The Cause of Carbon Isotope Minimum Events on Glacial Terminations

See allHide authors and affiliations

Science  19 Apr 2002:
Vol. 296, Issue 5567, pp. 522-525
DOI: 10.1126/science.1069401


The occurrence of carbon isotope minima at the beginning of glacial terminations is a common feature of planktic foraminifera carbon isotopic records from the Indo-Pacific, sub-Antarctic, and South Atlantic. We use the δ13C record of a thermocline-dwelling foraminifera, Neogloboquadrina dutertrei, and surface temperature estimates from the eastern equatorial Pacific to demonstrate that the onset of δ13C minimum events and the initiation of Southern Ocean warming occurred simultaneously. Timing agreement between the marine record and the δ13C minimum in an Antarctic atmospheric record suggests that the deglacial events were a response to the breakdown of surface water stratification, renewed Circumpolar Deep Water upwelling, and advection of low δ13C waters to the convergence zone at the sub-Antarctic front. On the basis of age agreement between the absolute δ13C minimum in surface records and the shift from low to high δ13C in the deep South Atlantic, we suggest that the δ13C rise that marks the end of the carbon isotope minima was due to the resumption of North Atlantic Deep Water influence in the Southern Ocean.

A persistent feature of planktic and intermediate-depth benthic foraminifera carbon isotope records from tropical through Southern high latitudes is the large (0.3 to 1.2‰), rapid negative δ13C excursions that are observed on deglaciations (1–5). Because these carbon isotope events occur when the northern hemisphere ice sheets are collapsing, it is difficult to explain them via the transfer of isotopically light carbon from the terrestrial to oceanic carbon reservoirs. Rather, since the terrestrial biosphere was already expanding and sequestering 12C-rich CO2 into biomass, and the glacial oceans were preconditioned with a low δ13C signature from remineralized terrestrial carbon during ice sheet growth (6), one would expect the δ13C of the oceanic carbon reservoir to increase at the end of glacials. The extensive distribution of these δ13C minima in the tropical Indo-Pacific, south Atlantic and sub-Antarctic and the rate at which the full signal appears in different basins, suggests the signal source originates from an oceanic region with direct connection to the different ocean basins (5).

Ninnemann and Charles (5) argued that these carbon isotope minima could not be a whole ocean signal because they are absent from north Atlantic planktonic records. Rather, they suggested that the tropical Indo-Pacific surface δ13C minima were due to the transfer of a preformed δ13C signal from the sub-Antarctic via Antarctic Intermediate Water (AAIW) or sub-Antarctic Mode Water (SAMW) (2, 7), with subsequent propagation through the low-latitude thermocline. They further proposed that the signal source was a change in gas exchange fractionation across the air-sea interface as Southern Ocean temperatures warmed at the end of glacials. Because a decrease in13C/12C fractionation would accompany a general postglacial oceanic surface temperature (SST) increase, (5) predicted that the δ13C of atmospheric CO2should rise as oceanic δ13C decreased. However, δ13C data from Late Quaternary packrat middens (8) and CO2 from the Taylor Dome ice core (9) show that atmospheric δ13C initially decreased as the deglaciation began, and did not begin to increase until ∼3 thousand years (ky) later. These atmospheric proxies are not consistent with the proposed equilibration mechanism, requiring us to explore alternatives to explain the termination δ13C events.

Site TR163-19 is located on the Cocos Ridge in the eastern equatorial Pacific (EEP) (2°15.5′N, 90°57.1′W, 2348 m water depth), just north of the cold upwelling water that characterizes much of this region (10). Here, the subsurface thermocline waters are thought to be strongly influenced by sub-Antarctic water masses via the equatorial undercurrent (EUC) (11, 12) and the region is not complicated by changes in the position of deep ocean currents with different isotopic composition as is the case in the Atlantic Ocean (13). The source of the EUC and waters upwelling from the Peru-Chile undercurrent is likely SAMW, which forms north of the sub-Antarctic front and is the major precursor to AAIW (14). It is thought that SAMW ventilates the Pacific Ocean as AAIW by subduction and northward advection into the Pacific subtropical gyre from its primary source in the southeast Pacific (15). The dense component of SAMW subsequently flows through the Drake Passage where it is slightly modified, finally becoming the main core of AAIW in the Atlantic and Indian Oceans. In the southwest Pacific, older AAIW flows into the Pacific subtropical gyre where it also contributes to the EUC. The nutrient and δ13C content of Pacific AAIW is set by the chemistry of upwelled circumpolar deep water (CPDW) and partial equilibration with atmospheric CO2 (16) as waters advect into the SAMW source region. On the basis of CFC-11 content, part of the AAIW source to the Pacific EUC has a component that is <25 years old (17), suggesting that changes in water column chemistry at the sub-Antarctic front will be recorded in the chemistry of EEP foraminifera within a century or less.

