Deglacial and Holocene sea-ice and climate dynamics in the Bransfield Strait, northern Antarctic Peninsula

The reconstruction of past sea-ice distribution in the Southern Ocean is crucial for an improved understanding of ice–ocean–atmosphere feedbacks and the evaluation of Earth system and Antarctic ice sheet models. The Antarctic Peninsula (AP) has been experiencing a warming since the start of regular monitoring of the atmospheric temperature in the 1950s. The associated decrease in sea-ice cover contrasts the trend of growing sea-ice extent in East Antarctica. To reveal the long-term sea-ice history at the northern Antarctic Peninsula (NAP) under changing climate conditions, we examined a marine sediment core from the eastern basin of the Bransfield Strait covering the last Deglacial and the Holocene. For sea-ice reconstructions, we focused on the specific sea-ice biomarker lipid IPSO₂₅, a highly branched isoprenoid (HBI), and sea-ice diatoms, whereas a phytoplankton-derived HBI triene (C_25:3) and warmer open-ocean diatom assemblages reflect predominantly ice-free conditions. We further reconstruct ocean temperatures using glycerol dialkyl glycerol tetraethers (GDGTs) and diatom assemblages and compare our sea-ice and temperature records with published marine sediment and ice core data. A maximum ice cover is observed during the Antarctic Cold Reversal 13 800–13 000 years before present (13.8–13 ka), while seasonally ice-free conditions permitting (summer) phytoplankton productivity are reconstructed for the late Deglacial and the Early Holocene from 13 to 8.3 ka. An overall decreasing sea-ice trend throughout the Middle Holocene coincides with summer ocean warming and increasing phytoplankton productivity. The Late Holocene is characterized by highly variable winter sea-ice concentrations and a sustained decline in the duration and/or concentration of spring sea ice. Overall diverging trends in GDGT-based TEX${}_{\mathrm{86}}^{\mathrm{L}}$ and RI-OH' subsurface ocean temperatures (SOTs) are found to be linked to opposing spring and summer insolation trends, respectively.

1 Introduction

Sea ice significantly affects the global climate system through its impact on the atmosphere–ocean exchange of heat and gas, the physical and chemical properties of the water masses, ocean circulation, primary production, and biogeochemical cycles (Chisholm, 2000; Vancoppenolle et al., 2013). Sea-ice cover limits evaporation, affects precipitation and increases the reflection of solar radiation due to a high albedo (Allison et al., 1982; Butterworth and Miller, 2016; Turner et al., 2017). When sea ice forms, cold and dense brines develop, contributing to the formation of intermediate and deep waters (Nicholls et al., 2009). Importantly, the downwelling of these dense-water masses can prevent warm currents from reaching the continental shelf and stimulating basal melt of Antarctic ice shelves, with implications for the stability of ice sheets and global sea level (Cook et al., 2016; Escutia et al., 2019; Etourneau et al., 2019; Hellmer et al., 2012; Huss and Farinotti, 2014). During the spring season, sea-ice melting boosts marine primary production by seeding algal cells, releasing nutrients and promoting ocean stratification and a shallow mixed-layer depth (Arrigo et al., 1997; Vernet et al., 2008). In addition, nutrient supply can be locally enhanced by wind-driven upwelling activity along the sea-ice edge (Alexander and Niebauer, 1981). Enhanced carbon fixation through this sea-ice-stimulated biological pump hence leads to an increase of biological material transport and organic carbon export to the ocean floor, thus lowering surface pCO₂ (Han et al., 2019; Kim et al., 2004; Schofield et al., 2018; Wefer et al., 1988).

Since satellite-based sea-ice data became available in 1979, fast and profound changes have been observed both in the Arctic and in West Antarctica and have been ascribed to anthropogenic global warming (IPCC, 2021). The western Antarctic Peninsula (WAP), in particular, is experiencing a rapid warming of the atmosphere (Carrasco et al., 2021; Vaughan et al., 2003) and the ocean (Cook et al., 2016). This is accompanied by rapidly retreating glaciers and ice shelves (Cook et al., 2016; Rignot et al., 2019) and by a remarkable loss of sea-ice cover in the adjacent seas (Parkinson and Cavalieri, 2012).

For an assessment of the region's past sensitivity to climate change, the Deglacial and Holocene climate history of the Antarctic Peninsula (AP) has been studied extensively. The Deglacial, the transition from the Last Glacial Maximum (LGM; Clark et al., 2012) to the Holocene, is characterized by a rapid warming punctuated by a distinct cold event, the so-called Antarctic Cold Reversal (ACR) from 14.7 to 13 ka (EPICA Community Members, 2004; Mulvaney et al., 2012; Pedro et al., 2016). This drastic cooling of both atmosphere and ocean temperatures in the high southern latitudes is well reflected in stable isotope records of Antarctic ice cores and within marine sediments (Blunier and Brook, 2001; Domack et al., 2001; Jouzel et al., 1995; Morigi et al., 2003; Stenni et al., 2001). From the Deglacial towards the Middle Holocene, the Antarctic Peninsula Ice Sheet (APIS) retreated rapidly from the outer shelf to its modern configuration with high melt water discharge (Bentley et al., 2014). Several marine and lacustrine Holocene climate records reveal that the timing of both hydrological and environmental changes was highly variable across the AP (Allen et al., 2010; Ingólfsson et al., 2003; Minzoni et al., 2015; Roseby et al., 2022; Sjunneskog and Taylor, 2002; Totten et al., 2022). An overall consensus, however, is that WAP ocean temperatures were, in comparison to the Deglacial or the Late Holocene, warmer during the Early and Middle Holocene, i.e. between 12 and 4 ka (Shevenell et al., 2011). In contrast, marine sediment records show multiple different climate patterns for the Late Holocene around the AP, including a continuous Neoglacial cooling (Etourneau et al., 2013). Knowledge of past Southern Ocean sea-ice variability is crucial to accurately model climate feedbacks (Crosta et al., 2022). For periods beyond the satellite era, information on past sea-ice conditions is based on proxies from marine sediments, ice cores (e.g. Bracegirdle et al., 2015, 2019; Crosta et al., 2022; Escutia et al., 2019; Thomas et al., 2019) and snow petrel stomach oil deposits (McClymont et al., 2022). At present, most climate models do not only fail to reproduce observed sea-ice trends of the satellite era; simulated sea-ice conditions for both glacial and interglacial periods also often disagree with geological proxies (Green et al., 2022; Lhardy et al., 2021; Roche et al., 2012). Ice-core-based sea-ice reconstructions primarily use the concentrations of sea salt sodium (WAIS Divide Project Members, 2015). However, since sea salt aerosols might be overprinted by the highly variable wind direction and meteorological conditions in Antarctica, sea salt records may not sufficiently reflect regional sea-ice conditions (Thomas et al., 2019). Although marine sediment records usually have a lower temporal resolution than ice cores, marine proxy reconstructions can resolve regional and – depending on the spatial distribution of sediment cores – large-scale changes in sea-ice conditions, as well as sea surface and subsurface ocean temperature, primary productivity, and marine ecology (Hillaire-Marcel and de Vernal, 2007). In addition to commonly used geochemical, lithological and microfossil proxies (e.g. ice-rafted debris (IRD), diatom assemblages and total organic carbon), new approaches focus on specific organic biomarkers – highly branched isoprenoids (HBIs) – as proxies to distinguish between open marine and seasonally sea-ice-covered environments. The di-unsaturated HBI IPSO₂₅ (Ice Proxy for the Southern Ocean, C_25:2; Belt et al., 2016; Massé et al., 2011) that is produced by sea-ice algae and deposited on the ocean floor after the sea-ice melt in spring has already been applied in Antarctic sea-ice reconstructions (e.g. Barbara et al., 2013; Denis et al., 2010; Etourneau et al., 2013). Following the phytoplankton–IP₂₅ sea-ice index (PIP₂₅) approach for the Arctic (Müller et al., 2011), IPSO₂₅ has been combined with phytoplankton-derived HBI trienes and/or sterols to determine the phytoplankton–IPSO₂₅ sea-ice index PIPSO₂₅ (Vorrath et al., 2019), which has been successfully evaluated with recent Antarctic spring sea-ice concentrations (Lamping et al., 2021). Other studies applied PIPSO₂₅ and examined its potential for sea-ice reconstructions over the industrial era (Vorrath et al., 2020) and the Deglacial and Holocene time intervals in the Amundsen Sea (Lamping et al., 2020). Combining these new molecular proxies with the classical diatom assemblage approach and/or geochemical ice core proxies provides a thorough assessment of past sea-ice conditions.

