Introduction

The end of the last glacial was marked by several dramatic changes in the climate system. Global temperatures rose by ~ 4 °CRef.1, ice sheets retreated globally2,3, while the concentration of atmospheric CO2 increased by ~ 90 ppm across the last glacial transition4. Several studies imply that changes in the strength and geometry of the Atlantic Meridional Overturning Circulation (AMOC) coincided with transient climate perturbations, punctuating the deglacial warming5. Concomitant with the Northern Hemisphere (NH) cold intervals, Heinrich Stadial 1 (HS1; 17.5–14.7 ka) and the Younger Dryas (YD; 12.9–11.7 ka), the AMOC weakened/shoaled in response to changes in buoyancy forcing5. Simultaneously, the Southern Hemisphere (SH) experienced several—antiphased—modifications in the ocean–atmosphere system6.

Thus far, the formation and dynamics of North Atlantic Deep Water (NADW) and more broadly the AMOC were in the focus of studies reconstructing past changes in the global overturning circulation. Notwithstanding the importance of NADW-formation for the ventilation of the oceans subsurface as a major contributor to the global Thermohaline Circulation (THC), it is equally crucial to understand the evolution of its main counterpart, the Southern Ocean (SO), where bottom- and intermediate-waters are formed and advected to all ocean basins. Anderson et al.7 showed that SO upwelling significantly increased ~ 18,000 years before present (ka) at the beginning of the last glacial termination. Collapsing Antarctic ice sheets2,3 and rising atmospheric CO2Ref.4 accompanied enhanced upwelling south of the Antarctic Polar Front, highlighting the importance of Southern Ocean overturning in propelling the global climate system out of the last glacial8,9,10,11,12. In the Pacific sector of the SO however, our knowledge related to circulation patterns on glacial-interglacial and millennial timescales, remains fragmentary.

Modeling studies suggested that, depending on boundary conditions, changes in the AMOC may directly impact Pacific overturning via atmospheric teleconnections13, raising the following questions: (i) can the bipolar seesaw hypothesis14, based on Atlantic data also explain the reconstructed deglacial changes in the Pacific overturning circulation in general, and the South Pacific overturning circulation (SPOC) specifically? (ii) did the SPOC react to changes in AMOC, or did it evolve independently?

Luo et al.15 analyzed the influence of deep-water circulation on the spatial distribution of 231Pa and 230Th in the Pacific Ocean (Fig. 1). Based on their findings, we aim to test the applicability of 231Pa/230Th as a paleo circulation proxy in the SW-Pacific (Fig. 2), in the context of gaining further insights into the questions outlined above.

Figure 1
figure 1

Holocene 231Pa/230Th from the Pacific (grey dots)15, the SW-Pacific (orange triangles15; red squares, this study), and the W Atlantic (blue diamonds; Table S1). Decreasing 231Pa/230Th with water depth in the W Atlantic (blue) is interpreted as an imprint of AMOC, while in the Pacific (grey) a similar correlation is observed only for the SW region (orang, red).

Full size image
Figure 2
figure 2

Research area. Colored dots—core locations; SAF—Subtropical Front (red line); APF—Antarctic Polar Front (black line)68; Green shading—opal belt. Vertical grey line—section used in Fig. 6. Map created with GeoMapApp 3.6.12 (http://www.geomapapp.org).

Full size image

Generally, Pacific seawater 231Pa/230Th-ratios are modulated by the longer residence time of deep-waters and thus reflect higher boundary scavenging intensity at the continental margins when compared to the Atlantic basin16,17,18. Accordingly, 231Pa/230Th has yet mainly been applied to reconstruct spatio-temporal changes in the MOC regimes of Atlantic and not in the Pacific basins. An active meridional overturning cell induces a general decrease in sedimentary 231Pa/230Th with water-depth as observed in the Atlantic Ocean19,20,21. The reported depth-dependent decrease is a consequence of the difference in particle reactivity between the two radionuclides with the relatively less particle-reactive 231Pa, which is more prone to be advected, while 230Th is preferentially exported and deposited into the underlying sediments, by reversible particle scavenging. Recently Luo et al.15 examined to which extent the manifestation of ocean circulation can be recorded in core top sediments based on a compilation of > 250 231Pa/230Th measurements, covering large swaths of the Pacific Ocean (Fig. 1). The basin wide data-distribution underlines the anticipated predominant influence of particle fluxes and boundary scavenging sedimentary 231Pa/230Th. Yet, regional subdivisions of the data set reveal that in the central gyres and the Southwest Pacific region a discernable influence of SPOC is recorded by sedimentary 231Pa/230Th values. The vertical attenuation in 231Pa/230Th values along water depth in these regions, similar to the Atlantic, is interpreted as indicative of the influence of the overturning circulation15. Building on these findings, we measured 231Pa/230Th downcore profiles based on five sediment cores retrieved from the SW Pacific along a depth transect ranging between 835 and 4339 m back to ~ 30 ka to provide evidence for glacial and deglacial variations in PMOC dynamics.

