Radiocarbon evidence for the stability of polar ocean overturning during the Holocene

Proxy-based studies have linked the pre-industrial atmospheric pCO2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p_{{\rm{CO}}_{2}}$$\end{document} rise of ∼20 ppmv in the mid- to late Holocene to an inferred increase in the Southern Ocean overturning and associated biogeochemical changes. However, the history of polar ocean overturning and ventilation through the Holocene remains poorly constrained, leaving important gaps in the assessment of the feedbacks between changes in ocean circulation and the carbon cycle in a warm climate state. The deep-ocean radiocarbon content, which provides a measure of ventilation, responds to circulation changes on centennial to millennial time scales. Here we present absolutely dated deep-sea coral radiocarbon records from the Drake Passage, between South America and Antarctica, and Reykjanes Ridge, south of Iceland, over the Holocene. Our data suggest that ventilation in the Antarctic circumpolar waters and North Atlantic Deep Water is surprisingly invariant within proxy uncertainties at our sampling resolution. Our findings indicate that long-term, large-scale polar ocean overturning has not been disturbed to a level resolvable by radiocarbon and is probably not responsible for the millennial atmosphere pCO2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p_{{\rm{CO}}_{2}}$$\end{document} evolution through the Holocene. Instead, continuous nutrient and carbon redistribution within the water column following deglaciation, as well as changes in land organic carbon stock, might have regulated atmospheric CO2 budget during this period. Overturning circulation that mixes surface and deep water was invariant over the Holocene, suggesting a limited role in rising CO2 during this time, according to deep-sea coral radiocarbon records.

Article https://doi.org/10.1038/s41561-023-01214-2 approach of reconstructing 14 C evolution at different sites across the Drake Passage and Reykjanes Ridge, south of Iceland, through coupled U-series dating and 14 C analysis of deep-sea corals. The deep-sea corals of this study were recovered from Burdwood Bank, Cape Horn, Sars Seamount and the Shackleton Fracture Zone in the Drake Passage at a depth of ~0.3-1.9 km and from the Reykjanes Ridge at a depth of ~1.3-1.4 km ( Fig. 1 and Methods).
The reconstructed Δ 14 C (defined as (F 14 C × e (calender age/8,267) − 1) × 1,000, where F 14 C is the fraction modern of the sample with blank and δ 13 C correction) of different water masses in our records shows a gradual decrease over the past 10,000 years or so, following the trend of atmospheric radiocarbon reconstructed by IntCal20 (ref. 18), particularly for the samples from the Reykjanes Ridge (Fig. 2). These data also indicate distinctive Δ 14 C values between different sites, with better 14 C ventilated signatures at shallower and northern sites of the Drake Passage. Intriguingly, the shallow depths at Sars Seamount show distinctively enriched 14 C signatures at ~9.6 ka, which are reproducible from two different samples from the same sampling site. To interpret radiocarbon in terms of overturning changes, it is necessary to account for the contribution of changing atmospheric Δ 14 C on the initial 14 C content of deep waters at the time of their formation, as well as the atmospheric p CO 2 effect that impacts the air-sea carbon isotope exchange efficiency 19 . We therefore projected our 14 C data to the Marine20 calibration curve 20 (t proj ) as a proxy of deep-water ventilation age (Methods). Note that Marine20 represents the average 14 C concentration of the In the North Atlantic, the history of Holocene overturning has been inferred from multiple approaches, such as reconstructions on bottom water flow speed 8,9 , deep-water transport flux 10,11 and mid-to high-latitude salinity/temperature anomalies [12][13][14] . Nevertheless, existing reconstructions exhibit quite divergent trends, which might partly result from the fact that the Atlantic Meridional Overturning Circulation (AMOC) has different dynamic regimens and pathways 15 . In the Antarctic circumpolar system, detailed overturning and ventilation reconstructions over the Holocene have been hampered by the lack of suitable sedimentary archives. Moreover, consensus has not been reached on the position of the Southern Hemisphere westerlies 2,3,5,16 , which impacts the Southern Ocean overturning and ventilation. The relationship between atmospheric p CO 2 and polar ocean overturning during the Holocene thus remains elusive.

