Figures

top

Introduction

Atmospheric CO₂ levels have fluctuated across Quaternary glacial-interglacial cycles. During the shift from the Last Glacial Maximum (LGM) to the Holocene, an ∼80 ppm CO₂ increase was accompanied by several degrees of global mean temperature increase, >100 m of sea level rise, and a collapse of continental ice sheets (Shakun et al., 2012

Shakun, J.D., Clark, P.U., He, F., Marcott, S.A., Mix, A.C., Liu, Z., Otto-Bliesner, B., Schmittner, A., Bard, E. (2012) Global warming preceded by increasing carbon dioxide concentrations during the last deglaciation. Nature 484, 49. https://doi.org/10.1038/nature10915

). Understanding the mechanism(s) controlling these climatic changes is critical for many reasons, foremost being concern about climatic effects of anthropogenic CO₂ emissions (>120 ppm) that have already exceeded the increase from the LGM to Holocene.

To explain large pCO₂ swings on kyr timescales, most studies invoke carbon movement between the atmosphere and ocean. While originally an increase in productivity was proposed as a means of marine carbon sequestration (Broecker, 1982

Broecker, W.S. (1982) Glacial to interglacial changes in ocean chemistry. Progress in Oceanography 11, 151–197. https://doi.org/10.1016/0079-6611(82)90007-6

), subsequent work invoked a more efficient biological pump (i.e.  nutrients more completely utilised), particularly at high latitudes (Sarmiento and Toggweiler, 1984

Sarmiento, J.L., Toggweiler, J.R. (1984) A new model for the role of the oceans in determining atmospheric pCO2. Nature 308, 621. https://doi.org/10.1038/308621a0

). When coupled with ocean circulation changes, a more efficient biological pump would prevent carbon leakage from the deep ocean to atmosphere, namely in the Southern Ocean (Sigman and Boyle, 2000

Sigman, D.M., Boyle, E.A. (2000) Glacial/interglacial variations in atmospheric carbon dioxide. Nature 407, 859. https://doi.org/10.1038/35038000

), thereby lowering pCO₂. Although the processes responsible for an increase in biological pump efficiency remain debated, the consensus is that such variations can explain much of the pCO₂ variations across glacial cycles (Sigman et al., 2010

Sigman, D.M., Hain, M.P., Haug, G.H. (2010) The polar ocean and glacial cycles in atmospheric CO2 concentration. Nature 466, 47. https://doi.org/10.1038/nature09149

).

Importantly, this mechanism implies a stoichiometric increase in oxygen consumption from respiration in the water column, suggesting an oxygen deficit in glacial deep waters (Keir, 1988

Keir, R.S. (1988) On the Late Pleistocene ocean geochemistry and circulation. Paleoceanography 3, 413–445. https://doi.org/10.1029/PA003i004p00413

). Studies of pelagic settings have consistently found evidence of glacial oxygen depletion, but not benthic anoxia (reviewed in Jacobel et al., 2020

Jacobel, A.W., Anderson, R.F., Jaccard, S.L., McManus, J.F., Pavia, F.J., Winckler, G. (2020) Deep Pacific storage of respired carbon during the last ice age: Perspectives from bottom water oxygen reconstructions. Quaternary Science Reviews 230, 106065. https://doi.org/10.1016/j.quascirev.2019.106065

). In contrast, studies of continental margins – which host the majority of anoxic seafloor today – paint a more complicated picture. In some currently anoxic settings, bottom waters were more oxygenated in glacial time (e.g., Cariaco Basin; Yarincik et al., 2000

Yarincik, K.M., Murray, R.W., Lyons, T.W., Peterson, L.C., Haug, G.H. (2000) Oxygenation history of bottom waters in the Cariaco Basin, Venezuela, over the past 578,000 years: Results from redox‐sensitive metals (Mo, V, Mn, and Fe). Paleoceanography 15, 593–604. https://doi.org/10.1029/1999PA000401

). Other sites show the opposite, experiencing oxygen depletion (e.g., Namibian margin; Riedinger et al., 2021