In the vicinity of Site TR163-19, EUC waters are found below ∼75 m depth (12) where a strong halocline, thermocline, and carbonate chemistry chemocline define the boundary between the mixed layer and underlying EUC (Fig. 1) (18, 19). The nonspinose species, Neogloboquadrina dutertrei, is a common inhabitant of the EEP thermocline with a preferred depth habitat of 60 to 150 m (19). As such,N. dutertrei records a subsurface thermocline signal derived from the EUC that is distinct from the mixed layer record ofGlobigerinoides ruber at Site TR163-19 (Fig. 2, A and B).

Figure 1

Water column hydrographic data from the IRONEX cruise, Station 8, November 1993, 92°W, 1°N (18) and planktic foraminifera habitat distribution (19) near Site TR163-19 in the EEP. Globigerinoides ruber resides in the mixed layer whereas Neogloboquadrina dutertrei inhabits the deeper thermocline whose chemistry is controlled by the EUC. Predicted δ18Ocalcite was calculated from the low lightOrbulina universa relationship of Bemis et al.(36) using salinity and temperature data from Station 8 in (18). Measurements of δ13CDIC are from the Panama Basin to the east of TR163-19 (19). These gradients reflect the strength of the near-surface pycnocline and chemocline which separate the habitats of G. ruber andN. dutertrei.

Figure 2

Measured shell δ18O (A) and DIC-normalized shell δ13C (B) for G. ruber (250 to 350 μm fraction) andN. dutertrei (>500 μm shell length) at Site TR163-19 during the last 150 ky. G. ruber δ18O and Mg/Ca-derived SST data (B) were published previously (10). Note that on both Terminations, the initiation of the δ13C minimum events occurs at the start of mixed layer warming in the EEP, as indicated by G. ruber Mg/Ca. To account for biologically and environmentally controlled offsets (e.g., vital effects) between the two species, the raw δ13C data here have been normalized to δ13CDIC using offset corrections of +0.94 and –0.50‰ for G. ruber andN. dutertrei, respectively (37). The chronology for TR163-19 for the last 28 ky has been modified from the original correlation of the G. ruber oxygen isotope record to the SPECMAP chronology and a core-top radiocarbon age (10) using the two new AMS dates presented in the text. We sampled G. ruber at 5-cm resolution (10), equivalent to a potential resolution of about 2 ky. We sampled N. dutertreiat 10 cm resolution throughout the core with the exception of the last 50 ky and Termination II, which were sampled every 5 cm. The δ18O and δ13C differences support a continuous mixed layer and thermocline habitat for these two species. Note that the δ13C minima on glacial Terminations I and II are absent from the G. ruber record, indicating the geochemical environment of the mixed layer is distinct from the EUC dominated thermocline (Fig. 1). Spero et al. (37) suggest the surface δ13C signal is not in equilibrium with the atmosphere and may be controlled by a combination of surface productivity changes and advection of surface waters from outside the region.

Lea et al. (10) argued that within the 2 ky resolution of the G. ruber samples analyzed at TR163-19, Mg/Ca derived SST and the Vostok ice core deuterium record (20), a proxy for Antarctic temperature, are coherent with no discernible phase lag for the past 260 ky. This implies that in the eastern tropical Pacific, SST changes synchronously with temperatures over Antarctica and the Southern Ocean. Model results (21) and observations from Greenland ice cores (22) show that high-latitude warming and sea ice melt back can occur on time scales of decades. If these interpretations and models are correct, then geochemical signals transmitted from the Southern Ocean into the EEP via the EUC-SAMW/AAIW connection should be synchronous with Southern Ocean circulation changes south of the sub-Antarctic front.