Here, we present a marine sediment record covering the past 13.8 kyr and reconstruct Deglacial and Holocene environmental conditions in the eastern Bransfield Strait at the NAP. Our study is based on a multiproxy approach focusing on the sea-ice biomarker IPSO₂₅, an open-ocean marine phytoplankton biomarker (HBI triene), and on glycerol dialkyl glycerol tetraether lipids (GDGTs) for subsurface ocean temperatures (SOTs). Additional estimates of primary productivity, winter sea-ice coverage (WSI) and summer sea surface temperature (SSST) come from bulk-sediment organic carbon, biogenic silica contents and diatom assemblages using transfer functions, respectively. In an intercomparison, we evaluate the different approaches to reconstruct sea-ice conditions and ocean temperatures. We discuss our proxy results with regard to other marine sediment and ice core records, providing further insight into the environmental dynamics at the Antarctic Peninsula across the Deglacial and the Holocene.

2 Material and methods

2.1 Study area

The Bransfield Strait is located between the NAP and the South Shetland Islands (SSIs; Fig. 1a), comprising a trough (>2000 m) between a narrow shelf to the north (SSIs) and a broad shelf area to the south (AP; Fig. 1b). The shelf areas were affected by intense ice sheet dynamics during the last glaciation (Canals and Amblas, 2016b; Ingólfsson et al., 2003), leaving ice sheet grounding lines and glacial troughs on the seafloor (Canals et al., 2016; Canals and Amblas, 2016a).

Figure 1 (a) Overview map with modern oceanography in the study area (Hofmann et al., 1996; Sangrà et al., 2011). ACC: Antarctic Circumpolar Current; BSW: Bellingshausen Sea water; CDW: Circumpolar Deep Water; WSW: Weddell Sea water; PF: Peninsula Front. (b) Bathymetric features in the Bransfield Strait with the location of sediment core PS97/072-1 (red star) and other sediment records discussed in the text (green), and the CTD station (purple cross) where (c) the vertical profile of ocean temperature and salinity (cruise POWELL2020, CTD 007 ($\mathrm{62}{}^{\circ }{\mathrm{09.075}}^{\prime }$ S, $\mathrm{56}{}^{\circ }{\mathrm{37.09}}^{\prime }$ W) from 27 January 2020) shows a clear stratification of the upper 100 m of the water column. It indicates that surface waters are dominated by the BSW, while the basin is filled with WSW. Maps were done with QGIS 3.0 (QGIS Development Team, 2018), and the bathymetry was taken from GEBCO_14 from 2015.

The modern Bransfield Basin is influenced by complex oceanic current systems. Cold (<0 ^∘C) and relatively salty Weddell Sea water (WSW) enters from the east, flows alongshore the peninsula and fills the Bransfield Strait basins below 150 m water depth. In the western part of the Bransfield Strait, the WSW mixes with warmer Bellingshausen Sea water (BSW; 0–50 m water depth) and Circumpolar Deep Water (CDR; 200–550 m water depth; Collares et al., 2018; Sangrà et al., 2011, 2017), which are transported in a branch of the Antarctic Circumpolar Current (ACC) over the Anvers Shelf. BSW and WSW form the Peninsula Front that runs parallel to the Antarctic mainland (Sangrà et al., 2011, 2017). The interplay of currents leads to a pronounced pycnocline within the upper 20 m of the water column in summer, accompanied by a steep temperature gradient in the upper 100 m, as observed in hydrographic profiles from the Bransfield Basin that show a dominance of WSW below 200 m (see Fig. 1c and Sangrà et al., 2011). Modern sea-ice conditions at the core site in the eastern Bransfield Strait are characterized by a mean winter sea-ice concentration of ca. 50 %, which declines to 18 % and less than 2 % sea-ice concentration during spring and summer, respectively (see Vorrath et al., 2019). While atmospheric temperatures have shown a rising trend since the 1950s (Carrasco et al., 2021), ocean temperatures are increasingly influenced by warm-water intrusions and higher sea surface temperatures (Martinson and McKee, 2012; Meredith and King, 2005). At the core site, mean annual sea surface temperatures are −0.6 ^∘C and are up to 0.8 ^∘C during summer (World Ocean Atlas, WOA, 18; Boyer et al., 2018; Locarnini et al., 2018).

