Deglacial Carbon Escape From the Northern Rim of the Southern Ocean

The Southern Ocean regulates atmospheric CO2 and Earth's climate as a critical region for air‐sea gas exchange, delicately poised between being a CO2 source and sink. Here, we estimate how long a water mass has remained isolated from the atmosphere and utilize 14C/12C ratios (Δ14C) to trace the pathway and escape route of carbon sequestered in the deep ocean through the mixed layer to the atmosphere. The position of our core at the northern margin of the Southern Indian Ocean, tracks latitudinal shifts of the Southern Ocean frontal zones across the deglaciation. Our results suggest an expanded glacial Antarctic region trapped CO2, whereas deglacial expansion of the subantarctic permitted ventilation of the trapped CO2, contributing to a rapid atmospheric CO2 rise. We identify frontal positions as a key factor balancing CO2 outgassing versus sequestration in a region currently responsible for nearly half of global ocean CO2 uptake.


Introduction
Over at least the last ∼800,000 years (Bereiter et al., 2015), rising temperatures have been linked to increasing atmospheric CO 2 concentrations, and global climate feedbacks that flip the switch between glacial and interglacial climate states.Cold climate conditions during glacial maximums are associated with a transfer of CO 2 from the atmosphere into the deep ocean resulting from altered global ocean overturning circulation (Broecker & Barker, 2007;Curry & Oppo, 2005;McManus et al., 2004) and more efficient biological transfer of carbon from the surface to deep ocean (Kohfeld, 2005).The transition to the warmer climate conditions of interglaciations rapidly releases this stored carbon to the surface ocean as ocean circulation returns to modern configuration (Anderson et al., 2009;Broecker & Barker, 2007).There is an abundance of evidence for a respired CO 2 reservoir in the glacial deep ocean (Anderson et al., 2019;Curry & Oppo, 2005;Hoogakker et al., 2018;Skinner et al., 2017) and the timing of the deglacial increase in atmospheric CO 2 is well-constrained (Marcott et al., 2014).However, evidence for the dynamic interactions that transfer carbon stored in the ocean to the atmosphere is lacking.The Indian Ocean sector of the Southern Ocean (Figure 1) has an abundance of bathymetric features that enhance eddy mixing, fostering air sea exchange within the Southern Ocean's Antarctic Circumpolar Current (ACC; Brady et al., 2021;Rintoul, 2018) making it a crucial location for examining millennial scale changes in deep ocean uptake and release of CO 2 .In today's Southern Ocean, atmosphere-ocean CO 2 exchange is associated with seasonal CO 2 outgassing through wind-driven upwelling of old, carbon-rich deep waters (Gray et al., 2018) that have accumulated respired CO 2 as they remained out of contact with the atmosphere on their journey along the global ocean conveyor belt (Talley et al., 2011).The deep-water radiocarbon content (Δ 14 C) of dissolved CO 2 decays as a water mass remains out of contact with the atmosphere.In the southern Indian Ocean, southward-flowing Indian Deep Water (IDW) delivers aged, 14 C-depleted, CO 2 -rich, waters supplying Upper Circumpolar Deep Water (UCDW), which upwells to the north of the Polar Front (PF) and to the south of the Antarctic Subantarctic Front (SAF) in the Polar Frontal Zone (PFZ; Key et al., 2004;Paterne et al., 2019).Wind stress nudges this water northward toward the Subantarctic Zone (SAZ) where it becomes more buoyant through warming and freshening while it ventilates CO 2 to the atmosphere and acquires atmospheric 14 C (Key et al., 2004;Talley et al., 2011).These dynamics create the shallow water masses of the Southern Ocean overturning circulation; Antarctic Intermediate Water (AAIW) and Subantarctic Mode Water (SAMW; Figure 2; Talley et al., 2011), which trace the 14 C balance of carbon sequestration and carbon release as influenced by surface ocean equilibration with the atmosphere and the input of deep, 14 C-depleted waters to the mixed layer of the surface ocean.