Riedinger, N., Scholz, F., Abshire, M.L., Zabel, M. (2021) Persistent deep water anoxia in the eastern South Atlantic during the last ice age. Proceedings of the National Academy of Sciences. Proceedings of the National Academy of Sciences 118, e2107034118. https://doi.org/10.1073/pnas.2107034118

). Further complicating things, glacioeustatic sea level changes altered the area of anoxic deposition via marine incursion into lacustrine settings during deglaciation, with the Black Sea providing a notable example (Arthur and Dean, 1998

Arthur, M.A., Dean, W.E. (1998) Organic‐matter production and preservation and evolution of anoxia in the Holocene Black Sea. Paleoceanography 13, 395–411. https://doi.org/10.1029/98PA01161

). Therefore, across glacial cycles the competing effects of productivity, sea level, ocean circulation, and temperature may have resulted in different changes in benthic anoxia across sites, making individual localities poor approximators of the global prevalence of seafloor anoxia. These confounding factors thwart global extrapolations of local records to such an extent that an understanding of the magnitude and sign of changes in seafloor anoxia across glacial cycles is still lacking.

Uranium isotopes (²³⁸U/²³⁵U, expressed in delta notation as δ²³⁸U) are a well established quantitative tracer of global ocean redox. Uranium (U) has a long marine residence time (∼400 kyr; Ku et al., 1977

Ku, T.-L., Knauss, K.G., Mathieu, G.G. (1977) Uranium in open ocean: concentration and isotopic composition. Deep Sea Research 24, 1005–1017. https://doi.org/10.1016/0146-6291(77)90571-9

), making it well mixed and isotopically homogenous in seawater (Cheng et al., 2000

Cheng, H., Adkins, J.F., Edwards, R.L., Boyle, E.A. (2000) U-Th dating of deep-sea corals. Geochimica et Cosmochimica Acta 64, 2401–2416. https://doi.org/10.1016/S0016-7037(99)00422-6

; Tissot and Dauphas, 2015

Tissot, F.L.H., Dauphas, N. (2015) Uranium isotopic compositions of the crust and ocean: Age corrections, U budget and global extent of modern anoxia. Geochimica et Cosmochimica Acta 167, 113–143. https://doi.org/10.1016/j.gca.2015.06.034

; Kipp et al., 2022

Kipp, M.A., Li, H., Ellwood, M.J., John, S.G., Middag, R., Adkins, J.F., Tissot, F.L.H. (2022) ²³⁸U, ²³⁵U and ²³⁴U in seawater and deep-sea corals: A high-precision reappraisal. Geochimica et Cosmochimica Acta 336, 231–248. https://doi.org/10.1016/j.gca.2022.09.018

). Furthermore, marine U isotope mass balance is dominated by preferential ²³⁸U burial in reducing sediments (Tissot and Dauphas, 2015

Tissot, F.L.H., Dauphas, N. (2015) Uranium isotopic compositions of the crust and ocean: Age corrections, U budget and global extent of modern anoxia. Geochimica et Cosmochimica Acta 167, 113–143. https://doi.org/10.1016/j.gca.2015.06.034

; Andersen et al., 2016

Andersen, M.B., Vance, D., Morford, J.L., Bura-Nakić, E., Breitenbach, S.F., Och, L. (2016) Closing in on the marine ²³⁸U/²³⁵U budget. Chemical Geology 420, 11–22. https://doi.org/10.1016/j.chemgeo.2015.10.041

). Thus, expansion of the “reduced” sink (i.e.  expansion of anoxic seafloor area) will decrease seawater δ²³⁸U (δ²³⁸U_sw). If one can access a record of δ²³⁸U_sw through time, this can be equated to seafloor anoxia via isotope mass balance (e.g., Kipp and Tissot, 2022

Kipp, M.A., Tissot, F.L.H. (2022) Inverse methods for consistent quantification of seafloor anoxia using uranium isotope data from marine sediments. Earth and Planetary Science Letters 577, 117240. https://doi.org/10.1016/j.epsl.2021.117240

