y, starting 200k y in thepast, through the present day and 100k y in to thefuture. From Berger and Loutre, (1992).

Variations in summer insolation in middleto high latitudes are thought to play a par-ticularly important role in the growth andretreat of ice sheets: melting occurs only dur-ing a short time during the summer and icesurface temperature is largely determinedby insolation. Thus cool summers in thenorthern hemisphere, where today most ofthe Earth’s land mass is located, allow snowand ice to persist through to the next winter.In this way large ice sheets can developover hundreds to thousands of years.Conversely, warmer summers shrink icesheets by melting more ice than canaccumulate during the winter.

Figure 12.22 shows insolation variationsas a function of latitude and seasonsduring various phases of Earth’s orbit.These can be calculated very accurately,as was first systematically carried out byMilankovitch. Note that fluctuations oforder 30 W m−2 occur in middle to high

288 12. CLIMATE AND CLIMATE VARIABILITY

FIGURE 12.22. Insolation at the top of the atmosphere computed using the orbital solution of Berger and Loutre(1992). (a) Daily average intensity in W m−2 contoured against latitude and month, indicating average conditionsover the last two million years. (b) Modern insolation plotted as an anomaly from average conditions. (c) Insolationaveraged during each maximum of obliquity over the last two million years and shown as an anomaly from averageconditions. (d) Similar to (c) but for when Earth is closest to the Sun during northern hemisphere summer solstice.

latitudes, a significant signal comparable,for example, to the radiative forcing dueto clouds.

Astronomical forcing is an immenselyappealing mechanism, offering a seeminglysimple explanation of climate variability ontimescales of tens to hundreds of thousandsof years. It is widely applied in an attempt torationalize the paleorecord. One of the mostconvincing pieces of evidence of astronom-ical periods showing up in the paleorecordare the fluctuations in δ18O of calcite foundin North Atlantic deep sea cores shown inFig. 12.18 (left) over the past 2.5M y. Asdiscussed previously, an oscillation with aperiod of about 40k y, that of obliquity, canbe seen by eye for the first 2M y of therecord. However, the 100k y cycles at theend of the record (see also Fig. 12.19), which

are signatures of massive glacial-interglacialcycles, may have little directly to do withorbital forcing, which has very little powerat this period. Perhaps the 100K y cycleis being set by internal dynamics of theice sheets, almost independently of orbitalforcing. Whatever the extent of orbital forc-ing, it must be significantly amplified bypositive feedbacks involving some or allof the following: water vapor, ice-albedointeractions, clouds, ocean circulation, inter-nal ice-sheet dynamics, among many otherprocesses.

In summary, many theories have beenput forward to account for the shape andperiod of oscillations of the kind seen in, forexample, Figs. 12.18 and 12.19, but none canaccount for the observed record and none isgenerally accepted.

12.3. PALEOCLIMATE 289

Abrupt Climate Change

As we have seen, over Earth history theclimate of the planet has been in markedlydifferent states, ranging from a ‘‘green-house’’ to an ‘‘icehouse.’’ Moreover thepaleorecord suggests that there have beenvery rapid oscillations between glacial andinterglacial conditions. For example, the50k y record of δ18O shown in Fig. 12.18(right), taken from an ice core in Green-land, reveal many large rightward spikes onmillennial timescales (indicating frequentabrupt transitions to warmer followed bya return to colder conditions). These arecalled Dansgaard-Oeschger events (or D-Ofor short, after the geochemists Willi Dans-gaard and Hans Oeschger who first notedthem) and correspond to abrupt warmingsof Greenland by 5–10◦C, followed by grad-ual cooling and then an abrupt drop tocold conditions again. They were probablyconfined to the N. Atlantic and are lessextreme than the difference between glacialand interglacial states. Scientists have alsofound evidence of millennial timescalefluctuations in the extent of ice-rafteddebris deposited in sediments in the NorthAtlantic—known as Heinrich events (afterthe marine geologist Hartmut Heinrich).They are thought to be the signature of inter-mittent advance and retreat of the sea-iceedge. D-O and Heinrich events are examplesof what are called ‘‘abrupt’’ climate changes,because they occur on timescales very muchshorter (10, 100, 1000 y) than that of exter-nal climate forcings, such as Milankovitchcycles, but long compared to the seasons. Itis important to realize then that the LGMwas not just much colder than today, butthat it repeatedly and intermittently swungbetween frigid and milder climates in just afew decades. Indeed such erratic behavior isa feature of the last 100k y of climate history.

