The apparent ubiquity of abrupt millennial-scale climate

Transcription

1 Radiocarbon evidence for alternating northern and southern sources of ventilation of the deep Atlantic carbon pool during the last deglaciation Luke C. Skinner a,1, Claire Waelbroeck b, Adam E. Scrivner a, and Stewart J. Fallon c a Godwin Laboratory for Palaeoclimate Research, Department of Earth Sciences, University of Cambridge, Cambridge CB2 3EQ, United Kingdom; b Laboratoire des Sciences du Climat et de l Environment/Institut Pierre Simon Laplace, Laboratoire Centre National de la Recherche Scientifique Commissariat à l Energie Atomique Université de Versailles Saint-Quentin, F Gif-sur-Yvette Cedex, France; and c Research School of Earth Sciences, Australian National University, Canberra, ACT 0200, Australia Edited by Mark H. Thiemens, University of California, San Diego, La Jolla, CA, and approved March 4, 2014 (received for review January 14, 2014) Recent theories for glacial interglacial climate transitions call on millennial climate perturbations that purged the deep sea of sequestered carbon dioxide via a bipolar ventilation seesaw. However, the viability of this hypothesis has been contested, and robust evidence in its support is lacking. Here we present a record of North Atlantic deep-water radiocarbon ventilation, which we compare with similar data from the Southern Ocean. A striking coherence in ventilation changes is found, with extremely high ventilation ages prevailing across the deep Atlantic during the last glacial period. The data also reveal two reversals in the ventilation gradient between the deep North Atlantic and Southern Ocean during Heinrich Stadial 1 and the Younger Dryas. These coincided with periods of sustained atmospheric CO 2 rise and appear to have been driven by enhanced ocean atmosphere exchange, primarily in the Southern Ocean. These results confirm the operation of a bipolar ventilation seesaw during deglaciation and underline the contribution of abrupt regional climate anomalies to longer-term global climate transitions. ocean circulation carbon cycle abrupt change The apparent ubiquity of abrupt millennial-scale climate anomalies on Pleistocene glacial interglacial transitions (1) has prompted some to suggest these events may represent a necessary component of the deglacial process, primarily through their effect on atmospheric CO 2 (2 4). According to this hypothesis (hereafter referred to as the millennial purge hypothesis), incipient deglaciation triggers a transient collapse of the North Atlantic overturning circulation, as well as a warming around Antarctica via the thermal bipolar seesaw mechanism (5). This, in turn, is proposed to drive the release of previously sequestered marine CO 2 to the atmosphere via increased upwelling and/or air sea exchange in the Southern Ocean, which would have resulted from changes in the density stratification (6), sea ice cover (4) or wind forcing (2) around Antarctica. Despite the promise of this new theory for deglacial CO 2 release, and despite emerging data in its support (4, 7, 8), it rests on two key premises that have yet to be confirmed. The first premise is the existence before deglaciation of a deeply sequestered, and hence radiocarbon-depleted (9), marine carbon pool of significant extent. The existence of such a carbon pool continues to be questioned (e.g., ref. 10). The second premise is a necessary (or at least conditional) alternation between northern and southern Atlantic ventilation anomalies, broadly in line with the bipolar seesaw in deep ocean ventilation that was initially proposed by Broecker (11). Direct evidence for this is still lacking. At this time, the strongest support for the bipolar ventilation seesaw of ref. 11 comes from Southern Ocean opal productivity pulses that occur in time with the deglacial stadial events Heinrich Stadial 1 (HS1) and the Younger Dryas (YD) (2). However, even these data strictly remain mute on the anticorrelation or otherwise of southern versus northern (hemisphere) ventilation of the Atlantic interior. Indeed, most other available proxy reconstructions have been interpreted as showing that Southern Ocean ventilation (as inferred, often ambiguously, from radiocarbon, stable carbon isotopes, carbonate preservation, or water mass sourcing) was reduced during North Atlantic stadials when CO 2 was increasing (e.g., refs ), in apparent conflict with the millennial purge hypothesis. Results To assess possible links between North and South Atlantic ventilation, we generated a continuous record of deep-water radiocarbon ventilation from the Northeast Atlantic, which we compare with similar data from the Atlantic sector of the sub- Antarctic Southern Ocean (4). Deep-water radiocarbon ventilation records specifically constrain the extent of isotopic equilibration between the deep-ocean and atmospheric carbon pools and therefore bear directly on the role of the ocean circulation on ocean atmosphere carbon exchange. Here, radiocarbon measurements have been performed on paired samples of rigorously cleaned (15) monospecific planktonic and mixed benthic foraminifera from core MD K (37 48 N, W; 3,146 m). The site of MD K, on the Iberian Margin in the Northeast Atlantic, is currently bathed in northward recirculating Northeast Atlantic Deep Water, which includes 47% Lower Deep Water (derived