Synchronous centennial abrupt events in the ocean and atmosphere during the last deglaciation

See allHide authors and affiliations

Science  25 Sep 2015:
Vol. 349, Issue 6255, pp. 1537-1541
DOI: 10.1126/science.aac6159

Flushing the deep ocean

Have changes in ocean circulation contributed to the sudden increases in the concentration of atmospheric carbon dioxide that occurred during the last deglaciation? Chen et al. provide a high-resolution radiocarbon record for that time, derived from deep sea corals. This record shows that two deep ocean “flushing” events were accompanied by abrupt rises in carbon dioxide and Northern Hemispheric warming. There is a clear connection between these ocean processes and the atmosphere during this interval.

Science, this issue p. 1537


Antarctic ice-core data reveal that the atmosphere experienced abrupt centennial increases in CO2 concentration during the last deglaciation (~18 thousand to 11 thousand years ago). Establishing the role of ocean circulation in these changes requires high-resolution, accurately dated marine records. Here, we report radiocarbon data from uranium-thorium–dated deep-sea corals in the Equatorial Atlantic and Drake Passage over the past 25,000 years. Two major deglacial radiocarbon shifts occurred in phase with centennial atmospheric CO2 rises at 14.8 thousand and 11.7 thousand years ago. We interpret these radiocarbon-enriched signals to represent two short-lived (less than 500 years) “overshoot” events, with Atlantic meridional overturning stronger than that of the modern era. These results provide compelling evidence for a close coupling of ocean circulation and centennial climate events during the last deglaciation.

Paleo-records have shown that warming during the transition from the Last Glacial Maximum (LGM) [~22 thousand to 18 thousand years ago (ka)] to the Holocene occurred in several abrupt events, which were not synchronous between hemispheres (1). The warming in the Southern Hemisphere was accompanied by millennial-scale atmospheric CO2 concentration increases during the Younger Dryas (YD) (12.9 to 11.5 ka) and Heinrich Stadial 1 (HS1) (~18 to 14.6 ka) (2). Together, the timing of interhemispheric temperature and CO2 changes at millennial scales point to a critical role for Atlantic Meridional Overturning Circulation (AMOC) through its “seesaw” behavior (3, 4). Reduced AMOC strength decreased heat transport from the south to the north during the YD and HS1 (5). At the same time, increased Southern Ocean upwelling likely enhanced the release of CO2 (6, 7). Recently, a new, high-resolution Antarctic ice-core record (8) has revealed three abrupt centennial-scale atmospheric CO2 increases superimposed on the millennial-scale deglacial CO2 rise, each of 10 to 15 parts per million by volume (ppmv), contributing a substantial portion of the total 90 ppmv deglacial CO2 increase. Within the constraints of the ice-age and gas-age offsets, the timings of the latter two of these centennial changes are coincident with abrupt Northern Hemisphere warming at the end of the YD and HS1. These two abrupt centennial CO2 rises have been interpreted as being driven from the north by reinvigoration of AMOC (8).

In order to establish direct links between the atmosphere and ocean at centennial time scales, it is necessary to have well-dated, high-resolution marine records that are comparable with ice-core records. Deep-sea fossil corals have the particular advantage that they can be precisely dated with U-series disequilibrium methods (9). The aragonite skeletons of scleractinian corals also record the radiocarbon (14C) content of dissolved inorganic carbon (DIC) at the time of growth, so that coupled 14C/12C analysis and U-series dating of deep-sea corals provides the reconstruction of past deep-ocean 14C/12C ratios. Radiocarbon is produced in the upper atmosphere by cosmic ray-induced nuclear reaction and has a decay half-life of 5730 years. Once introduced into the deep sea from the surface ocean, it is isolated from the atmosphere and decays away. Variability of 14C in the deep ocean thus provides a proxy that is related to the isolation and geometry of deepwater masses and the rate of deep circulation both in the modern and the geological past. In this study, we have generated a detailed deglacial radiocarbon history in the Equatorial Atlantic mainly at depths from 750 to 2100 m [Intermediate/Deep waters (EAI/DW)] and at locations within modern-day Southern Ocean Upper Circumpolar Deep Water (UCDW) (700 to 1800 m) from the Drake Passage on an absolute time scale based on deep-sea corals. We use these data to put new constraints on the millennial to centennial mechanisms connecting the Atlantic, the Southern Ocean, and the atmosphere.