). Carbonate sediments (review in Zhang et al., 2020

Zhang, F., Lenton, T.M., del Rey, A., Romaniello, S.J., Chen, X., Planavsky, N.J., Clarkson, M.O., Dahl, T.W., Lau, K.V., Wang, W., Li, Z., Zhao, M., Isson, T., Algeo, T.J., Anbar, A.D. (2020) Uranium isotopes in marine carbonates as a global ocean paleoredox proxy: A critical review. Geochimica et Cosmochimica Acta 287, 27–49. https://doi.org/10.1016/j.gca.2020.05.011

), and biological carbonate precipitates in particular (Chen et al., 2018a

Chen, X., Romaniello, S.J., Herrmann, A.D., Samankassou, E., Anbar, A.D. (2018a) Biological effects on uranium isotope fractionation (²³⁸U/²³⁵U) in primary biogenic carbonates. Geochimica et Cosmochimica Acta 240, 1–10. https://doi.org/10.1016/j.gca.2018.08.028

; Kipp et al., 2022

Kipp, M.A., Li, H., Ellwood, M.J., John, S.G., Middag, R., Adkins, J.F., Tissot, F.L.H. (2022) ²³⁸U, ²³⁵U and ²³⁴U in seawater and deep-sea corals: A high-precision reappraisal. Geochimica et Cosmochimica Acta 336, 231–248. https://doi.org/10.1016/j.gca.2022.09.018

), have been shown to record δ²³⁸U_sw, and have been extensively targeted in palaeo-redox studies.

Here, we use the δ²³⁸U redox proxy to quantify changes in seafloor anoxia across glacial cycles. We leverage a well preserved, high resolution biological carbonate archive: cold-water scleractinian corals. These corals are faithful archives of δ²³⁸U_sw (Kipp et al., 2022

Kipp, M.A., Li, H., Ellwood, M.J., John, S.G., Middag, R., Adkins, J.F., Tissot, F.L.H. (2022) ²³⁸U, ²³⁵U and ²³⁴U in seawater and deep-sea corals: A high-precision reappraisal. Geochimica et Cosmochimica Acta 336, 231–248. https://doi.org/10.1016/j.gca.2022.09.018

) and their fossil record spans >200 kyr. We performed physical and chemical cleaning tests to ensure removal of exogenous U from fossil corals, then analysed a large fossil dataset. Our analyses (n = 98) reveal a narrow range of δ²³⁸U values, which via isotope mass balance implies little change in anoxic seafloor area across the LGM. We compare this record to other redox proxies and make the argument that while anoxic area changed little across glacial time, suboxia likely became more prevalent in pelagic deep ocean settings.

top

Results

The 32 subsamples of eight corals subjected to variable cleaning treatments have nearly identical δ²³⁸U values, between −0.402 ± 0.021 ‰ and −0.361 ± 0.029 ‰ (Fig. 1). All modern and fossil corals (Fig. 2) have δ²³⁸U values within uncertainty of the salinity-normalised global average modern seawater value (−0.379 ± 0.023 ‰; Tissot and Dauphas, 2015

Tissot, F.L.H., Dauphas, N. (2015) Uranium isotopic compositions of the crust and ocean: Age corrections, U budget and global extent of modern anoxia. Geochimica et Cosmochimica Acta 167, 113–143. https://doi.org/10.1016/j.gca.2015.06.034

; Kipp et al., 2022

Kipp, M.A., Li, H., Ellwood, M.J., John, S.G., Middag, R., Adkins, J.F., Tissot, F.L.H. (2022) ²³⁸U, ²³⁵U and ²³⁴U in seawater and deep-sea corals: A high-precision reappraisal. Geochimica et Cosmochimica Acta 336, 231–248. https://doi.org/10.1016/j.gca.2022.09.018

). In contrast, δ²³⁴U_i was anomalously high in one sample (+184.16 ‰) and low in another (+112.51 ‰) (Fig. 1). The δ²³⁸U stability of the coral dataset contrasts with the considerable isotopic variability seen in coeval carbonate sediments (Chen et al., 2018b

