sign in sign up
<<
Home , Ice core

RESEARCH ARTICLE

iniferal calcite extends from marine (MIS) 50, 1550 thousand ago (ka), to the (Fig. 1A). It shows features in com- Evolution of Ocean Temperature mon with the LR04 stack with stable values (except for the anomalous MIS 5, 9, and 11) and lower glacial values be- and Volume Through the fore MIS 22. There are, however, differences between our Mid-Pleistocene Transition d18O record and the LR04 stack, in particular over the period thought to encompass the MPT. Ben- 18 H. Elderfield,* P. Ferretti,† M. Greaves, S. Crowhurst, I. N. McCave, D. Hodell, A. M. Piotrowski thic d OC shows an abrupt increase from about 950 ka to 870 ka, whereas LR04 shows a more Earth’s climate underwent a fundamental change between 1250 and 700 thousand years ago, gradual increase to values 0.5 per mil (‰) lighter the mid-Pleistocene transition (MPT), when the dominant periodicity of climate cycles changed from than for ODP 1123. This is not an artifact of ben- 41 thousand to 100 thousand years in the absence of substantial change in orbital forcing. Over this thic species analyzed, because we have veri- , an increase occurred in the amplitude of change of deep-ocean foraminiferal isotopic fied this observation by measuring d18Oalsoon ratios, traditionally interpretedasdefiningthemainrhythmofice ages although containing large the epibenthic species Cibicidoides wuellerstorfi effects of changes in deep-ocean temperature. We have separated the effects of decreasing temperature (fig. S3A). and increasing global ice volume on oxygen isotope ratios. Our results suggest that the MPT was What is the importance of this difference? initiated by an abrupt increase in Antarctic ice volume 900 thousand years ago. We see no evidence LR04 is thought to represent a global average from of a pattern of gradual cooling, but near-freezing temperatures occur at every glacial maximum. which regional variability has been removed. Ne- vertheless, LR04 is not an area-weighted stack he deep-ocean sediment oxygen isotopic a low-resolution benthic temperature record for and is biased to the Ocean and eastern 18 composition record (d OC), commonly the past 440 thousand years (ky) (15). We now equatorial Pacific. Only one site samples the ma- Tconsidered to be a record of global ice have generated a 1.5-million- record of deep- jor flow to over half the volume of the world volume and thus a of ice ages, also is sea temperature and sea level with a resolution ocean and that is ODP 1123. The LR04 stack is