Radiocarbon evolution reconstructed from deep-sea corals
Radiocarbon has been widely used as a ventilation and overturning proxy because it is dissolved in the surface ocean through air-sea exchange and is introduced into depths by deep-water formation 17 . As 14 C is radioactive, precise and accurate calendar ages are crucial for reliable reconstruction of 14  non-polar surface waters (between 40° S and 50° N in the Atlantic or 40° N in the Pacific). t proj in this case provides a measure of the time lag for the entrainment of surface waters from the non-polar surface region to the polar deep waters. Our calculated t proj of the Drake Passage shows that ventilation ages have a rather small variability in the well-resolved records, with the exception of those reconstructed at Sars Seamount over the Holocene (Methods). The sites with only a few coral samples across the Holocene (Burdwood Bank, Cape Horn and the Shackleton Fracture Zone) also show limited variability of t proj , which fits well into the modern 14 C ventilation structure in the Southern Ocean 21 . t proj at Sars Seamount indicates that ventilation was better during the early Holocene (10.5-7.5 ka) than the later Holocene. Strongly ventilated signatures at ~9.6 ka have been recorded by corals at a water depth of 692 m, with radiocarbon contents as high as those typically only observed in thermocline waters at Burdwood Bank and Cape Horn. The ventilation age at Sars Seamount of ~9.6 ka appears to be even younger than that of the northern site (Burdwood Bank) at similar depths during the early Holocene (Fig. 3). During ~6.8-1.6 ka, available data from a water depth of 647-981 m at Sars Seamount show that t proj was fairly stable at 643 ± 103 years (n = 6; 2σ; Fig. 3). In general, the standard deviations of t proj at each site are close to or smaller than the 2σ uncertainties of Marine20 (~110-150 years in the Holocene). Similarly, records from the Reykjanes Ridge site exhibit essentially invariant ventilation ages (−20 ± 82 years; n = 17; 2σ).

Impact of climate forcing on polar ocean overturning
Previous reconstructions of radiocarbon from the near-shore Northeast Atlantic indicate well-ventilated mid-depth waters during the Holocene similar to the pre-bomb modern time 22,23 (Fig. 3). However, those records exhibit larger short-term variability compared with our samples from Reykjanes Ridge, which sits in the centre of the deep subpolar gyre. The present-day deep waters recorded by Reykjanes Ridge corals at water depths of ~1.3-1.4 km were primarily formed via the Labrador Sea Water (LSW) convection with mixing of Northeast Atlantic Deep Water 24 . Northeast Atlantic Deep Water itself reflects mixing between LSW, Iceland-Scotland Overflow Water and modified Antarctic bottom waters 25 . Therefore, our studied site at Reykjanes Ridge represents an ideal location for recording average North Atlantic Deep Water (NADW) signatures. Compensating feedbacks might have existed in the process of NADW formation through subpolar gyre modulation on the intensity of LSW convection and Iceland-Scotland Overflow Water 13,26 . For example, during weak LSW convection in response to enhanced melt water flux, the mixing of such fresh components into the polar North Atlantic might be decreased, thus enhancing deep-water formation in the Norwegian Sea 13,26 . Our North Atlantic deep-sea coral Δ 14 C evolution is essentially identical to that of Marine20, suggesting that deep waters overflowing the Reykjanes Ridge had not experienced any prolonged time of isolation (that is, multi-centennial) after sinking from the surface at our sampled resolution. Notably, the difference in the average t proj between 9.7-7.9 and 5.8-2.2 ka is only 5 ± 82 years (n = 10; 2σ). This exceptionally constant t proj suggests that the long-term strength of North Atlantic overturning circulation was invariant during the Holocene. These results are also consistent with independent proxy studies from high-resolution 231 Pa/ 230 Th measurements of North Atlantic detrital sediment cores 10,11 through this time period (Fig. 3c).
