Introduction

On glacial-interglacial timescales, the carbon cycle is profoundly linked to the global Thermohaline Circulation or Atlantic Meridional Overturning Circulation (AMOC), with small changes in the capacity of oceanic carbon sequestration resulting in major changes of atmospheric CO2-levels. Strong support for this process comes from the co-evolution of deglacial changes in ocean carbon with atmospheric CO21, Δ14C (corrected 14C activity)2, and atmospheric δ13C-values3. These patterns indicate that vast amounts of CO2 were released from a reservoir (or reservoirs) with low 14C and δ13C values during the last deglaciation. Evidence for the existence of such reservoirs has been found in all sectors of the Southern Ocean4,5,6,7,8, the North Pacific9, as well as northern permafrost soils10,11. However, evidence for the pathway of sequestered carbon from the deep-waters to the surface and ultimately the atmosphere is limited. Upwelling of Circumpolar Deep Waters (CDW) in the Antarctic and Subantarctic Zones of the Southern Ocean is the most likely pathway for the carbon-rich waters to debouch carbon to the atmosphere. These upwelled waters are subsequently subducted and exported northward as intermediate- and mode-waters (Fig. 1).

Figure 1
figure 1

Schematic view of Southwest Pacific research area and oceanographic features important to this study. Purple dot—location of PS75/104-1; teal dot Bay of Plenty core locations used by Allen et al.39,52; AAIW Antarctic intermediate water; orange line—STF/subtropical front; light blue line—SAF/subantarctic front; dark blue line—APF/Antarctic polar front. Fronts according to Orsi et al.71. AAIW path (grey arrows) according to Bostock et al.12. Map created with GeoMapApp version 3.6.12 (https://www.geomapapp.org).

Full size image

Any change in the sequestration and cycling of CO2 in the ocean will affect the marine inorganic carbon system. In this respect, reconstructions of water mass carbon chemistry (carbonate ion concentrations; [CO32−]) can yield important insights into the effect of past changes in dissolved inorganic carbon (DIC), and the oceanic carbon reservoir.

Here we present reconstructions of [CO32−] and δ13C on benthic foraminifers (Cibicidoides wuellerstorfi and C. dispars) from an intermediate-water record recovered at 44.4°S offshore New Zealand (PS75/104–1; 835 m; Fig. 1). PS75/104-1 is bathed by Antarctic Intermediate Water (AAIW)12, and thus changes in the [CO32−] will reflect changes in DIC that could provide clues to air-sea gas exchange occurring upstream in the Southern Ocean prior to the subduction of these waters to intermediate depths. Our study presents evidence for deglacial changes in the Pacific DIC pool, highlighting a pathway of sequestered CO2 from the ocean to the atmosphere during Heinrich Stadial 1 (HS1; ~ 18–14.5 thousand years before present (ka)) and the Younger Dryas (YD; ~ 12.9–11.6 ka).

Results

Site PS75/104 is located about 150 km to the east of the South Island of New Zealand, south of the Subtropical Front (Fig. 1). We conducted B/Ca and stable carbon isotope measurements on benthic foraminifers to infer past AAIW changes in carbon chemistry of the past ~ 22 thousand years (kyr). The age model was developed, using a high resolution 14C-based record of 56 individual dates13. Age uncertainties reported by Küssner et al.13 range between 15 and 190 years and are too small to show in Fig. 2. Thus, we have high confidence in the timing of the records reported here. While four samples might be affected by Zoophycos bioturbation13 (Fig. 2), excluding these samples does not impact our age model, the general trends seen in our proxy records, or our interpretations.

Figure 2
figure 2

(a) δ13C (green) and carbonate ion data [CO32−] (purple) of core PS75/104-1. Black dots indicate samples that might be affected by bioturbation. Black triangles—radiocarbon dates8,13. Error bars—1σ. (b) Uncalibrated B/Ca data of PS75/104-1. Filled symbols—C. wuellerstorfi; empty symbols—C. dispars. Blue shading—HS1 Heinrich Stadial 1, YD Younger Dryas, ACR Antarctic Cold Reversal. P I/II—CO2 pulses as shown in Fig. 3.