from Antarctic Bottom Water) (Fig. S1). The chronology for core MD K is based on the alignment of local surface temperature trends [recorded in foraminiferal δ 18 O and Mg/Ca measurements (16)] to the uranium-series dated speleothem records from Hulu Cave (17 19) (Fig. S2). This alignment is Significance This study sheds light on the mechanisms of deglacial atmospheric CO 2 rise and, more specifically, on the hypothesized role of a bipolar seesaw in deep Atlantic ventilation. Comparing new high-resolution radiocarbon reconstructions from the Northeast Atlantic with existing data from the Southern Ocean, we show that a bipolar ventilation seesaw did indeed operate during the last deglaciation. Whereas today the deep Atlantic s carbon pool is flushed from the north by North Atlantic Deep Water export, it was flushed instead from the south during Heinrich Stadial 1 and the Younger Dryas, in time with sustained atmospheric CO 2 rise. Author contributions: L.C.S. designed research; L.C.S. performed research; A.E.S. performed sample preparations and analyses; S.J.F. performed accelerator mass spectrometry analyses; L.C.S. and C.W. analyzed data; and L.C.S., C.W., A.E.S., and S.J.F. wrote the paper. The authors declare no conflict of interest. This article is a PNAS Direct Submission. 1 To whom correspondence should be addressed. lcs32@cam.ac.uk. This article contains supporting information online at /pnas /-/DCSupplemental PNAS April 15, 2014 vol. 111 no. 15

2 A (Holocene) YD B-A HS1 (Pleniglacial) B C Fig. 1. (A) Surface reservoir ages in MD K (solid black stars with 1 sigma dating uncertainties and dashed b-spline, with shaded range of possible values implied by different viable calendar chronologies) compared with other Northeast Atlantic surface reservoir age estimates [open circles (21) and gray stars (22), with solid black 5-point smoothed b-spline]. (B) MD K benthic planktonic age offsets (B-P, gray open diamonds and line) and deep-water reservoir ages (i.e., benthic atmosphere age offsets, B-Atm) (solid black diamonds and line, with shaded max/min range). (C) Planktonic δ 18 O in core MD K (solid black line), shown on two distinct calendar chronologies that encompass the range of reservoir ages in A and B (shaded range) compared with the NGRIP δ 18 O record (fine gray line) (16, 35). All uncertainties are 1σ. Vertical lines show the timing of the YD, Bølling Allerød, and HS1 [onset based on dust content in Greenland ice (36)]. in near-perfect agreement with the North Greenland Ice Core Project (NGRIP) ice-core chronology (Fig. 1C), although the timing of events between the LGM and HS1 is not obviously constrained by the NGRIP event stratigraphy. For this reason, we consider a range of possible calendar age constraints that permit the construction of a best guess and a maximum/minimum bounding range of sediment age-depth models (20) for core MD K (see Supporting Information). These age models are constructed to be consistent with the various possible speleothem calendar age-constraints while also taking into account the down-core radiocarbon dating constraints. Surface- and deepwater reservoir ages are then derived from the offset between down-core foraminifer radiocarbon ages in MD K (on the range of possible calendar age-scales) versus contemporary atmospheric radiocarbon ages recorded in the Hulu Cave deposits (19). The resulting surface reservoir ages are in excellent agreement with previous estimates from the Northeast Atlantic (21, 22) (Fig. 1A). These results demonstrate a degree of coherence between regional ocean atmosphere radiocarbon disequilibration (in the subsurface habitat of planktonic foraminifera) and the general climatic trends of the North Atlantic region, which may result from a combination of changes in the ventilation of the thermocline, the thickness of the mixed layer (e.g., the presence of a seasonal/perennial halocline), the presence of a radiocarbon-depleted subsurface water mass (e.g., from the Nordic Seas), and/or local upwelling effects. The significant variability of surface reservoir ages in this context (and others like it) has clear implications for marine radiocarbon-based chronologies (21);however,italsoservesasareminderthattheapriori assumption of a constant surface reservoir age begs the question of air sea CO 2 exchange efficiency across the uppermost surface ocean under glacial climate conditions. This, in turn, has implications for our interpretation of the atmospheric radiocarbon (Δ 14 C atm ) record (9) and for deep-water radiocarbon reconstructions (e.g., refs. 4, 22). Whereas the benthic planktonic (B-P) ventilation ages in MD K already suggest a significant increase (by 1,000 y) in the age of deep water filling the Northeast Atlantic during the last glacial period, benthic atmosphere (B-Atm) ages indicate an increase, versus the atmosphere, that is up to 2.5 times larger. The radiocarbon reservoir age of the deep northeast Atlantic may thus have increased to between 2,250 and 3,400 y during the last glacial maximum. These results confirm the existence of a glacial marine carbon reservoir that was at least as radiocarbon-depleted as the oldest deep-water masses in the modern ocean, which extended from the deep North Atlantic (22) to the deep Southern Ocean (4, 7) and likely also contributed to a greater volume of the deep ocean downstream