Results

Here we reconstruct changes in the SPOC from a depth transect of five sediment cores from the New Zealand Margin and the East Pacific Rise (Fig. 2). These sediment cores are bathed by the major Southwest Pacific deep-water masses, including Antarctic Intermediate Water (AAIW) the Upper and Lower Circumpolar Deep Water (UCDW and LCDW) as well as the Antarctic Bottom Water (AABW).

To assess temporal variations in the dynamic of SPOC, we make use of the sedimentary ratio of 231Pa and 230Th, which based on the recent results by Luo et al.15 (Fig. 1), suggest that the Southwest Pacific is an area sensitive to circulation driven changes in 231Pa/230Th, as well as published ΔΔ14C-records10 (Figs. 3, 4, 5). In the open ocean, the residence time of 231Pa is about 10-times higher than of 230ThRef.22. Oceanic circulation thus results in the enhanced advection of 231Pa and hence 231Pa/230Th values below the production ratio (0.093)23. Consequently, if deep-water circulation weakens, sedimentary 231Pa/230Th increases toward the production ratio21,24. As biogenic opal is widely known to decrease the residence-time of 231Pa by preferentially removing it from the water column25, we analyzed the opal contents along with 231Pa/230Th (PANGAEA). The amount of biogenic opal in most of our samples is very low (< 3 wt%) and—more importantly—does not correlate with the pattern of 231Pa/230Th, implying that a significant impact on the scavenging behavior of 231Pa at our core locations remains improbable26. However, we consider the contribution of biogenic opal export production poleward (i.e. upstream) of the core locations on the local 231Pa/230Th signal (supplementary information).

Figure 3
figure 3

Patterns of deep-water 231Pa/230Th. Dashed pink line indicates time interval when influences by the advancing glacial opal belt cannot be ruled out (supplementary information). Filled symbols—samples with parallel 14C measurements. Please note that 14C-Age error bars are sometimes smaller than the symbols used.

Full size image
Figure 4
figure 4

Evolution of glacial-interglacial ocean–atmosphere patterns (A) North Atlantic records (brown) OCE326-GGC5Ref.5 and ODP 1063Ref.28. (B) Southwest Pacific 231Pa/230Th records (this study). The dashed line indicates the 231Pa/230Th production ratio. (C) Atmospheric CO2—red69. (D) Deep-water ΔΔ14CRef.10. (E) Smoothed West Antarctic δ18ORef.42. (F) Scotia Sea Iceberg rafted debris (IBRD)—green3; South Atlantic opal flux—black7. ACR—Antarctic Cold Reversal; HS—Heinrich Stadial; LGM—Last Glacial Maximum.

Full size image
Figure 5
figure 5

Comparison of 231Pa/230Th-data (this study) to published radiocarbon records, measured on the—same sediment cores10,56. Colors as indicated in Fig. 1. Black lines—ΔΔ14C; Please note that some additional 14C-data were added for SO213-82-1 (A) as indicated in Table S2. The dashed lines indicate the 231Pa/230Th production ratio.

Full size image

Reaching back ~ 35,000 years, our dataset displays coherent variations in the sedimentary 231Pa/230Th values. The general trends of the independent 231Pa/230Th-data and ΔΔ14C-records10, with bulges in the records in the glacial and deglacial sections imply a certain level of consistency throughout time and space (Fig. 4). Glacial 231Pa/230Th-ratios are generally higher and more variable than during the Holocene (Fig. 3), as are the ΔΔ14C-records of this region10. The most noticeable feature is a transient increase in 231Pa/230Th values between ~ 20 and 18 ka in cores PS75/104-1, SO213-76-2, and SO213-82-1. With a maximum of 0.12, the signal recorded by SO213-82-1 is particularly salient, as it significantly exceeds the production ratio. After ~ 17 ka, we report a constant decrease in 231Pa/230Th values gradually declining toward Holocene values in all sediment records (Fig. 3).