Chen, X., Romaniello, S.J., Herrmann, A.D., Hardisty, D., Gill, B.C., Anbar, A.D. (2018b) Diagenetic effects on uranium isotope fractionation in carbonate sediments from the Bahamas. Geochimica et Cosmochimica Acta 237, 294–311. https://doi.org/10.1016/j.gca.2018.06.026

; Tissot et al., 2018

Tissot, F.L.H., Chen, C., Go, B., Naziemiec, M., Healy, G., Bekker, A., Swart, P.K., Dauphas, N. (2018) Controls of eustasy and diagenesis on the ²³⁸U/²³⁵U of carbonates and evolution of the seawater (²³⁴U/²³⁸U) during the last 1.4 Myr. Geochimica et Cosmochimica Acta 242, 233–265. https://doi.org/10.1016/j.gca.2018.08.022

).

Full size image

Full size image

top

Discussion

Impact of physical and chemical cleaning on coral U isotope ratios. As small δ²³⁸U fluctuations could imply significant changes in seafloor anoxia, we examined the sensitivity of coral U isotope analyses to chemical cleaning. Over time, fossil deep-sea corals develop FeMn oxide coatings, which adsorb trace elements and potentially contaminate skeletal analyses if not removed. FeMn coatings are particularly problematic for U-series dating, as their Th content is much higher than coral skeletons (Cheng et al., 2000

Cheng, H., Adkins, J.F., Edwards, R.L., Boyle, E.A. (2000) U-Th dating of deep-sea corals. Geochimica et Cosmochimica Acta 64, 2401–2416. https://doi.org/10.1016/S0016-7037(99)00422-6

).

There is a long history of chemical cleaning of corals before trace element analysis (e.g., Shen and Boyle, 1988

Shen, G.T., Boyle, E.A. (1988) Determination of lead, cadmium and other trace metals in annually-banded corals. Chemical Geology 67, 47–62. https://doi.org/10.1016/0009-2541(88)90005-8

; Lomitschka and Mangini, 1999

Lomitschka, M., Mangini, A. (1999) Precise Th/U-dating of small and heavily coated samples of deep sea corals. Earth and Planetary Science Letters 170, 391–401. https://doi.org/10.1016/S0012-821X(99)00117-X

; Cheng et al., 2000

Cheng, H., Adkins, J.F., Edwards, R.L., Boyle, E.A. (2000) U-Th dating of deep-sea corals. Geochimica et Cosmochimica Acta 64, 2401–2416. https://doi.org/10.1016/S0016-7037(99)00422-6

). We consolidated some of these practices into a series of tests (Table S-5; Supplementary Information) that were applied to subsamples (septa with some thecal material) from eight corals of varying ages. The tests spanned minimal to intensive oxidative cleaning; reductive cleaning was not studied here.

The δ²³⁸U and δ²³⁴U_i values were not affected by cleaning (Fig. 1). There are several possible reasons for this. First, all subsamples were subjected to extensive physical cleaning with a Dremel tool and/or scalpel. This removes most exogenous material (FeMn coatings, organics, detritus, secondary carbonate), leaving little to be removed via chemical cleaning (confirmed by elemental analyses; Fig. S-3). Second, U concentrations are similar in crust and coral material (Cheng et al., 2000

Cheng, H., Adkins, J.F., Edwards, R.L., Boyle, E.A. (2000) U-Th dating of deep-sea corals. Geochimica et Cosmochimica Acta 64, 2401–2416. https://doi.org/10.1016/S0016-7037(99)00422-6

). Thus, by physically removing most of the coating, the signal is overwhelmed by coral U. This is in contrast to Th, which is ∼10⁴ times more concentrated in coatings (Cheng et al., 2000

Cheng, H., Adkins, J.F., Edwards, R.L., Boyle, E.A. (2000) U-Th dating of deep-sea corals. Geochimica et Cosmochimica Acta 64, 2401–2416. https://doi.org/10.1016/S0016-7037(99)00422-6

). Third, because we aimed to recover 100–1000 ng U for high precision δ²³⁸U analysis, we digested large samples (10s to 100s mg). This gives a lower surface area to volume ratio than samples cut from surface layers of septa, buffering against exogenous U contributions. Overall, we concluded that U isotope analyses of fossil corals need not include chemical cleaning, and only performed physical cleaning for the remaining analyses.