of these basins. Recent estimates suggest that 25 30% of the modern ocean interior >1,500 m in depth is sourced from the North Atlantic (roughly >40 N) and that essentially the remainder ( 56 61%) is sourced from the Southern Ocean (roughly >40 S) (23, 24). If the surface reservoir age changes recorded at MD K and MD (4) were broadly representative of their wider North Atlantic and Southern Ocean regions [which remains unproven but is not implausible, given the regional consistency of modern and paleo reservoir age estimates (4, 7, 21, 22)], a circulation geometry similar to today s would already require a significant change in the marine radiocarbon inventory during the late glacial period. EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES Skinner et al. PNAS April 15, 2014 vol. 111 no

3 A (Holocene) YD B-A HS1 (Pleniglacial) B C Fig. 2. Ventilation records from the North Atlantic (MD K) and Southern Ocean (MD ). (A) Surface reservoir ages, as for Fig. 1. (B) Deep-water reservoir ages (i.e., benthic-atmosphere age offsets; B-Atm). (C) Benthic-planktonic age offsets (B-P) compared with the NGRIP δ 18 O record (fine gray line) (35). Data from MD K are shown by black symbols and heavy lines; data from MD are shown by gray symbols and heavy lines. B-A, Bølling Allerød. Fig. 2 compares the surface- and deep-water radiocarbon ventilation history of the northeast Atlantic (this study, MD K; 3,146 m) and the sub-antarctic Atlantic (MD ; 3,777 m) (4). For consistency, the sub-antarctic records are shown here referenced to the same Hulu atmospheric radiocarbon values as used for MD K (see Supporting Information). The coherence between the surface and deep-water records in each hemisphere is striking, with similar patterns of variability and amplitudes of change exhibited at each location for all three measures of radiocarbon ventilation (surface reservoir ages, B-P age offsets, and deep-water reservoir ages; Fig. 2 A C). However, important differences between the North Atlantic and Southern Ocean records emerge during HS1, and the YD in particular, when the radiocarbon ventilation gradient between the two sites collapsed. This is apparent in both surface- and deep-water reservoir ages and is expressed as a reversal of the north south radiocarbon ventilation gradient on the best guess MD K calendar chronology. Benthic planktonic offsets (Fig. 2B) also exhibit a clear reversal in the apparent ventilation gradient, although without a return to the normal north south gradient seen in the B-Atm and surface reservoir age records during the Bølling Allerød interstadial. Discussion These results confirm the operation of a bipolar ventilation seesaw across the last deglaciation, whereby the radiocarbon ventilation of the deep Southern Ocean increased during HS1 and the YD to a level commensurate with, or even above, that observed concurrently in the deep North Atlantic. This is further supported by the antiphase variability in shallow subsurface reservoir ages observed in each hemisphere (Fig. 2A). Fig. 3 shows how the reversal of the ventilation gradient between the North and South Atlantic corresponded with periods of sharply rising atmospheric CO 2 (25), as well as periods of inferred Atlantic meridional overturning circulation collapse (26). New planktonic δ 13 C measurements from core MD in the Southern Ocean (Supporting Information and Fig. 3B) further demonstrate a close correspondence between increased nutrient supply to the surface ocean and pulses in opal accumulation from across the Southern Ocean (2), as well as pulses in the ventilation age of the deep Southern Ocean (4), a strong indication that enhanced upwelling was indeed the driver of the observed export productivity pulses. This inference is further supported by silicon and nitrogen isotope evidence for enhanced nutrient supply to the surface ocean during HS1 and the YD (27). Together with the radiocarbon data from the North Atlantic, these findings provide strong evidence of enhanced ventilation specifically of the ocean interior s carbon pool, and specifically via the Southern Ocean in time with rapid atmospheric CO 2 rise during HS1 and the YD. Indeed, the sequence of events illustrated in Fig. 3 is essentially as required by the millennial purge hypothesis that has been advanced to explain the rapid and pulsed rise of atmospheric CO 2 across the last deglaciation (2, 3). According to the revised dating of the Antarctic ice cores (28), our data and chronologies indicate an overlap between the first indication of deep ocean hydrographic/circulation change in the Southern Ocean (at 19.7 ± 0.6 ka) and the initiation of Antarctic temperature and CO 2 rise (at 18.6 ± 1.9 ka; full absolute uncertainty). Given the chronological uncertainties, the apparent lag of 1,000 y between the marine and atmospheric changes is not strictly significant, although superficially it might be seen as corroborating a proposed delay in the response of atmospheric CO 2 versus the onset of hydrographic change in the Southern Ocean (29). The suggested delay in the carbon cycle response would remain unexplained, and although our data cannot confirm or refute the speculation that it represents a threshold dependence of CO 2 rise on overturning circulation change (29), it is notable that a qualitatively similar threshold response does Skinner et al.