Primary production in the Bransfield Strait is mainly driven by the mixing of water masses at the fronts (Gonçalves-Araujo et al., 2015), mixed-layer depth and upwelling (Sangrà et al., 2011), sea-ice dynamics (Vernet et al., 2008), and iron availability (Klunder et al., 2014). High concentrations of chlorophyll a and diatoms are distributed north of the Peninsula Front (PF) and at the SSIs, while lower production and communities of plankton nanoflagellates are found between the Peninsula Front and the WAP (Gonçalves-Araujo et al., 2015). Further, changes in coastal primary production are driven by upwelling, elevated iron availability and the nutrient release and surface water stratification generated by melting sea ice in the austral spring (Vernet et al., 2008). A robust link between marine primary production in surface waters and the sediment composition at the underlying ocean floor is reflected in high concentrations of total organic carbon (TOC), pigments, sterols and diatoms (Cárdenas et al., 2019) and is supported by studies confirming high fluxes of sinking particles (Kim et al., 2004; Wefer et al., 1988). In the study area, particle flux is highly variable, with seasonal peaks occurring in late spring, which accounts for 85 % of the total flux (Ducklow et al., 2008). Lithologically, the sediments consist mainly of terrigenous silt and clay, with varying amounts of diatom mud and ooze and of sand (Cádiz Hernández, 2019; Lamy, 2016; Wu et al., 2019).

2.2 Sediment samples and age model

Piston core PS97/072-1 ($\mathrm{62}{}^{\circ }{\mathrm{0.39}}^{\prime }$ S, $\mathrm{56}{}^{\circ }{\mathrm{3.86}}^{\prime }$ W; 1993 m water depth, 1583 cm in length) was recovered in the eastern Bransfield Strait basin during R/V Polarstern cruise PS97 (Lamy, 2016) (Fig. 1). The sediment is dominated by silt with thin layers of sand and clay and traces of volcanic ash. Single pebbles are present below 630 cm. The core is disturbed below 1015 cm depth, and we only considered samples from above this level for our analyses. Sampling for different analytical approaches was done at the Alfred Wegener Institute (AWI), where the samples were stored frozen in glass vials (for biomarker analysis) and at 4 ^∘C in plastic bags (for micropaleontology).

The age model of core PS97/072-1 is based on radiocarbon dating of eight benthic foraminiferal and mollusc fragment samples with the mini carbon-dating system (MICADAS) available at AWI (Mollenhauer et al., 2021). From the conventional ¹⁴C age, we subtracted a reservoir age based on modelling by Butzin et al. (2017) and also subtracted an estimated ventilation age of 1200 years that we derived from ages between paired planktic and benthic foraminifera of 1000 to 1500 years found off Chile (Siani et al., 2013) and the Kerguelen Islands (Gottschalk et al., 2020) to account for the considerable water depth of our site (see Table S1 in the Supplement). After the subtraction, we calibrated the ages with the calibration curve IntCal20 (Reimer et al., 2020) to calendar years before present (cal BP) with Calib 7.1 (Stuiver et al., 2018). To estimate the age of the core top. TOC and biogenic opal data of the piston core were matched with data from a multicore from the same sampling site that has been previously dated via ²¹⁰Pb (Vorrath et al., 2020; Sect. S2). Ages of sediments below the oldest radiocarbon date (868.5 cm; 12.04 ka) were extrapolated assuming a constant sedimentation rate. We applied the Bayesian age modelling tool hummingage, a freely available tool developed at AWI that has been successfully applied in previous studies (e.g. Ronge et al., 2021). As the lack of age constraints between 12 and 6 ka may introduce chronological uncertainties, we only focus on overall trends reflected in our data and refrain from detailed allocations of known climatic events in this older time period.

2.3 Organic geochemical analyses of piston core PS97/072-1

For the analyses of the bulk organic geochemical composition and biomarkers, 334 sediment samples were freeze-dried and homogenized in an agate mortar. Prior to sediment homogenization, coarse grains were separated using a sieve (500 µm mesh size). Total carbon (C) and nitrogen (N) were measured with a CNS analyser (Elementar Vario EL III, error of standards and duplicates <5 %). Total organic carbon (TOC) was measured on 0.1 g of acidified samples (500 µL HCl) and determined in a carbon–sulfur determinator (CS-800, ELTRA, standard error <0.6 %). To identify the source of TOC, measurements of stable carbon isotopes of bulk organic matter were done at Universität Hamburg (UHH), Germany, and at Washington State University (WSU), USA. At UHH, the samples were acidified three times with 100 µL 1 N HCl and dried on a hotplate. High-temperature combustion was done in an Elementar CHNOS Vario isotope elemental analyser at 950 ^∘C, and the analysis was conducted with an Elementar IsoPrime 100 isotope ratio mass spectrometer. We calibrated the pure tank CO₂ with the International Atomic Energy Agency reference standards IAEA-CH6 and IAEA-CH7. These and two other standards (IVA Sediment and Sucrose) acted as internal standards in the measurement. The error of continuous standard duplicates was <0.2 ‰, and it was <0.06 ‰ for sample duplicates. At WSU, 100 mg of freeze-dried sediment samples was used. An elemental analyser coupled with an IsoPrime isotope ratio mass spectrometer (IRMS) was used, with a precision of 0.1 ‰. The running standard was a protein hydrolysate calibrated against National Institute of Standards and Technology (NIST) standards. Isotope ratios are expressed in units per millilitre (‰). δ¹³C values are expressed in ‰ against Vienna PeeDee Belemnite (VPDB).

Biogenic opal was estimated on 327 samples following the alkaline extraction procedure described by Mortlock and Froelich (1989) but using 0.5 M NaOH as a digestion solution (Müller and Schneider, 1993). Extraction and analysis by molybdate-blue spectrophotometry were conducted at the University of Concepción, Chile. Values are expressed as biogenic opal by multiplying the Si (%) by 2.4 (Mortlock and Froelich, 1989). Opal values could be overestimated by 2 %–2.5 %, since we did not correct for the release of extractable Si from coexisting clay minerals (Schlüter and Rickert, 1998). The instrumental precision was ±0.5 % (error of duplicates ≤3 %). Details on the methodology used can be found in Cárdenas et al. (2019).