Full size image

During the late glacial and early deglaciation (18–17 ka) B/Ca values measured on C. dispars show a high variability, averaging at ~ 184 µmol/mol, similar to the nearest measurement on C. wuellerstorfi of 191.4 µmol/mol (Fig. 2), confirming that C. dispars can be used in the analysis of B/Ca. During the early deglacial (HS1), B/Ca values increase until ~ 14.3 ka, reaching a maximum of 256 µmol/mol at 15.8 ka. During the Antarctic Cold Reversal (ACR) values decrease to 183.3 µmol/mol (similar to the glacial values). A second significant increase to about 228.3 µmol/mol occurs between ~ 12.9 and 11.1 ka, followed by a progressive, yet variable decrease within the Holocene (Fig. 2).

Carbonate ion concentrations of core PS75/104-1 range from 31.6 µmol/kg up to 119.1 µmol/kg (Fig. 2). The most pronounced increase in [CO32−] (60.3–109.8 µmol/kg) occurred between 17.2 and 14.5 ka and is paralleled by an equally pronounced rise in δ13C benthic from ~ 1.25 to 1.6‰. Both records display a second, yet less pronounced, increase between 13.55 and 10.52 ka (Fig. 2). During the Holocene, both records diverge, with increasing δ13C and decreasing [CO32−]. Following the YD disturbance to the inorganic carbon system, carbonate compensation drives back the system back to its original state14, while δ13C continues to increase15.

Discussion

Changes in Southern Ocean deep-water ventilation are often used to explain the two deglacial pulses of rising atmospheric CO25,7,8,16,17 (Fig. 3). However, while being an indicator for deep-water residence time, radiocarbon alone does not allow for the direct analysis of the past oceanic carbon pool and changes in DIC. Barring lateral water mass transport, the primary driver for changes in δ13CDIC and [CO32−] is the biological pump, the export of organic matter from the surface into deeper water masses, and its subsequent degradation is a primary driver for changes in δ13CDIC and [CO32−]14,18. With progressive export of carbon (CO2), the biological carbon pump increases the DIC content of a given water mass, while decreasing its δ13C and [CO32−] values. Depending on the balance between CO2 sequestration via the biological carbon pump, circulation and ventilation via the upwelling of deep water masses, the circumpolar Southern Ocean can alter between a carbon sink or source.

Figure 3
figure 3

Southern Ocean proxy records in relation to atmospheric records. (a) PS75/104-1 δ13C (green). (b) Southwest Pacific intermediate-water [CO32−] records. PS75/104-1 (purple; this study), RR0503-7939 (yellow). Red triangle—modern [CO32−]sat 49.48 µmol/kg at depth of PS75/104-158. (c) Southwest Pacific mid depth [CO32–] records. RR0503-83 (teal), RR0503-12539 (dark-red). (d) Drake Passage coral δ11B data34 (orange). (e) South Atlantic upwelling and opal flux50 (black). (f) Atmospheric δ13C3 (maroon) and CO2 splice1 (blue). Blue shading—deglacial pulses in atmospheric CO2. P I first pulse—Heinrich Stadial 1, P II second pulse—Younger Dryas, SP South Pacific, SA South Atlantic, DP drake passage.

Full size image

Several processes might have resulted in the pronounced transient increase in PS75/104-1 [CO32−] and δ13C during HS1 (Fig. 3), such as increased surface export, release of carbon-rich fluids from pockmarks, or the discharge of CO2 from an oceanic carbon reservoir that built-up during the preceding glacial and fed into the formation area of SW Pacific AAIW. In this case the area, where surface waters are influenced by both the upwelling of deep waters south of the APF, and then air-sea exchange processes as it flows north due to Ekman transport and subducted as intermediate waters. During glacial times, a combination of multiple climatic factors enhanced the ability of the Southern Ocean to sequester CO2. Lower surface temperatures allowed for increased uptake of CO219, while higher fluxes of iron-rich dust20 resulted in increased primary productivity as a result of iron fertilization of the nutrient-rich Subantarctic Sector of the Southern Ocean21. This led to enhanced export of carbon to the deep ocean via the biological pump in the Atlantic sector. However, there is no clear evidence of increased productivity in the Subantarctic Sector of the SW Pacific22,23,24,25, while it was decreased in the Antarctic Sector23,24. 230Th fluxes of biogenic matter in our research area have shown no significant local change in export production off New Zealand since the LGM25. Thus, we assume that changes in productivity and surface export did not play a dominant role in driving SW-Pacific AAIW [CO32−] at our core site. Yet, as AAIW chemistry is also influenced by regional, zonal, and meridional processes, we acknowledge the further need for additional investigations to test the feasibility of our interpretation. The distinct anti-phased pattern of atmospheric δ13C (Fig. 3f)3 and PS75/104-1 suggests that AAIW δ13C is not driven by the atmosphere via air-sea gas exchange. An additional factor that potentially influenced water mass carbon chemistry in the SW Pacific might have been the release of CO2 from pockmarks that are documented in the research area26. During the late glacial and early deglacial, carbon-rich fluids were probably released from the seafloor off New Zealand and might have contributed to extremely low CDW and AAIW 14C values8,26. The injection of CO2 would also affect [CO32−] of their respective water masses (Fig. S1). However, as the records of PS75/104-1 14C8 does not point toward a deglacial influence of 14C-dead CO2 from pockmarks26 (Fig. 4) and as the transient increase in [CO32−] is interpreted as a loss of CO2 from a water mass14,27, we expect that the release of carbon-rich fluids did not play a role in the evolution of AAIW [CO32−] during the time interval covered by our study.