The Equatorial Atlantic coral samples (5 to 15°N) were recovered from depths of 750 to 2800 m from the Sierra Leone Rise, the Mid-Atlantic Ridge, and Researchers Ridge (10). The modern hydrography is mainly composed of North Atlantic Deep Water (NADW), Antarctic Intermediate Water (AAIW) (core depth of 700 to 800 m), and a lesser contribution from subtropical surface waters (fig. S1) (10). Samples of interest with ages less than 25 thousand years (ky) were selected and dated precisely with isotope-dilution methods (9). Radiocarbon analyses were made on samples with U-Th ages that passed our screening criteria (10). We have calculated Δ14C, ΔΔ14C, and B-Atmosphere age (11) for each sample based on IntCal13 (12). The deglacial 14C evolution of UCDW has been reported before, but with lower sampling resolution (13). In this study, we have filled important gaps in the earlier record and increased the number of samples from 31 to 55, allowing a comparison of 14C ventilation between the Southern Ocean and EAI/DW at submillennial time scales (Fig. 1, B and D).

Fig. 1 Radiocarbon variability in the EAI/DW and Drake Passage reconstructed from deep-sea corals.

(A and B) Reconstructed relative 14C activity (Δ14C) of the (A) EAI/DW and (B) Drake Passage. (C and D) The Δ14C offset (ΔΔ14C) of the sample from the contemporary atmosphere of (C) EAI/DW and (D) Drake Passage. Included are 2σ error ellipses reconstructed from deep-sea corals (11). Part of the Drake Passage data have been reported in (13), and the new data of this study are reported in tables S3 and S4. The EAI/DW records have been divided into five different layers according to the sample depths. The black line in (A) represents the IntCal13 (12) atmosphere radiocarbon curve (±2σ uncertainties, gray lines). The red star in (C) represents the sample from 2.8 km depth. One LGM data point from 1097 m depth is not shown because it lies directly on the atmospheric curve and has not been replicated (fig. S3). Before Present (BP) is defined as before 1950 AD.

Holocene Δ14C values of EAI/DW corals agree well with Δ14C of the modern-day seawater DIC (fig. S2). Strong advection of 14C-enriched NADW and relatively 14C-depleted AAIW in the modern Atlantic results in increasing Δ14C with depth between 1000 and 2000 m (14), which is reflected in the coral 14C reconstruction (fig. S2). The deepest sample, recovered from 2800 m water depth, has a calendar age of 10.4 ky and has almost the same Δ14C as that of other early Holocene samples in the EAI/DW (Fig. 1C, star).

The Δ14C values of the shallow EAI/DW layers (750 to 1162 m) (Fig. 1A) decreased from ~365 per mil (‰) during the early LGM to ~–110‰ in the late Holocene, with a slightly larger offset from the contemporary atmosphere (ΔΔ14C) in the glacial period (~–140‰ to –200‰ around 25 to 18 ka) than in the Holocene (~–100‰). At the end of HS1 and YD, the 14C gradient between deep and shallow depths was eroded (Fig. 1C). One glacial coral has a Δ14C similar to that of the contemporary atmosphere (fig. S3 and table S2). However, without more data to confirm the 14C-enriched signature, we do not interpret this 14C result as a signature of ventilation in the EAI/DW.

New 14C data from the Drake Passage (Fig. 1, B and D) support previous findings (13) and further constrain the timing of 14C evolution toward a smaller ΔΔ14C during late HS1. A rapid return to a relatively 14C-depleted condition during the early Bølling-Allerød (B-A) (14.6 to 12.9 ka) was followed by a second abrupt ΔΔ14C increase during the YD-to-Holocene transition (Fig. 1D) (13).

Observations from nutrient proxies indicate that a strong chemical gradient existed between 2000 and 2500 m depth in the Atlantic during the LGM (15). Glacial reconstructions also show 14C-depleted signatures in the deep Atlantic, thus supporting a more isolated deep ocean (1618). However, data in the intermediate ocean are more challenging to interpret. For example, geostrophic reconstructions from the Florida Straits point to a reduced Gulf Stream (19), whereas evidence from sedimentary Pa/Th ratios suggests that glacial Atlantic upper-ocean circulation (<2000 m) was at least as strong as the modern deep overturning (20). Our 14C data (Fig. 2C and fig. S4) from intermediate depths (750 to 1492 m) as well as a published record from thermocline waters (500 to 600 m) (21) indicate that the upper Equatorial Atlantic during the glacial period was filled with 14C-enriched water similar to that of the modern era, supporting relatively strong circulation in the upper ocean. However, the 14C age differences between UCDW and the shallow layer (750 to 1162 m) of EAI/DW were much larger during LGM (~800 to 1000 years) than the present (~500 years) (Fig. 2C). Mechanisms that could maintain these large glacial 14C gradients might be related to water mass mixing and advection rates, as well as more extreme 14C end-member compositions during the glacial period (10).

