Pliocene to modern Southern Ocean diatom biostratigraphy revised using samples from IODP Expedition 382

Biostratigraphy is frequently used to generate age models and is significant to understanding the rate and timing of Cenozoic climate change. Records from the Southern Ocean (SO) are particularly valuable in understanding the past behavior of the Antarctic Ice Sheet, whereby clues to this behavior can be gained from the presence and composition of preserved microfossils. Diatoms, a nearly ubiquitous group of microalgae that make cell walls out of opal, preserve well in Southern Ocean sediments and have been used extensively in Southern Ocean biostratigraphy. Here, we present an updated diatom biostratigraphy of the Southern Ocean extending 3.3 Myr from sediments recovered during International Ocean Discovery Program (IODP) Expedition 382 “Iceberg Alley” Site U1537. Furthermore, we compare a tuned age model to a paleomagnetic-based age model to provide two independent estimates of ages of these datums with quantified uncertainty. The high sedimentation rate found at Site U1537 allows detailed age assessment, allowing the generation of more finely tuned age models in Southern Ocean sediments.

1 Introduction

Biostratigraphy is used extensively as a technique to assign ages to sediments and sedimentary rocks. In Quaternary and Neogene sediments, ages are often assigned via correlation to benthic δ¹⁸O (commonly, the LR04 stack; Lisiecki and Raymo, 2005) or Antarctic ice cores (e.g., Weber et al., 2022). Accurately assigning ages to sedimentary records of past climate change is vital to understanding the rate of ice sheet mass loss over the coming decades and centuries. Records from the Southern Ocean (SO) are particularly useful in understanding the behavior of the Antarctic Ice Sheet (e.g., Konfirst et al., 2012; Bailey et al., 2022; Warnock et al., 2022). Carbonate microfossils are often used in biostratigraphic and orbitally tuned age models; however, they rarely preserve in Southern Ocean sediments. Diatoms, single-celled marine algae found in nearly all habitats on Earth with light and water, which make their cell walls out of hydrated amorphous silica (opal, SiO₂ ⋅ nH₂O), preserve well in Southern Ocean sediments (Warnock and Scherer, 2015).

Diatoms have been used extensively as Southern Ocean biostratigraphic markers. Early work began with the Deep Sea Drilling Project (DSDP), establishing significant biostratigraphic markers and zones (e.g., McCollum, 1975; Gombos, 1983). Refinements of biostratigraphic markers in both the Paleogene (Barron, 2003) and the Neogene (e.g., Gersonde and Burckle, 1990; Winter and Iwai, 2002; Kato et al., 2016) were made possible with additional Southern Ocean drilling under the Ocean Drilling Project (ODP). Southern Ocean drilling and related diatom-based biostratigraphic zonations were published under the Integrated Ocean Drilling Program/International Ocean Discovery Program (IODP; e.g., Whitehead and Bohaty, 2003). In addition to these deep-water sites, drilling on the continent allowed further refinement of biostratigraphy during the Cape Roberts Project (CRP; e.g., Harwood and Bohaty, 2001; Olney et al., 2007) and the Antarctic Drilling Project (ANDRILL; e.g., Winter et al., 2012; Sjunneskog and Winter, 2012). While these studies have provided useful biostratigraphic information for understanding Cenozoic climate change, they are constrained by recovery breaks of the drilled sediments and low sedimentation rates that provide limited detail on the associated time zones captured. A significant advance in SO diatom biostratigraphy came from the utilization of constrained optimization to produce an idealized biostratigraphy from all available Southern Ocean data, rather than one site at a time (Cody et al., 2008, 2012). While these studies provided a valuable new framework for utilizing SO diatoms, they are based on previous biostratigraphic studies and are therefore also associated with low recovery and low sedimentation rate intervals. Here, we present an updated Southern Ocean diatom biostratigraphy spanning the last 3.3 Myr from a Scotia Sea contourite deposit. We compare an age model based on climate correlation to the EPICA Dome C (EDC) (0–800 ka) and LR04 stack (800–1500 ka) (Weber et al., 2022) to a paleomagnetic age model (0–3300 ka) (Reilly et al., 2021) and discuss the uncertainties associated with each approach. This study offers new perspective in providing a biostratigraphy from a deep-sea area with high sedimentation rates and fairly continuous accumulation (Weber et al., 2021). The sedimentation rate is 6.1 cm kyr⁻¹ in the lower portions of the core and increases to 14.8 cm kyr⁻¹ upcore where opal accumulation is higher.

