Productivity and Dissolved Oxygen Controls on the Southern Ocean Deep‐Sea Benthos During the Antarctic Cold Reversal

The Antarctic Cold Reversal (ACR; 14.7 to 13 thousand years ago; ka) phase of the last deglaciation saw a pause in the rise of atmospheric CO2 and Antarctic temperature, that contrasted with warming in the North. A reexpansion of sea ice and a northward shift in the position of the westerly winds in the Southern Ocean are well‐documented, but the response of deep‐sea biota and the primary drivers of habitat viability remain unclear. Here, we present a new perspective on ecological changes in the deglacial Southern Ocean, including multifaunal benthic assemblage (foraminifera and cold‐water corals) and coral geochemical data (Ba/Ca and δ11B) from the Drake Passage. Our records show that, during the ACR, peak abundances of thick‐walled benthic foraminifera Uvigerina bifurcata and corals are observed at shallow depths in the sub‐Antarctic (∼300 m), while coral populations at greater depths and further south diminished. Our ecological and geochemical data indicate that habitat shifts were dictated by (a) a northward migration of food supply (primary production) into the sub‐Antarctic Zone and (b) poorly oxygenated seawater at depth during this Antarctic cooling interval.

productivity in the sub-Antarctic Zone as fronts shift north, including a notable peak in total organic carbon mass accumulation rates on the Falkland Plateau (Core GC528; Figure 1d; Roberts et al., 2017) and more subtle pauses in the deglacial decline in opal flux and nitrate utilization at this time Martínez-García et al., 2014). Yet, despite the many records documenting changes in primary production (Anderson et al., 2002;Martínez-García et al., 2014) and shifts toward surface-water biotic assemblages adapted to cooler conditions (Figures 1c and 1e; Barker et al., 2009;Bianchi & Gersonde, 2004), relatively little is known about associated impact on benthic marine organisms in the Southern Ocean during the ACR.
With their important roles in carbon-cycling and deep-sea food webs (Gooday et al., 1992) and the creation of habitat for biodiversity hotspots (Hovland, 2008;Roberts et al., 2006), the distribution of benthic foraminifera and cold-water corals in the past is of high paleoceanographic significance. Benthic foraminiferal assemblage data also harbor a great deal of paleoenvironmental information as different species bear an affinity to particular environmental conditions (e.g., flux of raining organic carbon or oxygen content (a) Compiled and smoothed Antarctic ice core temperature change relative to modern (Parrenin et al., 2013), atmospheric CO 2 concentration (Bereiter et al., 2014), and sea ice extent (ssNa; Buizert et al., 2015). Red arrows correspond to centennial-scale CO 2 rise events referred to in the text and Figure 7. (b) Fluorescent organic matter tryptophan-like component (fOM TRYLIS) from Patriot Hills ice core on Antarctica (Fogwill et al., 2020). (c) South East Atlantic planktonic foraminiferal assemblage from sediment core TN057-21 (Barker et al., 2009) for warm and cool (Neogloboquadrina pachyderma (sinistral) and Turborotalita quinqueloba) indicator taxa. (d) Total organic carbon mass accumulation rate of core GC528 Falkland Plateau (Roberts et al., 2017). (e) Fragilariopsis curta and F. cylindrus diatom abundance in core PS2090-1 (Bianchi & Gersonde, 2004). (f) Antarctic Zone opal flux from site TN057-13-4PC (∼2,800m water depth; Anderson et al., 2009). (g) Biogenic Ba flux from core TN057-13-4PC, ∼2,800-m water depth within the Antarctic Zone  10.1029/2021PA004288 3 of 17 of seawater; Gooday, 2003). Cold-water coral habitat migration has been previously suggested at ∼50°S during the ACR in the Drake Passage (Margolin et al., 2014) and Tasmania (Thiagarajan et al., 2013), yet these studies were based on limited coral specimens and low precision dating techniques. Furthermore, without paired geochemical proxy data, the dominant drivers of any changes in benthic ecology remain unclear (e.g., circulation driven primary production and food supply, dissolved oxygen, seawater pH).
Cold-water corals offer a variety of paleoceanographic proxy potential, including tracers for circulation, productivity, and seawater pH. Radiocarbon data, for instance, can be corrected for postmortal decay using uranium-thorium dating, and used to reconstruct changes in seawater ventilation. Published radiocarbon data from Tasmania Seamounts and Drake Passage corals (Hines et al., 2015;Li et al., 2020) suggest that initial destratification of the Southern Ocean during HS1 halted briefly during maximum Antarctic cooling in the late ACR (∼13.5 ka). Ba/Ca ratios in cold-water corals record the dissolved barium concentration in seawater ([Ba] sw ; Anagnostou et al., 2011;Spooner et al., 2018), which is controlled by adsorption/absorption onto organic particulates in the upper ocean and release during remineralization at depth (Cardinal et al., 2005), resulting in a nutrient-like seawater profile (Chan et al., 1977). Ba/Ca data may thus provide insights into changes in upwelling of deep waters and/ or surface primary productivity that may act as a food source for benthic ecosystems. The boron isotopic composition ( 11 B/ 10 B ratio expressed as δ 11 B relative to NIST SRM 951 in parts per thousand) of cold-water corals increases as a function of seawater pH and has been used to document carbon release from the deep Southern Ocean during the last deglaciation . Coral δ 11 B data can therefore be used to assess whether large changes in pH (and associated shifts in saturation state) were responsible for changes in benthic calcifier abundance.
Here, we use well-dated biotic and paleoceanographic records to answer the questions: (a) What impact did a return to more glaciated conditions have on benthic ecology in the Southern Ocean during the ACR? (b) What were the primary drivers of deep-water habitat changes for marine calcifiers? We provide new benthic foraminifera and cold-water scleractinian coral abundance data and geochemical reconstructions from sites across the Drake Passage, spanning the Antarctic, Polar Frontm and sub-Antarctic Zones and occupying water depths from 300 to 1,750 m ( Figure 2). More than 600 new coral radiometric dates have been added to initial coral abundance compilations (Burke et al., 2010;Margolin et al., 2014) and many low precision "reconnaissance" dates have now been improved by isotope-dilution U-Th techniques (see Section 2). These new data allow us to present precisely dated, coral abundance data from across the Drake Passage. Furthermore, we provide new records of productivity and pH from coral Ba/Ca and δ 11 B measurements for direct comparison with published coral geochemical records (Burke & Robinson, 2012;Chen et al., 2015;Li et al., 2020). When viewed in conjunction with benthic foraminifera assemblage data, these results provide unique insight into the likely causes of biotic change during the ACR.