In the Southern Ocean, the transport of deep waters towards the surface is related to the upwelling of Upper Circumpolar Deep Water (UCDW) and the lower overturning cell linked to Antarctic Bottom Water upwelling 24 . At shallower depths, the upwelled UCDW is ventilated through air-sea gas exchange, mixes with well-ventilated subtropical waters and transforms into intermediate and mode waters 24 . The Drake Passage is a unique geographical location of the Southern Ocean where the meridional extent of the Antarctic Circumpolar Current (ACC) converges between South America and the Antarctic Peninsula. What then caused the distinct changes in t proj at Sars Seamount but not the other sites during the early Holocene? Sars Seamount is located close to the modern polar front that approximately divides the downwelling of well-ventilated surface waters and the upwelling of relatively poorly ventilated UCDW in the upper ocean 24 (Fig. 4a). This unique location makes Sars Seamount more sensitive to shifts in the polar front and conditions in the Antarctic zone compared with the other locations. We suggest that better ventilation at Sars Seamount is consistent with a more poleward position of the polar front during the early Holocene compared with the later period. In this case, downwelling of the upper ocean waters would occur at more southern latitudes along with poleward polar front migration, thus leading to enhanced ventilation at Sars Seamount (Fig. 4). Indeed, recent studies based on radiolarian and diatom assemblages from the Indian sector of the Southern Ocean 27,28 have indicated poleward polar front migration by a few degrees during the early Holocene.
However, strong ventilation recorded by Sars Seamount during ~9.6 ka is even better than the northern sites at similar depths, requiring increased mixing of 14 C-enriched water from the south as well. Higher ice-rafted debris flux from the Antarctic to the Scotia Sea ( Fig. 3d) during the early Holocene may reflect higher fresh water input to the Antarctic Zone. These (sub)millennial forcings could potentially result in a transient response of the Southern Ocean ventilation and contribute to the higher radiocarbon content observed at Sars Seamount. For example, transient input of western Antarctic ice-sheet melt associated with rapid deglaciation at ~9.6 ka (ref. 29) could result in strong stratification and thus a longer residence time and enhanced air-sea gas exchange of the Antarctic surface waters, which could then be mixed on isopycnals down to the depths of the Sars Seamount site  Fig. 1. ±2σ error ellipses are also shown. IntCal20 represents the Northern Hemisphere atmospheric 14 C age calibration curve 18 , whereas Marine20 represents the non-polar marine 14 C age calibration curve 20 . The shadings of Marine20 represent ±2σ uncertainty.
Article https://doi.org/10.1038/s41561-023-01214-2 ( Fig. 4c). Nevertheless, the near-constant ventilation gradient between records from the North Atlantic and the Southern Ocean, as well as the lack of any long-term trend in t proj at each site except Sars Seamount, provide strong support that Southern Ocean overturning remained stable without prolonged, large-scale disturbance. Likewise, reported coral Nd isotope data 30 also show limited variability at each site of the Drake Passage through the Holocene, with the exception of two data points in the middle Holocene. These two radiogenic isotope signatures recorded in Sars Seamount corals (869 m) could be related to zonal mixing of more radiogenic waters from the Pacific rather than changes in large-scale meridional overturning. Given that the t proj of the deep UCDW, as represented by the deepest samples from Burdwood Bank (Fig. 1), is 870 ± 67 years (n = 10; 2σ; that is, <10% variability), the variability of the mixing proportion between well-ventilated North Atlantic and poorly ventilated Pacific endmembers should also have been similarly small during the Holocene. Our study could not exclude the possibility of strong short-term AMOC slowdown following major melt water pulses during the Holocene (for example, the 8.2 ka event 31 and other interglacials 32 ). Given the rapid sea-level rise and thus ice-sheet decay in the early Holocene, our study supports assertions derived from a conceptual framework, which suggests that AMOC reaches a stable strong mode once atmospheric CO 2 levels approach pre-industrial levels (that is, regardless of the changes in Northern Hemisphere ice-sheet volume) 33,34 .