“heavily contaminated” (1, 2) by the effect of (≤1000 ky) sufficient to define their behavior over widely used, but to discuss the implications for on September 24, 2012 deep-water temperature variability. The contribu- the period of the MPT, including a comparison changing global ice volume we need to delve 18 tions of ice volume and temperature to d OC with the climate of the last 800 ky derived from further. Because of the possible bias in the stack, have been debated for more than 40 years [e.g., ice-core records (16) and information on changes it makes more sense to compare our site 1123 (1–5)], and definition of the trajectories that each in ocean carbon chemistry. record with a few other high-quality records from followed during the mid-Pleistocene transition The site studied—Ocean Drilling Program (ODP) the world oceans. The more abrupt change in 18 (MPT) has been elusive. Here, we present a high- leg 181, site 1123—is located on the Chatham Rise, d OC is seen in some records but not others (fig. resolution record of deep-water temperature, which east of New Zealand at 41°47.15′S, 171°29.94′W, S3, B to D). Some records show a stepped in- 18 we constructed based on foraminiferal Mg/Ca at a depth of 3290 m (fig. S2). It lies under the crease in d OC over the period 1000 to 600 ka. 18 paleothermometry (see supplementary text and Deep Western Boundary Current (DWBC) that If interpreted as ice volume, d OC shows that www.sciencemag.org fig. S1). Only one previous deep-sea temperature feeds the Pacific and comprises Lower Circum- the first large was at MIS 16 in the record extending over the MPT and obtained in- polar Deep Water (CDW), which forms from dense Atlantic but at MIS 22 in the Pacific, as indicated dependently of oxygen exists, and this waters sinking around , particularly by ODP 1123. It is likely that the difference of is at orbital-scale resolution from the North At- cold waters from the Weddell and Ross seas and individual sites from each other and from LR04 lantic using Mg/Ca ratios from benthic forami- Adelie Coast. The modern hydrography of the reflects changes in hydrography, smoothed in nifera (6). However, it has been recognized site shows a slight contribution from North At- the stack. 18 recently that Mg/Ca ratios of epifaunal benthic lantic Deep Water (NADW) expressed by a small The benthic d OC record has a temperature foraminifera are affected by saturation salinity anomaly marker (17). More than half of component and a seawater d18O component, and Downloaded from effects (7, 8), requiring a “regional calibration,” the flux of cold bottom water entering the major the latter may reflect a combination of a global and, consequently, the reliability of this approach basins of the world ocean does so through the ice volume signal and a local hydrographic signal has been both challenged (9) and defended (10). Southwest Pacific Ocean, as the Pacific DWBC relating, for example, to changes in proportions The increasing interest in applying inverse mod- (17, 18). CDW is drawn to the surface by up- of colder low d18O AABW versus warmer high els to deconvolve temperature and ice volume in welling with descending return flows south of the d18O NADW. If our signal may reflect “local” climate records [e.g., (11–14)] adds to the interest upwelling as Antarctic Bottom Water (AABW) hydrographic changes, those changes are related in separating these signals from the oceanic and north of it as Antarctic Intermediate Water. to Antarctica because deep-water masses at site sediment record using geochemical methods. The connection of this site to the regions where 1123 originate from waters sinking around Ant- Previously, we carried out a calibration study deep Antarctic water masses form enables com- arctica. Detailed records in addition to those in using the shallow-infaunal benthic foraminifera parison with Southern Ocean and Antarctic atmo- Fig. 1 at key hydrographic locations would be Uvigerina spp., showing that it is little affected by spheric records and is important for determining required to fully assess variability in the global bottom-water carbonate saturation, and generated the role of the deep ocean in climate evolution. seawater d18O signal. Our assessment is that the 18 Benthic d OC, deep-sea temperature, and hydrographic component at the site of 1123 is 18 Godwin Laboratory for Palaeoclimate Research, Department Antarctic temperature. We determined d OC and small: (i) benthic isotopic depth profiles at the of Earth Sciences, University of Cambridge, Downing Street, Mg/Ca on 1650 and 1485 samples of Uvigerina site of ODP 123 are related to the structure of Cambridge CB2 3EQ, UK. spp. (supplementary text, table S1, and Fig. 1). Data water masses at present and inferred for the past, *To whom correspondence should be addressed. E-mail: are also archived at doi:10.1594/PANGAEA.786205 with no apparent changes in the depths of water- [email protected] d18 † The O data were correlated to an orbitally tuned mass boundaries between glacial and intergla- Present address: Consiglio Nazionale delle Ricerche, Istituto 18 per la Dinamica dei Processi Ambientali (CNR-IDPA), Calle Larga benthic d Ostack,LR04(19), to provide an ini- cial states (17); (ii) estimates of oxygen isotopic 18 18 Santa Marta 2137, Venice I-30123, Italy. tial time scale. The d Orecordofbenthicforam- composition of seawater (d OW) from corals and

704 10 AUGUST 2012 VOL 337 SCIENCE www.sciencemag.org RESEARCH ARTICLE pore water modeling discussed below show global arctica from ice cores (16) (Fig. 1B). For this Glacial-interglacial model comparisons (23) values. Ideally, what is needed is a global array comparison, we converted Mg/Ca temperature suggest that Antarctic temperature changes rep- of seawater d18O records that could be stacked data to units of standard deviation and transposed resent global-scale temperature changes with a po- to produce a mean ocean d18Osignal. themtotheEDC3(EpicaDomeCicecore)time lar amplification factor of ~2: glacial-interglacial There is also considerable millennial and sub- scale (20). The pattern seen in the ice-core dD temperature change (Antarctic/Global) = 2.07. millennial scale variability in the records, more ( isotope-based temperature) record is The comparison of Antarctic and our deep-sea 18 in Mg/Ca than d OC (supplementary materials matched by the Mg/Ca–based deep-sea tempera- temperature data shows the same amplification andFig.1,AandB).Toaidincomparison,ox- ture: (i) glacial temperatures are similar throughout (Antarctic/deep sea) = 2.1 T 0.08 (fig. S5). There- ygen isotope and Mg/Ca data have been smoothed this deep-sea record, (ii) interglacial temperatures fore, it seems that the temperature of the bottom in the time domain using a Gaussian filter hav- are cooler prior to 450 ka, and (iii) interglacials water reflects the air temperature at southern high ing a total width of 5 ky. The range in Mg/Ca of 5.5, 7.5, 9.3, and 11.3 are exceptionally warmer. latitudes and thus Southern Ocean surface tem- Uvigerina spp. is narrow (~0.8 to 1.3 mmol/mol) The distinction between interglacial bottom- peratures, aided by upwelling and surface mix- but shows clearly a different overall pattern than water temperature before and after ~450 ka seen ing of CDW. d18O (Fig. 1B). in our record and in the ice-core atmospheric Ice volume and sea level. Given Mg/Ca, re- Mg/Ca was converted to temperature using record is also similar to a sub-Antarctic Atlantic arrangement of the paleotemperature equation 18 18 the sensitivity previously determined (15)of0.1 sector record of sea surface temperature using allows changes in the d OW (Dd OW)tobe mmol/mol per °C. This is based on 56 core top an alkenone (21) and of the oxygen iso- calculated [(15) and supplementary text). We samples that show a statistically significant cor- topes of that site (22). It is also recognized in made two tests of the methodology. First, we com- relation between Mg/Ca and bottom-water temper- model-based deconvolution of global deep-sea pared our results over the past 250 ky with esti- ature (fig. S4). Results show a temperature range temperature from the benthic d18O stack (11). mates of global sea level made from accurately from about 3°C in the warmest interglacial to a However, it is not seen in the North Atlantic ben- dated corals. The coral method is independent 18 reasonably constant value of about –1.5 to –2°C thic Mg/Ca record referred to earlier (6). It is un- of orbital-based d OC, whereas other sea-level for the glacials (Fig. 1B), which we discuss later, clear whether this reflects the linkage of that methods rely on comparison with benthic d18O. with an estimated error of about T 1°C (supporting site to hydrographic conditions associated with Considering that propagation of estimated tem- 18 materials). changing proportions of NADW and AABW or perature and d OC uncertainties results in an 18 There is a striking similarity between our is somehow a mixed signal related to a contri- error in d OW of T0.2 ‰, our results (fig. S6) deep-water temperature record and the 800-ky bution from carbonate as discussed earlier show excellent agreement with compilations of record of atmospheric temperature above Ant- (9, 10). coral data (24, 25). Second, it can be seen that on September 24, 2012