Comparison of coral and carbonate sediment records. Over the last 240 kyr, δ²³⁸U values in Bahamian carbonate sediment cores vary by >0.5 ‰ (Fig. 2; Romaniello et al., 2013

Romaniello, S.J., Herrmann, A.D., Anbar, A.D. (2013) Uranium concentrations and ²³⁸U/²³⁵U isotope ratios in modern carbonates from the Bahamas: Assessing a novel paleoredox proxy. Chemical Geology 362, 305–316. https://doi.org/10.1016/j.chemgeo.2013.10.002

; Chen et al., 2018b

Chen, X., Romaniello, S.J., Herrmann, A.D., Hardisty, D., Gill, B.C., Anbar, A.D. (2018b) Diagenetic effects on uranium isotope fractionation in carbonate sediments from the Bahamas. Geochimica et Cosmochimica Acta 237, 294–311. https://doi.org/10.1016/j.gca.2018.06.026

; Tissot et al., 2018

Tissot, F.L.H., Chen, C., Go, B., Naziemiec, M., Healy, G., Bekker, A., Swart, P.K., Dauphas, N. (2018) Controls of eustasy and diagenesis on the ²³⁸U/²³⁵U of carbonates and evolution of the seawater (²³⁴U/²³⁸U) during the last 1.4 Myr. Geochimica et Cosmochimica Acta 242, 233–265. https://doi.org/10.1016/j.gca.2018.08.022

). This variation, which is predominantly attributed to diagenetic ²³⁸U enrichment under reducing porewater conditions, has prevented the reconstruction of primary δ²³⁸U_sw in the recent past. The tight range of coral δ²³⁸U values (<0.05 ‰) observed here shows that δ²³⁸U_sw was invariant over glacial time. This serves as a baseline from which the magnitude of diagenetic δ²³⁸U offsets can be quantified. This record of constant δ²³⁸U_sw, coupled to the noisy record from Bahamian sediments, reinforces the conclusion that δ²³⁸U_sw reconstructions using carbonate sediments are subject to considerable uncertainty due to not only the magnitude, but more so the variability of diagenetic alteration through stratigraphic sections (Kipp and Tissot, 2022

Kipp, M.A., Tissot, F.L.H. (2022) Inverse methods for consistent quantification of seafloor anoxia using uranium isotope data from marine sediments. Earth and Planetary Science Letters 577, 117240. https://doi.org/10.1016/j.epsl.2021.117240

).

Model constraints on anoxic and suboxic seafloor area. Our δ²³⁸U record allows us to constrain changes in the area of anoxic and suboxic seafloor during glacial time. Using the inverse isotope mass balance model of Kipp and Tissot (2022)

Kipp, M.A., Tissot, F.L.H. (2022) Inverse methods for consistent quantification of seafloor anoxia using uranium isotope data from marine sediments. Earth and Planetary Science Letters 577, 117240. https://doi.org/10.1016/j.epsl.2021.117240

, we explored the last 60 kyr of our record where data density is highest. We find that the median reconstructed extent of seafloor anoxia (f_anox) through the 60 kyr record is within uncertainty of the modern value (Fig. 3a), with 84 % confidence that f_anox never exceeded 0.3 % (the upper limit on modern estimate; grey shading in Fig. 3a).