Site information

The sediments used for this analysis were recovered as part of IODP Expedition 382, “Iceberg Alley and Subantarctic Ice and Ocean Dynamics”, which drilled the Dove (sites U1536 and U1537) and Pirie (Site U1538) basins of the Scotia Sea (Fig. 1). Site U1537 is located at 59°6.65^′ S, 40°54.37^′ W with a water depth of 3713 m (Weber et al., 2021). Icebergs calved from around Antarctica are directed towards the Dove Basin, giving this area the name “Iceberg Alley” (Anderson and Andrews, 1999). These icebergs release terrigenous sediment and nutrients when they melt via interaction with the relatively warmer water of the Antarctic Circumpolar Current (ACC; Budge and Long, 2018). Site U1537 was drilled on a contourite deposit affected by Antarctic Bottom Water (Pérez et al., 2021). The recovered sediments are comprised of diatom ooze and diatom-bearing silty clay (Weber et al., 2021). High sedimentation rates up to 100 cm kyr⁻¹ make this area ideal for recovering high-resolution records of Antarctic and Southern Ocean change (Weber et al., 2021) and provide the opportunity to refine biostratigraphic datums. Samples were taken from holes A and D of Site U1537.

Figure 1 Map. This figure shows the site location and the locations of a selection of other significant sites. Purple shading indicates the proportion sea-ice coverage during the year, based on data from 2002 to 2011 (Spreen et al., 2005). Colored lines indicate the position of the southern boundary of the Antarctic Circumpolar Current (white) and the Polar Front (green) (Orsi et al., 1995). Map created with Quantarctica (Matsuoka et al., 2021) using ETOPO1, IBSCO, and RAMP2 data (Amante and Eakins, 2009; Arndt et al., 2013; Liu et al., 2016). AP: Antarctic Peninsula; WAIS: West Antarctic Ice Sheet; EAIS: East Antarctic Ice Sheet; EDC: EPICA Dome C.

2 Methods

Samples utilized for this study were collected aboard the D/V JOIDES Resolution during Expedition 382. For preliminary shipboard diatom biostratigraphic analysis, each core catcher was sampled with a toothpick and prepared by smearing sediment in a small quantity of water across the slide (i.e., a “smear slide”). Slides were set with Norland Optical Adhesive No. 61. In addition to shipboard sampling, U1537 paleomagnetic cube samples were also toothpick-sampled to refine the resolution of the diatom datums, providing material at a resolution of about 1 per section, or ∼ 1.5 m. However, higher-resolution sampling was used to narrow down datum positions when possible. Samples were taken from Hole A for cores 1 to 31 and from Hole D for cores 34 to 50. Smear slides were used for these samples as well. A total of 124 microslides were made. All slide depths can be found in Table S1 in the Supplement. Diatom species identifications follow those of Weber et al. (2021) and Warnock et al. (2022). Three full transects (44 mm each) were counted on each slide, yielding between 100 and 300 valves. Much like biostratigraphic markers, taxonomic concepts for SO diatoms were developed from scientific drilling on the continental shelf and deeper water by DSDP, ODP, and IODP (McCollum, 1975; Schrader, 1976; Gombos, 1976; Ciesielski, 1983; Gersonde and Burckle, 1990; Gersonde, 1991; Gersonde et al., 1990; Fenner, 1991; Baldauf and Barron, 1991; Harwood and Maruyama, 1992; Mahood and Barron, 1996; Gersonde and Bárcena, 1998; Censarek and Gersonde, 2002; Iwai and Winter, 2002; Zielinski and Gersonde, 2002; Arney et al., 2003; Bohaty et al., 2003; Whitehead and Bohaty, 2003). Efforts closer to the continent have provided additional taxonomic clarity, especially for shallow-water- and sea-ice-related species (Harwood, 1986, 1989; Winter and Harwood, 1997; Bohaty et al., 1998; Harwood et al., 1998; Scherer et al., 2000; Sjunneskog and Scherer, 2005; Olney et al., 2007, 2009; Sjunneskog et al., 2012; Winter et al., 2012). All species identifications and data collections were conducted at 1000× magnification, and data were primarily collected on a presence/absence basis.