Fig. 1. Records on the 2.5 2.5 A LR04 of (A) 9 18 1 5 7 11 13 15 17 19 21 25 31 47 The LR04 benthic d O 29 35 37 43 49 19 stack ( )(red);ODP 45 18 3.5 27 39 41 3.5 1123 benthic d O (origi- 33 naldatainblueopensym- 23 bol and smoothed record (‰) 3 with a Gaussian filter hav- www.sciencemag.org calcite 4.5 4.5 ing a span of 5 ky as a O line in blue); (B)Record 18 on the EDC3 chronology ODP 1123 1123 Gint LR04 (20) of ODP 1123 ben- 5.5 5.5 B thic Mg/Ca temperature 0 200 400 600 800 1000 1200 1400 (original data in green and 5.77 3.01 smoothed record with a 2.32 I 2.07 5-ky Gaussian filter red) Downloaded from and Antarctic temperature -1.12 i 1.12 (16) (blue). Data were con- -4.58 0.017 verted to units of stan- 3 dard deviation. Ordinate -8.02 -0.77 axes show mean temper- D Temperature (°C) G atures (in red and blue) -11.5 -1.75 Mg/Ca Temperature (°C) and also values of mean -14.9 D Temperature T3s°C. MIS is labeled for ODP 1123 ODP Gint orientationinFig.1Aand in shaded panels in Fig. C -0.5 1B. (C)Icevolume(d18O I of seawater) derived from 0

ODP 1123 Mg/Ca and (‰) 18 d OC (blue) compared 0.5 with the simulation of seawater et al 11 O g Bintanja .( )(red) 18 andrecordfromtheRed 1.0 Sea (28)(green). G ODP 1123 Red Sea simulation

0 200 400 600 800 1000 1200 1400 Age (ka)