Full size image

A few takeaways emerge from this stringent redox constraint. First, it invalidates models of glacial ocean chemistry (e.g., Broecker, 1982

Broecker, W.S. (1982) Glacial to interglacial changes in ocean chemistry. Progress in Oceanography 11, 151–197. https://doi.org/10.1016/0079-6611(82)90007-6

) that imply deep ocean anoxia due to a stronger biological pump and slower ocean circulation. While such models have fallen out of favour and been superseded by models invoking deep ocean suboxia, our finding strengthens that consensus. To further illustrate an upper limit on redox change, we performed forward model tests where f_anox (modern = 0.2 %) was increased to 1 %, 2 % and 5 % during the LGM (25 to 18 ka). These yield δ²³⁸U_sw trajectories (Fig. 4a) that increasingly deviate from the coral δ²³⁸U record, firmly suggesting that f_anox did not exceed 1 % during glacial time. Compared to “anoxic events” in Earth’s past, which the δ²³⁸U proxy implies have experienced f_anox of up to tens of percent over 100s of kyr (Zhang et al., 2020

Zhang, F., Lenton, T.M., del Rey, A., Romaniello, S.J., Chen, X., Planavsky, N.J., Clarkson, M.O., Dahl, T.W., Lau, K.V., Wang, W., Li, Z., Zhao, M., Isson, T., Algeo, T.J., Anbar, A.D. (2020) Uranium isotopes in marine carbonates as a global ocean paleoredox proxy: A critical review. Geochimica et Cosmochimica Acta 287, 27–49. https://doi.org/10.1016/j.gca.2020.05.011

; Kipp and Tissot, 2022

Kipp, M.A., Tissot, F.L.H. (2022) Inverse methods for consistent quantification of seafloor anoxia using uranium isotope data from marine sediments. Earth and Planetary Science Letters 577, 117240. https://doi.org/10.1016/j.epsl.2021.117240

), the glacial ocean experienced a much subtler redox change that did not impact δ²³⁸U_sw.

Full size image

Second, glacial-interglacial dynamics in individual low-oxygen settings can be tested against this constraint. For instance, the Black Sea, which today represents roughly half of anoxic seafloor area, became isolated from the ocean during glacial time, with marine re-connection at ∼9 ka (Arthur and Dean, 1998

Arthur, M.A., Dean, W.E. (1998) Organic‐matter production and preservation and evolution of anoxia in the Holocene Black Sea. Paleoceanography 13, 395–411. https://doi.org/10.1029/98PA01161

). Its isolation and re-connection to the ocean could, in principle, impact f_anox. Model runs simulating the re-connection of the Black Sea, however, show no visible changes in δ²³⁸U_sw on the timescale available (Fig. 4b). The reason for the difference between this and the prior test (Fig. 4a) is that excursions to higher than modern f_anox result in rapid δ²³⁸U_sw changes (due to rapid U scavenging from seawater in anoxic settings), whereas deviations toward lower than modern f_anox have much smaller and slower effects (due to the slow trend toward crustal δ²³⁸U from the trickle of riverine U into the ocean).

Third, we considered how possible changes in global runoff during deglaciation might affect the U flux into seawater. We tested this by arbitrarily doubling riverine U input (J_riv) during deglaciation (as well as exploring a concurrent change in isotopic composition), finding that there would not be significant effects on δ²³⁸U_sw (Fig. 4d). The reason, as above, is that riverine U inputs are small compared to the size of the marine reservoir. Thus, even doubling J_riv would not significantly affect δ²³⁸U_sw on 10 kyr timescales.

Last, we considered a set of tests incorporating a third sink: “suboxic” sediments. While most work with the δ²³⁸U proxy uses a simplified scheme of anoxic vs. non-anoxic sinks, U burial occurs on a continuum that includes “suboxic” sediments (Morford and Emerson, 1999

Morford, J.L., Emerson, S. (1999) The geochemistry of redox sensitive trace metals in sediments. Geochimica et Cosmochimica Acta 63, 1735–1750. https://doi.org/10.1016/S0016-7037(99)00126-X

), i.e.  those with O₂ in bottom waters but O₂ depletion in shallow porewaters (we note that this definition of “suboxic” differs from some used in the literature; see, e.g., Canfield and Thamdrup, 2009

Canfield, D., E., Thamdrup, B. (2009) Towards a consistent classification scheme for geochemical environments, or, why we wish the term 'suboxic’ would go away. Geobiology 7, 385–392. https://doi.org/10.1111/j.1472-4669.2009.00214.x