Fig. 2 Radiocarbon age variability in the EAI/DW of the past 25 ky in comparison with other climate records.

(A) Δ14C of the atmosphere from the IntCal13 compilation (12). (B) New high-resolution CO2 concentration record from the West Antarctic Ice Sheet Divide ice core (WDC) (8). (C) B-Atmosphere age reconstructed from 14C data of EAI/DW and Drake Passage (700 to 1800 m) (13). For clarity, the 2σ error ellipses of Drake Passage records are not shown. The green arrow illustrates the modern radiocarbon age difference between EAI/DW and the atmosphere, and the gray arrow illustrates the modern radiocarbon age difference between UCDW and the atmosphere (14). The red star represents the sample from 2.8 km depth. (D) δ18O (an Asian monsoon index) record from the Hulu speleothem (35, 40). (E) 231Pa/230Th ratios (an AMOC strength index) of a subtropical North Atlantic deep sediment core (OCE326-GGC5, water depth 4550 m) (5). The age model has been revised on the basis of IntCal13 (12). (F) δ18O (atemperature index) record of the Northern Greenland (NGRIP) (41) and Antarctic (WDC) (8) ice cores. Gray lines in (F) represent five-point moving average. Light yellow bands indicate cold stadials, and light blue bands indicate periods of sudden centennial atmosphere CO2 increases.

In the early HS1 (18.0 to 16.0 ka), deep sedimentary Pa/Th ratios indicate a substantial reduction in AMOC rates (5), which coincided with the initiation of the deglacial atmospheric CO2 rise (8) as well as the decrease in atmospheric Δ14C (Fig. 2, A, B, and E) (12). During this time period, the mid-depth North Atlantic and Brazil Margin (1500 to 2000 m) both saw a major decrease in benthic δ13C (a measure of the ratio of stable isotopes 13C:12C) (22, 23), with a coincident decrease in North Atlantic [CO32–] concentration (~1800 m) (24). These changes in the carbon chemistry of the mid-depth Atlantic may have been due to the weakened AMOC and its associated effects (such as respired carbon accumulation in mid-depth waters) rather than upwelling in the deep South Atlantic (>2500 m) (22, 23, 25). Under this scenario, an additional, well-ventilated North Atlantic water mass at shallower depths (such as <1200 m) is required to explain the early HS1 Atlantic benthic δ13C and δ18O (a measure of the ratio of stable isotopes 18O:16O) data (23). In support of this hypothesis, our 14C data from 972 to 1162 m do not show any substantial decrease during the initial transition from glacial to the early HS1 (18 to 17.0 ka) (fig. S4). This result can be best explained by a persistent North Atlantic shallow overturning (26) fed by relatively 14C-enriched waters from the north, as observed during the LGM. There is also no evidence in the Equatorial Atlantic for the extremely 14C-depleted signals during HS1 that have been observed at other locations in intermediate waters of the Pacific, Atlantic, and Indian Oceans (2729). Given that extremely depleted 14C waters do not seem to pass through the intermediate depths of the Southern Ocean or Equatorial Atlantic (13, 21, 30), those 14C depletions are likely to be regional or localized features.

In the deeper waters of the Equatorial Atlantic (1827 to 2100 m), the early part of HS1 is characterized by a decreasing Δ14C trend from 17.4 to 16.2 ka (Fig. 1A). The B-atmosphere age at 17.4 ka was more than 400 14C years older than the intermediate waters (972 to 1162 m), which is in marked contrast with the Holocene (Fig. 2C) and is consistent with a reduced presence of 14C-rich NADW. From the LGM to the early HS1 (Fig. 1D), the ΔΔ14C of UCDW in the Drake Passage increased by ~90‰. However, by 16.6 ka this trend was reversed, with ΔΔ14C decreasing by ~80‰, which is consistent with increased mixing of 14C-depleted waters from below (13) and potentially associated with a reduction in AMOC. At the same time, between 18 and 16 ka, Greenland remained cold (31), whereas Antarctic δ18O indicates continuous warming under a background of increasing CO2 forcing (Fig. 2F). These changes in the early HS1 suggest a reduced/shoaled AMOC (Fig. 2E) and a comparatively low efficiency of heat transport from the Southern to the Northern Hemisphere.