The extraction, purification and identification of 137 samples to identify HBIs followed the analytical protocol published e.g. in Belt et al. (2014) and Vorrath et al. (2019). Prior to extraction, 40 µL 7-hexylnonadecane (7-HND; 0.0019 µg µL⁻¹) and 100 µL C₄₆ (0.0098 µg µL⁻¹) were added as internal standards. Lipids were extracted using ultra sonication and a mixture of CH₂Cl₂:MeOH ($v/v$ 2:1; 6 mL). HBIs and GDGTs were separated by means of open-column chromatography using SiO₂ as the stationary phase and hexane and CH₂Cl₂:MeOH ($v/v$ 1:1) as eluents. HBIs were analysed by means of an Agilent 7890B gas chromatograph (30 m DB 1MS column, 0.25 mm diameter, 0.250 µm film thickness) coupled to an Agilent 5977B mass spectrometer (MSD, 70 eV constant ionization potential, ion source temperature 230 ^∘C). The initial oven temperature of 60 ^∘C was held for 3 min, ramped to 325 ^∘C within 23 min and then held at 325 ^∘C for 16 min. HBIs were identified via comparison of their retention times (IPSO25 and HBI triene with RI 2084DB-1MS and 2046DB-1MS, respectively) and mass spectra with published mass spectra (Belt, 2018) and were quantified using the ratio of peak areas of individual HBIs ($m/z$ 346; $m/z$ 348) and the 7-HND ($m/z$ 266) standard and through a consideration of instrumental response factors. The error of duplicates was <1.4 % for IPSO₂₅ and <2.6 % for HBI trienes. The phytoplankton–IPSO₂₅ index (PIPSO₂₅) was calculated after Vorrath et al. (2019) as follows:

$$\begin{array}{}\text{(1)}& {\text{PIPSO}}_{\mathrm{25}}={\displaystyle \frac{{\text{IPSO}}_{\mathrm{25}}}{{\text{IPSO}}_{\mathrm{25}}+(c\times \text{phytoplankton marker})}}.\end{array}$$

The concentrations of the phytoplankton-derived HBI z-triene are at the same level as IPSO₂₅, and the c factor was hence set to 1 (Vorrath et al., 2019). To confirm the sea-ice origin of IPSO₂₅, the stable carbon isotope composition of IPSO₂₅ was examined in eight samples (with minimum 50 ng carbon) via GC-irm-MS at the GFZ Potsdam, Germany. The GC (7890N Agilent) equipped with an Ultra 1 column (50 m×0.2 mm diameter, 0.33 µm film thickness) was connected to a DeltaVPlus isotope ratio mass spectrometer through a modified GC IsoLink interface. Each sample was separated chromatographically using a temperature programme that started with an oven temperature of 80 ^∘C, which was held for 3 min, ramped to 250 ^∘C with 3 ${}^{\circ }\mathrm{C}{\min }^{-\mathrm{1}}$ and then ramped to 320 ^∘C with 5 ${}^{\circ }\mathrm{C}{\min }^{-\mathrm{1}}$ before finally reaching a temperature of 325 ^∘C with a ramp of 1 ${}^{\circ }\mathrm{C}{\min }^{-\mathrm{1}}$, which was then held for 15 min. The organic substances of the GC effluent stream were oxidized to CO₂ in the combustion furnace, held at 940 ^∘C on a $\mathrm{CuO}/\mathrm{Ni}/\mathrm{Pt}$ catalyst. Samples were measured in duplicate, and the standard deviation was ≤0.5 ‰. The quality of the isotope measurements was checked regularly (for each analysis) by measuring different n-alkane standards with known isotopic composition of n-C15, n-C20 and n-C25 (in equal concentration) and n-C16 to n-C30 (in various concentrations) provided by Campro Scientific, Germany, and Arndt Schimmelmann, Indiana University, USA.

GDGTs were re-dissolved in 120 µL hexane : isopropanol ($v/v$ 99:1) and filtered through polytetrafluoroethylene filters (0.45 µm in diameter) and analysed using high-performance liquid chromatography (HPLC, Agilent 1200 series HPLC system) coupled to a single-quadrupole mass spectrometer (MS, Agilent 6120 MSD) via an atmospheric pressure chemical ionization (APCI) interface. The individual GDGTs were separated at 30 ^∘C on a Prevail Cyano column (150 mm×2.1 mm, 3 µm). After injection of the sample (20 µL), it passed a 5 min isocratic elution with mobile phase A (hexane$/$2-propanol$/$chloroform; $\mathrm{98}:\mathrm{1}:\mathrm{1}$, flow rate 0.2 mL min⁻¹). The mobile phase B (hexane$/$2-propanol$/$chloroform; $\mathrm{89}:\mathrm{10}:\mathrm{1}$) was increased to 100 % in the following two steps: a linear increase to 10 % over 20 min followed by an increase to 100 % within 10 min. During the measurement, the column was cleaned after 7 min via backflushing (5 min, flow 0.6 mL min⁻¹) and was then re-equilibrated with solvent A (10 min, flow 0.2 mL min⁻¹). The conditions of the APCI were a nebulizer pressure of 50 psi, a vaporizer temperature and N₂-drying gas temperature of 350 ^∘C, a flow of 5 L min⁻¹, a capillary voltage of 4 kV, and a corona current of 5 µA. Following Liu et al. (2020), isoprenoid GDGTs (iGDGTs) and branched GDGTs (br-GDGTs) were detected by selective ion monitoring (SIM) of (M+H⁺) ions (dwell time 76 ms) using their molecular ions (GDGTs-1 ($m/z$ 1300), GDGTs-2 ($m/z$ 1298), GDGTs-3 ($m/z$ 1296), crenarchaeol ($m/z$ 1292), GDGTs-Ia ($m/z$ 1022), GDGTs-IIa ($m/z$ 1036) and GDGTs-IIIa ($m/z$ 1050)) and quantified in relation to the internal standard C₄₆ ($m/z$ 744). The hydroxylated GDGTs OH-GDGT-0 ($m/z$ 1318), OH-GDGT-1 ($m/z$ 1316) and OH-GDGT-2 ($m/z$ 1314) were quantified in the scans of their related GDGTs (Fietz et al., 2013). The standard deviation was 0.01 units of TEX${}_{\mathrm{86}}^{\mathrm{L}}$.

Kalanetra et al. (2009) showed that GDGT-producing Thaumarchaeota are abundant in subsurface marine waters in both Arctic and Antarctic regions. As Thaumarchaeota were found between water depths of 50 and 200 m in Antarctica (Kim et al., 2012), temperatures based on GDGTs are suggested to reflect sub-surface waters (Etourneau et al., 2013, 2019). Similarly, RI-OH'-based temperatures in Prydz Bay have also been interpreted to reflect subsurface water temperatures (Liu et al., 2020). We therefore consider our results to reflect subsurface ocean temperatures (SOTs). We calculated TEX${}_{\mathrm{86}}^{\mathrm{L}}$ after Kim et al. (2012) with the $m/z$ 1296 (GDGT-3), $m/z$ 1298 (GDGT-2) and $m/z$ 1300 (GDGT-1) as follows:

$$\begin{array}{}\text{(2)}& {\text{TEX}}_{\mathrm{86}}^{\mathrm{L}}=\log \left({\displaystyle \frac{\left[\text{GDGT-2}\right]}{\left[\text{GDGT-1}\right]+\left[\text{GDGT-2}\right]+\left[\text{GDGT-3}\right]}}\right),\end{array}$$

and calibrated with

$$\begin{array}{}\text{(3)}& \text{SOT}=\mathrm{50.8}\times {\text{TEX}}_{\mathrm{86}}^{\mathrm{L}}+\mathrm{36.1}\end{array}$$

(Kim et al., 2012).