Figure 4
figure 4

PS75/104-1 [CO32−] (purple line) vs. intermediate and deep-water ΔΔ14C15. Blue diamonds—PS75/104-1; blue squares—SO213-84-1; orange triangles—PS75/104-1. Purple and green shading—glacial and early deglacial interval of potential CO2-release from pockmarks26.

Full size image

The expansion of Antarctic sea ice toward the north28,29,30, a displacement of Southern Westerly Winds24,31, and increased stratification15,32,33,34, reduced deep-water ventilation and circulation5,7,8, reduced upwelling and release of CO2 in the Polar Frontal Zone, all allowed for the accumulation of carbon in the deep glacial Southern Ocean. Reconstructions of deep-water [CO32−] show that the deep South Atlantic was ~ 15 µmol/kg lower and stored ~ 30 gigatons of additional carbon during the Last Glacial Maximum (LGM) than during the Holocene35,36,37.

HS1 was marked by the most pronounced increase in atmospheric CO21 and was paralleled by a similar increase in PS75/104-1 [CO32−] of ~ 58 µmmol/kg (Figs. 2 and 3). Transient increases in [CO32−], as recorded by PS75/104-1, reflect a loss of CO2 and the subsequent return to the previous state of the marine carbon system via carbonate compensation14,27. The inverse relation of [CO32−] and water mass pCO238, imply that the observed [CO32−] rise is consistent with the HS1 release of CO2 via the Southern Ocean (Fig. 5b). The release of sequestered CO2 is furthermore supported by our record of benthic δ13C that closely parallels the increase in observed [CO32−] as well as the patterns observed in AAIW RR0503-7917,39 (Figs. 3 and 6) and Bay of Plenty δ18O and δ13C gradients40. This trend is in good agreement with the loss of metabolic (high 12C) CO2 via air-sea gas exchange in the formation area of AAIW40 (Fig. S1). However, other factors such as the thermodynamic effect, an overprint of the atmospheric signal41, changes in export production (EP), and changes in AAIW formation should be considered when interpreting our records. Air sea exchange under colder temperatures shifts the δ13C values toward higher values42,43,44. Thus, warming temperatures tend to shift the system in the direction of higher δ13C values39. Our HS1 and YD trends in δ13C and [CO32−] (Fig. S1) are more in line with the slope predicted for regenerated organic carbon39, and thus imply that carbon sequestration via the biological pump, and release via ventilation was the more dominant driver. Another process that was observed to have a pronounced influence, is the overprint of atmospheric δ13C values on surface and recently ventilated waters41. Several δ13C-records follow the atmospheric pattern, while minor differences can be attributed to different temperatures during air-sea exchange41. In contrast to these records, PS75/104-1 δ13C is antiphased to the atmospheric record3 during HS1 and the YD (Fig. 3). During the Holocene, however, we observe a strong correlation between atmospheric and AAIW δ13C values that imply an overprint as proposed by Lynch-Stieglitz et al.41. Any pronounced decrease in EP might also affect and increase both δ13C and [CO32−]. In the South Atlantic, a reduction in EP coincided with a HS1 increase in [CO32−], pointing to decreased biological productivity as a contributing factor to rising CO245. A similar process could presumably have driven or contributed to both pulses observed during HS1 and the YD. The analysis of 230Th normalized fluxes of biogenic opal, carbonate, and excess barium on a suite of sediment records from the SW-Pacific indicate no pronounced change in EP since the LGM25. Thus, while a contribution of changes in EP cannot be excluded, we assume that it would only have a subordinate effect on the patterns observed. Given that AAIW could integrate signals from broad Southern Ocean regions due to homogenization by Antarctic Circumpolar Current, we want to encourage further work. Another factor to consider is the formation of AAIW and its subduction toward our core location. If the formation of low [CO32−] AAIW is reduced, an increase in concentrations at our core site can be expected. The formation of AAIW is closely coupled to the applied wind stress46. During the glacial, stationary SW-Pacific ocean fronts47 in combination with a northward displacement of Southern Westerly Winds24, reduced the wind stress experienced in the formation area of SO AAIW. These processes in combination to changing salinity contrasts might have been coupled to the reduced the glacial subduction of SO AAIW (Ronge et al., 2015). Given these local effects, we expect only a subordinate role on the patterns recorded in PS75/104-1.