www.sciencemag.org SCIENCE VOL 337 10 AUGUST 2012 705 RESEARCH ARTICLE

18 18 the change in d OW since the Last Glacial Max- The Dd OW component of the record is 0.77 T by 700 ka, with low-frequency (i.e., 100 ky) pow- imum is ~1.2‰ for 120 m of sea-level change, 0.16 ‰ for MIS 24 to 50, compared with 1.09 T er emerging at about 1250 ka followed by a lull identical to the value of 1.1 T 0.1‰ derived from 0.17‰ over MIS 2 to 22 (fig. S8), equivalent to centered on 1000 ka, with reemergence of pow- porewater modeling at this Chatham Rise site a sea-level fall from ~60 to 90 m for the ob- er beginning at ~900 ka. Wavelet analysis of our 18 18 and similar to global estimates (26). This exercise liquity cycles to ~90 to 125 m for the eccentricity d OC, d OW, and Mg/Ca records using Sowas 18 goes some way toward addressing the issue dis- cycles. However, Dd OT shows no such con- software (29, 30) shows somewhat similar but 18 cussed earlier on the influence of local hydrog- trast and averages 0.80 T 0.14‰ over all gla- more complex patterns (Fig. 3). In the d OW 18 raphy on the site 1123 d OC record. cials sampled, equivalent to cooling of about record, the highest spectral power in the 100-ky One further factor in estimating sea level by 3°C to –2°C (Fig. 2C). band develops from 650 ka. Temperature gains this approach is the simplifying assumption of a This study emphasizes the point that benthic power in this band over 400 to 100 ka, whereas 18 18 18 constant d OW through a glacial cycle, whereas d O is not a direct proxy of ice volume or sea d OC is a “fusion” of the two with lower yet sig- 18 18 18 d Oofice(d OI), and d OW, will increase as level (1–3). Because the LR04 stack (19)isoften nificant power from 900 ka and intensified power the ocean is progressively depleted of 16O(27). taken as a proxy for ice volume, it is interesting from ~650 ka. 18 18 Assumption of d OW based on a constant d O that the temperature component of this record may The record of glacial deep-ocean temperatures 18 of ice has little impact on derivation of sea level approach as much as 50% of the range in d OC. over this period is similar to that which charac- 18 estimated from d OW records during the start of Timing of MPT events. With this assessment terizes glacials at this site throughout the Pleisto- a glacial cycle but has a larger impact at full of the relative contributions of temperature and cene. As shown above, rapid deep-ocean cooling 18 18 glaciation (fig. S7). Here, we have used d OI = ice volume to d OC, it is possible to address within each glacial-interglacial cycle preceded –40‰. On this basis, we have calculated the what was the sequence of events leading to the slower buildup of ice sheets, but there is no evi- 18 18 d OW and, assuming that this reflects global ice MPT.Areviewofd OC records (5) has shown dence of progressive cooling throughout the record; volume, derived a sea-level reconstruction since that the MPT began 1250 ka and was complete glacial to interglacial temperature change (DdT) 1550 ka and compared it with a recent sim- ulation and record of sea-level change (11, 28) (Fig. 1C). 2 2 18 First, we estimate the contributions of d OW 18 A and temperature to d OC, a subject of long- standing debate (1–4). Temperature has been 1 1

18 on September 24, 2012 scaled as d OT, and all data are plotted as “cold upwards” to show the evolution of temperature and ice volume from interglacial to glacial con- 0 0 ditions (Fig. 2, A and B). We see different behav- w c ior between 100,000-year (Fig. 2, A and B) and 2 2 41,000-year (Fig. 2, C and D) cycles, periodic- ities consistent with the tempo of variations in orbital eccentricity and obliquity. This exercise 18 1 1 shows that the increase in d OW (ice volume) dur-

ing eccentricity cycles proceeds slowly through- www.sciencemag.org out some glacial periods to a maximum value that 18 0 corresponds to a ~70% contribution to d OC, but 0 there are also periods of rapid change. In con- T c d18 trast, OT cools early in each glacial cycle by 0 200 400 600 800 1000 about 0.8‰. The contrast between the rapid tem- 2 2 perature decrease completed one-third of the way B 18 through each d OC cycle and the more grad- Downloaded from ual lagged increase in ice volume has two 1 1 implications. One is that, as suggested by the simulation of Bintanja et al.(11), rapid deep- ocean cooling preceded slower buildup of ice sheets. The other is that the near constancy of 0 0 18 w c glacial d OT (red dashed line in Fig. 2, A and B), 18 2 2 irrespective of the strength of individual d OC maxima, suggests that glacial cooling is limited by water’s freezing temperature such that glacial temperatures have remained nearly constant 1 1 since at least 1500 ka. This analysis also shows that the saw-tooth character of the oxygen iso- d18 tope cycles is a feature of the OW component 0 — 0 of the record slow ice-sheet buildup and rapid T c decay—whereas the temperature component re- sembles a square wave function resulting from 1000 1100 1200 1300 1400 1500 the limitation of the temperature fall during the Age (ka) early stages of ice-sheet growth. Fig. 2. Comparison of contributions of changes in ice volume, DdW (blue) and temperature, DT(red)to The records over obliquity cycles before about benthic d18O, DdC(green):(A) for 0 to 1000 ka and (B) for 1000 to 1500 ka. Data are plotted as “cold 1 million years ago show a much weaker role upwards.” Temperature converted to dT ‰ units using 0.25 ‰ per °C/0.1 mmol/molper°C.Histogram for ice volume because of smaller ice sheets. ofdatashowninfig.S8.