Age model

The orbitally tuned age model of Weber et al. (2022) was used to evaluate the timing of biostratigraphic events from 0 to 1.5 Ma, the furthest it has been extended. This age model was generated via correlation of the U1537 magnetic susceptibility record to the EPICA Dome C (EDC) dust record for the last 0.8 Myr and to the LR04 benthic δ¹⁸O record for 0.8–1.5 Ma (Weber et al., 2022; hereafter W22), with uncertainty in tie points propagated using the Undatable age–depth modeling tool (Lougheed and Obrochta, 2019). The age model was tuned to the LR04 benthic stack in order to provide as detailed an age model as possible, due to the clear correlations between the benthic stack and physical properties of the core from Site U1537 (Lisiecki and Raymo, 2005; Weber et al., 2022). The paleomagnetic age model of Reilly et al. (2021; hereafter R21) and its uncertainty estimates were used to assess biostratigraphic datums from 0 to 3.3 Ma. From 0–1.5 Ma, the R21 age model requires fewer assumptions than W22 about how U1537 climate signals relate to regional and global climate, but it also has significantly larger uncertainty. Both age models are in agreement around the magnetic reversals, and the W22 age model falls within the R22 uncertainty structure (Fig. 2). The R21 age model is extended linearly to 3.549 Ma in order to include key diatom taxa using the Gauss–Gilbert Chron boundary, as identified by Reilly et al. (2021); however, the U1537 stratigraphy between ∼ 3.3 and 3.6 Ma becomes more complex, and accumulation rates are likely variable (Reilly et al., 2021; Weber et al, 2021).

Figure 2 Age–depth model comparison. This figure compares the age–depth models between the paleomagnetically based results of Reilly et al. (2021) versus the orbitally tuned results of Weber et al. (2022). The Weber model is preferred for the most recent 1400 kyr due to increased accuracy and number of tie points. In panel (a), black dots represent reversal tie points in Reilly et al. (2021), and the gray shading represents the error. The yellow dots represent tie points in the Weber et al. (2022) model, and associated red bars are the errors in that model. In panel (b), the errors in age models are compared between Reilly et al. (2021; black columns) and Weber et al. (2022; red bars). mcd: meters composite depth.

Figure 3 Biostratigraphic range model. This figure shows the age range for the diatoms utilized in this study.

3 Results

The age models of Weber et al. (2022) and Reilly et al. (2021) differ more with increasing distance from each paleomagnetic reversal, as Weber et al. (2022) allows a variable sedimentation rate, whereas the median age of the Reilly et al. (2021) uncertainty structure is similar to assuming linear sedimentation rates between reversals (Fig. 2). A linear sedimentation rate lasting for 773 kyr, i.e., the length of the Brunhes Chron (Gradstein et al., 2020), is highly unlikely due to glacial–interglacial variations in sediment focusing and supply in the Scotia Sea (Sprenk et al., 2013) and due to a long-term increase in accumulation rates at Site U1537 since 3.3 Ma (Weber et al., 2022; Reilly et al., 2021). The age model of Weber et al. (2022) shows a strong correlation with the well-dated EPICA Dome C (EDC) ice core dust record over the last 0.8 Myr (Lambert et al., 2008; Weber et al., 2022), allowing confidence in its assigned ages over that interval. Only the age model of Reilly et al. (2021) is available prior to 1.5 Ma; therefore biostratigraphic events are only dated using the age model of Reilly et al. (2021) for the interval from 1.5 to 3.6 Ma (to 340.1 m composite depth (mcd)). Paleomagnetic reversals are more common during this interval than from 0–0.8 Ma, resulting in lower uncertainty. Table 1 gives the timing of biostratigraphic events as determined by this analysis. Figure 3 provides a range chart for the analyzed species. Zonal definitions follow modern schemes, the most recent in the Southern Ocean being Winter et al. (2012), which was developed from the scheme of Winter and Iwai (2002). Some biozone names from older works, e.g., Zielinski and Gersonde (2002), were not used in Winter et al. (2012) and so are not used here. Furthermore, both Winter and Iwai (2002) and Winter et al. (2012) utilized samples taken closer to the continent than those of Zielinski and Gersonde (2002). Given that Site U1537 is much nearer the continent than the sites utilized by Gersonde and Zielinski, Winter et al. (2012) is used here. Zonal definitions and ages are as follows.