4 A B C D EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES Fig. 3. (A) North south offsets in deep-water reservoir ages (solid circles and heavy black line, with shaded range) compared with reconstructed North Atlantic overturning circulation strength derived from excess protactinium-231 to thorium-230 ratio ( 231 Pa/ 230 Th xs ) measurements [heavy gray line (26)]. (B) Southern Ocean upwelling, derived from planktonic δ 13 C measured in core MD (solid black circles and line), and opal flux measured in core TNO57-13PC4 (2) (solid gray line). (C) The rate of change of B-Atm ventilation ages in MD (D) Atmospheric CO 2 concentrations from the European Project for Ice Coring in Antarctica ice-core, placed on the AICC2012 age-scale (28). Vertical lines indicate the timing of rapid CO 2 rise, broadly coinciding with HS1 and the YD. appear in numerical model simulations (6). Determining the veracity of this lag and its implications for the triggering mechanisms of the bipolar ventilation seesaw and CO 2 release during HS1 will be an important future research goal. It is important to note that if our observations confirm the operation of a ventilation seesaw in the deep Atlantic, they only do so in relation to ventilation as defined as a means of introducing radiocarbon (i.e., water with a dissolved inorganic carbon pool that has equilibrated with the atmosphere) into the ocean interior. This definition of ventilation is of particular relevance to ocean atmosphere carbon exchange but is strictly not identical to definitions based on, for example, stable carbon isotope fractionation (14), carbonate preservation (12), or the direction/rate of mass transport (26) in the ocean interior. The suggestion of continued deep-water export from the North Atlantic to the Southern Ocean during late HS1, based on neodymium isotopes (13, 30), underlines this fact and may imply the existence of an aged (radiocarbon-depleted) water mass of North Atlantic origin. The results presented here confirm two necessary and hitherto contested aspects of the millennial purge hypothesis for deglacial CO 2 rise; namely, the existence of a deeply sequestered carbon pool in the glacial ocean and the operation of a bipolar ventilation seesaw in the Atlantic. However, although these results indicate that the millennial purge hypothesis for deglacial CO 2 rise is indeed viable, they do not yet prove that this mechanism was necessary for late Pleistocene deglaciations (1, 3), in which global field insolation anomalies and albedo feedbacks will have played leading roles. Nevertheless, our findings indicate that via their carbon cycle effects, these potentially stochastic millennial events might have played a critical role in shaping the character and exact timing of Pleistocene deglaciations, the predictability of which would therefore be extremely limited on the millennial time frame. Materials and Methods Mixed benthic foraminifera (excluding agglutinated and broken shells) and monospecific samples of the planktonic foraminifer Globigerina bulloides or Neogloboquadrina pachyderma were picked from 1-cm slices of core MD K, supplementing previous radiocarbon dates reported by Skinner and Shackleton (31). Samples were cleaned according to the Mg/Ca cleaning method of ref. 15 before drying and sealing in evacuated septum blood vials for hydrolysis in 0.5 ml dry phosphoric acid at 60 C. Carbon dioxide evolved from the samples was graphitized at the Research School of Earth Sciences (Australian National University), using a standard hydrogen/ironcatalyst protocol (32). Samples were graphitized in parallel with Iceland Spar Skinner et al. PNAS April 15, 2014 vol. 111 no

5 calcite backgrounds, as well as primary and secondary standards for normalization and quality control. Pressed graphite targets were analyzed by single-stage accelerator mass spectrometry at the Australian National University (33). An additional suite of (similarly cleaned) samples were graphitized and analyzed by accelerator mass spectrometry at the Natural Environment Research Council/Scottish Universities Environmental Research Centre (SUERC) radiocarbon facility. Radiocarbon ages are reported according to the standard protocol of ref. 34. ACKNOWLEDGMENTS. We are grateful to K. James at the Australian National University radiocarbon laboratory and S. Moreton at the SUERC/Natural Environment Research Council (NERC) radiocarbon laboratory for technical assistance. We are also grateful for the critical and constructive comments of two anonymous reviewers. This work was supported by the Royal Society United Kingdom through a University Research Fellowship (to L.C.S.), and by NERC radiocarbon Grant C.W. is supported financially by the Centre National de la Recherche Scientifique, Institut National des Sciences d l Univers. This is Laboratiore des Sciences du Climat et de l Environment Contribution Barker S, et al. (2011) 800,000 years of abrupt climate variability. Science 334(6054): Anderson RF, et al. (2009) Wind-driven upwelling in the Southern Ocean and the deglacial rise in atmospheric CO2. Science 323(5920): Denton