; for isotope mass balance purposes, “suboxic” simply means a sink with intermediate rate constant and isotopic effect). The expressed U isotopic fractionation during burial in “suboxic” settings is a matter of open debate; while sensitive to factors such as sedimentation rate and organic export (e.g., Lau et al., 2020

Lau, K.V., Lyons, T.W., Maher, K. (2020) Uranium reduction and isotopic fractionation in reducing sediments: Insights from reactive transport modeling. Geochimica et Cosmochimica Acta 287, 65–92. https://doi.org/10.1016/j.gca.2020.01.021

), we here adopted Δ_subox = 0.1 ‰ (Tissot and Dauphas, 2015

Tissot, F.L.H., Dauphas, N. (2015) Uranium isotopic compositions of the crust and ocean: Age corrections, U budget and global extent of modern anoxia. Geochimica et Cosmochimica Acta 167, 113–143. https://doi.org/10.1016/j.gca.2015.06.034

; Andersen et al., 2016

Andersen, M.B., Vance, D., Morford, J.L., Bura-Nakić, E., Breitenbach, S.F., Och, L. (2016) Closing in on the marine ²³⁸U/²³⁵U budget. Chemical Geology 420, 11–22. https://doi.org/10.1016/j.chemgeo.2015.10.041

) and a scavenging rate between those of anoxic and oxic sediments. We performed an inversion where the suboxic sink was added and f_anox was held at the modern value. We found that a wide range of f_subox values was consistent with the δ²³⁸U record (Fig. 3b). The reason for the lack of constraint is that this sink has little leverage on δ²³⁸U_sw due to the small isotopic effect and lower scavenging coefficient than anoxic settings. While both of those values are uncertain, this test reveals an important point: suboxic settings may have become more prevalent during glacial time without leaving an imprint on δ²³⁸U_sw. Below, we consider evidence that supports an expansion of glacial deep ocean suboxia and reconcile it with our δ²³⁸U record.

Expansion of deep ocean suboxia, but not anoxia. Several lines of evidence point to lower oxygen in glacial deep waters; here we focus on two proxies relevant to our δ²³⁸U dataset: authigenic uranium (aU) and thallium isotopes (ϵ²⁰⁵Tl). Authigenic uranium enrichments have long been recognised in glacial marine sediments (e.g., Chase et al., 2001

Chase, Z., Anderson, R.F., Fleisher, M.Q. (2001) Evidence from authigenic uranium for increased productivity of the glacial Subantarctic Ocean. Paleoceanography 16, 468–478. https://doi.org/10.1029/2000PA000542

). While aU is higher under more reducing conditions (and can be achieved under suboxia, not just anoxia), it is also sensitive to export production and sedimentation rate (McManus et al., 2005

McManus, J., Berelson, W.M., Klinkhammer, G.P., Hammond, D.E., Holm, C. (2005) Authigenic uranium: relationship to oxygen penetration depth and organic carbon rain. Geochimica et Cosmochimica Acta 69, 95–108. https://doi.org/10.1016/j.gca.2004.06.023

). Compilations of aU for the last glacial show strong inter-site differences, related to productivity and sedimentation rate, as well as changes in ocean circulation. On top of this heterogeneity is “burndown” of aU peaks during oxidation of older sediments. Despite these complexities, basin scale compilations consistently show greater aU at the LGM than Holocene (e.g., North Pacific, Jaccard and Galbraith, 2013

Jaccard, S.L., Galbraith, E.D. (2013) Direct ventilation of the North Pacific did not reach the deep ocean during the last deglaciation. Geophysical Research Letters 40, 199–203. https://doi.org/10.1029/2012GL054118

; Equatorial Pacific, Jacobel et al., 2020

Jacobel, A.W., Anderson, R.F., Jaccard, S.L., McManus, J.F., Pavia, F.J., Winckler, G. (2020) Deep Pacific storage of respired carbon during the last ice age: Perspectives from bottom water oxygen reconstructions. Quaternary Science Reviews 230, 106065. https://doi.org/10.1016/j.quascirev.2019.106065

; North Atlantic, Zhou and McManus, 2023

Zhou, Y., McManus, J.F. (2023) Authigenic uranium deposition in the glacial North Atlantic: Implications for changes in oxygenation, carbon storage, and deep water-mass geometry. Quaternary Science Reviews 300, 107914. https://doi.org/10.1016/j.quascirev.2022.107914

).