Comparison of the N. dutertreiδ13C record with the Mg/Ca SST reconstruction fromG. ruber (10) (Fig. 2B) shows that on both glacial Termination I and II, the intervals immediately prior to the initiation of the δ13C minimum events coincide with the final interval of cool glacial SST. Because the SST rise and δ13C decrease occur in the same intervals of one core, there is no ambiguity about the relative timing of the events. Radiocarbon dating of the sediment interval just prior to the δ13C decrease and SST rise yields a 14C age of 16,630 ± 50 years before the present (70 cm), equivalent to 19.8 ± 0.3 calendar ky (23). The δ13C minimum itself occurs at a 14C age of 13,250 ± 40 (53 cm), equivalent to 15.9 ± 0.2 calendar ky (Fig. 3A). A similar age has been obtained for the δ13C minimum in the N. dutertrei record from another EEP site on the Carnegie Ridge (core TR163-31B, 3°37′S, 83°58′W, 3210 m) (24, 25), which is overlain by the Peru current. The duration of the δ13C decrease at TR163-19 is ∼4 ky, after which δ13C gradually increases until the mid- Holocene.

Figure 3

(A) TR163-19 N. dutertreiδ13C and G. ruber SST across Termination I. AMS ages (upper arrowheads) correspond to the intervals prior to SST increase and δ13C decrease and the N. dutertrei δ13C minimum. (B) pCO2 concentration and δ13C of CO2 from the Taylor Dome ice core record (9). Note that the δ13C minimum is coincident in the two records, but the onset of the event is older in the sediment record.

Although a number of age models exist for the Vostok ice core, the age of the pre-event interval in TR163-19, 19.8 cal ky, agrees well with the timing of initial Antarctic warming according to the atmospheric δ18O age model (26). The age of the absolute δ13C minimum agrees with the timing of this event in South Atlantic upwelled waters off Namibia (13.2 14C ky) (4) and the initial appearance of high δ13C NADW in the South Atlantic (13.1 ky) (13). It also corresponds to the age of the high to low δ13C transition in the deep Caribbean that is thought to reflect the shift in source waters from Glacial North Atlantic Intermediate Water to AAIW (27, 28). The agreement among these 14C dates clearly links the δ13C minimum event to a combination of Southern Ocean warming and changes in thermohaline circulation.

Nitrogen isotopes and other geochemical tracers from Southern Ocean sediments have been used to argue for reduced nutrient supply, increased nitrogen utilization efficiency and increased surface stratification south of the Antarctic polar front during the Last Glacial Maximum (LGM) (29). François et al.(29) reasoned that unlike today, where unstratified waters of the Southern Ocean south of the polar front represent one of the major areas of CO2 efflux to the atmosphere, the isolation of this CO2 “leak” zone via stratification could help explain the low pCO2 levels during the LGM. Modeling results have reinforced this hypothesis, suggesting that LGM pCO2 levels could result from reduced deepwater ventilation (30) associated with a northward shift of the polar easterlies that produced intermediate rather than deep water upwelling (31), a combination of winter sea-ice coverage and ice-induced stratification during the summer (32), or some other mechanism (33). If correct, the δ13CDIC of late glacial lower CPDW should have decreased (29) while that of AAIW and SAMW would have been higher because of decreased nutrient content (31). In this regard, very negative glacial deep water δ13C values are present in benthic foraminifera from the Atlantic sector of the sub-Antarctic Southern Ocean (13, 34).

If we can assume that the Southern Ocean was affected by reduced deepwater ventilation at the end of the LGM, then several events should have occurred virtually simultaneously once the Antarctic continent began to warm, sea ice melted back, and deep mixing and convection of lower CPDW was reestablished. First, upwelling and subsequent northward advection of low δ13C waters to the sub-Antarctic front would have transmitted a low δ13C signal into the SAMW/AAIW source region. Simultaneously, atmospheric CO2 concentrations would rise as supersaturated deep waters evolve CO2 into the atmosphere, and the δ13C of atmospheric CO2 would decrease in response to the chemistry of this new outgassing source (32). Because these events should have occurred simultaneously, the initiation of the δ13C decrease would coincide with changes in atmospheric CO2 (Fig. 3B) and Antarctic temperature (9).