Figure 5
figure 5

Schematic South Pacific overturning and carbon cycling. (a) Last Glacial Maximum. Sluggish CDW circulation with separated lower and upper cells. Lower DIC-rich cell with very depleted ΔΔ14C values8. (b) Heinrich Stadial 1 (HS1) and Younger Dryas (YD) scenario. Progressive increase in deep-water overturning results in the release of CO2 from DIC-rich and high pCO2 CDW. This release causes the transient rises in AAIW [CO32−] observed in downstream PS75/104-1 indicated by purple dot. Green arrow—input of freshwater into the upper cell15, red arrow—input of highly saline brine into the lower cell33.

Full size image
Figure 6
figure 6

Intermediate- to deep-water δ13C records (four-point running average) from the Bay of Plenty (RR0503 cores)17 and Chatham Rise8. The pattern and general trends of the shallowest core PS75/104-1 (this study) are very similar to Chatham Rise core SO213-84-115 and Bay of Plenty record RR0503-7917. Blue shading—deglacial pulses in atmospheric CO2. P I first pulse—Heinrich Stadial 1, P II second pulse—Younger Dryas.

Full size image

The HS1 pulse of increasing atmospheric CO2 was accompanied by the most dramatic drop of atmospheric δ13C as reconstructed Antarctic ice cores3 (Fig. 3f). Thus, the release of CO2 via the Southern Ocean, as implied by PS75/104-1 and Bay of Plenty records39,40, illustrates a likely mechanism that can account for the coevolution of atmospheric CO2 and δ13C-values (Fig. 3)1,3.

Throughout the ACR (14.5–12.9 ka)48, when atmospheric CO2 plateaued1 and its δ13C briefly returned to higher values3, intermediate-water [CO32−] and δ13C return to lower, glacial-like values (Fig. 3). This suggests that expanding sea ice during the ACR disrupted the communication of upwelling deep waters with the atmosphere, before being incorporated into AAIW. These findings contrast 14C reconstructions from deep sea corals, in the South Tasman Sea bathed by AAIW16. The zonal asymmetry might be explained by regional differences in the movement of Southern Ocean fronts that are less constrained by sea floor topography in the Indo-Pacific sector16.

During the YD there is evidence in the ice cores for a second pulse of increasing CO2 and decreasing δ13C, this is again accompanied by an increasing intermediate-water [CO32−] and δ13C in PS75/104-1 (Fig. 3). Thus, our AAIW record at PS75/104-1 points to a strong mechanistic link between Southern Ocean ventilation and atmospheric CO2 that was active in the SW Pacific.

However, to understand the importance of this Southern Ocean pathway of CO2 in the deglacial carbon system, we have to address several important questions: Which key regions contributed to the two-pulse, deglacial rise in atmospheric CO27,34,39,49,50,51,52? What were the mechanisms and reservoirs that resulted in the release of CO210,11,53,54, and which role did these have on the patterns observed in intermediate waters off New Zealand (PS75/104-1)?