Discussion

The bathymetric transect of sediment cores presented here comprise records from different water depths and thus, different water masses and circulation regimes. The shallowest record PS75/104-1, is bathed by AAIW recently formed in the Antarctic Polar Zone, an area of high opal production (Fig. 1). This implies that changes in opal export production, upstream of our core location might have a significant impact on the 231Pa/230Th pattern locally27. Thus, we interpret this record as well as the older part of PS75/059-2 (> 18 ka), when the main area of high opal production migrated northwards, reflecting a combination of opal productivity, and (to a lesser extent) ocean circulation (see supplementary information for more details, Fig. S1). In a similar way to PS75/104-1, AABW record SO213-76-2 might have been influenced by waters recently formed in an area of high opal production (Fig. S1). However, as this core location is also under the influence of an admixture of LCDW, we assume that this effect is less severe than in our AAIW record. Mid-depth records SO213-82-1 and PS75/100-4 are bathed by southbound CDW/PDW (Circumpolar/Pacific Deep Water). It is expected that the concentration of 231Pa builds up relatively to 230Th along increasing travel time as a function of PMOC strength. Thus, changes in the upstream export of 231Pa are able to exert a dominant influence on the SW-Pacific 231Pa/230Th ratios as shown by Luo et al.15. Holocene ratios of all sediment cores (Fig. 3) fall within the modern budget of the SW-Pacific17, ranging between ~ 0.06 and 0.08.

During the last glacial, the deep-water cores indicate low ΔΔ14C-values10 and 231Pa/230Th-values close to or even higher than the production ratio. In general, we observe generally higher 231Pa/230Th-ratios during the glacial, compared to the Holocene, indicative of either weaker SPOC, higher glacial particle fluxes or most likely a combination of both (Fig. 1)15.

The deglacial trends of the mid-depth 231Pa/230Th-profiles (SO213-82-1 and PS75/100-4) are reminiscent of the North Atlantic 231Pa/230Th-records from the Bermuda Rise5,28. The Southwest 231Pa/230Th patterns are in good agreement with other studies reconstructing circulation and ventilation in the Pacific10,29,30,31,32, Drake Passage33, South Atlantic8, and South Indian Ocean34.

The comparison of 231Pa/230Th from the CDW cores (SO213-82-1 and PS75/100-4) and Atlantic 231Pa/230Th, reveals certain similarities but also striking differences on millennial-timescales. The Bermuda Rise records5,28 show prominent phases of an increased 231Pa/230Th during HS2, HS1 and YD, interrupting the generally constant Glacial and Holocene 231Pa/230Th-baseline (Fig. 4A). Our Southwest Pacific mid-depth records however, show an evolution from more variable, high glacial to gradually decreasing lower Holocene values (Fig. 4B), interrupted by a transient event of elevated 231Pa/230Th-values (~ 20–14 ka). In order to investigate if this general glacial to Holocene trend reflects a glacial reduction in upstream removal of 231Pa, we turned to several sediment records along the Equatorial East Pacific Rise. These records show a consistent trend from lower glacial, to increased Holocene values that approach and exceed the production ratio between ~ 17 ka and 11 ka (Fig. S2)35. It is thus plausible that the decrease in 231Pa/230Th as observed in our records (Fig. 3), represents diminished export of 231Pa from the north after ~ 17 ka. It is plausible to assume that the equatorial records35 were themselves also influenced by upstream processes, which is in good agreement with the processes outlined by Luo et al.15. Nevertheless, it is important to note, that these records are probably not exactly upstream of our core locations but provide the only available approximation of an upstream signal from 231Pa/230Th downcore profiles from the literature to date.

Despite a sufficient temporal resolution, we did not observe any changes during HS2 as manifested in Atlantic sediment cores (Fig. 3). Depending on the boundary conditions, HS2 might not have been associated with any sizeable variation in the SPOC13. An additional feature of our records is the inverse evolution of 231Pa/230Th and ΔΔ14C throughout the end of the last glacial (Figs. 3, 4, 5). At ~ 20 ka, our data show increased 231Pa/230Th-ratios, accompanied by enhanced ventilation (ΔΔ14C)10 (Figs. 4, 5). Appearing contradictory at first, we argue that both patterns are likely the result of increasing water mass advection. Upon SPOC increase, excess 231Pa gets transported via PDW/CDW downstream to our core locations15, increasing the 231Pa/230Th-ratio far above the production ratio (Figs. 4, 6B).