Benthic Foraminifera Assemblage Counts
Sediment core GC528 (53.01°S, 58.04°W, 598-m water depth) was collected on Cruise JR244 from the Falkland Plateau within the sub-Antarctic Zone (∼150 km north of Burdwood Bank; Figure 2); a site currently bathed in Antarctic Intermediate Water. We adopt the radiocarbon-based age model for this core from Roberts et al. (2016). This includes 14 radiocarbon measurements of Uvigerina bifurcata between 66-cm Location of previously published ice core (Fogwill et al., 2020), sediment core (Anderson et al., 2009;Jaccard et al., 2016), and coral records (Hines et al., 2015) referred to in data figures. (c) Meridional section (yellow line in panel a), grouping of samples in this study by water depth and modern phosphate concentration and seawater pH (Olsen et al., 2016). AZ, PFZ, and SAZ denote the Antarctic, Polar Front, and sub-Antarctic Zones, respectively. and 229-cm core depth (11-18 ka) with 14 C reservoir corrections based on nearby coral measurements by Burke and Robinson (2012). The benthic foraminiferal assemblage was determined on the >63-μm size fraction of 218 core samples. This broad size fraction was targeted so as to include the smaller Alabaminella weddellensis and Epistominella exigua species in count data (Gooday, 1993). A minimum of 300 individuals were counted per sample; samples containing significantly >600 individuals were split. Correspondence Analysis of count data between 8 and 22 ka (54-cm and 321-cm core depth) was conducted to establish the contrasting assemblages before, during, and after the ACR. One sample within a suspected dissolution horizon at 19 ka (see Section 4.1.1) dominated by agglutinated foraminifera (∼80%) was omitted to remove its weighting on axes.
Today, Shackleton Fracture Zone lies south of the Polar Front in the Antarctic Zone, Sars and Interim Seamounts are located in the Polar Front Zone, and Cape Horn and Burdwood Bank are in the more northerly sub-Antarctic Zone ( Figure 2). The shallowest coral samples come from depths of 334 m on Burdwood Bank that are currently bathed in sub-Antarctic Mode Waters. Corals were also collected at Burdwood Bank from 700 to 1,520 m, at water depths corresponding to modern day Antarctic Intermediate Water. Corals recovered from depths 695 to 1,200 m further south at Sars Seamount are currently bathed in Upper Circumpolar Deep Water while deeper samples at this site sit within Lower Circumpolar Deep Water (1,300-1,750 m). Samples from the nearby Interim Seamount are grouped with Sars samples for the purposes of this discussion.
More than 1,500 coral samples were initially "reconnaissance" dated by radiocarbon and U-series laser ablation techniques (typically ±650 years age control; this study and published data from Burke et al., 2010;Margolin et al., 2014;Spooner et al., 2016). This technique is based solely on measured 230 Th/ 238 U ratios and therefore must assume that the initial 234 U/ 238 U of the sample is similar to modern seawater and that the initial 230 Th is negligible (Spooner et al., 2016). Previous age distribution assessments from these sites by Margolin et al. (2014) only included around half the number of specimens in our study and of those only 59 had high-precision dates (Burke & Robinson, 2012), thus interpretations of benthic population changes were heavily reliant on these less precise reconnaissance dates-with the age resolution being similar to the duration of the ACR. Since those initial studies, where sufficient well-preserved material permitted, deglacial age samples have been precisely dated using U-Th isotope-dilution techniques (Chen et al., 2015;Cheng et al., 2000), typically offering age uncertainties better than ±250 years for deglacial aged samples (2 SD). This study includes 604 corals newly dated by "reconnaissance" techniques and a further 124 new high-precision isotope-dilution dates that are combined with previous coral dating studies (Burke & Robinson, 2012;Chen et al., 2015;Li et al., 2020). The result is that >47% of these coral samples within the deglacial interval now have precise dates. While reconnaissance dates are still shown for samples that lack precise dates in coral abundance plots, the vast majority of specimens used for δ 11 B and Ba/Ca records have precise dates assigned by isotope-dilution techniques.