GH, et al. (2010) The last glacial termination. Science 328(5986): Skinner LC, Fallon S, Waelbroeck C, Michel E, Barker S (2010) Ventilation of the deep Southern Ocean and deglacial CO2 rise. Science 328(5982): Schmittner A, Saenko OA, Weaver AJ (2003) Coupling of the hemispheres in observations and simulations of glacial climate change. Quat Sci Rev 22: Schmittner A, Galbraith ED (2008) Glacial greenhouse-gas fluctuations controlled by ocean circulation changes. Nature 456(7220): Burke A, Robinson LF (2012) The Southern Ocean s role in carbon exchange during the last deglaciation. Science 335(6068): Marchitto TM, Lehman SJ, Ortiz JD, Flückiger J, van Geen A (2007) Marine radiocarbon evidence for the mechanism of deglacial atmospheric CO2 rise. Science 316(5830): Broecker W, Barker S (2007) A 190 permil drop in atmospher s Delta-14C during the Mystery Interval (17.5 to 14.5 kyr). Earth Planet Sci Lett 256: Hain MP, Sigman DM, Haug GH (February 17, 2011) Shortcomings of the isolated abyssal reservoir model for deglacial radiocarbon changes in the mid-depth Indo- Pacific Ocean. Geophys Res Lett, /2010GL Broecker WS (1998) Palaeocean circulation during the last deglaciation: A bipolar seesaw? Paleoceanography 13: Barker S, Knorr G, Vautravers M, Diz P, Skinner LC (2010) Extreme deepening of the Atlantic overturning circulation during deglaciation. Nat Geosci 3: Piotrowski AM, Goldstein SL, Hemming SR, Fairbanks RG (2004) Intensification and variability of ocean thermohaline circulation through the last deglaciation. Earth Planet Sci Lett 225: Waelbroeck C, et al. (2011) The timing of deglacial circulation changes in the Atlantic. Paleoceanography 26:PA Barker S, Greaves M, Elderfield H (2003) A study of cleaning procedures used for foraminiferal Mg/Ca paleothermometry. Geochem Geophys Geosyst 4(9): Skinner LC, Elderfield H (2005) Constraining ecological and biological bias in planktonic foraminiferal Mg/Ca and d18occ: A multi-species approach to proxy calibration testing. Paleoceanography 20: Wang YJ, et al. (2001) A high-resolution absolute-dated late Pleistocene Monsoon record from Hulu Cave, China. Science 294(5550): Zhao K, Wang Y, Edwards L, Cheng H, Liu D (2010) High-resolution stalagmite d18o records of Asian monsoon changes in central and southern China spanning the MIS 3/2 transition. Earth Planet Sci Lett 298: Southon J, Noronha AL, Cheng H, Edwards RL, Wang Y (2012) A high-resolution record of atmospheric C-14 based on Hulu Cave speleothem H82. Quat Sci Rev 33: Parnell AC, Haslett J, Allen JRM, Buck CE, Huntley B (2008) A flexible approach to assessing synchroneity of past events using Bayesian reconstructions of sedimentation history. Quat Sci Rev 27: Waelbroeck C, et al. (2001) The timing of the last deglaciation in North Atlantic climate records. Nature 412(6848): Thornalley DJR, Barker S, Broecker WS, Elderfield H, McCave IN (2011) The deglacial evolution of North Atlantic deep convection. Science 331(6014): Gebbie G, Huybers P (2010) Total matrix intercomparison: A method for determining the geometry of water-mass pathways. J Phys Oceanogr 40(8): Khatiwala S, Primeau F, Holzer M (2012) Ventilation of the deep ocean constrained with tracer observations and implications for radiocarbon estimates of ideal mean age. Earth Planet Sci Lett : Monnin E, et al. (2001) Atmospheric CO2 concentrations over the last glacial termination. Science 291(5501): McManus JF, Francois R, Gherardi J-M, Keigwin LD, Brown-Leger S (2004) Collapse and rapid resumption of Atlantic meridional circulation linked to deglacial climate changes. Nature 428(6985): Horn MG, Beucher CP, Robinson RS, Brzezinski MA (2011) Southern ocean nitrogen and silicon dynamics during the last deglaciation. Earth Planet Sci Lett 310(3-4): Veres D, et al. (2013) The Antarctic ice core chronology (AICC2012): An optimized multi-parameter and multi-site dating approach for the last 120 thousand years. Clim. Past 9(4): Shakun JD, et al. (2012) Global warming preceded by increasing carbon dioxide concentrations during the last deglaciation. Nature 484(7392): Skinner L, et al. (April 29, 2013) North Atlantic versus Southern Ocean contributions to a deglacial surge in deep ocean ventilation. Geology, /G Skinner LC, Shackleton NJ (2004) Rapid transient changes in Northeast Atlantic deepwater ventilation-age across Termination I. Paleoceanography 19: Vogel JS, Southon JR, Nelson DE, Brown TA (1984) Performance of catalytically condensed carbon for use in accelerator mass spectrometry. Nucl Instrum Methods Phys Res B 5: Fallon SJ, Fifield LK, Chappell JM (2010) The next chapter in radiocarbon dating at the Australian National University: Status report on the single stage AMS. (Translated from English). Nucl Instrum Methods Phys Res B 268(7-8): Stuiver M, Polach HA (1977) Reporting of 14C data. Radiocarbon 19(3): Svensson A, et al. (2008) A 60,000 year Greenland stratigraphic ice core chronology. Clim. Past 4: Mayewski PA, et al. (1997) Major features and forcing of northern hemisphere atmospheric circulation using a 110,000-year-long glaciochemical series. J Geophys Res 102(C12): Skinner et al.