Figure 6
figure 6

South Pacific overturning and ventilation. (A) Last Glacial. Diminished SPOC with separated deep- and intermediate-water cells. SO213-76-2 (green dot) under the influence of LCDW and AABW (B) LGM. Increasing SPOC transports excess 231Pa downstream toward the CDW core-locations. (C) Deglacial Transition. Progressive increase in SPOC results in the release of CO2 and the transport of warm deep-waters onto the shelf regions. Influx of excess 231Pa from the upstream regions diminishes. Disintegration of ice-shelves and release of Iceberg Rafted Debris (black dots). (D) Holocene pattern of the SPOC. Colored dots—sediment records as shown in Fig. 2. Empty circles—no data in this time slice. White lines—assumed isopycnals, the thicker the lines, the stronger the vertical stratification. Black arrows—water mass circulation, thickness of the central arrow indicates relative SPOC strength. Color shading of the CDW-cell indicates its ventilation state according to Ronge et al. (2016). Dark blue—high CO2, low ΔΔ14C.

Full size image

Hence, at this water-depth, increased 231Pa/230Th-values may not reflect a SPOC slow-down15 but rather imply its reinvigoration. This is in good agreement with the ΔΔ14C-records that indicate more active overturning and ventilation during this time period10. Thus, the evolution of both proxies (231Pa/230Th and ΔΔ14C) can be explained with the same mechanism, an increase in SPOC that transported excess 231Pa to our core location, while also leading to an increase in ΔΔ14CRef.10. From ~ 19 ka on, PS75/104-1 (AAIW) recorded lower 231Pa/230Th-ratios, while the mid-depth cores still experienced elevated values (Fig. 3). This short interval was likely caused by increasing opal production in the Antarctic Zone of the Southern Ocean7 that stripped 231Pa from the water before reaching the downstream location. Subsequent to this peak, 231Pa/230Th-values rapidly increased and paralleled the pattern observed in both mid-depth cores (Fig. 3). Following this drastic rise in 231Pa/230Th, we observed decreasing values during HS1 (PS75/104-1; PS75/100-4; SO213-82-1; Figs. 3, 4). With the continuous deglacial SPOC-strengthening, the supply of excess 231PaRef.15 progressively abated, so that 231Pa/230Th-ratios approached modern-like values toward the Holocene.

While the AMOC plummeted, South Pacific 231Pa/230Th- and ΔΔ14C-data (Fig. 4) indicate a progressive evolution of the SPOC toward modern values, starting as early as ~ 20 ka, which culminated during HS1 (Fig. 4). Other records from the Southwest Pacific also corroborate this timing, featuring a significant decrease of εNd as a result of deep-water destratification and enhanced mixing11.

The significantly different patterns of AMOC vs. SPOC bring us back to our initial questions. Did NH changes force the South Pacific via bipolar teleconnections, or were climatic changes on the SH the driving factors?

The declining ventilation and circulation of mid-depth SO waters8,10,29 paralleled peak glacial SH climate such as low temperatures, changes in the density structure of intermediate- and bottom-waters, expanded sea ice, and displaced or weakened Southern Westerly Winds (SWW)36,37,38,39. According to the age models of our sediment cores, the mid-depth SPOC recovered faster from the HS1 disturbance than the AMOC (Fig. 4). Hence, the initial impulse that triggered the increase in deglacial SPOC and upwelling7 must have arisen from the SO or the SH and not via NADW. During this period, glacial climate conditions reversed and gradually exposed the upwelling area of CDW to the surface. The combination of different parameters such as reduced northward heat advection and shifts in the Intertropical Convergence Zone and wind belts increased Southern Ocean upwelling and CO2-release40,41. In addition, local changes in orbital forcing, are considered to be an important factor driving early deglacial changes in the West Antarctic and Pacific sector40. Hence, we argue that the SPOC increase that preceded the end of the LGM (Fig. 6A) was triggered by processes centered in the SH and not by changes in NADW dynamics. In this respect, independent records of Antarctic Ice Sheet retreat3 and the early West Antarctic warming phase40,42 are consistent with the timing observed in our records (Fig. 4).