Cold-Water Coral Ba/Ca and δ 11 B Analyses
Deglacial coral samples from Burdwood Bank and Sars/Interim Seamounts were selected for Ba/Ca analysis to examine changes in circulation and organic matter flux within the sub-Antarctic and Polar Front Zones (water depth range: 300-1,750 m). The Burdwood Bank site, with coral samples present during the late ACR, was primarily targeted for boron isotope analysis in this study. These new δ 11 B data are supplemented by existing data from Sars/Interim sites , to which we add one new boron isotope measurement from a newly dated coral from Interim Seamount (13.6 ka; late ACR). Whole "S1" septa and attached theca were taken from cup corals using a rotary cutting tool for geochemical analysis. This tool was further used to remove surficial oxide coatings and any chalky white altered carbonate. This type of visual inspection and physical removal of diagenetically altered material from deglacial specimens is shown to yield robust U isotope results reducing the effects of open-system behavior (Burke & Robinson, 2012;Chen et al., 2016). Microstructures within cold-water corals are known to exhibit highly contrasting trace metal and boron isotopes values (Blamart et al., 2007;Gagnon et al., 2007;Stewart et al., 2016), therefore where sufficient sample material allowed, multiple subsamples (typically duplicates) were measured to characterize microstructural bias. Coral fragments were finely crushed using a pestle and mortar before a 5 mg aliquot of the homogenous powder was taken. A warm 1% H 2 O 2 (buffered in NH 4 OH) oxidative clean and weak acid polish (0.0005 M HNO 3 ) was performed on powders (Boyle, 1981;Rae et al., 2011). Samples were then dissolved in distilled 0.5 M HNO 3 .
All coral trace metal and boron isotope analyses were carried out at the Universities of Bristol or St Andrews. An aliquot of the dissolved sample was analyzed by ICPMS using well-characterized, matrix-matched, synthetic standard solutions to give Ba/Ca ratios. Repeat analysis of NIST RM 8301 (Coral) (n = 19) yielded analytical precision of <±2% (2 SD). Small analytical offsets between the labs are quantified and corrected using NIST RM 8301 (Coral) results to bring values into line with interlaboratory comparison results following Stewart et al. (2020). To this end, Ba/Ca ratios of samples measured in Bristol and St Andrews were adjusted by 0.95 and 1.03, respectively.
For boron isotope analysis, a further aliquot of the dissolved sample containing ∼40 ng B was separated from the carbonate matrix using 20 μl Amberlite IRA 743 boron-specific anionic exchange resin microcolumns (Foster, 2008). The δ 11 B of purified boron samples were measured by (MC)-ICPMS against NIST SRM 951 at Bristol and St Andrews following identical protocols (Foster, 2008;Rae et al., 2011Rae et al., , 2018. Interlaboratory comparison studies reveal no systematic analytical offset between these institutions or other laboratories using this method (Gutjahr et al., 2020;Stewart et al., 2020). Samples, blanks, and standards were introduced to the instrument in a 0.5 M HNO 3 and 0.3 M HF acid matrix to aid B wash out (Zeebe & Rae, 2020). Full procedural uncertainty is assessed using repeat measurements of NIST RM 8301 (Coral) consistency standard (Stewart et al., 2020) yielding average δ 11 B values of 24.24 ± 0.15‰ at Bristol (2 SD; n = 67) and 24.12 ± 0.12‰ at St Andrews (2 SD; n = 15). Total procedural blanks (n = 21) were on average <0.13% of the sample boron loaded onto columns.

Estimating Seawater [Ba] and pH
Seawater barium concentration ([Ba] sw ; nmol/kg) is estimated from coral Ba/Ca using the cold-water coral calibration of Spooner et al. (2018) (Equation 2). Many of the modern corals included in the Spooner et al. (2018) calibration are from the same sites used in our study Scleractinian corals do not make their carbonate skeleton from ions derived directly from seawater, rather they calcify from a semirestricted extracellular calcification fluid (Al-Horani et al., 2003). The chemistry of this calcification fluid is modified through ion pumping mechanisms (Ca-ATPase pump) that transport in Ca 2+ ions at the expense of H + to make conditions more favorable for calcification. This "pH up-regulation" means that internal pH is routinely higher (∼pH 9) than external seawater (Al-Horani et al., 2003;Gagnon et al., 2021). This results in coral δ 11 B values that are higher (cold-water corals ∼+11‰ higher; [McCulloch et al., 2012]) than that predicted by the simplified relationship between δ 11 B borate and seawater pH described in Equation 3 (Zeebe & Wolf-Gladrow, 2001); see also full expression in Rae (2018) where α B (1.027) is the fractionation factor between boric acid and borate (Klochko et al., 2006), δ 11 B sw (39.61) is the δ 11 B of seawater (Foster et al., 2010), and * B p E K is the dissociation constant of the two boron species (8.82 for temperature salinity and pressure of 3°C, 35 psu, and 1,000 m, respectively, typical of sites in this study).
Scleractinian coral δ 11 B varies as a function of ambient seawater pH, providing the basis for deep water pH reconstruction from these archives . A secondary influence of dissolved inorganic carbon on coral δ 11 B has also been observed in culture, that has the same sense of change as pH, with lower pH, carbon-rich waters yielding lower δ 11 B (Gagnon et al., 2021). Due to biological vital effects, taxa-specific coral δ 11 B versus pH calibrations are required to give seawater pH estimates (Anagnostou et al., 2012;Martin et al., 2016;McCulloch et al., 2012;Stewart et al., 2016). Cold-water coral samples used in this study include Desmophyllum, Flabellum, Caryophyllia, and Balanophyllia. Calibrations for Desmophyllum dianthus and Caryophyllia smithii (Anagnostou et al., 2012;McCulloch et al., 2012;Rae et al., 2018) reveal that these taxa (both family Caryophylliidae) upregulate internal pH by a similar amount and broadly lie on a common calibration line (Supporting Information S1). By contrast, Balanophyllia cultured under laboratory conditions yield similar δ 11 B-pH sensitivity (slope of regression) to the well-studied taxon Desmophyllum, but with δ 11 B values ∼1.5‰ lower for a given seawater pH (Gagnon et al., 2021) (Supporting Information S1). In light of these contrasting vital effects and the additional minor dependency of coral δ 11 B on dissolved inorganic carbon, in this study, we adopt a simple approach to our assessment of relative changes in seawater pH, presenting raw coral δ 11 B, but with a +1.5‰ adjustment factor applied to Balanophyllia δ 11 B results (all figures and discussion of Balanophyllia δ 11 B refer to adjusted values).