While global extrapolation from these local records is uncertain, we conducted a sensitivity test to determine whether our δ²³⁸U record precludes a glacial expansion of seafloor suboxia. We assigned f_subox = 10, 40 and 90 % during the last glacial and varied Δ_subox between 0.1 and 0.2 ‰. We see that f_subox up to ∼40 % has little impact on δ²³⁸U_sw (Fig. 4c), meaning changes in f_subox – that potentially gave rise to observed aU enrichments – are allowable given the δ²³⁸U data. This is consistent with Figure 3b and highlights the insensitivity of δ²³⁸U to suboxia on glacial timescales.

We also consider a new, purportedly global record of glacial ocean redox: Tl isotopes in reducing sediments. Cores TN041-8PG/8JPC from the Oman margin revealed a ϵ²⁰⁵Tl trend (Fig. 3) that suggests a shift from more reducing to more oxidising deep ocean conditions during the last deglaciation (Wang et al. 2024

Wang, Y., Costa, K.M., Lu, W., Hines, S.K.V., Nielsen, S.G. (2024) Global oceanic oxygenation controlled by the Southern Ocean through the last deglaciation. Science Advances 10, eadk2506. https://doi.org/10.1126/sciadv.adk2506

). Marine Tl isotope mass balance is mainly affected by Tl adsorption to Mn oxides in oxygenated waters. Thus, glacial expansion of deep ocean suboxia could have diminished the impact of this sink, causing higher ϵ²⁰⁵Tl. As demonstrated above, this would have little effect on δ²³⁸U_sw.

top

Conclusions

Our data imply minimal change in anoxic seafloor area during the last ice age, but allow significant changes in seafloor suboxia. In view of other proxy records, expansion of suboxic sediment area appears probable, particularly in the pelagic deep ocean. Though our record cannot quantify pelagic suboxia, recent efforts to do so (e.g., Jacobel et al., 2020

Jacobel, A.W., Anderson, R.F., Jaccard, S.L., McManus, J.F., Pavia, F.J., Winckler, G. (2020) Deep Pacific storage of respired carbon during the last ice age: Perspectives from bottom water oxygen reconstructions. Quaternary Science Reviews 230, 106065. https://doi.org/10.1016/j.quascirev.2019.106065

) have concluded that observed water column oxygen deficits imply stoichiometric amounts of dissolved inorganic carbon storage that can explain much of the glacial CO₂ drawdown. Our data are consistent with these estimates, and thus are consistent with the biological pump playing an important role in glacial marine carbon sequestration.

As for continental shelves, our results require that settings with local expansion of glacial anoxia, or even euxinia (e.g., Namibian shelf), were mostly balanced by locations that became better oxygenated (e.g., Cariaco Basin), or disconnected from the ocean (e.g., Black Sea). Modest change in f_anox is possible, but more likely toward lower than higher values. Our forward model experiments imply a strict upper limit on f_anox of ∼1 %, but our inverse analyses of the coral δ²³⁸U timeseries make a probabilistic case for even less variability (<16 % likelihood of f_anox >0.3 % at any point in the last 60 kyr). Future work could corroborate these estimates with more thorough compilations of varved sediment area through time.

top

Acknowledgements

This work was supported by a Postdoctoral Fellowship in Geobiology from the Agouron Institute to MAK, a Caltech WAVE Fellowship to AG, as well as NSF grant MGG-2054892, a Packard Fellowship, a research award from the Heritage Medical Research Institute, and start-up funds (provided by Caltech) to FLHT.

Editor: Gavin Foster

top

References

top

Supplementary Information

The Supplementary Information includes:

• Samples
• Methods
• Tables S-1 to S-9
• Figures S-1 to S-3
• Supplementary Information References

Download the Supplementary Information (PDF)