706 10 AUGUST 2012 VOL 337 SCIENCE www.sciencemag.org RESEARCH ARTICLE

18 is constant at about 0.8‰ over every glacial cy- mate system varied between three end-member and d OW lags Mg/Ca by ~21 ky with, as ex- 18 18 cle except for the warm interglacials. Tempera- climate states: lukewarm interglacials (i), mild gla- pected, d OC intermediate between d OW and ture decline cannot therefore be the unique cause cials (g), and full glacials (G) [after Paillard (31)]. temperature; see supplementary text and fig. S9 of increased ice volume across the MPT. How- It is not until 400 ka during the mid-Brunhes for discussion of the phase relationship between ever, a characteristic of the eccentricity cycles, event that full interglacial conditions (I) emerged, orbital eccentricity and Mg/Ca (temperature).The less so for obliquity, is the persistence of near- most vividly expressed in the deep-sea temper- 41-ky bandpass filtered data show, as in (33), in- freezing temperatures over most of the periods ature record. phase relationships between all measured param- of ice growth; the longer slow growth of ice Although suggestions that the MPT was a eters, whereas the 100-ky bandpass filtered data sheets occurred when cold temperatures per- gradual transition are shown in a number of ben- shows considerable variability in phase (fig. S9). sisted for longer. thic d18O records, our evidence for a more abrupt Over the period from ~1100 to 750 ka, where 18 In contrast to temperature, the change in ice transition comes from seawater d OW (Fig. 1C). 100-ky power is weak, the lag increases to the 18 18 volume is late within the MPT period, and its We attribute this difference to hydrographic con- extent that d OC and d OW are out of phase 18 timing coincides with the so-called 900-ka event tributions to d OC. with dT at 900 ka, after which there is a transition at the onset of MIS 22 (Figs. 1C and 4A). Be- What conditions may have led to the abrupt to strong 100-ky cycles from ~700 ka. Thus, the fore this event, the record oscillates between mild increase in continental ice volume during MIS “lock in” to a late Pleistocene 100-ky phase rela- glacials (g) and lukewarm interglacials (i), sug- 22? Examination of the sea-level (ice volume) tionship occurs at around 750 ka, with some var- gesting that the climate system varied between record across the MPT (Fig. 4A) shows that the iation before then. two quasistable climate states (31, 32). After MIS critical step in ice-volume variation was asso- The notion that the Antarctic ice sheet was 25, ice volume increased sufficiently to weaken ciated with the suppression of melting in MIS inactive throughout the Pleistocene is not sup- interglacial MIS 23 and initiated the onset of 23, followed by renewed ice growth in MIS 22 ported by our work or by records emanating full glacial conditions (G) beginning with MIS to yield a very large ice sheet with 120 m of sea- from the margins of Antarctica and some mod- 22 (~900 ka). MIS 22 marks an abrupt increase level lowering (Fig. 4A). This period is associated els (34–37). Raymo et al. have proposed a more 18 in glacial d OW values, indicating a transition with an anomalously low Southern Hemisphere active East Antarctic ice sheet before 1 Ma with to maximum ice volume that has characterized summer insolation across the minor melting MIS its expansion and development of the present less- most full glacial (G) stages since ~900 ka. Our 23. We suggest that this configuration suppressed active regime at the MPT (34). Their proposal— results suggest that the MPT represents an abrupt melting and allowed larger ice-sheet growth that that at the MPT, “marine-based ice-sheet margins reorganization of the climate system as opposed led to the first prolonged 100-ky glacial cycle replaced terrestrial ice margins around East Ant- to a long-term trend toward increased ice volume comprising MIS 22 to 24. Because the low sum- arctica, resulting in a shift to in-phase behavior on September 24, 2012 and colder temperatures as previously thought mer insolation occurred in the Southern Hemi- of northern and southern ice sheets” (34)—is (5). The duration of the glacial cycle lengthens sphere, we suggest that the cause of ice volume borne out by drill cores on the edge of the Ross with MIS 22 to 24, but the strong increase in the increase at 900 ka lay in Antarctica. Sea (35, 37). Pollard and De Conto (36) model 100-ky power is delayed until ~650 ka (MIS Cross-spectral analysis of our records shows an increase in ice vol- 18 18 7 3 16), which is where power in d OW reaches its that at eccentricity periods, d OC lags Mg/Ca ume of ~10 km (~17 m of sea level) across the maximum (Fig. 3). From 900 to 400 ka, the cli- (temperature) by ~12 ky, as found earlier (15, 33), MIS 25 to 22 transition. www.sciencemag.org Downloaded from

Fig. 3. Wavelet plots using Sowas software (29, 30). These plots make 18 it clear that the 100-ky cyclicity is most powerful in d OW over 700 to 300 ka, and in the temperature 400 to 100 ka, with the calcite d18Oa “fusion” of the two.