Benthic Foraminifera
Our deglacial benthic foraminiferal assemblage data from core GC528 are shown in Figure 3. Both percentage ( Figure 3a) and raw benthic foraminiferal abundance count data (Figures 3b-3d) from this core suggest that three taxa dominate the assemblage at different times during the deglaciation: Fursenkoina fusiformis prior to the ACR (∼80%; 22-14.5 ka), U. bifurcata during the ACR (∼80%; 14.5-13 ka), and E. exigua after the ACR (∼30%; 13-8 ka). The ACR also features several short-lived (<100-year duration) peaks in abundance of E. exigua (Figure 3d). The peak in relative abundance of thick-walled infaunal foraminifer Uvigerina during the ACR (Figure 3c) is coincident with the peak in total organic carbon (TOC) mass accumulation rates in the same core ( Figure 1d; Roberts et al., 2017), and with higher total benthic foraminiferal abundance ( Figure 3e) and diversity (Figure 3f) compared to the preceding HS1 event (∼16 ka).
We confirm these inferred deglacial shifts in foraminiferal assemblage through a Correspondence Analysis approach (Figure 4). Once results were plotted on their principal axes, broad groupings were defined that encompass >95% of the respective age points (pre-ACR, during-ACR, and post-ACR) and their neighboring taxa. While there is some overlap between assemblages, the broad-scale assemblage shifts from before (Assemblage 1 dominated by F. fusiformis; 22-14.5 ka), during (Assemblage 2 dominated by U. bifurcata; 14.5-13 ka), and after (Assemblage 3 dominated by E. exigua; 13-8 ka) the ACR appear robust (Figure 4). Alternatively, when plotted over time, the shift in sign (negative to positive) of principal component 1 (Axis 1) shows that the majority of the variance in the benthic foraminiferal data are described by a shift from Assemblage 1 to Assemblage 3 during the deglaciation (Figure 3g). The negative shift in principal component 2 (Axis 2) during the ACR interval shows that the next most important change in assemblage was a shift from Assemblages 1 and 3 to Assemblage 2 at ∼14 ka (Figure 3h).

Cold-Water Corals
Our new radiometric dates for Drake Passage cold-water corals are shown in Figures 5a-5d. Once these data are combined with previously published dates, grouped by depth interval, and plotted as histograms, changes in coral abundance over time become clear ( Figure 6). Similar to Uvigerina abundance, we observe a peak in coral (absolute) abundance at ∼300-m water depth at the nearby Burdwood Bank site during the ACR (Figure 6a), where the genus Balanophyllia dominates the assemblage (74% of individuals; Supporting Information S1). By contrast, at deeper depths (>700 m) on Burdwood Bank, we see a distinct demise in coral populations, identified by a gap in occurrence (Figure 6a), with no recovered corals dated between the ages of 14 and 13.2 ka despite dedicated analytical effort. With Balanophyllia also present at deeper depths both before and after this gap in occurrence, this event appears to have impacted all coral genera (Supporting Information S1). A similar decrease in coral abundance also occurs across the Cape Horn, Sars/Interim 7 of 17 Seamount, and Shackleton Fracture Zone sites with just 10 corals found within this entire 800-year interval (Figures 6b-6d). With a greater number of coral specimens recovered from Cape Horn and Sars/Interim sites, a clear decrease in coral abundance is observable from the peak in abundance at the end of HS1 (14.5 ka; up to 15 individuals per 200 years bin of the histograms) to near zero by the middle of the ACR. At the Sars/Interim sites, coral numbers recover initially at the deeper sites (∼12 ka), followed by a return of shallower water specimens later in the YD (∼11.5 ka; Figure 6c).
We adopt a statistical approach to assess the likelihood that the 800-year gap in coral occurrence during the late ACR (14.0-13.2 ka) is merely an artifact of sampling. Three hundred and ten corals were recovered from deep waters (>500 m) at all sites with ages between 16.0 and 11.5 ka (late HS1, ACR, and YD intervals). We run multiple Monte Carlo simulations (10,000) to estimate the probability that a random distribution of these corals would yield 10 or fewer corals within any 800-year interval between 16.0 and 11.5 ka. In >99% of simulations there were >10 corals in 800-year bins suggesting that the gap in coral occurrence is unlikely  (Figure 4) over time showing assemblage shifts. A potential dissolution event (∼19 ka; low diversity, low total number, high proportion of thick-walled Uvigerina) is highlighted in yellow. This contrasts with the peak in Uvigerina during the ACR (∼14 ka) that is associated with high total benthic foraminiferal abundance and diversity. a sampling artifact. This result is robust (95% confidence) even if we limit the sample set to only the most well-dated corals in this interval (age uncertainty <±400 years; n = 158; i.e., isotope-dilution dates and only the best laser reconnaissance analyses) and include those samples that fall within age uncertainty of the ACR gap.