Here, we provide a record of changes in the relative water mass age of surface and AAIW waters based on the Δ 14 C of paired planktic (Globigerina inflata; lower mixed layer-dwelling) and mixed benthic foraminifera from a marine sediment core (TT1811-50GGC; 38.344°S, 77.715°E; 1,116 m, Figure 1) collected between Île Amsterdam and Île St. Paul, just north of the northern boundary of the Southern Ocean (defined by the modern Subtropical Front; STF; Talley et al., 2011).This site sits at the depth of modern AAIW (Figure 2) and is ideally positioned to monitor climate-driven shifts in the Southern Ocean frontal system and the latitudinal expansion and contraction of the Southern Ocean for the last ∼40,000 years (Figure 1).

Materials and Methods
Marine sediment core TT1811-50GGC was collected by the R.V. Thomas G. Thompson in late-2018 from 1,116 m water depth along the southeast Indian Ridge between Île Amsterdam and Île St. Paul (38.344°S,77.715°E).Seventy-five sample depths were analyzed for planktic and benthic radiocarbon content (Figure S1 in Supporting  3,166 m;Bharti et al., 2022).Frontal definitions from Belkin and Gordon (1996).
Information S1).The lower mixed layer dwelling planktic foraminifera Globorotalia inflata was selected for radiocarbon analyses as it is persistent species throughout the core and is abundant during the austral spring (King & Howard, 2005).In this region, the spring mixed layer is deep (200-500 m on average) and poorly stratified (Sallée et al., 2006).Mixed benthic for radiocarbon analyses were selected with care taken to exclude benthic species with deep (>2 cm) infaunal depth habitats (Globobulimina, Chilostomella, etc.;McCorkle et al., 1990), porcelaneous textures (Pyrgo; Magana et al., 2010), and agglutinated compositions.Prior to analysis of radiocarbon content at the Keck Carbon Cycle Accelerator Mass Spectrometer Facility at the University of California, Irvine, 10% of each sample was leached using weak HCl to remove authigenic carbonates and graphitized per Santos et al. (2007).
The oxygen isotopic composition (δ 18 O) of the mixed-layer planktic foraminifera species Globigerina bulloides, the thermocline dwelling species Globorotalia truncatulanoides, and the benthic foraminifera Cibicidoides lobatulus were analyzed on a Finnigan MAT-253 Plus Isotope Ratio Mass Spectrometer (IRMS) with a Kiel IV device at the City of New York's Advanced Science Research Center (ASRC) and at the University of Florida Light Stable Isotope Mass Spectrometry Laboratory in the Department of Geological Sciences with a Kiel III carbonate preparation device coupled with a Finnigan-Mat 252 isotope ratio mass spectrometer.Long term precision of repeated δ 18 O measurement of the NBS-19 standard across the measurement period was ±0.05‰ VPDB and ±0.06‰, respectively.
The age model for TT1811-50GGC was constrained by identifying changes persistent in both the oxygen isotope record of G. bulloides and G. truncatulanoides that were coeval with Antarctic temperature and δ 18 O events in the EPICA (European Project for Ice Coring in Antarctica) Dome C (EDC; Buizert et al., 2015;Buizert et al., 2021;Lorius et al., 1979) and Dome Fuji (Fujita et al., 2015;Kawamura et al., 2007;Watanabe et al., 2003) ice cores synchronized to the layer counted West Antarctic Ice Sheet Divide ice core (WD2014; Figure S2 in Supporting Information S1; Buizert et al., 2018).The rapid thermal equilibration of circum-Antarctic surface waters and Antarctic air temperatures is known to be transmitted to the EPICA Dome C ice core through an atmospheric connection.This understanding is the basis for the widespread use of tying δ 18 O and Sea Surface Temperature (SST) records to Antarctic ice cores (i.e., Gottschalk et al., 2020;Rose et al., 2010;Skinner et al., 2010).To address any possible local salinity effect on the planktic δ 18 O, tie-points were verified with co-occurring changes in thermocline (G.truncatulanoides) and benthic (C.lobatulus) δ 18 O (Figure S2 in Supporting Information S1).The slight variations in the timing of events in the δ 18 O records fall within our estimated age model and reservoir age error.Initial reservoir ages were determined by correcting the G. inflata 14 C-ages to calendar ages using the preindustrial reservoir age estimated for the Kerguelen Islands (ΔR = 442 ± 101; Paterne et al., 2019) from the marine reservoir correction database (Reimer & Reimer, 2001) with an error of ±101 added to allow for the documented preindustrial range of 14 C variability (Paterne et al., 2019).Using the initial alignment of the 14 Ccalibrated ages, the match between surface ocean (G.bulloides) and thermocline (G.truncatulanoides) δ 18 O with Holocene temperature and δ 18 O EDC events was reassessed.Surface ocean δ 18 O values that showed an EDC match within the Holocene reservoir age error window were then aligned to the EDC matches.As the early Holocene portion of the core has less variability in Antarctic temperature and δ 18 O, the 14 C-calibrated ages were used for age control (Figure S2; Table S1 in Supporting Information S1).