Table 1 This table provides age/depth comparisons for selected diatom species from site U1537. Abbreviations follow those used in the text. Age errors were calculated following the methods of Lougheed and Obrochta (2019), Reilly et al. (2021), and Weber et al. (2022). mcd: meters composite depth.

Zonal definitions

•  

  Thalassiosira lentiginosa partial range zone

•  

  Author: McCollum (1975), renamed by Kellogg and Kellogg (1986), updated by Zielinski and Gersonde (2002) modified by Winter et al. (2012)

•  

  Definition of top: last occurrence (LO) of Thalassiosira lentiginosa (extant)

•  

  Definition of base: LO of Actinocyclus ingens

•  

  Age: 0–0.527 Ma

•  

  Biostratigraphic events contained within this zone: LO of R. leventerae at 0.121 Ma; LO of Hemidiscus karstenii at 0.252 Ma; LO of Rouxia constricta at 0.426 Ma.

•  

  Actinocyclus ingens partial range zone

•  

  Author: Gersonde and Burckle (1990), updated by Winter et al., 2012, who defined base of this zone as the LO R. antarctica. Here, we redefine the base back to the LO of Thalassiosira kolbei due to the order of observed biostratigraphic events at Site U1537.

•  

  Definition of top: LO A. ingens

•  

  Definition of base: LO Th. kolbei

•  

  Age: 0.527–1.261 Ma

•  

  Biostratigraphic events contained within this zone: LO of Rhizosolenia harwoodii at 0.534 Ma; first occurrence (FO) of Th. antarctica at 0.551 Ma; LO of Th. elliptipora at 0.588 Ma; FO of Porosira glacialis at 0.602 Ma; LO of Th. fasciculata at 0.928 Ma; FO of R. constricta at 1.074 Ma; FO of Fragilariopsis separanda at 1.242 Ma.

•  

  Thalassiosira kolbei partial range zone

•  

  Author: Ciesielski (1983), renamed by Baldauf and Barron (1991), replaced with Rouxia antarctica partial range zone by Winter et al. (2012)

•  

  Definition of top: LO Th. kolbei

•  

  Definition of base: LO Rouxia antarctica

•  

  Age: 1.261–1.637 Ma

•  

  Biostratigraphic events contained within this zone: LO of F. barronii at 1.438 Ma; LO of F. robusta at 1.438 Ma; LO of Shionodiscus tetraoestrupii var. reimeri at 1.465 Ma; FO of F. obliquecostata at 1.630 Ma.

•  

  Rouxia antarctica partial range zone

•  

  Author: Sjunneskog and Winter (2012)

•  

  Definition of top: LO R. antarctica

•  

  Definition of base: LO A. fasciculatus

•  

  Age: 1.637–1.670 Ma

•  

  Biostratigraphic events contained within this zone: LO of Actinocyclus karstenii at 1.648 Ma; LO of F. bohatyi at 1.670 Ma; LO of Th. inura at 1.670 Ma; LO of Th. torokina at 1.670 Ma.

•  

  Actinocyclus fasciculatus/Act. maccollumii Concurrent range zone

•  

  Author: Winter et al. (2012)

•  

  Definition of top: LO A. fasciculatus

•  

  Definition of base: FO A. maccollumii

•  

  Age: 1.670–2.791 Ma

•  

  Biostratigraphic events contained within this zone: LO of R. diploneides at 2.079 Ma; LO of R. naviculoides at 2.079 Ma; LO of Th. vulnifica at 2.098 Ma; FO of Shionodiscus gracilis at 2.211 Ma; FO of Th. ritscheri at 2.211 Ma; FO of Th. scotia at 2.211 Ma; FO of F. kerguelensis at 2.127 Ma; LO of Th. insigna at 2.244 Ma; LO of F. interfrigidaria at 2.404 Ma; FO of A. fasciculatus at 2.454 Ma; FO of R. leventerae at 2.483 Ma; LO of F. weaveri at 2.487 Ma; LO of Th. complicata at 2.657 Ma; FO of F. curta at 2.687 Ma.