In combination with shifting Southern Westerly Winds24, the deglacial retreat of Antarctic sea ice29 resulted in an intensification of upwelling of Circumpolar Deep Water throughout the Southern Ocean (Fig. 3e)50. Deep-water records from the Southern Ocean show significant perturbations during HS1 and the early deglacial. Radiocarbon values from the Atlantic, Indian, and Pacific Sectors of the Southern Ocean indicate an increase in deep-water ventilation that reflects the renewed contact of the deep and shallow overturning cells and thus exchange of the glacial carbon pool with surface waters and the atmosphere5,7,8. Simultaneously, a rise in CDW [CO32−]39,52 and coral-derived δ11B34 reflect the loss of CO2 from this lower cell, coeval with increasing deep-water ventilation throughout HS1 (Fig. 3). Increasing pH in the lower overturning cell and decreasing pH in the upper cell indicate a release of CO2 from the deep Drake Passage during HS134. However, our understanding of the lower cell carbon chemistry in other sectors of the Southern Ocean is still poorly constrained. In this respect, our AAIW data from the Pacific Sector suggest a close link between sea ice and westerly winds24,29, upwelling intensity50, deep-water ventilation and carbonate chemistry8,39,52, and atmospheric CO2 (Figs. 3 and 5). Intensified air-sea gas exchange triggered a transient increase in [CO32−] in the formation area of AAIW. The signal from this process was subsequently exported from the formation area of SW Pacific AAIW12 toward the core location of PS75/104-1 (Figs. 1 and 5). During HS1, the abrupt reduction in Bay of Plenty intermediate-water Δδ13C (663–1165 m) suggests the loss of CO2 from SW Pacific AAIW40. Radiocarbon reconstructions from a southeasterly bathymetric transect off New Zealand likewise identified the ventilation from upwelling deep-waters in the formation area of Southern Ocean AAIW8. HS1 upwelling and ventilation of carbon-rich deep waters did not affect the (14C) ventilation of PS75/104-18 (Fig. 4), while [CO32−] in our record as well as nearby AAIW record RR0503-7939 show a significant excursion to higher values. Collectively this argues for a loss of CO2 in the formation area of AAIW.

In combination with other Southwest Pacific records that display a similar HS 1 pattern8,15,17,39,52 our AAIW data highlight the importance the Southern Ocean’s Pacific pathway had on the HS1 atmospheric CO2 increase. The majority of AAIW reaching our core location is formed directly to the south as so called SO AAIW12. Nevertheless, given the similar pattern evident in the Bay of Plenty records39 it could also be due to an upstream contribution from the SE Pacific (the primary formation region of AAIW) or the Indo Pacific via the Antarctic Circumpolar Current7,16.

Following HS1, the southern hemispheric ACR was marked by a reduction in Southern Ocean upwelling rates, however less pronounced than during the LGM (Fig. 3e)50. In PS75/104-1 [CO32−] and δ13C values rapidly decreased during the ACR (Fig. 2). The decrease in opal flux (suggested to be an indicator of upwelling of carbon-rich deep-waters)50 and an enhanced winter and spring sea ice cover55, reduced the carbon loss in the formation area of AAIW. Throughout the Austral summer and autumn, ACR sea-ice and biological feedbacks increased the sequestration of CO2 in the high southern latitudes55. In combination, these factors provide a likely scenario as the mechanism for the ACR trends seen in our record and highlight the fact that AAIW off New Zealand (PS75/104-1) can trace upstream changes in the Antarctic Zone.

Following the ACR, during the YD AAIW [CO32−] (this study) and opal flux suggest a reinvigorated upwelling in the Antarctic zone of the Southern Ocean (Fig. 3e)50. This is supported by other records from the Southern Ocean. In the South Atlantic, two records from 4276 m56 and 4981 m57 point toward a progressive deepening in the erosion of the deep ocean carbon pool. Similar to the South Atlantic57, it is likely that the ventilating water masses came from below ~ 4300 m. Radiocarbon-based reconstructions of deep-water ventilation show that down to this depth, deep-water ventilation reached modern-like values at the end of HS18. There is also evidence for a Southern Ocean contribution to the second pulse in atmospheric CO2 from rapidly decreasing pH values of the lower cell in the Drake Passage (Fig. 3d)34, and the Southern Indian Ocean off the Kerguelen Archipelago where steepening isohalines and isopycnals decreased stratification and allowed for a resumption of deep-water ventilation7. Other records of deep- and intermediate water [CO32−]39,52, and radiocarbon8 from the SW Pacific, lack sufficient resolution across the YD. Our study provides more data points and significantly improved chronological constraints during this time interval13.