Decoupling between circulation and biogeochemical cycles
The stability of the millennial polar ocean overturning leads us to suggest that the long-term Holocene p CO 2 evolution was not the result of changing ocean overturning circulation. In particular, during the major phase of rising atmospheric p CO 2 (for example, 7-2 ka), none of our deep records show any sign of ventilation changes (Fig. 3), suggesting that overturning in the North Atlantic and Southern Ocean did not drive changes in oceanic carbon release during this period. While our study does not provide additional constraints on oceanic biogeochemical cycles, proxies suggest that carbon and nutrient distribution within the water column was not in a steady state over the Holocene. For example, deep oceanic [CO 3 2− ] content shows a prominent decrease in the early Holocene (Fig. 3b), indicating ocean alkalinity removal and the release of carbon to the atmosphere 35 . Foraminiferal boron isotopes also suggest a high Δp CO 2 relative to the contemporaneous atmosphere in surface waters of the Subantarctic Zone and Eastern Equatorial Pacific in the early Holocene, potentially as a result of carbon release via continuous Circumpolar Deep Water upwelling and intermediate water advection 36 . In contrast, nitrogen isotope (for example, Fig. 3e) and productivity records from the northern Antarctic Zone indicate a gradually decreased nutrient utilization rate and increased nutrient supply towards the late Holocene 6 , inferred to be associated with obliquity-driven enhanced westerlies and Southern Ocean overturning 37 . We surmise that the δ 15 N decrease in diatom or coral-bound organic matter observed in the northern Antarctic Zone is probably the result of nutrient redistribution within the ocean basins following deglaciation, rather than increased physical overturning. It has long been hypothesized and modelled that the redistribution of major denitrification locations (for example, between the continental shelf sediments and the deep sea) could affect global biogeochemical cycles 38 . The sites of denitrification and corresponding N 2 fixation could change as a result of sea-level rise 39 and oxygen availability 40 in the upper ocean, which are not directly dependent on meridional overturning circulation. For example, a decrease in bulk sediment δ 15 N values of the eastern tropical Pacific over the past 10,000 years may partly reflect decreased denitrification in response to a weakened oxygen minimum zone 41 . In turn, decreased denitrification in the eastern Pacific water column might effectively increase the nitrate concentration and decrease the δ 15   Ocean. Other factors, such as changes in the efficiency of nutrient recycling in the upper water column (for example, associated with the depth of organic particle remineralization or ecological community structure) also warrant investigation 42 , which is beyond the scope of our study. These nutrient redistribution processes do not require large changes in oceanic N inventory over the Holocene. Future work should target records from more southern latitudes of the Antarctic Zone to quantitatively understand the contribution of these processes to the Holocene biogeochemical cycles. Other than changes in oceanic biogeochemical cycles, variability in land carbon inventory is also probably important, as supported by transient carbon cycle modelling 43 , as well as archaeological and ecological evidence (for example, refs. 44,45) during the Holocene. Closing the Holocene atmosphere carbon budget would require a better understanding of natural and human-associated changes in terrestrial organic carbon stock 46,47 . In essence, if orbital parameters alone are sufficient to predict CO 2 evolution during the Holocene, we would expect the atmosphere p CO 2 to decrease continuously like its closest analogue, Marine Isotope Stage 19c (ref. 48). While such a trend is not observed, the exact mechanism of Holocene atmosphere p CO 2 evolution remains an open question. In any case, our deep-sea coral 14 C data based on precise, absolute U-series ages provide tight constraints on the stability of the polar ocean overturning and thus demonstrate a clear decoupling between physical ocean circulation and atmospheric p CO 2 in the Earth's most recent interglacial period.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41561-023-01214-2.

Samples and site descriptions
The Drake Passage samples were dredged from a number of sites (Burdwood Bank, Cape Horn, Sars Seamount and the Shackleton Fracture Zone) during RVIB Nathaniel B. Palmer Cruises 0805 and 1103 in 2008 and 2011, respectively. Sixteen Holocene samples have been reported previously in discussions of the deglacial Southern Ocean ventilation 52 . In this Article, we provide a spatially and temporally detailed 14 C ventilation picture of the Holocene by presenting 61 new Drake Passage data points. Our study thus avoids ambiguities in assessing ventilation changes caused by linking deep-sea coral samples from different depths and sites. These Holocene samples were selected based on reconnaissance dating 53,54 of more than 1,500 samples and the preservation of coral skeletons. The species of the samples are mostly solitary coral Desmophyllum with a few other genera such as Flabellum, Balanophyllia and Caryophyllia.