Primary concerns when analyzing past oceanic radiocarbon variability are the development of a rigorous age model and sedimentary or diagenetic influences on radiocarbon age.These concerns are often related.The presence of ash layers or terrestrial sourced plant material can provide an independent constraint on sediment age (Sikes et al., 2000;Zhao et al., 2018).Distinct thin ash layers and dispersed ash are present in TT1811-50GGC as shallow as 60 cm (Figure S3 in Supporting Information S1), but these ashes have not been linked to a dated terrestrial event and may be submarine in origin.However, the changes in core color associated with the presence of ash and/or changes in deposition provide useful sedimentary contacts that can confirm minimal mixing.Bioturbation and mixing can homogenize sedimentary contacts or result in a mottled rather than sharp basal contact.The presence of sharp basal contacts in 50GGC suggests that mixing and bioturbation was relatively minimal in the deglacial and glacial portions of the core (Figure S3 in Supporting Information S1).Also, changes in sediment color and sedimentation rate are roughly contemporaneous with changes in benthic assemblage and Cibicidoides lobatulus oxygen isotopic composition (Figure S3 in Supporting Information S1) suggesting a primarily climatological rather than sedimentary control.
The final chronology was calculated using the Bayesian analysis age modeling program Bacon (version 2.5.7;Blaauw & Christen, 2011;Figure S4 in Supporting Information S1) in the open-source statistical software R (version 4.2.1;R Core Team, 2022).Uncertainty in the EDC WD2014 age model was propagated with a matching error before reconstructing the Bayesian accumulation histories.The median offset between the foraminifera ages and the atmosphere (Reservoir ages) were estimated using the Reservoir Age offset package (ResAge; Soulet, 2015; Figure S5 in Supporting Information S1) with the Southern Hemisphere SHCal20 reference record (Hogg et al., 2020) in R (legacy version 3.3.1;R Core Team, 2022).Measured Δ 14 C values were corrected for 14 C decay that has occurred in the time after the forams died and were initially deposited to estimate the initial radiocarbon content (Δ 14 C 0 ) of the sample (Adkins & Boyle, 1997).

Glacial
The Holocene Δ 14 C in our core reflects the modern surface ocean Δ 14 C equilibration dynamics (Figure 3b), with mean Holocene planktic reservoir ages of ∼940 years (Figure 3d) mirroring the modern reservoir ages documented in bivalves from Kerguelen Island (∼920 years; Paterne et al., 2019).The modern latitudinal surface ocean 14 C gradient tracks the progressive CO 2 release (Key et al., 2004) and this pattern allows us to use planktic 14 C to track the intensity of past equilibration, while our benthic 14 C traces the reservoir age of the upwelled UCDW, which appears to have been ∼6,000 years in the LGM.However, latitudinal shifts in Southern Ocean frontal boundaries have been documented to have occurred in this region (Bard & Rickaby, 2009;Civel-Mazens et al., 2021;Howard & Prell, 1992;Kohfeld et al., 2013;Moros et al., 2021;Shetye et al., 2014;Thöle et al., 2022), complicating interpretations.