Thus, while the main part of the atmospheric CO2-increase during the YD is thought to be due to the thawing permafrost soils on the northern hemisphere10,11, our data can now point to southern hemispheric contribution of CO2 at this time via outgassing from the Southern Ocean (Fig. 5b). During the Holocene AAIW records RR0503-7939 and PS75/104-1 begin to diverge (Fig. 3). The youngest values of both records agree with modern [CO32−] data39,58, likely the result of two different sources and pathways of AAIW at the core sites; Tasman AAIW at the Bay of Plenty site, and SO AAIW at our Chatham Rise site, as defined by Bostock et al.12.

During the last deglacial, the history of AAIW [CO32−] and δ13C in our record closely trace Southern Ocean upwelling rates from opal flux50, as well as atmospheric δ13C3 and CO21 (Fig. 3). Between ~ 18 and 11 ka, changes in the extent of Antarctic sea ice cover59 and the meridional shift of the southern westerly winds54 modulated the upwelling rates of CDW50. In combination with changes in the efficiency of the biological carbon pump53, the increased communication of CO2-rich deep-waters via Southern Ocean upwelling was the main driver of early deglacial atmospheric CO2. Both transient peaks, observed in PS75/104-1 during HS1 and the YD, are indicative of a loss in CO214,27 in the formation area of AAIW. The very close relationship between AAIW [CO32−] and δ13C with atmospheric patterns (Figs. 3 and 5) highlights the Southern Oceans role on the deglacial climate. Deglacial deep-water records of Bay of Plenty [CO32−]39 and δ13C17 (Figs. 3c and 6), and Bounty Trough ΔΔ14C8 (Fig. 4) indicate a progressive change in water mass properties, indicative of circulation induced shifts in water mass mixing and/or loss of respired CO2 through ventilation39.

During time periods with cold northern hemispheric stadial conditions (HS1 or YD), the bipolar seesaw hypothesis60 argues for a reduction in the efficiency of the AMOC. A diminished AMOC results in the reduced export of heat from the southern hemisphere to the northern hemisphere, ultimately triggering a decrease in Antarctic sea ice that contributed to Southern Ocean release of CO2. The warmer northern hemispheric Bølling-Allerød period again resulted in a strengthening of North Atlantic Deep Water formation and the AMOC61. This period roughly correlates to the ACR that saw an increase in Antarctic sea ice, a northward displacement of southern westerly winds and reduced upwelling of CDW50. As our data show, the ACR was also marked by a return to glacial-like [CO32−] and δ13C (Fig. 3) and thus reduced air-sea gas exchange in the Southern Ocean, the area of AAIW formation. The well constrained temporal evolution of AAIW [CO32−] and δ13C throughout the entire deglacial period (Fig. 3) provides important new insight into the key role, the Southern Ocean played in the two-step rise of atmospheric CO2.

Conclusions

Our investigation of foraminifer-based [CO32−] and δ13C records on an intermediate water core off New Zealand highlight the role SW-Pacific AAIW played during the deglacial rise in atmospheric CO2. In conclusion, we propose that:

  1. 1.

    Reconstructed [CO32−] and δ13C trends point to a potential release of respired CO2 through ventilation in the upwelling region of circumpolar deep water.

  2. 2.

    Our findings agree with previous studies from the region that indicated that the mid-depth Pacific acted as a reservoir for CO2 during the last glacial8,39,40,52,62.

  3. 3.

    The observed transient rises in [CO32−] and δ13C during HS1 are consistent with the release of CO2 during this interval. While this interpretation is not unambiguous, it adds to a growing set of studies that indicate a similar process8,39,40,50,52,63.

  4. 4.

    In addition to northern hemisphere sources10,11, the YD rise in atmospheric CO2, might have experienced a contribution of released CO2 from the South Pacific as well.

  5. 5.

    Throughout the Holocene, AAIW δ13C probably experienced an overprint from atmospheric values41.

  6. 6.