Coral Geochemical Records
Replicate Ba/Ca measurements in these corals generally agreed within 5% of one another, thus internal skeletal variability in Ba has little influence on our record (Figure 5e). Mean Ba/Ca ratios in corals from both deep and shallower depths at the Sars/Interim Seamount sites are generally high (>15 μmol/mol) during the YD and HS1 (Figure 7d), equating to [Ba] sw values comparable with modern measurements at these sites (∼100 nmol/kg; [Spooner et al., 2018]). We find similarly high Ba/Ca ratios in the few Sars samples available in the early and latest ACR (i.e., outside of the occurrence "gap"). Samples from Burdwood Bank have generally lower Ba/Ca ratios (∼10 μmol/mol) in both shallow and deep waters throughout the deglacial interval, with the exception being corals at ∼300 m within the latter part of ACR interval (∼13.3 ka), and one coral at ∼14 ka that lived at 1,400-m depth, exhibiting Ba/Ca values similar to those at the more southerly deep Drake Passage sites (∼20 μmol/mol; [Ba] sw ∼120 nmol/kg).
The majority of our δ 11 B replicates from the same coral fall within 0.5‰ of one another, though some samples have higher variability (∼1‰; Figure 5f). Average δ 11 B measurements from corals are generally high (>26‰) during the initial phase of the ACR (∼13.8-14.6 ka), including our new data from Burdwood Bank and new and published data from the Sars/Interim Seamount sites (Rae et al., 2018; Figure 7e). During the 9 of 17 latter half of the ACR (∼13.2-13.8 ka), at the same time as the coral abundance gap at depth and the Ba/ Ca peak in Burdwood Bank samples, there is some hint of lower or more variable values. We also note that a single Desmophyllum sample at 16.3 ka records low δ 11 B (<25‰, inferred seawater pH < 7.7; Figure 7e), similar to the low δ 11 B excursions at the Sars/Interim Seamounts at the end of YD and HS1 (14.7 and 11.5 ka; Rae et al., 2018). As this event is represented by a single coral we refrain from using it in a detailed discussion; however, we note that it has a temporal correspondence with the 16.3 ka centennial CO 2 rise event (Marcott et al., 2014) and the associated Southern Ocean ventilation anomalies revealed by coral radiocarbon in Li et al. (2020).

Discussion
By combining records of multifaunal biotic change and coral geochemistry during the last deglaciation we seek to come to a consensus view of the both the timing of benthic habitat change in the Southern Ocean and its potential paleoceanographic drivers. In this discussion, we first reason that recorded shifts in benthic Figure 5. New radiometric dates and geochemistry of corals from the Drake Passage in this study. (a-d) Cold-water coral occurrence with depth from Burdwood Bank, Cape Horn, Sars/Interim Seamount, and Shackleton Fracture Zone, respectively. Larger points are from this study. Full age distributions (outside of the deglacial interval) are shown in Supporting Information S1. New coral (e) Ba/Ca and (f) δ 11 B results in this study. Symbol shape in panels e and f denotes coral genus with Balanophyllia δ 11 B data adjusted by +1.5‰ (see methods). Error bars show 2σ age uncertainties.
foraminiferal assemblage are a true representation of faunal exchange and are not merely a result of variable preservation in core GC528. Second, we argue that peaks in U. bifurcata and shallow cold-water abundance in the sub-Antarctic Zone during the ACR are the result of the northward shift in centers of primary production. Finally, by considering all evidence from benthic foraminiferal abundance and coral geochemistry data sets we posit that depletion of oxygen in seawater was the principal cause of the disappearance of corals at deeper and more southerly Drake Passage sites during the ACR.