We can tackle the issue of climate related north-south shifts in Southern Ocean fronts by leveraging the AAIW-tosurface vertical 14 C gradient (Figure 2) in our core to better reconstruct the depth and latitude of the northward flowing tongue of AAIW through time in the context of regional records of latitudinal SST gradients, faunal distributions, and productivity.A shoaling and/or poleward shift of the AAIW tongue relative to our site would Geophysical Research Letters 10.1029/2023GL106413 increase the vertical 14 C gradient by moving the locus of the gradient minimum (Figure S6 in Supporting Information S1), resulting in a larger offset between the benthic and planktic 14 C (B-P offset).A B-P offset approaching zero indicates a deepening of AAIW convection and/or a equatorward shift of the Southern Ocean fronts would place the zone of AAIW convection toward our site.A moderate B-P gradient of ∼1,170 years in the Holocene reflects the modern location of TT1811-50GGC within the mid-point of AAIW (Figure 2).Our B-P results prior to the LGM (from ∼30 to 22 ka) mirror the late Holocene AAIW-to-surface vertical ocean 14 C configuration, with similar surface ocean and AAIW reservoir ages (Figure 3d).This suggests that the ventilation of shallow water masses prior to the LGM were not largely different from today and places UCDW upwelling firmly south of our core location from ∼30 to 22 ka.
These results contrast with published results from a nearby site at 3,615 m water depth that sits in Lower Circumpolar Deep Water (LCDW) and indicates that from 30 to 22 ka LCDW was ∼1,000 years older than present (Figure 3f; Gottschalk et al., 2020;Ronge et al., 2020).Today, LCDW is formed as the ACC mixes with southward flowing North Atlantic Deep Water and northward flowing Antarctic Bottom Water (AABW) that forms along the continental margin of Antarctica, subducts, and moves northward at abyssal depths (Figure 2; Talley et al., 2011).Older LCDW during the ∼7 ka before the LGM likely stemmed from isolation of LCDW from the atmosphere while at the surface as a result of expanded winter sea ice (Ferrari et al., 2014;Gersonde et al., 2005;WAIS Divide Project Members, 2015), allowing these waters to retain an aged 14 C value.Recirculation of this older LCDW as AABW northward would account for the aging of IDW and UCDW prior to the LGM (Figures 3e and 3f; Bharti et al., 2022;Ronge et al., 2020).
Paired SST and biological productivity evidence from South of Africa indicate the STF had reached its northernmost position prior to 30 ka and remained there relatively constantly from 30 to 18 ka (Bard & Rickaby, 2009) whereas evidence from cores near our site indicate that the SAF and PF continued to move northwards from 30 ka into the LGM (Civel-Mazens et al., 2021).This paints a picture of an expanded PFZ and a progressively narrowing SAZ.Today, deep winter mixing of SAMW in the SAZ reventilates recently subducted waters (Koch-Larrouy et al., 2010), but a tightened belt of SAZ would have contributed to a reconfigured Southern Ocean that may have not aligned with the Westerly winds, an arrangement that would have restricted CO 2 outgassing (Russell et al., 2006;Toggweiler et al., 2006).We hypothesize this was instrumental to keeping atmospheric CO 2 low (Marcott et al., 2014) during the LGM (Figures 3a and 4a).