Deep- to bottom-water sediment core SO213-76-2 (4339 m) however, differs from the deglacial pattern of the other records, as it marks an interval of decreasing 231Pa/230Th from ~ 19 on (Figs. 4B, 5). This interval is in very good agreement to the increasing ΔΔ14C-values measured on the same sediment core10 that also show a similar shift at ~ 20 ka (Fig. 5). Today, SO213-76-2 is influenced by LCDW and AABW29. During the LGM however, the core location of SO213-76-2 was probably only exposed to LCDW. We argue that during peak glacial times, AABW was too dense to be exported to the north of the Pacific Antarctic Ridge, in a similar manner as it was observed in the Atlantic sector of the SO43. In the Pacific sector, processes associated with the early Southern Hemisphere warming42 began to erode the deep stratification at ~ 20ka11. This erosion allowed AABW to reach the core location and thus reduce the influence of excess CDW 231Pa on SO212-76-2.

With our new data, we might also be able to add to a debate of Pacific studies that argued for or against a glacial reduction in Pacific overturning11,44,45,46,47. Our results are consistent with the notion of enhanced glacial carbon storage within mid-depth South Pacific deep-waters10,45. In combination with more complete surface nutrient consumption36,48 the reported slowdown in glacial SPOC might account for the sequestration of carbon in the deep sea along with a progressive decay of its 14C-content. However, to confine the sequestered carbon into the deeper ocean, climatic conditions must have changed in parallel. The stratification of the Southern Ocean’s water column was intensified by an increased glacial density gradient11,37,39, Antarctic sea ice noticeably expanded to the north49, while the SWW were displaced toward the north38. Ultimately, all factors significantly hampered the exchange of deep-waters and the atmosphere. In combination, stratification, expanding sea ice, and displaced winds reduced the upwelling of Circumpolar Deep Waters (CDW) during the glacial, thus are likely the driving parameters behind the observed slowdown of the SPOC. This glacial slowdown extended the residence time of deep-waters within the ocean’s interior by several thousand years8,10,33,34, increased 231Pa/230Th-ratios due to boundary scavenging15, and ultimately allowed for the progressive accumulation of carbon (CO2) in the glacial ocean.

At the final phase of the LGM, we argue that South Pacific overturning progressively increased (Fig. 6). Possibly supported by atmospheric teleconnections, the increase in SPOC sparked the re-ventilation of Southwest Pacific deep-waters, leading North Atlantic processes by almost 2000 years (Fig. 4). This mechanism transported excess 231Pa toward the South Pacific, resulting in the observed transient increase in 231Pa/230Th. Ultimately, this process is probably linked to the upwelling of carbon-rich deep-water and culminated in the transfer of old, 14C-depleted CO2 to the atmosphere (Fig. 6C). In addition, the re-invigoration of the SPOC is also a likely process, which carried warm CDW onto the shelf regions, fostered the early retreat of Antarctic ice shelves2,3 (Figs. 4B,F, 6C). Hence, our reconstructions of South Pacific overturning present a physical link between increasing deep-water ΔΔ14CRefs.8,10,33,34, declining atmospheric Δ14CRef.50, rising atmospheric CO2-levels4, retreating Antarctic ice shelves3, and rising global sea level. However, as mentioned above, several reconstructions of benthic δ13C values and sortable silt44 and foraminiferal εNdRef.47 argue against a pronounced decrease in glacial deep Pacific overturning. To a large extent, these studies cover water masses deeper than 3000 m. As our main signal was observed in a water depth of 2066 m, that is also in the range of the so-called floating glacial carbon pool10,51, this discrepancy might be explained by the different hydrographic sections sampled. However, to obtain a more comprehensive overview in past SPOC changes and 231Pa/230Th budgets, we stress the need for more Pacific downcore records.

Conclusions

Our analysis of downcore 231Pa/230Th-records based on SW-Pacific sediment cores allowed us to investigate the applicability of this proxy with respect to Pacific core top 231Pa/230Th-data15. In addition, when interpreted as a circulation signal, our new 231Pa/230Th data set sheds light on the impact changes in the SPOC had on the glacial Pacific carbon pool, and it’s deglacial erosion, and the release of CO2. In particular, we conclude that:

  1. 1.

    Following the transient maximum of excess 231Pa from the north, all records show a decrease toward Holocene values.

  2. 2.

    Mid-depth Holocene values are considerably lower than glacial values, and are indicative of a strengthened SPOC, compared to the glacial.

  3. 3.

    The 231Pa/230Th-proxy can be a suitable tool for the investigation of paleo-circulation in the SW-Pacific, if downstream transport of 231Pa is accounted for.