For the calculation of temperatures based on hydroxylated GDGTs, we followed the approach of Lü et al (2015) as follows:

$$\begin{array}{}\text{(4)}& \begin{array}{rl}& \text{RI-OH'}=\\ & {\displaystyle \frac{\left[\text{OH-GDGT-1}\right]+\mathrm{2}\times \left[\text{OH-GDGT-2}\right]}{\left[\text{OH-GDGT-0}\right]+\left[\text{OH-GDGT-1}\right]+\left[\text{OH-GDGT-2}\right]}},\end{array}\end{array}$$

and calibrated it with

$$\begin{array}{}\text{(5)}& \text{SOT}=(\text{RI-OH'}-\mathrm{0.1})/\mathrm{0.0382}.\end{array}$$

For the branched and isoprenoid tetraether (BIT) index for indicating terrestrial organic matter (Hopmans et al., 2004), we used crenarchaeol ($m/z$ 1292) and the branched GDGTs and calculated it as follows:

$$\begin{array}{}\text{(6)}& \text{BIT}={\displaystyle \frac{\left[\text{GDGT-Ia}\right]+\left[\text{GDGT-IIa}\right]+\left[\text{GDGT-IIIa}\right]}{\left\{\begin{array}{c}\left[\text{Crenarchaeol}\right]+\left[\text{GDGT-Ia}\right]\\ +\left[\text{GDGT-IIa}\right]+\left[\text{GDGT-IIIa}\right]\end{array}\right\}}}.\end{array}$$

2.4 Diatom analyses

We selected a set of 76 samples for the analysis of diatom assemblages. At first, the sampling resolution was every 40–50 cm; thereafter, and based on the biogenic opal results, the resolution was increased (every 8 cm) at intervals with high variability. Freeze-dried samples (20–120 mg) were treated with hydrogen peroxide and sodium pyrophosphate to remove organic matter and clays, respectively, and were washed several times with deionized water (DI) water until reaching a neutral pH. The treated samples were then settled for 6 h in B-KER2 settling chambers to promote an even distribution of settled particles (Scherer, 1994; Schrader and Gersonde, 1978; Warnock and Scherer, 2015). Once the samples were dry, the quantitative slides were mounted with a Norland mounting medium (refraction index of 1.56). Diatom valves per slide were counted across traverses (at least 400 valves per slide) using an Axioskop 2 Plus and an Olympus BX60 at a magnification of ×1000. The counting procedure and the definition of counting units followed those of Schrader and Gersonde (1978). We performed two sets of counts, with and without Chaetoceros resting spores. Diatoms were identified to the species or species group level and, if applicable, to the variety or form level following the taxonomy described by e.g. Gersonde and Zielinski (2000), Armand and Zielinski (2001), Esper et al. (2010) and Esper and Gersonde (2014a, b). Diatom analyses were done by the same investigator at the University of Concepción, Chile, and at Colgate University, USA.

Because diatom distribution in the Southern Ocean is directly associated with the temperature zonation and the frontal systems of the ACC (Cárdenas et al., 2019; Esper et al., 2010; Esper and Gersonde, 2014a, b; Zielinski and Gersonde, 1997), diatom species were grouped into ecological assemblages reflecting (i) seasonal sea ice (associated with temperatures −1.8 to 0 ^∘C), (ii) a cold open ocean (associated with the maximum sea-ice extent in winter and temperatures between 1 and 4 ^∘C), (iii) a warmer open ocean with temperatures between 4 and 14 ^∘C and (iv) benthic–epiphytic habitats (Buffen et al., 2007; Cárdenas et al., 2019). Additionally, a group of reworked diatoms was identified (the specific group composition is described in detail in Sect. S3). A Spearman principal component analysis (PCA) was applied to the diatom assemblages to differentiate their temporal distributions.

For estimation of winter sea-ice (WSI) concentrations, we applied the transfer function MAT-D274/28/4an to the total diatom counts (including Chaetoceros resting spores). The transfer function comprises 274 reference samples with 28 diatom taxa or taxa groups and considers an average of four analogues (Esper and Gersonde, 2014a). Further, the transfer function retrieved the four lowest squared chord distances as a measure for assemblage similarity for each sample depth, which does not automatically equal the four closest geographical reference samples. However, analysis of the geographical location of the retrieved analogues identified four major source regions for the sediment core, with 38.9 % of all reference samples coming from the seasonal sea-ice zone of the Scotia Sea, 33.2 % coming from the summer sea-ice zone of the coastal Amundsen Embayment, 18.4 % coming from the summer sea-ice zone of the coastal Weddell Sea and 8.9 % coming from the summer sea-ice zone of the Ross Sea. Thus, about 72 % of all retrieved samples represent Atlantic Ocean environment, and about 60 % of all retrieved samples represent a polar coastal environment similar to the region of the analysed core. The WSI renders sea-ice concentrations in a 1^∘ by 1^∘ grid for the September average of the period 1981 to 2010 (Reynolds et al., 2002, 2007). The threshold between the open ocean and the sea-ice covered area is set at 15 % of the sea-ice concentration (Zwally et al., 2002), and the average sea-ice edge is defined at 40 % (Gersonde et al., 2005; Gloersen et al., 1993). The estimation of summer sea surface temperature (SSST) came from the transfer function IKM-D336/29/3q, comprising 336 reference samples (Pacific, Atlantic and Indian Southern Ocean) with 29 diatom taxa and 3 factors (Esper and Gersonde, 2014b). The calculations for WSI were done with the software R (R Core Team, 2012) using the packages Vegan (Oksanen et al., 2012) and Analogue (Simpson and Oksanen, 2012).

3 Results

Based on our age model, sediment core PS97/072-1 covers the last 13.8 kyr with a mean sedimentation rate of 67 cm ka⁻¹ and a temporal resolution ranging between 50 and 150 years per sample interval. We note a higher sedimentation rate of 95 cm ka⁻¹ between 5.5 and 3 ka and a few short intervals of lower (19 cm ka⁻¹) and higher (190 cm ka⁻¹) sedimentation (Fig. 2).