Preservation Bias
Benthic foraminiferal assemblage data have the potential to be biased by variable preservation. Thin-walled specimens are more vulnerable to dissolution compared to thick-walled taxa; thus, a dissolution interval can lead to an increase in the proportion thick-walled benthic foraminifera within the fossil assemblage, without an actual increase in their true abundance within the living population at that time. An example of a potential dissolution event is present in GC525 at ∼19 ka where a peak in the percentage of the thick-walled taxon U. bifurcata (Figure 3a) occurs during a low in total benthic foraminiferal abundance and diversity (Figures 3e and 3f) and is not associated with a peak in U. bifurcata abundance per gram (Figure 3c). The proportion of noncalcareous, agglutinated, foraminifera is also high within this interval, further supporting a transient carbonate dissolution event. By contrast, the general assemblage shift from F. fusiformis dominated before the ACR, to U. bifurcata dominated during ACR, and E. exigua dominated after the ACR observed from relative benthic foraminiferal abundance data (Figure 3a) is also robust in the absolute counts (Figures 3b-3d). Moreover, the double peak in U. bifurcata abundance during the ACR occurs during maxima in benthic foraminiferal total abundance and diversity. This result suggests that these wholescale shifts in foraminiferal abundance are not an artifact of variable preservation and that the ACR represents a true increase in abundance of Uvigerina.  (Burke & Robinson, 2012;Chen et al., 2015;Li et al., 2020;Margolin et al., 2014).

Deglacial Benthic Foraminiferal Abundance
The changes in benthic foraminifera we document in GC528 from F. fusiformis, to U. bifurcata, to E. exigua dominated assemblages before, during, and after the ACR, respectively, are likely to represent distinct regional transitions of organic carbon flux to the seafloor, impacting benthic ecology. On the West Antarctic Peninsula shelf, the species F. fusiformis is typically associated with high primary productivity resulting from seasonal ice melt (Peck et al., 2015). High F. fusiformis abundance on the Falkland Plateau during HS1 and LGM likely represents seasonal productivity peaks associated with nutrient-rich meltwater injections and/or seasonal sea ice melt. In shelf settings such as this, the genus Uvigerina is also associated with high primary productivity regimes (100-300 g m −2 yr −1 ) where organic carbon fluxes to the seafloor are elevated Figure 7. Chemistry of fossil Southern Ocean cold-water corals from the last deglacial. (a) Benthic to contemporary atmosphere 14 C age offset (B-atmosphere) recorded in corals south of Tasmania (Hines et al., 2015) and (b) and (c) Drake Passage corals used in this study (Burke & Robinson, 2012;Chen et al., 2015;Li et al., 2020). For clarity radiocarbon error ellipses are not shown. (d) Coral Ba/Ca (this study; replicates averaged), converted to [Ba] sw using calibration by Spooner et al. (2018). (e) Coral δ 11 B pH proxy (this study; replicates averaged; Balanophyllia δ 11 B data adjusted by +1.5‰ [see Section 2]). Sars/Interim δ 11 B data from Rae et al. (2018). Red arrows highlight the deglacial low δ 11 B events (<25‰) at 16.3 (Burdwood Bank; this study), 14.7, and 11.5 ka (Sars/Interim; Rae et al., 2018) discussed in the main text. (f) Nitrogen isotopes measured in corals from Sars/Interim sites (Li et al., 2020;Wang et al., 2017). Vertical green bar marks the late ACR gap in cold-water coral occurrence highlighted in Figure 6.
(2-20 g m −2 yr −1 ) (Gooday, 2003). Uvigerina are also able to withstand lower oxygen conditions owing to its ability to denitrify (Piña-Ochoa et al., 2010). As a thick-walled foraminifer resistant to winnowing by currents, Uvigerina is also found in high relative abundance in strong flow speed environments, with U. bifurcata in particular associated with the Falkland Current (Harloff & Mackensen, 1997). Within GC528, the close association of U. bifurcata with high TOC mass accumulation rates (Figure 1d) and increases in the δ 13 C and C/N ratios of TOC (Roberts et al., 2017), suggests that the shift to Assemblage 2 is most likely a result of northward displacement of centers of surface-water primary production into this region and potentially enhanced current speeds across on the Falkland Plateau during the ACR. Furthermore, this interpretation is consistent with the transient peaks in E. exigua also seen at this time: E. exigua is classed as an opportunist species adapted to fluctuating nutrient supply and preferring strong seasonality in open ocean waters (Barker et al., 2009;Gooday, 1993). These interludes of E. exigua abundance may therefore also support claims that the ACR was a time of enhanced seasonality in Southern Ocean Sea ice extent (Fogwill et al., 2020). The abrupt decline in U. bifurcata at ∼13 ka suggests a rapid transition at the end of the ACR from an extremely high primary productivity regime on the Falkland Plateau to conditions approaching the modern.