Throughout HS1, the water column of the equatorial Atlantic was characterized by low-Δ14C deep waters underlying shallower, more 14C-enriched waters. Although we find no evidence for enhanced 14C ventilation in the 1827-to-2100-m layer, our records show that the Drake Passage records and Equatorial Atlantic mid-depth (1296 to 1612 m) waters started to shift toward a higher Δ14C value as compared with that of the early HS1 (Fig. 2C). Therefore, it is unlikely that this shift was caused by upwelling of 14C-enriched waters from below, but rather from increased lateral advection of a relatively well-ventilated intermediate water mass. We cannot distinguish the exact source of ventilation [for example, Northern Component Water (NCW) from the North Atlantic, Southern Component Water (SCW) from the Southern Ocean, or both] here, but the better ventilation of mid-depth waters in late HS1 is a distinct feature of our records. Foraminifera Nd isotopes from the tropical Atlantic have previously highlighted a two-phase HS1 water-mass provenance shift at intermediate water depths (~1000 m) (32). In contrast, Pa/Th ratios and Nd isotopes from the deep subtropical North Atlantic (5, 33) did not show a mid HS1 shift, which is consistent with our observation that the deeper Equatorial Atlantic remained in a poorly 14C-ventilated condition throughout HS1. Together, these results suggest that increased 14C ventilation and the potential increase in AMOC strength in the late HS1 might have been restricted to the upper ocean (such as <2000 m), whereas the abyssal Atlantic remained less affected (5, 22). This mid-HS1 shift in 14C ventilation was accompanied by major reorganization of the atmosphere and the North Atlantic climate system (10, 31, 34, 35). At around the same time, a reversal in the early HS1 temperature decline at Northern Greenland (NGRIP) initiates Northern Hemisphere deglacial warming (Fig. 3B).

Fig. 3 Comparison of new 14C records with ice-core records during three key intervals of the last deglaciation.

(A to C) Detailed comparison during (A) 11.0 to 12.5 ka, (B) 15.8 to 13.8 ka, and (C) 17.8 to 15.8 ka between B-Atmosphere ages of EAI/DW and UCDW (colors and symbols are as in Fig. 2C, with 2σ error ellipses), reconstructed temperature of North Greenland NGRIP (thin purple line) (31), and high-resolution atmosphere CO2 concentration from WDC (thin blue line) (8). Abrupt centennial CO2 increases are highlighted with light blue bands. Dashed lines in (A) and (B) mark the return of UCDW to reduced 14C ventilation condition. The arrows indicate periods of AMOC overshoot.

Deglacial warming in the Northern Hemisphere is characterized by two abrupt warming events of ~10°C at Northern Greenland, occurring at the HS1-to–B-A transition and YD-to-Holocene transition (31), which were synchronous with the rapid intensification of Asian monsoons (Fig. 2D) (35). During both events, CO2 increased by ~12 ppmv (Figs. 2B and 3) (8), and two coincident abrupt resumptions of deep AMOC are indicated by sedimentary Pa/Th ratios (Fig. 3E) (5).

The two distinct centennial-scale events toward a 14C-enriched water column in both EAI/DW and the Southern Ocean (14.8 to 14.6 ka and 11.7 to 11.5 ka) (Fig. 3) provide strong support for an AMOC-related mechanism driving the abrupt increases in atmosphere CO2 concentration (8) and Northern Hemisphere warming (31). The atmospheric CO2 concentration increased at the same time as the 14C gradient was eroding, suggesting that CO2 was released into the atmosphere when excess respired carbon in the deep ocean was being flushed out by newly formed, high-Δ14C NCW. Meanwhile, changes in the solubility of CO2 may have played a role in modulating atmospheric CO2 as deep waters became warmer (less soluble) and fresher (more soluble). In contrast, the first centennial CO2 increase at ~16.3 ka occurred during a period of reduced AMOC and with no notable changes in North Greenland temperature (Fig. 3C). Therefore, mechanisms other than sudden AMOC changes are likely to dominate this first centennial CO2 increase, such as a rapid shift in ocean fronts driving the degassing of CO2 from mid-depth waters of the Southern Ocean (36).

At the beginning of the B-A and Holocene, the Drake Passage records were even more 14C-enriched than in the modern day, by some 400 to 500 14C years at ~14.3 and ~11.3 ka, respectively (Fig. 2C and fig. S4). Each of the two high-14C peaks in the Drake Passage are well constrained by coral samples from similar depths and locations (13), adding confidence that these 14C shifts reflected changes in deep-ocean circulation. We propose that rapid and deepened advection of well-ventilated NCW homogenized the 14C composition of the water column during these two pronounced “flushing” events. The presence of NCW at abyssal depths during the B-A and Holocene is also supported by the distinctive unradiogenic Nd isotopic shift and enriched 14C signatures in the deep North Atlantic (17, 33).