Figure 2 Age–depth model for sediment core PS97/072-1 based on eight ¹⁴C-dated calcite samples (black) with error bars and mean sedimentation rates (cm ka⁻¹, dashed blue line). The core top age (red) was estimated as 0.05 ka from matching with the ²¹⁰Pb-dated multicore PS97/072-2 (Vorrath et al., 2020; see Sect. S2).

Organic geochemical bulk parameters (TOC and biogenic opal), concentrations of HBIs (IPSO₂₅ and C_25:3 HBI triene) and diatom species of warmer open-ocean conditions and sea ice-assemblages of piston core PS97/072-1 are summarized in Fig. 3 (additional data can be found in Sect. S4). TOC increases from very low values of 0.1 wt % at 13.7 ka to an average concentration of ∼0.8 wt % between 9.9 ka and the top of the core, with recurring short-lived minima down to 0.03 wt % during the Middle and Late Holocene (Fig. 3f). Some of these TOC minima occur within thin sandy layers of volcanic ash. Biogenic opal shows a similar pattern, with minimum values in the lower part of the record (3.2 wt % at 13.0 ka) and increases throughout the Deglacial to Holocene, with average values of 30 wt % and a maximum of 54.4 wt % at 5.3 ka (Fig. 3e).

Figure 3 Overview of organic geochemical parameters and main diatom assemblages determined in sediment core PS97/072-1 used to characterize the environmental setting over the past 14 kyr. (a) Warm open-ocean diatom assemblage, (b) C_25:3 HBI triene, (c) sea-ice diatom assemblage, (d) IPSO₂₅, (e) biogenic opal and (f) TOC contents. Asterisks in (f) mark layers of volcanic ash, where ** can be linked to a tephra layer in a sediment core from the Bransfield Strait at 5.5 ka (Heroy et al., 2008). Black lines display running averages. Grey-shaded interval refers to the Antarctic Cold Reversal.

Between 13.8 and 13.4 ka, both IPSO₂₅ and HBI triene concentrations are close to or below the detection limit (0.1 µg g⁻¹ TOC). Throughout the record, the IPSO₂₅ concentration ranges between 0.1 to 31.5 µg g⁻¹ TOC, while the concentration of the HBI triene ranges between 0.1 and 6.6 µg g⁻¹ TOC (Fig. 3). IPSO₂₅ is absent before 13.5 ka and rises rapidly to maximum values of 31.5 µg g⁻¹ TOC at 12.8 ka. Subsequently, concentrations decrease steadily until 8.5 ka and then remain at a stable level of ∼4 µg g⁻¹ TOC, with a slightly decreasing trend to 1 µg g⁻¹ TOC between 3.0 ka and the present and smaller peaks of 10 µg g⁻¹ TOC at 6.0 and 3.0 ka. Only traces of the HBI triene occur until 13.0 ka, while its concentration increases up to 6.6 µg g⁻¹ TOC after 8.5 ka, with large fluctuations of more than 5 µg g⁻¹ TOC in the Middle Holocene and from 3.4 ka to the present.

The diatom composition has two contrasting groups indicating open-ocean conditions, a cold-water assemblage and a warm-water assemblage, and a seasonal sea-ice assemblage (Fig. 3; see Sect. S3). Although the group reflecting seasonal sea ice is present throughout the core (mostly >20 %), the highest contributions are seen before 13 ka and between 10.8 and 9.9 ka. The contribution of the warmer open-ocean assemblage is very low in the Deglacial and Early Holocene, rises to highest values in the Middle Holocene and remains around 10 % in the Late Holocene. A biplot of a principal component analysis (PCA) shows the relationship of the ecological groups for three time intervals, with a clear dominance of seasonal sea ice before 13 ka and warmer open-ocean conditions after 8.5 ka (Sects. S5 and S6).

Winter sea-ice concentration estimates based on diatom assemblages (WSI) and the PIPSO₂₅ index, as well as the content of IRD in PS97/072-1, are summarized in Fig. 4a–c. Reconstructed winter sea-ice concentrations (% WSI) derived from the MAT transfer function range from 80 % to 90 % during the ACR and the Deglacial (13.8–11 ka) and exhibit an overall decreasing trend over the Middle Holocene, with fluctuations reaching minimum sea-ice concentrations of ca. 65 % during the Middle and Late Holocene (Fig. 4a). PIPSO₂₅ values show a similar trend, indicating higher sea-ice cover during the ACR, the Deglacial and the Early Holocene (PIPSO₂₅>0.8) and a successive decline to 0.5 on average throughout the Middle and Late Holocene, with a distinct minimum at 0.5 ka (Fig. 4b). IRD (lithic particles and pebbles >500 µm) occurs frequently between 13.8 and 9 ka and is virtually absent in the younger part of the sediment core (Fig. 4c).

Figure 4 Sea-ice-related proxies in sediment core PS97/072-1 with (a) the diatom-based WSI, (b) the sea-ice index PIPSO₂₅ and (c) ice-rafted debris (IRD). For comparison, PIPSO₂₅ values of sediment core (d) JPC10 from the Palmer Deep station (Etourneau et al., 2013) and sea salt sodium (ssNa) records of (e) the EPICA ice core from Dronning Maud Land (EDML) ice core (Fischer et al., 2007) and (f) the Western Antarctic Ice Sheet (WAIS) ice core (WAIS Divide Project Members, 2015) are shown. Black lines display running averages. Grey shaded interval refers to the Antarctic Cold Reversal.

Figure 5 provides ocean temperature anomalies based on diatom assemblages (SSST) and GDGT-derived RI-OH' and TEX${}_{\mathrm{86}}^{\mathrm{L}}$ SOTs in core PS97/072-1 (Fig. 5b–d). Diatom-derived SSST estimates generally depict lower temperatures during the Deglacial and Early Holocene, accompanied by a shift to ca. 1 ^∘C warmer temperatures in the Middle and Late Holocene (Fig. 5b). A short cold event with an SSST decrease of ca. 1.5 ^∘C occurred around 3.1 ka. Similarly to SSSTs, RI-OH'-derived SOTs likewise reflect generally lower temperatures during the Deglacial and Early Holocene and 0.4 ^∘C warmer temperatures in the Middle and Late Holocene (Fig. 5c). TEX${}_{\mathrm{86}}^{\mathrm{L}}$-derived SOTs display an opposite trend to both SSST and RI-OH' SOTs, with peak temperatures during the Deglacial and an overall Holocene cooling towards present (Fig. 5e).