6 Supporting Information Skinner et al /pnas SI Materials and Methods Core Locations. The locations of cores MD K and MD are shown in Fig. S1. The site of MD K (37 48 N; W; 3,146 m), on the Iberian Margin in the Northeast Atlantic, is currently bathed in northward recirculating Northeast Atlantic Deep Water, a modified form of North Atlantic Deep Water that includes 47% Lower Deep Water (derived from Antarctic Bottom Water) (1). The site of MD ( S; W; 3,770 m), on the eastern flank of the mid-atlantic Ridge in the sub-antarctic South Atlantic, is currently bathed in a predominantly southward-flowing mixture of modified North Atlantic Deep Water and Lower Circumpolar Deep Water, which feeds into the core of eastward flowing Circumpolar Deep Water. Fig. S1 illustrates the locations of these cores within the field of modern (prebomb) seawater radiocarbon concentrations, as well as the estimated proportion of water that is currently derived from the surface North Atlantic (>40 N) versus the surface Southern Ocean (>40 S), based on the methods of ref. 2. Planktonic Stable Carbon Isotopes. Planktonic stable carbon isotopes were measured on samples of Globigerina bulloides, using 30 individuals handpicked from the μm size-fraction of core MD Stable isotope analyses were carried out using a VG SIRA mass spectrometer at the Godwin Laboratory, Cambridge, United Kingdom, reacted sequentially using an ISO- CARB preparation system. Measurements of δ 13 C were determined relative to the Vienna Peedee Belemnite standard; analytical precision is better than Radiocarbon Dates and Chronology. Radiocarbon dates from core MD K are listed in Datasets S1 and S2. Here we have supplemented radiocarbon dates previously reported in ref. 3. Where planktonic dates are not available from precisely the same depth interval as the benthic dates, interpolated planktonic dates have been used to calculate benthic planktonic (B-P) offsets. To obtain a record of surface- and deep-water reservoir age variability (i.e., the evolving offset between marine and atmosphere radiocarbon concentrations), radiocarbon dates from core MD K must first be placed on an independent calendar age-scale. It has been shown that past changes in Greenland temperatures, as recorded by ice-core precipitation isotope records, have been closely tracked by surface temperature changes on the Iberian Margin, as recorded by planktonic δ 18 O (4, 5) and Mg/Ca (6), as well as pollen transported from the Iberian Peninsula (7). This close coupling (which is also conceptually supported) provides a basis for aligning marine records from the Iberian Margin to the Greenland ice-core event stratigraphy. A similar alignment may also be performed with the event stratigraphy of the east Asian speleothem records (8 11). The advantage of this alignment is that it permits the transferral of the relatively precise and accurate radiometric speleothem agescale (12, 13). In the context of this study, a further advantage of adopting the speleothem chronostratigraphy is that it directly incorporates atmospheric radiocarbon measurements that have been performed in the Hulu speleothem H82 (11). If the Greenland and Asian speleothem event stratigraphies are indeed coupled, and if the independent chronologies associated with each of these stratigraphies are indeed correct, then an alignment of marine records to each of these event stratigraphies should be equivalent and should be consistent with further independent calendar age constraints [including uranium-series dated atmospheric radiocarbon estimates from corals, for example, (13)]. Despite small discrepancies between Asian speleothem ages and the most recent Greenland NGRIP-GICC05 ice-core chronology (14), it does appear that the two chronologies are very close indeed (15). The Greenland/Asian speleothem link is robust for the clearly identifiable Greenland Isotope Stages 1 4, for example, but it appears to break down in the interval comprising Heinrich Stadial 1 (HS1) and the Last Glacial Maximum (LGM), where the Greenland event stratigraphy itself becomes ambiguous, and the link between Greenland and Asian speleothem event stratigraphies is no longer clear. For this reason, we adopt what might be an optimum calendar chronology for core MD K, based on an alignment to the most highly resolved Asian speleothem records, using all possible tie-points (i.e., including within the HS1/LGM interval), while also considering an alternative chronology that ignores tie-points that cannot be clearly identified in both the speleothem and Greenland ice-core stratigraphies within the HS1/LGM interval. The two possible stratigraphic alignments are illustrated in Fig. S2. We use these two possible stratigraphic alignments, and their likely uncertainties (consisting of a uniform 50-y dating uncertainty plus a uniform 300-y correlation uncertainty), to derive best guess, maximum, and minimum age-scales for MD K. These age-scales (illustrated in Fig. S3) then provide the basis