Shallow Sub-Antarctic Zone Corals
We explore the possibility that the peak in Balanophyllia coral abundance at shallow depths on Burdwood Bank during the ACR may be linked to upwelling and plentiful food supply caused by a shift in Polar Front position toward the sub-Antarctic Zone. The rain of organic carbon through the water column is an important food source for all benthic marine calcifiers that live outside of the photic zone, not least cold-water corals (Roberts et al., 2006). The coral Balanophyllia in particular has been shown to maintain calcification and respiration rates, even under low pH conditions (such as those in the Southern Ocean), provided that food supply is high (Crook et al., 2013;Griffiths et al., 2019).
The abundant Balanophyllia corals living at shallow depths on Burdwood Bank between 13.8 and 13 ka also exhibit anomalously high Ba/Ca ratios. With supporting evidence for locally high productivity (e.g., our benthic foraminiferal assemblage data), it is possible that seafloor [Ba] was enhanced by a high flux of Ba-rich (organic) particulates during the ACR. Ba is actively scavenged from seawater as a result of barite formation down to 1,000-m water depth (Stroobants et al., 1991); hence, high particulate organic flux to the sea floor has the potential to reduce seawater [Ba] at these relatively shallow-water depths (∼300 m). However, it is not uncommon for upwelling of Ba-rich deep waters to overwhelm any depletion caused by scavenging (Montaggioni et al., 2006). Therefore, the increase in Ba/Ca values measured in these shallow corals, to values more typical of deeper/more poleward sites, suggests that shallow Burdwood Bank was influenced by upwelling of deeper, high-[Ba] deep waters during the late ACR. We consider it likely therefore, that the increased coral abundance (notably of Balanophyllia) at ∼300-m water depth at Burdwood Bank, along with the shift in benthic foraminifera assemblages on the Falkland Plateau, result from higher productivity driven by increased nutrient availability. Furthermore, the shift toward a Uvigerina dominated assemblage and evidence from sortable silt from core GC528 suggest that there was an increase in flow of water from the Pacific through the Drake Passage after 14 ka (Roberts et al., 2017). This strengthening of the bottom current would be consistent with enhanced regional upwelling and corals may have further benefited from optimal bottom current speeds for larval transport or prey capture (Purser et al., 2014).
We posit that changes in benthic ecology we document in the Drake Passage are unlikely a local effect and were seen throughout the Southern Ocean. In contrast to the high TOC flux in the sub-Antarctic Zone (Roberts et al., 2017), decreases in opal (Figure 1f; Anderson et al., 2009) and biogenic barium flux (Figure 1g; Jaccard et al., 2016) in the Antarctic Zone suggest that primary production decreased at more poleward sites during the ACR. This anticorrelation between peaks in organic matter/barium fluxes between sites north and south of the Polar Front forms the basis of the theory that the focus of primary production was displaced northward during Antarctic cool periods as a result of expanding sea ice (Anderson et al., 2002), an equatorward shift in the Southern Hemisphere westerly winds (Toggweiler, 2009), and the associated changes in buoyancy flux (Watson & Naveira Garabato, 2006). An increase in coral abundance similar to that we observe at Burdwood Bank is also seen at the shallowest sites on the Tasmanian Seamounts in the South Pacific (∼1,600 m; ∼45°S) between 13.5 and 14.5 ka (Thiagarajan et al., 2013). Synchronous increases in cold-water coral abundance in opposing sectors of the Southern Ocean therefore suggest that these changes in benthic ecology are the result of the wholesale displacement of the Polar Front northward.
Our benthic ecological records provide yet more evidence that centers of primary production moved further from the pole during the ACR. This result however makes the source of a transient peak in marine-aerosol fluorescent organic matter biomarkers (derived from nanoplankton, picoplankton, and picoeukaryotes) recorded in the Patriot Hills ice core on West Antarctica during this interval rather enigmatic (Figure 1b; Fogwill et al., 2020). To explain the occurrence of an organic-rich zone within this ice core, we suggest that marine-derived organics were likely sourced from the more distal sub-Antarctic Zone where productivity was high (rather than the immediate Antarctic Zone) or that marine aerosols in polar regions were simply more efficiently transported to the Patriot Hills during interval the ACR interval of cooling (e.g., Markle et al., 2018).