A series of frontal zones where the ACC is most enhanced are confined within the Drake Passage 24 . In the upper water column of the Drake Passage (water depth <2.0 km), the most prominent water mass is the UCDW, which is identified by its low oxygen content (<180 μmol kg −1 ) (ref. 24), with a Δ 14 C of about −140 to −160‰. Despite large Δ 14 C differences between the Pacific and Atlantic on the same isopycnal surfaces in the deep waters, the horizontal and vertical mixings have effectively homogenized 14 C once these waters have been steered into the ACC 21 . Note that our coral samples did not directly record Antarctic Bottom Water signatures. However, unlike conservative tracers used to study ocean circulation, 14 C is particularly sensitive to the aging of the deep waters due to radioactive decay. The upper North Atlantic overturning cell and lower Antarctic bottom water overturning cell are intertwined through the circumpolar deep waters. Today, the UCDW originates from mixing between deep waters from the Pacific, Indian and Atlantic oceans. Overlying the UCDW is the downwelling Antarctic Intermediate Water, which has lower salinity and a higher oxygen concentration, while the overlying mode water is characterized by its low nutrient concentration, such as silicon (for example, <10 μmol kg −1 ) (ref. 55). When the overturning rate in the Antarctic bottom cell slows down, the Pacific and Indian deep waters become aged and the 14 C signatures of UCDW and Antarctic Intermediate Water would change accordingly. Similarly, decreased transport of well-ventilated North Atlantic deep waters into the circumpolar deep waters would also cause aging of the 14 C recorded by Drake Passage corals. Therefore, 14 C recorded by our samples provides a unique tracer that can be used to monitor the rate of global meridional overturning circulation.
The North Atlantic samples were dredged from the Reykjanes mid-ocean ridge at a depth range of ~1,290-1,429 m during Celtic Explorer cruise CE0806 in 2008. These fossil samples are mostly framework-building corals (Lophelia pertusa and Madrepora oculata). The water mass where these corals were grown is part of the subpolar gyre circulation system, characterized by cyclonic re-circulation steered by the topographic boundaries in the mid-depths 56 . The modern ventilation of waters at our studied depth is mainly via Labrador Sea convection along with potential mixing of intermediate waters of subtropical origin 57 . Pre-bomb Δ 14 C distribution during the late nineteenth and early twentieth centuries of North Atlantic deep waters is helpful to understand the data of our study. The surface Δ 14 C along the northeast Atlantic coast is around −45 ± 5‰ based on shells of known age 58 21 . These waters might be shoaled to the subsurface along the North American continental margin 60 . Coral records suggest that convective mixing in the Labrador Sea can effectively homogenize the pre-bomb Δ 14 C signature (−67 ± 4‰) in the Northwest Atlantic at least to ~1 km depth 60 . As a result, the subsurface waters of the subpolar North Atlantic are well ventilated with relatively homogenous pre-bomb 14 C signatures. Therefore, 14 C in our Holocene samples can be understood as mixing between the 14 C-enriched surface North Atlantic waters and potentially 14 C-depleted intermediate waters of southern origin at periods of decreased AMOC.

U-series analytical methods
U-series dating of the deep-sea corals with the isotope spiking method was performed by the Bristol Isotope Group at the University of Bristol. Chen et al. 61 have described the procedure for column chemistry and mass spectrometry in the U-series analysis of deep-sea corals in detail, which we followed in this study. All samples were physically and chemically cleaned after cutting about 0.2-1.0 g of chunky aragonite from each of the deep-sea coral specimens. Approximately 0.2 g of each sample was dissolved in optima-grade HNO 3 and spiked with ~0.06 g of the 236 U-229 Th mixed spike. To facilitate column chemistry, U and Th in the sample were first co-precipitated with iron hydroxides and then dissolved and loaded to anion-exchange columns to further purify the U and Th fractions. U and Th isotopes were measured using the sample-standard bracketing method on a Neptune multi-collector inductively coupled plasma mass spectrometer. The typical internal precision for 234 U/ 238 U is ~0.7-1‰ and for 229 Th/ 230 Th it is ~0.9-1.5‰, with accuracy better than 1 and 2‰, respectively. To maintain consistency with previously published coral age data, we adopted the decay constants of the U-series nuclides from Cheng et al. 62 . Analytical errors and procedural blanks were propagated analytically into the isotope ratios of 234 U/ 238 U, 236 U/ 238 U and 229 Th/ 230 Th. An additional source of uncertainty for age determination is the initial 230 Th incorporated into the coral skeleton. The initial [ 230 U/ 232 Th] values of the Drake Passage and North Atlantic corals are assumed to be 37 ± 37 (2σ) and 14.8 ± 14.8 (2σ), respectively 51,63 . A Monte Carlo technique was then applied to propagate the errors of isotope ratios into the final age uncertainties.