A subsequent rapid return to a modern-like 14C water column after both flushing events, highlighted by the rapid decrease in Drake Passage Δ14C (Fig. 1B), adds weight to the idea of an “AMOC overshoot,” a transient stronger AMOC than that of the modern ocean. This AMOC overshoot during B-A has been previously proposed in (16) on the basis of benthic 14C data, in which it was described as a transient expansion of the NADW cell and a better ventilated deep ocean as compared with the modern ocean (16, 18). In contrast to but not necessarily at odds with those studies that suggest an overshoot throughout B-A (10), our data indicate that the return to poorer ventilation of the Southern Ocean occurred within 500 years after the start of B-A (14.6 ka) and Holocene (11.5 ka) (Fig. 3, A and B). The rapid return thus suggests that the timing of decline in peak AMOC strength occurred in less than 500 years, which is more consistent with AMOC predictions from modeling studies (16, 37). These results therefore provide evidence for the existence of short-lived deep-ocean flushing events during the last deglaciation.

After the AMOC overshoot in the early B-A, the water-column structure of the EAI/DW was similar to that of the Holocene (Fig. 2C) (16, 17), providing support for the modern-like advection in the Atlantic. At the start of YD, the EAI/DW water column is characterized by high B-Atmosphere ages in the deeper layer and low B-Atmosphere ages in the shallow layer (Fig. 2C and fig. S5). However, the 14C gradient during the YD was not as large as during the early HS1, probably because the 14C-depleted water that built up over the glacial period had been largely flushed out by the start of the B-A. Although the data are not well resolved during the Holocene, we do not see any evidence for substantial change in the water column structure of the EAI/DW during this time.

The millennial-scale atmosphere CO2 increase during YD and HS1 has been suggested to be caused by a southward shift of the westerlies, with increased upwelling in the Southern Ocean (4, 6). At the same time, reduction of AMOC would lead to a decreased efficiency of the ocean’s biological pump because NCW has lower preformed nutrients than that of SCW (25). These oceanic processes would perturb the deep-ocean alkalinity balance and facilitate further CO2 release on millennial scales (24). For the centennial-scale abrupt changes, the first event at 16.3 ka may have been related to a southward shift of ocean fronts in the Southern Ocean (36), with a mechanism similar to that of the millennial-scale CO2 rise. The processes driving the latter two CO2 rise events are likely to be very different; they were related to enhanced AMOC and increased ventilation of the deep ocean with newly formed waters, as supported by the data presented here. Although the terrestrial carbon reservoir may have contributed to carbon release at 14.6 ka (38), the in-phase relationship between EAI/DW ventilation and atmosphere CO2 concentration supports the fundamental role of ocean carbon release during the latter two events. Therefore, mechanisms that can counteract the effect of efficient nutrient utilization of NCW (for example, in contrast with the millennial mechanism) and that are faster than the alkalinity feedback are likely to play a more important role for these events. One of those mechanisms, could be an increase in the preformed nutrient content of NCW associated with a shoaled organic matter remineralization depth during abrupt warming (39), combined with a fast oceanic overturning that rapidly releases respired CO2 into the atmosphere.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S6

Tables S1 to S4

References (4254)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Δ14C is the deviation in per mil units of sample 14C activity from (preindustrial) modern atmosphere after correction for both age-integrated decay and isotope fractionation. ΔΔ14C is the Δ14C difference between sample and the contemporary atmosphere. “B-Atmosphere age” represents the 14C age difference between sample and the contemporary atmosphere.
  3. Acknowledgments: This study was funded by the European Research Council, the Philip Leverhulme Trust, the U.S. National Science Foundation (grants 0636787, 0944474, 0902957, and 1234664), and a Marie Curie Reintegration Grant. All the data reported in this paper are available in the supplementary materials. We acknowledge the crew and science parties of RRS James Cook cruise JC094 and RV Nathaniel B. Palmer cruise NBP1103 who made this study possible. We also thank J. F. McManus and K. R. Hendry for the helpful comments during the preparation of this manuscript and C. D. Coath, C. A. Taylor, S. Lucas, and C. Bertrand for help with sample preparation and analyses. Comments from two anonymous reviewers helped to improve the manuscript, inspiring us to look at the deglacial ventilation and circulation events from a more broadened view.
View Abstract

Stay Connected to Science

Navigate This Article