  4. 4.

    The hypothetical glacial SPOC reduction would have resulted in a relative increase in 231Pa/230Th-values in the SW-Pacific off New Zealand, and probably along the EPR compared to the Holocene.

  5. 5.

    With the SPOC reinvigoration at ~ 19 ka, all cores display a significant departure from glacial 231Pa/230Th-values toward higher levels.

    1. a.

      As shown in the study by Luo et al.15, SW-Pacific 231Pa/230Th is potentially sensitive to changes in Pacific overturning circulation.

    2. b.

      Based on this sensitivity, we argue that the deglacial onset of the SPOC presumably transported excess 231Pa from the north Pacific to our core locations via PDW/CDW, resulting in values exceeding the production ratio.

  6. 6.

    After ~ 17 ka, upstream records along the Equatorial EPR show a progressive removal of 231Pa. This upstream removal likely put an end to the transport of excess 231Pa to our cores, causing the progressive decrease in 231Pa/230Th as observed by us following this time interval.

  7. 7.

    The transient rise in 231Pa/230Th, during HS, ending earlier in the Pacific than in Atlantic records, is likely indicative of the transport of 14C-depleted and CO2-rich waters from the floating Pacific carbon pool10 toward the surface.

  8. 8.

    Ultimately, this process is a likely common driver between the rise of atmospheric CO2Ref.4, it’s drop in Δ14CRef.50, and the observed collapse of Antarctic glaciers3.

  9. 9.

    Changes in the SPOC did not simply react to AMOC changes via the bipolar seesaw, but were probably triggered by Southern Hemisphere changes in orbital forcing, shifting atmospheric systems, and CO2-release40,41.

  10. 10.

    Our findings highlight the need for additional downcore records from the South Pacific to obtain a more reliable 231Pa/230Th as well as better insight into past changes in SPOC.

Methods

Sediments and sample treatment

The sediment cores analyzed in our study were recovered during R/V Polarstern expedition ANT-XXVI/2 (PS75) in 2010 and R/V Sonne expedition SO213/2 in 2011 (Fig. 2). The sediments predominantly consist of foraminifer- and nannofossil-bearing muds with negligible quantities of biogenic opal52,53. Sedimentation rates vary between 3.5 and 22.5 cm ka−1Ref.10. Due to topographic constraints of the Chatham Rise and the East Pacific Rise, the core locations are not affected by changes in the position of the Subtropical and/or Subantarctic fronts54,55. The sediment cores form a bathymetric transect that covers the major Southwest Pacific water masses including Antarctic Intermediate Water (PS75/104-1, 835 m), Upper (SO213-82-1, 2066 m; PS75/100-4, 2489 m) and Lower Circumpolar Deep Water (PS75/059-2, 3616 m; SO213-76-2, 4339 m). For the geochemical analyses, we used three to four cubic centimeters of bulk sediment from the working halves, except for core PS75/100-4, where we had to sample the archive halves, as no material was left in the working halves. Unfortunately, PS75/100-4 was so intensively sampled in the deglacial interval that literally no material was left for our analyses of this time period.

Radiocarbon dating and age models

To better constrain the distinct HS1 pattern of SO213-82-1, we expanded the radiocarbon record of Ronge et al.10. Briefly, monospecific planktic foraminifera Globigerina bulloides and mixed benthic species (Cibicidoides spp. and Uvigerina spp), were separated and analyzed at the Alfred-Wegener-Institute’s MICADAS facility. Following the procedure previously outlined in Ronge et al.10, we added six data-points into the stratigraphy of core SO213-82-1 and updated the stratigraphy of PS75/104-1 according to Küssner et al.56.

However, as the stratigraphy of SO213-82-1 in Ronge et al.10 is based on a 14C-independent method, via the correlation to the EDC ice core record, we updated these age models using a 14C-related approach. To convert 14C-ages into calendar ages, we used the Calib 7.1 solution57 along with the IntCal13 calibration curve50 as well as the New Zealand Margin surface reservoir ages of Skinner et al.45. However, as the surface reservoir age reconstruction of Skinner et al.45 is insufficient to calibrate all our data points, we also used modelled 14C-ages58. A direct comparison of modelled58 and reconstructed45 surface reservoir ages reveals an offset between both methods, ranging from only 20 to about 300 years. To account for the ages of data points without a direct 14C-age, we applied two independent Bayesian approaches (see below). Using both methods, we are confident that our age models provide the necessary resolution to discern and characterize millennial-scale changes in deep ocean circulation.