•  

  Fragilariopsis interfrigidaria partial range zone

•  

  Author: McCollum, 1975, updated by Weaver and Gombos (1981), Harwood and Maruyama (1992), and Winter et al. (2012)

•  

  Definition of top: FO A. maccollumii

•  

  Definition of bottom: FO F. interfrigidaria

•  

  Age 2.791–3.549 Ma

•  

  Biostratigraphic events contained within this zone: FO of F. weaveri at 3.187; LO of F. praeinterfrigidaria at 3.354 Ma; LO of P. barboi at 3.412 Ma; LO of R. heteropolara at 3.491 Ma; FO of Th. elliptipora at 3.491 Ma; FO of A. ingens at 3.501 Ma; LO of F. fossilis at 3.501 Ma; FO of F. ritscheri at 3.501 Ma; FO of F. robusta at 3.501 Ma; FO of H. karstenii at 3.501 Ma; FO of Rh. harwoodii at 3.501 Ma; LO of Th. striata at 3.501 Ma; FO of Th. torokina at 3.501 Ma; FO of Th. tumida at 3.501 Ma; FO of P. barboi at 3.518 Ma; FO of Th. fasciculata at 3.518 Ma; FO of Th. lentiginosa at 3.518 Ma; FO of Th. oliverana at 3.518 Ma.

4 Discussion and conclusions

The high sedimentation rate of Site U1537 and continuous sedimentation allowed detailed refinement of Southern Ocean diatom biostratigraphic markers. The difference between this and other Southern Ocean diatom-based biostratigraphic models varies with age and geography. Gersonde and Bárcena (1998) updated Southern Ocean diatom biostratigraphy utilizing existing sediments and novel gravity core material. Differences between that age model and the one presented here are variable. Gersonde and Bárcena (1998) place the LO of A. ingens at 0.65 Ma, whereas we place it at 0.527 Ma. Often, the ages presented by Gersonde and Bárcena (1998) are ranges, whereas the ages presented here are specific. For example, they place the LO of Th. kolbei at 2.0–1.8 Ma, whereas we place it at 1.261 Ma. Other differences are smaller, e.g., the LO of Th. vulnifica at 2.6–2.5 Ma (Gersonde and Bárcena, 1998) compared to 2.550 Ma in this study.

Zielinski and Gersonde (2002) analyzed sediments from ODP Leg 177 in the Atlantic sector of the Southern Ocean. Ages tend to vary by 0.1 Myr or less. For example, we find the LO of H. karstenii to have occurred at 0.252 Ma, whereas Zielinski and Gersonde (2002) place it at 0.184 Ma utilizing sediment from ODP sites 1089 through 1094. We place the LO of Th. fasciculata at 0.928 Ma compared to 0.87 Ma (Zielinski and Gersonde, 2002).

Table 2 This table compares biostratigraphic datum ages from this study to those of Cody et al. (2008) and Crampton et al. (2016).

N/A – not available.

More recently still, Kato et al. (2024) published a diatom- and radiolarian-based biostratigraphy from the Atlantic sector of the Southern Ocean using materials from ODP Site 697. This paper provided ages for biostratigraphic events for the early and middle Pliocene. Their datums are generally within 0.2 Ma of ours, for example, the LO of F. praeinterfrigidaria at 3.18 Ma (Kato et al., 2024) compared to 3.354 Ma (this study). The LO of Th. lentiginosa is placed at 3.47 Ma by Kato et al. (2024) and at 3.518 Ma here. A large difference was found in the LO of Th. complicata, however. Kato et al. (2024) place the datum at 3.11 Ma, whereas this paper finds it at 2.657 Ma. Given the large discrepancy and the possibility of reworking in Iceberg Alley, the date provided by Kato et al. (2024) is likely more valid.