www.sciencemag.org SCIENCE VOL 337 10 AUGUST 2012 707 RESEARCH ARTICLE

Carbon chemistry and ocean circulation. The -50 -0.5 rapid increase in ice volume at MIS 22 is asso- 20 22 24 26 ciated with a large perturbation to ocean carbon A chemistry (38–40), as reflected in the most depleted benthic d13C values in the Pacific and 0 Atlantic over the past 5 million years (36–38) -1.0 This is also seen in the ODP 1123 record (Fig. 4A and Fig. 5). Because it is represented in many 50 records, some authors have argued that this de- C (‰) crease in oceanic d13C is a global event. It has 13 also been argued that benthic d13C values begin -1.5 to decline much earlier than at MIS 22, at about 100 1.5 Ma (41–43), reaching a minimum at 900 ka and increasing to about 500 ka. However, our 18 13 results suggest that the carbon isotope excursion Ow C sea level equivalent (m rel to modern) was associated with an abrupt reorganization 420 -2.0 of the climate system rather than a long-term 20 22 24 26 trend, characterized by the most depleted d13C B 400 values at the MPT within a record of otherwise 4

rather similar glacial values (Fig. 5). Therefore ) we will focus first on the extreme depletion in -2 380 2 d13C at the MPT and next consider its glacial- interglacial cycles. 13 360 We can dismiss the hypothesis that the d C 0 record for this site, which is based on an in- faunal foraminiferal species, is depleted relative 340 to bottom water and reflects productivity-related -2

changes in pore-water composition. This is be- on September 24, 2012 320 insolation 75° S

cause paired Cibicidoides-Uvigerina data (fig. Summer Insolation (W m S10) show a constant offset of ~0.9‰ between insolation 65° N -4 the epifaunal and infaunal species. The Uvigerina 300 spp. d13C record cannot therefore reflect changes 750 800 850 900 950 1000 Age (ka) in metabolic CO2 at the depth of this ODP site. 18 Given this, we can explore interpretations of Fig. 4. (A) Record over 750 to 1000 ka showing age of the MPT at MIS 22 inferred from d OW (sea the benthic d13C record, first exploiting the ice- level) (blue). Also shown is benthic d13C (red). MIS with boundaries taken from LR04 (19) are labeled volume record that we have derived. and shown in shaded panels for orientation. (B) Normalized orbital variation in energy received by Earth

Before 900 ka, sea-level lowstand during at eccentricity and obliquity (blue and green dashed lines) and Northern and Southern Hemisphere (red www.sciencemag.org glaciations was ~70 m below the present. MIS and blue lines) insolation frequencies for the period 1 Ma to 800 ka; note the antiphase relation 22 represents the first time in millions of years between obliquity and a wide Southern Hemisphere insolation peak at 920 to 900 ka. that sea level dropped to 120 m, exposing shelf break and upper-slope deposits (38). Marine or- ganic carbon (with d13Cof~–18‰) exposed on shelves and transferred to seawater would lower benthic d13C. The depletion in d13Ccompared with “normal” glacials (for example, MIS 20, Downloaded from 24, and 26) is about 0.2‰, equivalent to about 300 Pg of carbon. A test of this hypothesis is to examine the relationship between d13Cand 18 13 d OW. Ice volume and d C look very tightly coupled in Fig. 4A and Fig. 5. This strong link 18 13 between d OW and d C suggests that sea level and/or circulation is involved in glacial-interglacial 13 d C changes. This is a more likely cause than Fig. 5. Record of benthic d13Candd18Oofseawater.d13C at MIS 22 is most depleted of benthic d13C change in the terrestrial carbon reservoir result- values in the Pacific and Atlantic over the past 5 million years (42). ing from global aridity, as originally conceived by Shackleton (44). The 41-ky bandpass filtered data show (as above for other climate variables) graphic properties of site ODP 1123 (mostly in hydrographic tracer largely unaffected by changes 18 13 that d OW and d C are closely in phase (fig. addition to fluctuations in the marine C reservoir) to organic carbon systematics. Northern source S9) but the responses in the 100-ky band after affected by changes in global thermohaline cir- waters have a more negative eNd than southern 13 18 13 900 ka show a small of d Ctod OW. Taken culation with or without concomitant change source waters (46). If the d C changes reflect at face value, this does not support the shelf in nutrients (44, 45). As discussed above, differ- changes in proportions of water masses, we would 18 hypothesis over 100-ky time scales. ences between d OC records of site 1123 and anticipate that they should be associated with The second interpretation, argued indirect- theLR04stackprovidesomeevidenceofasec- concomitant changes in eNd. We examined three ly from several avenues [summarized in (45)], ondary hydrographic control. We have tested periods: over the last deglaciation, the warm is that benthic d13C is controlled by the hydro- this hypothesis using Nd isotopes that act as a interglacial 100-ky cycles MIS 12 to 10, and