    C. dispars can be used for reconstructions, using our new calibration B/Ca = 2.27(Δ[CO32−]) + 152.37

Materials and methods

Sediments and sample treatment

We analyzed sediment core PS75/104-1 that was retrieved during expedition ANTXXVI/2 at S44° 46′ 9.012′′ E174° 31′ 31.8′′ in a water depth of 835 m (AAIW), using a BGR type piston corer. The core was split into an archive and a working half and subsequently sampled. All samples were frozen and freeze dried for 2–3 days. The dried samples were wet-sieved, using a 63 µm mesh sieve and subsequently dried at 50 °C for 2–3 days. The > 63 µm fraction was subdivided into the size fractions < 125 µm, 125–250 µm, 250–315 µm, 315–400 µm, and > 400 µm. Planktic and benthic foraminifers used were picked from the 250–315 and 315–400 µm fractions.

Age control

For the core interval between 0 and 0.18 m, the age model for PS75/104-1 is based on the results of Ronge et al.8. Below 0.18 m, we used the new age model provided by Küssner et al.13, based on the highly accurate plateau tuning technique and 56 planktic 14C ages. Thus, we have a very reliable age control on the deglacial interval discussed in this study (7.4–17.8 ka). Throughout this time interval, sedimentation rates range between 16 and 32 cm/ky13. Some depths of PS75/104-1 might be affected by pronounced Zoophycos burrows that were mapped by densely spaced 14C-samples and X-radiographies13. However, only four of our samples fall into these intervals (Fig. 2). Excluding these would not affect any of our interpretations. Hence, we are highly confident in the integrity of our records.

Stable carbon isotopes

To determine past water mass δ13C values we analyzed monospecific samples (2–4 specimens) of Cibicidoides wuellerstorfi. The measurements were conducted at the Alfred Wegener Institute Helmholtz Center for Polar and Marine Research in Bremerhaven, using Finnigan MAT 253 and 251 spectrometers, coupled to carbonate preparation devices Kiel II and Kiel IV, respectively. Based on an internal laboratory standard (Solnhofen limestone), the long-term precision over one year was better than 0.06 ‰. Isotope ratios are reported as ‰ deviations (δ) from the Vienna PeeDee Belemnite (VPDB) standard.

B/Ca measurements and [CO3 2−] calculation

B/Ca measurements were conducted on the 315–400 µm fraction of specimens of the epibenthic foraminifer species Cibicidoides wuellerstorfi and Cibicidoides dispars64, which showed no sign of alteration or secondary fillings. B/Ca analyses were conducted at the GEOMAR Helmholtz Center for Ocean Research in Kiel, using a Coherent GeoLasPro 193 nm Excimer laser ablation system, coupled to a Nu Instruments AttoM magnetic sector mass spectrometer. LA-ICP-MS is a well-established method for the analysis of foraminiferal calcite65,66,67. The analytical method we used for this study has been proven to be accurate and precise68 (instrument details given in the supplementary information). For each sample 3–6 specimens were analyzed on four 90 µm spots in the three oldest chambers on the umbilical side. Before and after each set of five specimens, the NIST615 standard69 was measured and used for calibration. Before beginning the analyses, each shell as well as the NIST615 standard were pre-ablated to prevent any surface contamination effects. Samples with ratios of Mn/Ca > 0.2 mmol/mol and Al/Ca > 0.4 mmol/mol were discarded from the dataset. For our calculations, we applied the calibration of Yu et al.56: B/Ca = 1.14(Δ[CO32−]) + 176.6 for C. wuellerstorfi. For C. dispars, provide a new calibration B/Ca = 2.27(Δ[CO32−]) + 152.37 (Supplementary text and Figs. 7 and S2). Modern [CO32−]sat of 49.48 µmol/kg was derived from GLODAP v258. [CO32−]sat can be affected by changes in salinity (S), bottom water temperature (BWT), and pressure (P). Based on Yu and Elderfield70, we assume an insignificant impact of glacial-interglacial changes in S, BWT, or P on our downcore [CO32−]. While probably more pronounced in intermediate-waters, than deep-waters, changes in S, BWT, and P are within the uncertainty of the proxy’s calibration39,52.

Figure 7
figure 7

Calibration of core-top B/Ca (1σ-error bars) against pre-industrial water mass Δ[CO32−]6. Dashed lines show the ± 15 µmol/mol uncertainty envelope, calculated according to Yu and Elderfield70.

Full size image