Radiocarbon analysis and data report
Approximately 15-20 mg of cleaned sample was weighed and leached by 0.1 N HCl to ~10 mg before graphitization in an automated graphitization device. The graphite target was analysed at the new MICADAS Accelerator Mass Spectrometer (AMS) facility at the University of Bristol. Before our sample analysis, the consistency of the coral 14 C data measured by the Bristol AMS facility had been checked with repeated measurements of coral samples that were previously analysed in the AMS laboratory of the University of California, Irvine 64 . The fossil corals with ages older than 100 ka yielded a 14 C age of 46-50 ka and were used as the procedural blank. All data reported in this study have been δ 13 C and blank corrected, as shown in Supplementary Table 1.
There are various ways to estimate the past oceanic 14 C ventilation based on combined U-series dating and 14 C analysis of deep corals. Δ 14 C is a straightforward measure of the actual 14 C content in past seawater, which is calculated as: Δ 14 C coral = (F 14 C × e (calender age/8,267) − 1) × 1,000. The projection age is also a common way to report deep-ocean 14 C data 65,66 . Changes in atmospheric 14 C levels take time to affect the surface ocean, and the mixing between the surface layer and the deeper ocean layers would result in effective signal damping 20 . This effect causes changes in atmospheric 14 C levels to be smoothed and shifted in phase in the surface ocean. When applying 14 C as a deep circulation tracer, it can be assumed that deep-water 14 C is sourced from surface waters. Therefore, we project our data to Marine20 to account for the effect of surface oceanic smoothing of atmospheric 14 C. Marine20 used for projection age calculation was obtained from a large set of simulations with various ocean-atmosphere-biosphere parameterizations of the global carbon cycle 20 . The simulations using the BICYCLE box model 67 were forced by IntCal20 atmosphere 14 C and ice-core p CO 2 data, incorporating effects related to changes in ocean mixing and air-sea gas exchange. Marine20 thus serves as the ideal candidate for initial Δ 14 C of surface source waters that supply the deep and polar oceans. The negative Nature Geoscience Article https://doi.org/10.1038/s41561-023-01214-2 projection age of the sample, in this case, would mean a higher Δ 14 C in the coral record than in the Marine20 curve of the same age, while it would take |t proj | for the sample 14 C decay trajectory to intersect with Marine20. The calculated mean Holocene t proj values are −20 ± 82 (n = 17; 2σ), 159 ± 54 (n = 6; 2σ), 335 ± 121 (n = 6; 2σ), 512 ± 92 (n = 14; 2σ) and 870 ± 67 years (n = 10; 2σ) for Reykjanes Ridge (depth: 1,290-1,429 m), Cape Horn (depth: 450 m), Burdwood Bank (depth: 816 m), Cape Horn (depth: 1,012 m) and Burdwood Bank (depth: 1,879 m), respectively. To avoid attributing modern water masses to past coral locations, we have not grouped samples by modern day water masses as applied in previous deep-sea coral 14 C studies 52 . Typically, a single data point (t proj ) of the deep-sea corals is thought to average seawater signatures over a few decades. The deep-sea corals grow much slower than the warm water corals. For example, a solitary deep-sea coral with a length of several centimetres has a typical life span of about a century 68 , while the sampling and homogenization of the septa of the coral aragonites would average out over a few decades.
We note that the version of the BICYCLE box model 67 used to generate Marine20 contains no circulation changes in the Holocene 20 . If the Holocene atmospheric CO 2 or 14 C variability is in part due to ocean circulation changes, the variability will be implicitly included in Marine20. Nevetheless, what makes our dataset unique is the records from multiple sites both in the polar North Atlantic and the Southern Ocean. A stable meridional overturning circulation during the Holocene is necessary not only to maintain the invariant t proj of each site projected to Marine20, but also to ensure stable t proj offsets of different sites in the Southern Ocean from that of NADW (Fig. 2f). The t proj offsets from Reykjanes Ridge for Cape Horn (depth: 450 m), Burdwood Bank (depth: 816 m), Cape Horn (depth: 1,012 m) and Burdwood Bank (depth: 1,879 m) exhibit small variabilities during the Holocene, with values of 179 ± 98, 355 ± 146, 532 ± 123 and 890 ± 102 years, respectively (2σ errors propagated). These variabilities are similar to the propagated analytical uncertainties from radiocarbon and U-Th measurements, suggesting that ventilation between these sites has remained constant within proxy uncertainties. In this regard, our conclusion on stable Holocene oceanic overturning does not rely on the accuracy of the Marine20 surface 14 C curve.