Deep-Dwelling and Polar Front Zone Corals
While the increased sub-Antarctic Zone primary productivity appears to favor shallow cold-water coral habitats at the ACR, we find a distinct minimum in coral abundance in the late ACR at deeper depths on Burdwood Bank, as well as across all depths at the more southerly Sars/Interim sites. In the following section, we show that our coral δ 11 B records reveal little evidence for large reductions in seawater pH during the ACR, and instead we suggest that low oxygen conditions were more likely responsible for the temporary loss of these cold-water coral habitat.
High surface-water primary production recorded during the ACR at the shallow Burdwood Bank coral and sediment core sites allows us to discount food supply as a limiting factor for coral occurrence at the deep Burdwood Bank site. In addition to food supply, however, it is thought that cold-water scleractinian corals tend to live in seawater that is near to or above the aragonite saturation depth (Ω > 0.9) (Guinotte et al., 2006;Thiagarajan et al., 2013) and which has dissolved oxygen content greater than ∼165 μmol/kg (Dodds et al., 2007;Thiagarajan et al., 2013). The individual impact of multiple stressors on cold-water coral physiology is difficult to establish, particularly for ancient samples (Gori et al., 2016). However, it seems unlikely that a decrease in saturation state is the leading control on the absence of corals in deeper, more southerly sites during the ACR. Although the coral gap precludes detailed δ 11 B-based assessment of pH at this time, the one available datum at ∼13.6 ka from Interim Seamount shows relatively high δ 11 B values. Furthermore, previous δ 11 B reconstructions show that Southern Ocean cold-water corals can, to some extent, tolerate low pH conditions, for example during the LGM and during centennial-scale events during the last deglaciation . We note also that the high Ba values in shallow Burdwood Bank Balanophyllia in the early ACR are associated with relatively high δ 11 B, suggesting some decoupling of Ba and carbonate chemistry. This might result from the relatively deep regeneration profile of Ba, and/or from a combination of high Ba and buffering of pH by carbonate dissolution in waters input from the Pacific, so although the Ba data suggest high availability of nutrients this does not necessarily imply the influence of carbon-rich (low pH) waters from depth. Whatever the exact explanation, as there are no low δ 11 B values (e.g., <25‰) recorded in available samples during the late ACR we infer that any decrease in pH associated with a northward shift in organic matter flux was likely modest and within the uncertainty of bulk coral δ 11 B-pH proxy capabilities (i.e., <0.2 pH units; Anagnostou et al., 2012;Stewart et al., 2016). We therefore suggest that low pH and saturation state may not have been the dominant factor limiting deep and southerly coral growth during the late ACR.
In lieu of large changes in pH or saturation state, we draw on several lines of evidence for regionally low oxygen conditions at this time. The thick-walled shallow infaunal Uvigerina genus is readily abundant in low oxygen bottom water conditions (Boersma, 1986;Kaiho, 1994) and is often found to dominate in oxygen minimum zones (Kender et al., 2019;Schumacher et al., 2007). The shift toward a Uvigerina dominated assemblage during the ACR on the Falkland Plateau is therefore suggestive of high primary productivity and is consistent with low dissolved oxygen. Further evidence for a return to widespread lower oxygen conditions in the Southern Ocean region is provided by a collapse in epifaunal-infaunal (bottom water to pore water) benthic foraminiferal carbon isotope gradients during the ACR Haddam et al., 2020).
Many paleoceanographic records also support a regional increase in stratification of seawater that would make the Southern Ocean more prone to widespread oxygen depletion at depth. Coral radiocarbon records show an increase in B-atmosphere age ( 14 C age difference between sample and the contemporary atmosphere) caused by a relative decrease in 14 C at the shallowest site (∼300 m) on Burdwood Bank between 14.7 and 13.5 ka (Figure 7b) toward values similar to those seen at the deeper and more southerly sites (Figure 7c) immediately before and after the ACR (Burke & Robinson, 2012;Chen et al., 2015;Li et al., 2020). This radiocarbon depletion at shallow depths further implies that this site had limited exchange with the atmosphere and was more influenced by deeper, older, waters at this time. Other lines of evidence suggest that the impacts of this return to more stratified conditions were felt far beyond the Drake Passage. There is a sharp 500-year increase in B-atmosphere age measured in corals from the Tasmania Seamounts (Figure 7a; Hines et al., 2015) also supporting increased isolation and influence of older waters over this time period. Stable silicon isotope ratios from the Indian Ocean show a striking increase during the ACR, indicative of enhanced silicic acid utilization relative to supply (Dumont et al., 2020). More stratified conditions and more efficient nutrient utilization in the Polar Frontal Zone are also inferred from the increase in the available cold-water coral nitrogen isotope data (Figure 7f; Li et al., 2020;Wang et al., 2017). Furthermore, planktonic foraminiferal δ 11 B-based surface-water pH records (also from 44°S) suggest the sub-Antarctic Zone briefly transitioned from a net source of carbon during HS1 to a net sink between 14 and 13 ka (Martínez-Botí et al., 2015). A reduction in dissolved oxygen content of seawater is therefore a likely driver of low coral abundance at this time and was likely caused by (a) an increase in sea ice extent that decreased air-sea gas exchange at more southerly sites and (b) remineralization of enhanced organic carbon export to depth in the sub-Antarctic Zone (Figure 8; Buizert et al., 2015;Stephens & Keeling, 2000).

Conclusions
We show that significant changes in benthic ecology occurred during the Antarctic Cold Reversal within the Drake Passage ( Figure 8). Increased abundance of corals, a shift toward a Uvigerina dominated benthic foraminifera assemblage, and elevated coral Ba/Ca ratios at more northerly and shallow sites during the ACR were most likely driven by enhanced surface-water nutrient supply, primary production, and thus food supply to the benthos in the sub-Antarctic Zone. We also find a contemporaneous gap in coral occurrence at more southerly/deeper sites during the ACR. We suggest this loss of habitat was the result lower dissolved oxygen levels at depth facilitated by extended ice cover and seawater stratification in the Antarctic Zone. Thus, shifts in food supply and changes in oxygen content of seawater during the ACR had profound impacts on the viability of benthic habitats for marine calcifiers both north and south of the Polar Front.

Data Availability Statement
Data sets for this research are available at https://pangaea.de/ https://doi.pangaea.de/10.1594/PAN-GAEA.924088, https://doi.pangaea.de/10.1594/PANGAEA.924091, and https://doi.pangaea.de/10.1594/ PANGAEA.924097. Figure 8. Schematic changes in the Southern Ocean during southern hemisphere warming intervals (Heinrich Stadial 1 and YD) and during cooling associated with the ACR (∼14 ka). Sea ice expansion during the ACR suppressed air-sea gas exchange and deep-water CO 2 released and resulted in a northward shift of the Southern Hemisphere Westerlies (SHW) and upwelling of nutrient and Ba-rich waters. This shifted the centers of primary productivity toward the sub-Antarctic Zone, stimulating shallow-water coral growth, thick-walled benthic foraminiferal (Uvigerina) occurrence, and organic carbon accumulation. At the same time, increased stratification and reduced oxygen concentrations inhibited coral occurrence at depth and further south.