Figure 5 A comparison of (a) December insolation (Laskar et al., 2004), (b) diatom-based SSST, (c) RI-OH'-derived SOT, (d) September insolation (Laskar et al., 2004), (e) TEX${}_{\mathrm{86}}^{\mathrm{L}}$–SOT of sediment core PS97/072-1 and temperature reconstructions, (f) TEX${}_{\mathrm{86}}^{\mathrm{L}}$ from JPC10 at Palmer Deep (Etourneau et al., 2013), (g) TEX₈₆ from ODP1098 at Palmer Deep (Shevenell et al., 2011), and (h) ice core stable isotope record of WAIS Divide (WAIS Divide Project Members, 2013). Ocean temperatures are displayed as anomalies with respect to the mean of the individual SOT and SSST values of the entire record. Black lines display running averages. Grey shaded area refers to the Antarctic Cold Reversal.

4 Discussion

4.1 The late Deglacial (13.8 to 11.7 ka)

In the oldest part of our sediment record, covering the later part of the last Deglacial from 13.8 until 11.7 ka, we observe a remarkable environmental change indicated by large shifts in the TOC, biomarker and diatom records (Fig. 3). The very low concentrations of HBIs (Fig. 3b and d), TOC (Fig. 3f) and biogenic opal (Fig. 3e) between 13.8 and 13.5 ka suggest that primary production of phytoplankton and also sea-ice algae synthesizing IPSO₂₅ were low, while sea-ice-related diatom species show the highest contribution of 73 % (Fig. 3c), albeit with very low concentrations (see online resource). The highest WSI concentrations and PIPSO₂₅ values (Fig. 4a and b) point towards a heavy sea-ice cover and are well in line with peak sea salt sodium (ssNa) concentrations in the EPICA ice core from Dronning Maud Land (EDML) and Western Antarctic Ice Sheet (WAIS) ice core records, referring to an extended sea-ice cover until 13 ka (Fig. 4; EPICA Community Members, 2006; Fischer et al., 2007; WAIS Divide Project Members, 2015). We note that, for the interpretation of PIPSO₂₅ values, changes in both IPSO₂₅ and HBI triene concentrations need to be evaluated carefully to reliably deduce information on sea-ice conditions. High PIPSO₂₅ values may refer to an extended sea-ice cover that lasts until summer (thus hampering phytoplankton productivity and HBI triene synthesis), whereas low PIPSO₂₅ values point to a reduced sea-ice cover in terms of duration (in spring) and/or sea-ice concentration. The near absence of IPSO₂₅, the HBI triene and warm open-ocean diatom species between 13.8 and 13.5 ka evidence a permanent, potentially perennial ice cover or at least sea ice that was too thick to allow photosynthesis of sea-ice algae inhabiting the sea ice. Similarly, Lamping et al. (2020) related the absence of IPSO₂₅ and phytoplankton-derived dinosterol in sediments in the western Amundsen Sea to the re-advancement of a floating ice shelf canopy during the ACR. At the PS97/072-1 core site in the eastern Bransfield Strait, both the presence of perennial sea ice and an ice shelf tongue extending from the APIS could explain the lack of indicators of phytoplankton productivity and IPSO₂₅-synthesizing ice algae. We hence assume that the very low absolute concentrations of sea-ice-associated diatoms result from lateral transport underneath the ice or reworking of sediments older than 13.5 ka. The abrupt increase in IPSO₂₅ concentrations at 13.5 ka  may indicate the retreat or thinning of such an ice canopy, permitting sea-ice algae growth during spring and a subsequent increase in primary production reflected in rapidly rising HBI triene concentrations since 13 ka (Fig. 3b and d). Such a transition from a perennial floating ice canopy to conditions characterized by (seasonal) sea-ice cover is also reported by Milliken et al. (2009) for the nearby Maxwell Bay (King George Island, SSIs) between 14 and 10 ka. Interestingly, a prominent decrease in sea-ice-associated diatoms between 13 and 12 ka (Fig. 3c) is not mirrored by the still-high WSI concentrations. This discrepancy could relate to a weaker preservation potential of certain diatoms reflecting seasonal sea ice (e.g. Synedropsis sp., Nitzschia stellata) that are not considered within the transfer function to estimate WSI, which highlights the need to examine silica dissolution effects for the interpretation of diatom records.

With regard to the ocean temperatures recorded at core site PS97/072-1, we note that the overall cool deglacial temperatures derived from diatom data (SSST) and hydroxylated GDGTs (RI-OH') seem to be linked to low summer insolation (Fig. 5a), whereas higher TEX${}_{\mathrm{86}}^{\mathrm{L}}$ temperatures seem to be associated with high spring insolation (Fig. 5d). The impact of seasonality on GDGT-based ocean temperature estimates is still under debate and would require further improvements in regional calibrations. The observation of maximum abundances of Thaumarchaeota species (producing isoGDGTs applied to determine TEX${}_{\mathrm{86}}^{\mathrm{L}}$) in Antarctic coastal waters during spring (Kalanetra et al., 2009; Murray et al., 1998), however, seems to support our interpretation and also helps to explain the divergent trends in TEX${}_{\mathrm{86}}^{\mathrm{L}}$ and RI-OH'-derived SOT estimates, as the latter proxy might also be sourced by other archaea species that probably grow mostly during the summer season.

While the ACR lasts from 14.7 to 13 ka (Pedro et al., 2016), as indicated by e.g. the WAIS Divide ice core records (Fig. 5i, WAIS Divide Project Members, 2013), our sediment record shows that cold conditions with an extended sea-ice cover, limiting summer phytoplankton productivity (Fig. 4a and b) in the eastern Bransfield Strait, lasted until ca. 11 ka. Further, the Deglacial and Early Holocene IRD content (Fig. 4c; including the presence of single large pebbles) in core PS97/072-1 points to the frequent occurrence of icebergs, evidencing the overall ice sheet disintegration along the WAP that occurred around 14 ka at the SSIs and promoted seasonally open-marine conditions at Anvers-Hugo Trough at 13.6 ka (middle WAP shelf) and at 12.9 ka in Palmer Deep (inner WAP shelf), respectively (Domack et al., 2001; Domack, 2002; Jones et al., 2022; Milliken et al., 2009; Roseby et al., 2022). At our core site, rising RI-OH' SOTs and a slight decrease in PIPSO₂₅ values characterize the late Deglacial between 13 and 11.7 ka (Figs. 4b and 5c). A prominent decline in large-scale sea-ice cover is also reflected in the decreasing ssNa concentrations in the EDML and WAIS ice cores between 13 and 11.7 ka (Fig. 4e and f), likely related to a distinct atmospheric warming, as reflected in ice core stable water isotopes (Fig. 5h).