The general shift from colder, dustierconditions to warmer, less dusty conditionsover the last 10k y or so seen in Fig. 12.23,is generally interpreted as the result oforbital-scale changes in obliquity and pre-cession. Obliquity reached a maximum

10k y ago (see Fig. 12.21), enhancing theseasonal cycle and producing a maximumof summer insolation at all latitudes inthe northern hemisphere (Fig. 12.22c), somaking it less likely that ice survives thesummer. Atmospheric CO2 concentrationsmay also have played a role (althoughit is not known to what extent they area cause or an effect), increasing from190 ppm to 280 ppm (Fig. 12.19). Thecombination of increased summer inso-lation and increased CO2 concentrationsprobably triggered melting of the mas-sive northern ice sheets, with ice-albedofeedbacks helping to amplify the shifts. Itis thought that huge inland lakes wereformed, many times the volume of thepresent Great Lakes, which may have inter-mittently and suddenly discharged into theArctic/Atlantic Ocean. As can be seen inFig. 12.23, the warming trend after the lastice age was not monotonic but involvedlarge, short-timescale excursions. Evidencefrom the deposits of pollen of the plantDryas octopetala, which thrives today in coldtundra in Scandinavia, tells us that 12k yago or so, warming after the LGM waspunctuated by a spell of bitter cold, a periodnow known as the Younger Dryas. Furtherevidence for this cold period, together withnumerous other fluctuations, come fromGreenland ice cores such as that shown inFig. 12.18 (right). Along with the longer-term trends, one observes (Fig. 12.23) spec-tacular, shorter-term shifts, of which theYounger Dryas is but one.

After the last ice age came to an end, theclimate warmed up dramatically to reachpresent day conditions around 10k y ago.Since then climate has settled in to a rela-tively quiescent mode up until the presentday. This period—the last 10k y—is knownas the Holocene. There was a warm climaticoptimum between 9k and 5k y ago, dur-ing which, for example, El Nino appearsto have been largely absent. The relativelybenign climate of the Holocene is per-haps the central reason for the explosionin the development of human social and

290 12. CLIMATE AND CLIMATE VARIABILITY

FIGURE 12.23. The transition from the Last Glacial Maximum to the relatively ice-free conditions of theHolocene took roughly ten thousand years. In certain regions this transition was punctuated by rapid climatevariations having timescales of decades to millennia. Shown is the GISP2 ice-core (Grootes and Stuiver, 1997) withshading indicating the return to glacial-like conditions, a period known as the Younger Dryas. The Younger Dryasis a prominent feature of many North Atlantic and European climate records and its presence can be detected inclimate records across much of the Northern Hemisphere.

economic structures, farming, and agricul-ture. Before the Holocene, agriculture wasperhaps impossible in much of NorthernEurope, because the variance in climate wasso great.

A commonly held view is that an impor-tant mechanism behind rapid climate shiftsis fluctuation of the ocean’s thermoha-line circulation discussed in Chapter 11.

The thermohaline circulation may havebeen sensitive to freshwater discharge frominland lakes formed from melting ice. Thedischarge of fresh water may have occurredintermittently and perhaps involved largevolumes of fresh water sufficient to alterthe surface salinity and hence buoyancyof the surface ocean. Let us return toFig. 11.28 (bottom), which shows the ocean’s

12.3. PALEOCLIMATE 291

meridional overturning circulation (MOC),with a deep-sinking branch in the northernNorth Atlantic. Warm, salty water is con-verted into colder, fresher water by heatloss to the atmosphere and fresh water sup-ply from precipitation and ice flow from theArctic. As discussed in Chapter 11, in thepresent climate the MOC carries heat pole-ward, helping to keep the North Atlanticice-free. But what might have happened if,for some reason, fresh water supply to polarconvection sites was increased, as was likelyin the melt after the LGM, reducing salinityand so making it more difficult for the oceanto overturn? One might expect the MOC todecrease in strength,12 with a concomitantreduction in the supply of heat to north-ern latitudes by ocean circulation, perhapsinducing cooling and accounting for theabrupt temperature fluctuations observedin the record.