for calculating a range of surface reservoir ages in MD K. This is done by determining the difference between radiocarbon ages in MD K (placed on the best guess, maximum, and minimum chronologies) versus concurrent atmospheric radiocarbon ages from the Intcal09 radiocarbon calibration curve (16) (0 10,670 y BP) and the Hulu H82 speleothem (11) (10,674 26,850 y BP). For surface reservoir ages, we take the most conservative approach of only inferring reservoir ages at tie-points (using interpolated planktonic radiocarbon dates where necessary). These reservoir-age constraints are used to correct down-core planktonic radiocarbon dates (interpolating where necessary), which are then used to derive a Bchron (17) age-depth model that is consistent with both the structure in the down-core radiocarbon dates and the calendar age constraints from the two alternative speleothem chronostratigraphies. Deep-water reservoir ages [i.e., benthic versus atmosphere radiocarbon age offsets, or benthic atmosphere (B-Atm)] are determined as the sum of the known down-core B-P offsets and interpolated surface reservoir ages, consistent with the best guess and bounding Bchron chronologies. It is important to stress that this approach is preferable to simply placing the sediment core on two alternative stratigraphic agemodels, based on linear interpolation between very sparse depthage constraints, and then deriving offsets between planktonic and benthic radiocarbon dates versus atmospheric ages from the Hulu H82 speleothem (the resultant B-Atm ventilation age records are illustrated in Fig. S3 for comparison). The reason for this is that the latter approach ignores additional information on the structure of the age depth relationship in MD K (i.e., from the succession of planktonic radiocarbon dates) in the intervals where there is no information on the likely age depth relationship from chronostratigraphic tie-points (e.g., when tie-points in the HS1/ LGM interval are ignored for the minimum age-model, as discussed earlier). Our approach essentially follows the premise that it is more reasonable to assume smooth/minimal changes in surface reservoir ages between exceedingly sparse chronostratigraphic constraints than to assume completely invariant sedimentation rates Skinner et al. 1of5

7 between exceedingly sparse chronostratigraphic constraints. Accordingly, the difference between the two approaches is maximized when the number of available chronostratigraphic tie-points is minimized. Finally, by plotting optimum reservoir age estimates within a possible range, we underline the proposition that the likelihood distribution of possible reservoir/ventilation ages within the maximum/minimum range is probably not uniform. This proposition will be correct to the extent that the best guess speleothem chronology is also closest to being correct. It is notable that the radiocarbon age versus depth relationship in MD K has a similar shape to the calendar age versus depth relationship defined by the best guess chronology (Fig. S3). This is not the case for the age scale that would derive from a linear interpolation that ignores the possible age constraints in the LGM/HS1 interval. This would suggest that the best guess chronology is indeed more likely to be correct than one that ignores age constraints during the HS1/LGM interval. Revised Ventilation Ages in MD For consistency, it is important the two radiocarbon records we compare are both referenced to the same atmospheric radiocarbon dataset. For MD K, placed on the speleothem age scale, we adopt the Intcal09 and Hulu Cave radiocarbon records (11, 16), whereas the reservoir ages, chronology, and radiocarbon ventilation estimates from core MD were originally derived using a splice of the Intcal04 (18) (0 26,000 y BP) and Cariaco Basin (>26,045 y BP) (12) datasets. Using the same calendar age constraints on core MD as originally published, we have reassessed the reservoir ages, the calibrated radiocarbon-age Bchron chronology, and the inferred B-Atm ventilation ages for MD , using the same Intcal09/Hulu atmospheric radiocarbon splice as used here for MD K. As shown in Fig. S4, this results in almost no change at all, except before 20,000 y BP, where the revised ventilation ages in MD are slightly older, yet still within the originally estimated uncertainty range (19). 