LGM
The most dramatic signal in our record is a ∼3,000 years increase in surface and AAIW reservoir ages during the LGM (∼22-18 ka; Figure 3d).Such old reservoir ages can only be the result of very old deep waters coming to the surface (Figure 4b).Equally old reservoir ages to the south and west of our site were recorded in UCDW during the LGM (Figure 3e; Ronge et al., 2020), while the presence of similarly aged IDW in the equatorial Indian Ocean at mid-depths (Bharti et al., 2022) point to IDW as the source of the old, 14 Cdepleted waters present in UCDW at these sites.Decreasing B-P 14 C offsets between AAIW and surface waters (Figure 3c) and an equatorward shift in the Southern Ocean fronts suggested by other studies (Bard & Rickaby, 2009;Civel-Mazens et al., 2021;Kohfeld et al., 2013;Li et al., 2010;Moros et al., 2021;Thöle et al., 2022) across this glaciation, place our core in the SAZ region of AAIW  (Bereiter et al., 2015;Marcott et al., 2014) is mirrored by changes in the (b) atmospheric radiocarbon content (Δ 14 C; Reimer et al., 2020) and the Δ 14 C of Southern Indian Ocean surface waters and Antarctic Intermediate Water (AAIW).(c) The AAIW 14 C offset from the surface ocean (B-P 14 C offset) and the (d) reservoir ages ( 14 C offset from the contemporaneous atmosphere) of Île St. Paul AAIW and surface waters are variable across the deglaciation.The reservoir error windows (see methods) are shown as green (G.inflata) and blue (mixed benthics) shaded regions.Regional comparison of (e) surface and shallow subsurface records along with (f) deep ocean records from the Kerguelen Plateau (2 = PS69/912-3&4; 3 = PS69/907-2; Ronge et al., 2020) Conrad Rise (4 = PS2606-6; Ronge et al., 2020), Australian-Antarctic Basin (5 = MD12-3396CQ; Gottschalk et al., 2020), and the northern Indian Ocean (6 = SS-172/4040; 7 = SS152-3828; Bharti et al., 2022).YD, Younger Dryas; HS1, Heinrich Stadial 1.

Geophysical Research Letters
10.1029/2023GL106413 convection.The continued shift of the PF and SAF to their most northerly positions (Bard & Rickaby, 2009;Civel-Mazens et al., 2021;Kohfeld et al., 2013;Li et al., 2010;Moros et al., 2021;Thöle et al., 2022) increasingly narrowed the SAZ and AZ, which appears to have caused upwelled UCDW to subduct as AAIW without allowing degassing or resetting the 14 C-depleted signature.We hypothesize this propagated an aged signal into the interior of the Indian Ocean where it may have recirculated as IDW of increasing age (Figure 4b).Deep water masses exchange through diapycnal mixing in which small scale turbulence mixes oppositely traveling water masses, entraining AAIW and AABW into IDW along density surfaces (de Lavergne et al., 2017).In the glacial Indian Ocean, this would have increased the reservoir age in IDW that was subsequently upwelled as UCDW and transferred back to AAIW as Earth descended into the LGM (22-18 ka; Figure 3e).The lack of a change in atmospheric Δ 14 C composition supports our interpretation that old CO 2 was recirculated into the mid depths of the Indian Ocean (Figure 3b) through the LGM, increasing the age of this regional reservoir until the start of Heinrich Stadial 1 (∼15-18 ka).Unchanged LCDW ventilation (Figure 3f; Gottschalk et al., 2020;Ronge et al., 2020) suggests these waters were mixing with AABW that was more ventilated than UCDW.Because CO 2 equilibrates more rapidly with the atmosphere than the isotopes of carbon, this could have resulted in decoupling of radiocarbon signatures and gas exchange.However, relatively stable LGM atmospheric CO 2 levels suggest that this was likely not a significant source of CO 2 to the atmosphere at this time.Model simulations have projected that reduced deep ocean ventilation was linked to expanded sea ice along with equatorward shifted Westerlies and Southern Ocean fronts in the LGM (Chen & Wang, 2021;Ferrari et al., 2014;Mandal et al., 2021;Menviel et al., 2015;Russell et al., 2006;Toggweiler et al., 2006).These new data are the first to clarify that the individual fronts did not move in sync or in parallel and identifies that the narrowing of the SAZ in the Indian Ocean was a significant process sequestering CO 2 .