Theories that invoke changes in theocean’s MOC as an explanation of abruptclimate change signals, although appealing,are not fully worked out. Sea ice, with itsvery strong albedo and insulating feedbacksthat dramatically affect atmospheric tem-perature, is a potential amplifier of climatechange. Moreover sea ice can also grow and(or) melt rapidly because of these positivefeedbacks and so is likely to be an importantfactor in abrupt climate change. A wind fieldchange could also account for the observedcorrelation between reconstructed Green-land temperatures and deep sea cores, withchanges in ocean circulation being drivendirectly by the wind. Moreover, the windfield is likely to be very sensitive to thepresence or absence of ice, because of itselevation, roughness, and albedo properties.

12.3.6. Global warming

Since the 1950s, scientists have been con-cerned about the increasing atmosphericconcentrations of CO2 brought about by

human activities (cf. Fig. 1.3). The problem,of course, is that the carbon locked up inthe oil and coal fields, the result of burialof tropical forests over tens of millionsof years, are very likely to be returnedto the atmosphere in a few centuries. Asalready mentioned, by the end of this cen-tury atmospheric CO2 concentrations arelikely to reach 600 ppm, not present on theEarth for perhaps 10M y (Fig. 12.14). Thereis concern that global warming will resultand indeed warming induced by humanactivity appears to be already underway.Figure 12.24 shows temperature reconstruc-tions of northern hemisphere surface airtemperature during the last 1100 y togetherwith the instrumental record over the past150 y or so. The spread between the recon-structions indicates a lower bound on theuncertainty in these estimates. Even aftertaking due note of uncertainty and that thetemperature scale is in tenths ◦C, the rapidrise in the late twentieth century is alarmingand, should it continue, cause for concern.

Global warming could occur gradually,over the course of a few centuries. However,some scientists speculate that the climatemight be pushed into a more erratic statethat could trigger abrupt change. If theatmosphere were to warm, so the argumentgoes, it would contain more water vapor,resulting in an enhancement of meridionalwater vapor transport, enhanced precipita-tion over the pole, a suppression of oceanconvection, a reduction in the intensity ofthe MOC, thence a reduction in the merid-ional ocean heat transport, and so an abruptcooling of the high-latitude climate.

Even though the possibility of imminentabrupt climate change is small and, as faras we know, even less likely to occur inwarm periods such as our own, it must betaken very seriously because the impactson the environment and humanity wouldbe so large, the more so if the transitionwere to be very abrupt. As we have seen,

12There is a commonly held misconception that a weakening of the Atlantic MOC is synonymous with a weakeningof the Gulf Stream. As discussed in detail in Chapter 10, the Gulf Stream is a wind-driven phenomenon whose strengthdepends on the wind and is not directly related to buoyancy supply.

292 12. CLIMATE AND CLIMATE VARIABILITY

FIGURE 12.24. Estimates of Northern Hemisphere surface air temperature during the last 1100 years. Temper-atures obtained from instruments (Jones and Moberg, 2003) are shown in black. Colored curves indicate differentproxy reconstructions of temperature. Proxies, such as tree rings, ice cores, and corals, are necessary for estimatingtemperature before widespread instrumental coverage, before about 1850. The spread between the reconstructionsindicates a lower-bound on the uncertainty in these estimates. All records have been smoothed using a 20-yearrunning average and adjusted to have zero-mean between 1900 and 1960.

the paleorecord suggests that such eventshave happened very rapidly in the past (ontimescales as short as a decade). Moreover,climate models support the idea that theocean’s MOC, with its coupling to ice andthe hydrological cycle, is a sensitive compo-nent of the climate system. We simply donot know the likelihood of an abrupt climateshift occuring in the future or, should it doso, the extent to which human activities mayhave played a role.

12.4. FURTHER READING

A good, basic discussion of the physicsof El Nino can be found in Philander (1990).A comprehensive introductory account of

climate over Earth history from the perspec-tive of the paleoclimate record is given inRuddiman (2001). Burroughs (2005) bringsa fascinating human perspective to hisaccount of climate change in prehistory.

12.5. PROBLEMS

1. Consider a homogeneous slab ofmaterial with a vertical diffusivity, kv,subject to a flux of heat through itsupper surface, which oscillates atfrequency ω given by Qnet = ReQωeiωt, where Qω sets the amplitude ofthe net heat flux at the surface. Solvethe following diffusion equation fortemperature variations within the slab,