1. van Aken H (2000) The hydrography of the mid-latitude northeast Atlantic Ocean I: The deep water masses. Deep Sea Res Part I Oceanogr Res Pap 47: Primeau F (2005) Characterizing transport between the surface mixed layer and the ocean interior with a forward adjoint global ocean transport model. J Phys Oceanogr 35: Skinner LC, Shackleton NJ (2004) Rapid transient changes in Northeast Atlantic deepwater ventilation-age across Termination I. Paleoceanography 19: Shackleton NJ, Hall MA, Vincent E (2000) Phase relationships between millennial-scale events 64,000-24,000 years ago. Paleoceanography 15(6): Skinner LC, Elderfield H (2007) Rapid fluctuations in the deep North Atlantic heat budget during the last glaciation. Paleoceanography 22:PA Skinner LC, Elderfield H (2005) Constraining ecological and biological bias in planktonic foraminiferal Mg/Ca and d18occ: A multi-species approach to proxy calibration testing. Paleoceanography 20: Margari V, et al. (2009) The penultimate glacial and the nature of millennial scale variability. Nat Geosci 3(2): Wang X, et al. (2006) Interhemispheric anti-phasing of rainfall during the last glacial period. Quat Sci Rev 25: Wang Y, et al. (2008) Millennial- and orbital-scale changes in the East Asian monsoon over the past 224,000 years. Nature 451(7182): Wang YJ, et al. (2001) A high-resolution absolute-dated late Pleistocene Monsoon record from Hulu Cave, China. Science 294(5550): Southon J, Noronha AL, Cheng H, Edwards RL, Wang Y (2012) A high-resolution record of atmospheric C-14 based on Hulu Cave speleothem H82. Quat Sci Rev 33: Hughen K, Southon J, Lehman S, Bertrand C, Turnbull J (2006) Cariaco Basin 14C calibration and activity record over the past 50,000 years updated from the Cariaco Basin. Quat Sci Rev 25: Skinner LC (2008) Revisiting the absolute calibration of the Greenland ice-core agescales. Clim. Past 4: Svensson A, et al. (2008) A 60,000 year Greenland stratigraphic ice core chronology. Clim. Past 4: Zhao K, Wang Y, Edwards L, Cheng H, Liu D (2010) High-resolution stalagmite d18o records of Asian monsoon changes in central and southern China spanning the MIS 3/2 transition. Earth Planet Sci Lett 298: Reimer PJ, et al. (2009) INTCAL09 and MARINE09 radiocarbon age calibration curves, 0-50,000 years cal BP. Radiocarbon 51(4): Parnell AC, Haslett J, Allen JRM, Buck CE, Huntley B (2008) A flexible approach to assessing synchroneity of past events using Bayesian reconstructions of sedimentation history. Quat Sci Rev 27: Reimer PJ, et al. (2004) INTCAL04 Terrestrial radiocarbon age calibration, 0-26 cal kyr BP. Radiocarbon 46(3): Skinner LC, Fallon S, Waelbroeck C, Michel E, Barker S (2010) Ventilation of the deep Southern Ocean and deglacial CO2 rise. Science 328(5982): Key RM, et al. (2004) A global ocean carbon climatology: Results from the Global Data Analysis Project (GLODAP). Global Biogeochem Cycles 18(GB4031):1 23. Skinner et al. 2of5

8 MD MD K S 60S 30S EQ 30N 60N 90N GLODAP pre-bomb radiocarbon (permil) Fig. S1. Location of cores discussed in the text; MD K (37 48 N, W; 3,146 m) and MD ( S; W, 3,770 m) are indicated by solid and open circles, respectively. Color shading indicates modern Global Data Analysis Project prebomb radiocarbon concentrations (20), and dashed contours indicate the proportion of water that last made contact with the surface Southern Ocean >40 S, based on the methods in ref. 2, courtesy of Primeau. Fig. S2. Alignment of planktonic δ 18 O measured in core MD K (6), with the Asian speleothem event stratigraphy (10, 11, 15). The Greenland (NGRIP) event stratigraphy on the GICC05 age scale is essentially identical (14), although it is difficult to identify the same tie-points as in the speleothem records within the HS1-LGM interval (crosses with dashed vertical drop-lines). Two bounding age scales are therefore considered in this study, one using all tie-points to the speleothem event stratigraphy (red line in top plot) and one that avoids using tie-points within the HS1-LGM interval (gray line in top plot). Skinner et al. 3of5

9 Fig. S3. (A) Best guess, maximum, and minimum likely surface reservoir ages in core MD K. (B) Deep ventilation ages (solid dark gray line indicates B-P offsets; solid black squares and line with 1σ radiocarbon dating uncertainties indicate best guess B-Atm offsets with maximum/minimum shaded range; dashed gray lines indicate alternative inferior ventilation histories as described in the text). (C) Benthic radiocarbon dates (black symbols) and planktonic radiocarbon dates (red symbols) compared with a range of Bchron calendar age depth models (solid black line indicates the best guess or 50% cumulative probability age, and the shaded area indicates maximum/minimum or % cumulative probability age range), as well as the speleothem-based calendar age constraints on which these Bchron age-models are based (colored crosses). The maximum Bchron age-scale in C is derived from the minimum surface reservoir ages in A, which ignore the calendar age constraints within the HS1-LGM interval (red crosses in C). All error bars represent 1σ. Fig. S4. Comparison of previously published ventilation age estimates from core MD (19) [red circles and dotted line; referenced to the Intcal04 and Cariaco Basin radiocarbon records (12, 18)], with the revised estimates presented in Fig. 2 [black crossed circles and solid line; referenced to the Intcal09 and Hulu speleothem atmospheric radiocarbon records instead (11, 16)]. The Hulu speleothem record only extends to 25,850 y BP. Skinner et al. 4of5

10 Other Supporting Information Files Dataset S1 (TXT) Dataset S2 (TXT) Skinner et al. 5of5