Deglaciation
Notably, the maximum offset to the atmosphere in both surface (∼4,400 years) and AAIW (∼5,000 years) waters (Figure 3d) occurred at the transition from the LGM to Heinrich Stadial 1 (∼18 ka) at our site (Figure 3d).This is coeval with a minimum in B-P 14 C (∼200 years; Figure 3c), suggesting the locus of UCDW upwelling and AAIW formation was positioned close to our core site.The presence of very old IDW at that time in the northern Indian Ocean (Figure 3f; Bharti et al., 2022) and the appearance of aged UCDW briefly to the south of our site, at Kerguelen (∼4,075 years; Figure 3e; Ronge et al., 2020), indicates these conditions marked the apex of the CO 2 sequestration event in the LGM.
Coinciding with Heinrich Stadial 1, the reservoir ages of AAIW and surface waters off Île St. Paul rapidly decreased, documenting the rapid, progressive loss of aged CO 2 from the SAZ in the Indian Ocean.A coeval decrease in IDW reservoir ages in the northern Indian Ocean (Figure 3f; Bharti et al., 2022) documents the flushing of old CO 2 from mid-depth waters (Figure 4c).During HS1, the Île St. Paul B-P 14 C offset increased (Figure 3c), which we interpret as a poleward shift of the SAF to the south of Île St. Paul and/or a shoaling of the depth of the AAIW tongue.Regionally, there is evidence for SAF retreat southwards (Moros et al., 2021) and warming sea surface temperatures (Gottschalk et al., 2020;Shukla et al., 2021;Thöle et al., 2022) across the southern Indian Ocean at this time, which we infer widened the SAZ and allowed the slow (decadal) equilibration of 14 C during the escape of CO 2 from the surface ocean.A widening SAZ appears to have opened a valve from the deep ocean to the atmosphere while deep winter mixing progressively ventilated AAIW and SAMW at our site.A 14 C-depleted signal is recorded briefly in shallow subsurface waters at Kerguelen (∼0.5 km water depth; Figure 3f; Ronge et al., 2020), verifying this effect was regional.
It is important to point out that the 14 C-depleted signal in both surface and AAIW at our site indicates this is the location of upwelling and likely not the site where full equilibration and ventilation occurred at this time.Mechanistically, bottom topography serves to steer the ACC and instigate upwelling in the Southern Ocean (Rintoul, 2018).The main flow of the ACC in the Indian Ocean is squeezed between the Kerguelen Plateau and the plateau where Île St. Paul sits, prompting topographic upwelling in the shadow of Kerguelen (Brady et al., 2021) and activating SAMW formation and ventilation (Figure 1) as waters in the SAZ are swept along in the ACC (Koch-Larrouy et al., 2010;Rintoul, 2018).Models show that today, water rises to the surface in the topographic upwelling domain created by the Kerguelen Plateau (Brady et al., 2021) and ventilates as they move eastward to the South Tasman Rise before sinking as SAMW (Koch-Larrouy et al., 2010).Deep winter mixing in the SAZ ventilates these waters as they transit across the Southern Ocean (Rintoul, 2018) and may be responsible for the progressive loss of the depleted 14 C signal in SAMW and AAIW seen off New Zealand (Rose et al., 2010) and the Chilean Margin (Siani et al., 2013) during the LGM and deglaciation.The verification of the SAZ as a locus of deglacial CO 2 degassing has been demonstrated previously by boron isotope records of surface ocean pH from the SAZ in the Pacific and South Atlantic (Martínez-Botí et al., 2015;Moy et al., 2019;Shao et al., 2019), which document the presence of CO 2 -rich waters during the first half of the deglaciation.
Our results imply that a northern migration of Southern Ocean fronts during and prior to the LGM (∼30-18 ka) may have altered the intersection of the ACC with the Kerguelen Plateau and generally reduced topographic driven turbulent mixing, thereby impacting overturning circulation by dampening deep-water mixing, upwelling, and the associated air sea exchange in this region (Ferrari et al., 2014).We suggest southward frontal migrations during the deglaciation (Anderson et al., 2009;Bostock et al., 2015;Moros et al., 2021;Sikes et al., 2009), would have re-invigorated Kerguelen-driven upwelling.Evidence suggests that while the SAF and PF began their trek southward at ∼22 ka, the STF remained northward until ∼18.5 ka (Moros et al., 2021;Sikes et al., 2009), contributing to an expanded SAZ and broad region allowing ventilation and SAMW formation (Koch-Larrouy et al., 2010).We hypothesize the combination of re-invigorated topographic-driven upwelling and a widened SAZ served to super charge ventilation in a broad swath of the Southern Ocean at the start of the deglaciation (Figure 4c).