When considering biostratigraphic schemes from the Ross Sea, the older the datum, the larger the difference between these data and previously published zonations (e.g., Winter et al., 2012). For example, Winter et al. (2012) place the LO of Actinocyclus ingens at 0.54 Ma, whereas we place it at 0.527 Ma. These small shifts in the timing of biostratigraphic events could represent local ecological factors, e.g., diachronous timing of ecological temperature and nutrient thresholds around the continent. Alternatively, these shifts could simply represent the refinement of the timescale afforded by the unusually high sedimentation rate at Site U1537. Given the error in dating techniques used to assign ages to diatom biostratigraphic markers and the relative scarcity of SO sediment cores, it is impossible to distinguish between these factors. However, a single circum-Antarctic diatom biostratigraphy, rather than regional biostratigraphies, is generally used to assign ages to sediment (Sjunneskog and Winter, 2012); thus, we argue for the use of most of the dates generated here due to the high sedimentation rate. While the Northern and Southern ACC have previously been divided biostratigraphically (Zielinski and Gersonde, 2002), other biostratigraphic schemes ignore this dichotomy (e.g., Winter et al., 2012) The scheme of Cody et al. (2012) based on constrained optimization (CONOP) does not utilize this dichotomy. Some discrepancies are larger; for example, the LOD of A. ingens is found at ∼ 0.38 Ma by Zielinski and Gersonde (2002) and Tolotti et al. (2018). Sampling resolution is similar between the studies, with sampling every 1.5 m. Discrepancies may reflect local ecological factors; however, sampling and sample counting are unlikely to have caused these differences.

It has been noted that SO diatom biostratigraphic dates are often based on poor-quality paleomagnetic records (Tauxe et al., 2012). When detailed paleomagnetic timescales are available, biostratigraphic datums can be well refined (Tauxe et al., 2012). The high sedimentation rate, the detailed and tuned age model over the upper 1.5 Myr (Weber et al., 2022), and the detailed high-quality magnetostratigraphic age model from 1.5 to 3.3 Ma make the ages presented here a significant improvement on previously established datums.

A circum-Antarctic diatom biostratigraphy, generated by using constrained optimization (CONOP) statistical techniques, has been developed and updated in recent years (Cody et al., 2008, 2012; Crampton et al., 2016). This novel approach seeks to confront regional disparity in the timing of established biostratigraphic events and represents a significant step forward in understanding the timing and pacing of Cenozoic climate events. However, given that this CONOP-based biostratigraphy was developed from previously existing records with existing issues around sedimentation rate, poor-quality magnetostratigraphy and gaps in core recovery mentioned above apply to the ages used in the novel statistical assessment. A more recent CONOP-based study (Crampton et al., 2016) assessed the relationship between orbital parameters, ice, and diatom biogeography and found that five major episodes of cooling over the past 15 million years have led to significant species turnover (Crampton et al., 2016). Table 2 compares the dates of Cody et al. (2008) and Crampton et al. (2016) to those found here.

Overwhelmingly, the dates found here are within the dates provided by the average range model of Cody et al. (2008) or fall within 100–200 kyr. Notable exceptions are the first occurrences of several species which Cody et al. (2008) and authors of the works used therein place millions of years before we do. This is mostly likely due to ecological or taphonomic factors at Site U1537; these species were not found in this habitat or were not preserved in the sediments here during the early parts of their temporal range. Due to the high sedimentation rate of Site U1537, the close match to the EPICA Dome C ice core dust record allowing precise orbital tuning (Weber et al., 2022), and the detailed paleomagnetic record in the lower portions of the core (Reilly et al., 2021), the dates provided here are ideal for biostratigraphic work, excepting the anomalous FO dates (specifically those of A. ingens, H. karstenii, Th. torokina, P. barboi, and Th. oliverana). Comparison to Crampton et al. (2016) is more enigmatic. Many dates are somewhat similar, falling within 200–300 kyr, e.g., the LO of Th. vulnifica, which is reported at 2.06 Ma by Crampton et al. (2016) and at 2.098 Ma here. In other cases, the differences are extreme, e.g., the FO of F. separanda, which was reported at 0.47 Ma by Crampton et al. (2016) and at 1.242 Ma here. While fossil reworking can lead to a false LO, these processes cannot alter an FO date. Therefore, we recommend using the scheme presented here for FO data presented in Crampton et al. (2016). Similarly to Cody et al. (2008), the oldest dates in Crampton et al. (2016) are recommended over those provided here due to sampling (e.g., FO datums for A. ingens, H. karstenii, Th. torokina, P. barboi, and Th. oliverana).