708 10 AUGUST 2012 VOL 337 SCIENCE www.sciencemag.org RESEARCH ARTICLE

Fig. 6. Records of eNd (red) and d13C (blue) for three periods within core ODP 1123: (A) 0 to 30 ka, (B) 300 to 500 ka, (C) 1150 to 1250 ka. All show that in warm periods, eNd is more negative, suggesting greater NADW influence.

the obliquity cycles MIS 38 to 35 (Fig. 6). In References and Notes 38. M. E. Raymo, D. W. Oppo, W. Curry, Paleoceanography each case, the records of eNd and d13Caresim- 1. J. Chappell, N. J. Shackleton, Nature 324, 137 12, 546 (1997). e (1986). 39. D. A. Hodell, K. A. Venz, C. D. Charles, U. S. Ninneman, ilar in form, with a small range in Nd from about 2. N. J. Shackleton, Science 289, 1897 (2000). Geochem. Geophys. Geosyst. 4, 1004 (2003). –7 at interglacial stages to –4 at glacial stages. 3. N. J. Shackleton, Nature 215, 15 (1967). 40. L. E. Lisiecki, Geophys. Res. Lett. 37, L21708 (2010). This implies that the 100-ky and 40-ky cycles 4. N. J. Shackleton, N. D. Opdyke, Quat. Res. 3,39 41. P. Wang, J. Tian, L. J. Lourens, Earth Planet. Sci. Lett. of d13C are associated with changes in deep- (1973). 290, 319 (2010). ocean circulation, perhaps a decrease in glacial 5. P. U. Clark et al., Quat. Sci. Rev. 25, 3150 (2006). 42. B. A. A. Hoogakker, E. J. Rohling, M. R. Palmer, T. Tyrrell, 6. S. Sosdian, Y. Rosenthal, Science 325, 306 (2009). R. G. Rothwell, Earth Planet. Sci. Lett. 248, 15 (2006). export of NADW relative to CDW coupled with 7. P. A. Martin et al., Earth Planet. Sci. Lett. 198, 193 43. K. A. Venz, D. A. Hodell, Palaeogeogr. Palaeoclimatol. changes in Southern Ocean deep-water ven- (2002). Palaeoecol. 182, 197 (2002). tilation (17, 33, 39). However, given the location 8. H. Elderfield, J. Yu, P. Anand, T. Kiefer, B. Nyland, 44. N. J. Shackleton, In Fate of Fossil Fuel CO2 in the Oceans, of ODP 1123, there is little scope for large var- Earth Planet. Sci. Lett. 250, 633 (2006). N. R. Anderson, A. Malahoff, Eds. (Plenum, New York, on September 24, 2012 9. J. Yu, W. S. Broecker, Science 328, 1480 (2010). 1977) pp. 401-427. iations in end members. 10. S. Sosdian, Y. Rosenthal, Science 328, 1480 (2010). 45. I. R. Hall, I. N. McCave, N. J. Shackleton, G. P. Weedon, Plots of the relationship between eNd and 11. R. Bintanja, R. S. W. van de Wal, J. Oerlemans, Nature S. E. Harris, Nature 412, 809 (2001). d13C data reveal that large changes in the oceanic 437, 125 (2005). 46. A. M. Piotrowski, S. L. Goldstein, S. R. Hemming, carbon reservoir occur in addition to those in 12. M. Siddall et al., Quat. Sci. Rev. 25, 3185 (2006). R. G. Fairbanks, Science 307, 1933 (2005). d13 e 13. E. M. Gasson et al., Rev. Geophys. 50, RG1005 47. B. Hönisch, N. G. Hemming, D. Archer, M. Siddall, C that are associated with changes in Nd (2012). J. F. McManus, Science 324, 1551 (2009). (fig. S11). Of the glacial-interglacial offset in 14. P. Kohler, H. Fischer, Clim. Past 2, 57 (2006). d13C of about 1‰ (Fig. 5), about half is asso- 15. H. Elderfield et al., Quat. Sci. Rev. 29, 160 (2010). Acknowledgments: This work was funded in part by The ciated with circulation and half with carbon res- 16. J. Jouzel et al., Science 317, 793 (2007). Natural Environment Research Council (UK) and The European 13 17. I. N. McCave, L. Carter, I. R. Hall, Quat. Sci. Rev. 27, Research Council (ERC grant 2010-ADG-267931) and is a d ‰ www.sciencemag.org ervoir. A whole-ocean decrease in Cof0.4 contribution to the European Project on Ocean Acidification 13 1886 (2008). would represent about half of the d C change 18. T. Whitworth III et al., Prog. Oceanogr. 