At the onset of the Antarctic Cold Reversal/Bølling Allerød (∼15 ka), the Île St. Paul benthic (AAIW) reservoir ages at our site had returned to late glacial values.In contrast, surface waters ages overshot their Holocene mean, reaching near-atmospheric values (<500 years; Figure 3d).The Antarctic Cold Reversal (∼15-13 ka; Figure 3d) was characterized by cooler, surface temperatures (Civel-Mazens et al., 2021;Moros et al., 2021;Sikes et al., 2009;WAIS Divide Project Members, 2015), expanded sea ice (Ferrari et al., 2014;Gersonde et al., 2005), and northward movement of Southern Ocean fronts (Civel-Mazens et al., 2021;Moros et al., 2021;Sikes et al., 2009).It is also associated with a pause in atmospheric CO 2 rise (Figure 3a; Marcott et al., 2014) and decreasing Δ 14 C (Figure 3b; Reimer et al., 2020).It is uncertain why our surface ages were very young, although well-ventilated mixed layer waters during the Antarctic Cold Reversal could have been achieved by an over invigorated AMOC (Barker et al., 2010).Similar to the early glaciation, the B-P offset at our site indicates there was a hiatus in ventilation of the deep carbon reservoir through AAIW.We suggest that much like during glaciation, the frontal positions in the Indian Ocean contributed to this hiatus and the pause in CO 2 rise (Marcott et al., 2014).
During the millennial period immediately following the Antarctic Cold Reversal (Younger Dryas; ∼13 to 11.7 ka), surface ocean reservoir ages at Île St. Paul site increase to Holocene-like values (Figure 3d), consistent with reservoir ages documented in UCDW-bathed cores to the south, off the Kerguelen plateau (Figure 3e; Ronge et al., 2020).However, at our site, Younger Dryas AAIW reservoir ages increased to values almost equal to the peak recorded at the glacial/HS1 transition and remained elevated until ∼8 ka (Figure 3d).This is echoed in the persistent old ages recorded in northern Indian Ocean IDW cores (Figure 3f; Bharti et al., 2022).Additionally, the B-P 14 C offset at our site remained elevated and increased to a maximum B-P offset during the early Holocene (Figure 3c) suggesting a shoaled and/or southward shift in the locus of AAIW convection.The fact that our AAIW values were similar to or greater than IDW in the northern Indian Ocean (Figure 3f; Bharti et al., 2022) suggests that there was continued transfer of the IDW 14 C-depleted signal southward to our site allowing its subsequent release from the Southern Ocean during the second half of the deglaciation and into the early Holocene.This Geophysical Research Letters 10.1029/2023GL106413 contrasts with the first half of the deglaciation when our results suggest CO 2 equilibration was farther downstream in the ACC.These results substantiate the involvement of Southern Ocean dynamics in the return of trapped CO 2 to the atmosphere throughout the deglaciation.

Conclusions
Our radiocarbon records emphasize the interplay of SAZ expansion and topographic mixing for enhancing CO 2 ventilation from the deep ocean to the atmosphere at the end of the last ice age.This work highlights the role that shifting fronts played in driving SAMW formation and the past modulation of oceanic CO 2 uptake and release from the Southern Ocean.Modern hydrographic studies are documenting signs that the modern ACC has been both strengthening (Freeman et al., 2016;Shi et al., 2021) and migrating poleward (Lee et al., 2023;Yamazaki et al., 2021) as our climate warms.Much like the deglaciation, the attendant thickening and strengthening of SAMW formation that has been documented in the modern Southern Indian Ocean (Lu et al., 2021) may indicate expansion of the SAZ that could foretell a Southern Ocean shift from a CO 2 sink to a CO 2 source as the globe continues to warm.