Two samples with prominent planktic age reversals (46–47 cm and 86–87 cm) were ignored for our age models.

Updating the age models resulted only in minor changes in the radiocarbon records provided by Ronge et al.10. A comparison of these records to our updated age models is shown in Figure S2. The inclusion of new ΔΔ14C data points furthermore improves the agreement of SO213-82-1 with other CDW records from the Pacific (MD97-2121)45 and the Atlantic (MD07-3076)8.

Bayesian age modelling

We applied the Bayesian age-depth model hummingage, which is developed at AWI and freely available at https://github.com/hummingbird-dev/hummingage. The GitHub repository provides source code for the R programming language and a Jupyter Notebook containing detailed explanations. Additionally, an easy-to-use web service is provided at https://hummingage.awi.de/ for applying hummingage online in the browser.

The age-depth model hummingage can be an alternative to the widely used Bacon59 method. In addition to hummingage we also applied Bacon to our data and show the comparison of both approaches in the supplements (supplementary information; Fig. S4).

Geochemistry

231Pa/230Th measurements

The concentrations of sedimentary 231Pa, 238U, 230Th, and 232Th were jointly measured by isotope dilution in a co-operation between AWI and Heidelberg University with contributions from the GEOMAR Kiel and using mass spectrometers at the AWI (Element 2), Heidelberg (Element 2, Neptune, iCap) and Kiel (Neptune).

The chemical separation and cleaning followed standard procedures60. The 233 Pa spike, milked from 237NpRef.61 was calibrated against the reference standard material UREM-11.

The ingrowth of excess 231Pa and 230Th, generated by the decay of 235U and 234U in the overlying water column (xs) at the time of deposition (Paxs,0 and Thxs,0, where 0 indicates the decay corrected values), has been calculated from the measured bulk concentrations22. Lithogenic contributions were corrected by applying a detrital 238U/232Th activity ratio of 0.5 based on a basin-wide average lithogenic background as suggested by Henderson and Anderson22 or Bourne et al.62. As well, a disequilibrium of 4% for 234U/238U in the lithogenic fraction has been considered to account for preferential 234U loss via the recoil-effect62.

We excluded samples SO213-76-2 200 cm and 314 cm from our interpretations. These samples contain high concentrations of volcanic glass that may have interfered with our measurements. Nevertheless, all measurements, including SO213-76-2 200 cm and 314 cm, are included in the PANGAEA datafiles (https://doi.pangaea.de/10.1594/PANGAEA.889934).

Opal measurements

As the presence of biogenic opal may affect the ratio of 231Pa and 230Th, by preferentially scavenging 231PaRef.25, we determined the sedimentary biogenic opal content of our samples following the method of Müller and Schneider63. For the analysis, we leached 20–100 mg of bulk sediment, using 100 ml of 1.0 M NaOH at 85 °C. The alkaline solution was then injected into the analyzer. In the analyzer, the leachate was treated with 0.088 M H2SO4, a molybdate reagent, oxalic acid and an ascorbic acid in order to form molybdate-blue complexes, which were analyzed using a photometer for dissolved silica. For most sediment cores, the average SiO2 content was well below 2%. Since these low biogenic opal concentrations would only marginally affect the sedimentary 231Pa/230Th-ratios26,64,65, and since the signals of both metrics did not covary, we dismiss any significant impact of biogenic opal in driving downcore changes in 231Pa/230Th. Rather, we interpret changes in sedimentary 231Pa/230Th as primarily reflecting temporal changes in overturning circulation.

Biogenic barium

Biogenic barium (BaBio) is considered to be a reliable proxy for paleoproductivity66. BaBio was calculated based on the difference of total Ba and lithogenic Ba (Balith), measured on our bulk samples via ICP-MS. To calculate Balith from 232Th (measured by ICP-MS via isotope dilution), we followed the method of Costa et al.67: Balith = 51.4 * 232Th.

To assess the role, vertical mass fluxes and paleoproductivity might have played on the 231Pa/230Th-ratios in the Southwest Pacific, we compared both, Thxs0 (Fig. S5A) and BaBio (Fig. S5B) to our 231Pa/230Th-records. Due to the weak correlation of these proxies, we are confident to interpret 231Pa/230Th in terms of paleocirculation (Figure S6).