43, 1 (1999). (EPOCA), which received funding from the European Community’s over glacial-interglacial cycles and would require 19. L. E. Lisiecki, M. E. Raymo, Paleoceanography 20, Seventh Framework Programme (FP7/2007-2013) under grant transfer of about 500 Pg of terrestrial carbon to PA1003 (2005). agreement no. 211384 and from the EPICA-MIS (Enhanced the oceans, together with an increase in atmo- 20. F. Parrenin et al., Clim. Past 3, 485 (2007). Palaeoreconstruction and Integrated Climate Analysis Through 21. A. Martínez-Garcia et al., Paleoceanography 24, PA1207 Marine and Ice Core Studies) program. Funding was also spheric CO2 of 45 parts per million by volume (2009). provided by a Royal Society Leverhulme Trust Senior Research before compensation. 22. D. A. Hodell, K. A. Venz, C. D. Charles, F. J. Sierro, Fellowship and a CORC-ARCHES (Consortium on the Ocean’s Importance. Most hypotheses account for in Earth’s Climate and Orbital Eccentricity: The Marine Role in Climate/Abrupt Studies) Visiting the origin of the MPT as a response to long-term Isotope Stage 11 Question, A. W. Droxler, R. Z. Poore, Scientist post at Lamont Doherty Earth Observatory of Downloaded from L. H. Burckle, Eds., AGU Geophys Monogr. 137, Columbia University, for which H.E. is grateful to W. Broecker. ocean cooling, perhaps because of lowering CO2 113–129 (2003). We thank J. Firth and colleagues at the IODP core repository, (5). Data of CO2, directly from ice cores and in- 23. V. Masson-Delmotte et al., Quat. Sci. Rev. 29, 113 College Station, for help in sampling ODP 1123. L. Booth and directly from the d11B proxy (47) (fig. S12), are (2010). C. Daunt, and additionally J. Clegg, J. Day, M. Hall, J. Nicholson, as yet too sparse to determine the respective 24. W. G. Thompson, S. L. Goldstein, Quat. Sci. Rev. 25, N. Roberts, J. Rolfe, J. Smith, S. Smith, and S. Sounef-Ureta 3207 (2006). provided crucial laboratory support in Cambridge. We thank roles of temperature and of the carbon system. 25. A. L. Thomas et al., Science 324, 1186 (2009). L. Booth and I. Weidle for their work on Uvigerina morphologies. We have defined the timing of initiation of the 26. J. F. Adkins, K. McIntyre, D. P. Schrag, Science 298, G. Bayon, E. Ponzeve, and N. Freslon provided help during MPT as an abrupt event centered on MIS 24 to 1769 (2002). eNd analysis at IFREMER (Institut français de recherche pour 22 (the 900-ka event). The descent into longer 27. A. Sima, A. Paul, M. Schulz, J. Oerlemans, Geophys. Res. l’exploitation de la mer) Brest, France. D. Maraun helped Lett. 33, L15706 (2006). with use of Sowas software, and F. Joos, H. Pälike, R. Rickaby, glacials, the MPT, appears to have begun in MIS 28. M. Siddall et al., Nature 423, 853 (2003). T. Stocker, and G. Weedon helped with discussions. The 23 by suppression of substantial melting of the 29. D. Maraun, J. Kurths, M. Holschneider, Phys. Rev. E Stat. reviews of the manuscript were very valuable, especially in d18 ice formed in MIS 24, perhaps due to season- Nonlin. Soft Matter Phys. 75, 016707 (2007). encouraging us to focus on interpretation of the OC record. ality effects originating in the Southern Hemi- 30. D. Maraun, J. Kurths, Nonlin. Processes Geophys. 11, 505 (2004). Supplementary Materials sphere, before the renewed ice growth in MIS 31. D. Paillard, Nature 391, 378 (1998). 22 yielded a very large ice sheet. We see no evi- www.sciencemag.org/cgi/content/full/337/6095/704/DC1 32. P. D. Ditlevsen, Paleoceanography 24, PA3204 (2009). Materials and Methods dence of a pattern of cooling since 1500 ka, but 33. K. T. Lawrence, S. Sosdian, H. E. White, Y. Rosenthal, Supplementary Text near-freezing temperatures occur at every gla- Earth Planet. Sci. Lett. 300, 329 (2010). Figs. S1 to S15 cial maximum. The records over obliquity cycles 34. M. E. Raymo, L. E. Lisiecki, K. H. Nisancioglu, Science Tables S1 to S3 before about 1 million years ago show a much 313, 492 (2006). References (48–64) 35. T. Naish et al., Nature 458, 322 (2009). weaker role for ice volume because of smaller 36. D. Pollard, R. M. DeConto, Nature 458, 329 (2009). 29 February 2012; accepted 25 June 2012 ice sheets. 37. R. McKay et al., Quat. Sci. Rev. 34, 93 (2012). 10.1126/science.1221294

www.sciencemag.org SCIENCE VOL 337 10 AUGUST 2012 709