We suggest that the δ13C minima on glacial terminations in EEP records are recording the postglacial expansion of deepwater ventilation and upwelling south of the polar front. The low δ13C signal derived from Southern Ocean deep water was transmitted into the Indo-Pacific and South Atlantic thermocline, where it was recorded as a decrease in foraminiferal δ13C (2). Because the event required ∼4 ky to attain minimum δ13C values, the establishment of modern mode upwelling may have occurred regionally at first, or could reflect a lag in the mixing time of the low latitude thermocline during a period of thermohaline circulation reorganization. We hypothesize that the δ13C of the waters feeding SAMW would have started to rise when North Atlantic thermohaline circulation reinitiated and NADW began contributing low nutrient, high δ13C to the SAMW source region (13). Contributing to this δ13C increase at the sub-Antarctic front would be the general reduction in13C/12C fractionation at the air-sea interface as SST increased (5), and the increase in12CO2 uptake as the terrestrial biosphere expanded.

The most compelling support for our hypothesis comes from the agreement in timing among events from different regions. The synchronous initiation of the δ13C minimum event and SST increase in the Cocos Ridge record and the agreement of the timing of these oceanic events with the start of Antarctic warming and pCO2 rise as recorded in ice cores links these events to the end-glacial warming of the Southern Ocean. Similarly, the age agreement of the absolute δ13C minimum at TR163-19 with changes in deep Caribbean and South Atlantic δ13C (13, 27) and the age of the minimum in surface waters of the South Atlantic (4) and Peru current (24, 25), links these geochemical changes to SAMW/AAIW and the reorganization of thermohaline circulation and NADW production in the North Atlantic.

If our hypothesis is correct, then two additional signals should be present in the geological record. First, in agreement with an earlier suggestion (32), the δ13C of atmospheric CO2 should decrease as pCO2 begins to rise on terminations. Both ice core (9) and terrestrial carbon data (8) show a ∼0.5‰ reduction in the δ13C of CO2 at the beginning of the last deglaciation (Fig. 3B) that is comparable in magnitude to the δ13C reduction recorded by N. dutertrei at TR163-19. Although the timing of the Taylor Dome pCO2 event is ∼2 ky later than the initiation of the N. dutertreiδ13C minimum event, the offset may reflect a combination of the age uncertainties at TR163-19 due to bioturbation and the models used to estimate the lock-in age of the gas relative to the ice. If, the atmospheric and oceanic δ13C events are coeval, then it may be possible to use them to link the marine carbon isotope record with ice core gas ages during deglaciations. Based on the existence of a larger marine event during Termination II (Fig. 2B), we would predict that a δ13C minimum should also be present in Antarctic pCO2 records at the end of the penultimate glaciation.

Second, records of benthic foraminifera from the Pacific should record the δ13C minima events at intermediate water depths, but it should be absent from depths that were bathed by CPDW during the glacial. Intermediate and deepwater benthic records from V19-27 (1373 m) and RC13-110 (3231 m) in the EEP (35) show that on both Terminations I and II the intermediate water record displays a clear δ13C minima spike while the deeper record displays a nearly continuous δ13C rise.

Previous studies have pointed out that benthic-planktic Δδ13C differences contain considerable Milankovich forcing and scale well to the Vostok pCO2 record (1). Perhaps a considerable component of the atmospheric pCO2–oceanic Δδ13C relationship is due to the strengthening and weakening of deep mixing south of the sub-Antarctic front, as the rate of upwelling, CO2 leakage to the atmosphere, and thermohaline circulation modified the vertical13C/12C distribution and changed with Milankovich forcing. In addition, the existence of deglacial δ13C minima in tropical surface water records (2,3) has been difficult to explain because the nutrient increase implied by the δ13C shift is not supported by evidence of increased upwelling in these presently nutrient-poor regions (2). Because the δ13C of atmospheric CO2 was lower at the onset of the deglaciation, tropical surface water δ13CDIC would have decreased via air-sea equilibration without an accompanying nutrient change. Such a mechanism is analogous to the invasion of low δ13C anthropogenic CO2 into the modern surface ocean (7). Taken as a whole, the timing and distribution of the deglacial carbon isotope minimum in tropical marine sediments is consistent with a Southern Ocean origin, with advection through intermediate waters and atmospheric equilibration providing the high-latitude–tropical connection.

  • * To whom correspondence should be addressed. E-mail: spero{at}


View Abstract

Stay Connected to Science

Navigate This Article