The ACR cooling and the subsequent late Deglacial warming may relate to inter-hemispheric teleconnections through a global reorganization of atmospheric and ocean circulation that is associated with the bipolar seesaw pattern of opposite climate trends between the Northern Hemisphere and the Southern Hemisphere (Anderson et al., 2009; Broecker, 1998; EPICA Community Members, 2006; Pedro et al., 2016; Stenni et al., 2011). While a northward shift of the southern westerlies during the ACR (Fletcher et al., 2021) promoted Antarctic sea-ice expansion and glacier readvance (potentially causing an ice cover over the PS97/072-1 core site), a cooling of the Northern Hemisphere with a southward shift of the intertropical convergence zone and the Southern Hemisphere westerlies (Lamy et al., 2007) resulted in intensified wind stress in the Drake Passage (Timmermann et al., 2007). This pattern would have increased upwelling that may have driven the continued ocean warming and sea-ice retreat in Antarctica towards the Holocene (Anderson et al., 2009).

4.2 Early Holocene warming from 11.7 to 8.2 ka

The Early Holocene from 11.7 to 8.2 ka is characterized by a progressively decreasing spring sea-ice cover inferred from declining PIPSO₂₅ values (Fig. 4b), as well as highly variable winter sea-ice cover with prominent shifts in sea-ice concentration (from 90 % to 65 %; Fig. 4a). These WSI fluctuations are not reflected in the sea ice diatom assemblage, which, similar to the biogenic opal content, follows an increasing trend until 10.5 ka (Fig. 3c and e). Increased accumulation of biogenic opal and a better preservation of (thin-walled) sea-ice-related diatoms that are not used for the transfer function may explain the mismatch between the WSI record and the sea-ice diatom assemblage. The increase in biogenic opal is further accompanied by rising TOC content, while concentrations of the HBI triene and warm open-ocean diatoms remain low, with an increase only after 9 ka, signalling higher phytoplankton productivity (Fig. 3a and b). Diatom-derived SSSTs exhibit marked fluctuations but remain relatively low until 8.2 ka (Fig. 5b). RI-OH' and TEX${}_{\mathrm{86}}^{\mathrm{L}}$ SOTs display diverging trends following the summer and spring insolation, respectively (Fig. 5). While PIPSO₂₅ values display a gradual decrease in sea-ice coverage, the WSI record suggests a highly variable sea-ice cover, with several distinct sea-ice minima between 11 and 10 ka and around 9 ka (Fig. 4a and b). These sea-ice minima may have resulted from punctuated warming events, e.g. around 10 ka, when SSST shows a short temperature peak, which might have led to a delayed sea-ice formation in autumn and winter (Fig. 5b). Another WSI minimum at 9 ka coincides with a major, final peak in IRD deposition at the core site (Fig. 4), evidencing iceberg discharge during episodes of peak AP ice sheet retreat and enhanced calving (Jones et al., 2022). As sea-ice melting may have been an important driver of ocean stratification, we suggest warmer, stratified surface waters with moderate production in summer, supported by increasing summer insolation (Fig. 5a). Ameliorating climate conditions, ice shelf retreat along the NAP and the establishment of modern-like ocean conditions after 9 ka have also been proposed for the western Bransfield Strait by Heroy et al. (2008) and are well in line with the rising concentrations of warm open-ocean diatoms and the phytoplankton-derived HBI triene at our core site after 9 ka (Fig. 3). The general decrease in spring sea-ice cover (reflected in declining PIPSO₂₅ values) may have been fostered by high spring and rising summer insolation (Fig. 5a and d), shortening the duration of sea-ice cover. Rising RI-OH' temperatures are consistent with the overall slight warming trend recorded in the WAIS Divide ice core (Fig. 5h), which has been shown to be mainly driven by increasing summer temperatures (Jones et al., 2022). The decreasing TEX${}_{\mathrm{86}}^{\mathrm{L}}$ SOT trend at core site PS97/072-1 corresponds to the declining TEX₈₆ temperatures reported for ODP site 1098 in Palmer Deep (Fig. 5g; Shevenell et al., 2011), though the latter displays a more pronounced temperature drop (of ca. 6 ^∘C) between 11.7 and 8.2 ka. These regional differences may relate to changing ocean circulation patterns, associated shifts in water mass distribution along the WAP and the local post-glacial environmental development during the Early Holocene. Deposition of laminated diatom oozes in the Anvers-Hugo Trough at the WAP middle shelf during the early Holocene, for example, documents episodes of extremely high productivity in response to a southward shift of the Southern Hemisphere westerlies and the advection of warm and nutrient-rich CDW (Circumpolar Deep Water) (Roseby et al., 2022). We propose that the eastern Bransfield Strait remained mainly inaccessible for CDW and BSW until further ice recession between 10 and 5 ka (Ó Cofaigh et al., 2014, and references therein) permitted advection of these water masses into the Bransfield Strait.

4.3 Middle Holocene from 8.2 until 4.2 ka

The Middle Holocene from 8.2 to 4.2 ka was a period of sea-ice retreat and minimum iceberg activity at the core site, indicated by decreasing WSI and PIPSO₂₅ values and virtually absent IRD (Fig. 4). Diatoms associated with warmer open-ocean conditions, peak HBI triene concentrations and maximum TOC, as well as biogenic opal contents (Fig. 3), indicate a high export production during the Middle Holocene. This higher primary productivity can be linked to a decrease in both winter and spring sea ice, indicated by WSI and PIPSO₂₅ minima, respectively (Fig. 4a and b), and also to elevated SSSTs and (summer) SOTs (Fig. 5b and c) promoting ice-free summer ocean conditions favourable for phytoplankton productivity. These Middle Holocene sea-ice conditions compare well with the modern situation at the core site, characterized by a seasonal decrease in sea-ice concentration from 50 % during winter to mainly ice-free summers (NSIDC; Cavalieri et al., 1996).

The continued retreat of the previously grounded APIS adjacent to the Bransfield Strait between 10 and 5 ka finally opened the passage for ACC waters to enter the Bransfield Strait from the west (Bentley et al., 2014; Ó Cofaigh et al., 2014). As a result, we suggest that sea-ice conditions at our core site were influenced by incursions of warmer ocean waters carried with the ACC (i.e. BSW and CDW), while cold-water inflow and sea-ice advection from the Weddell Sea were diminished due to the still-grounded ice sheet at the tip of the AP (Ó Cofaigh et al., 2014), leading to a shorter sea-ice season in the eastern Bransfield Strait. This shift towards a warmer, less-ice-covered ocean setting in the eastern Bransfield Strait is reflected in the transition from proximal to distal glaciomarine conditions in Maxwell Bay (Milliken et al., 2009) and may be associated with the Middle Holocene climatic optimum. This timing contrasts the notion of Heroy et al. (2008), who confined the Middle Holocene climatic optimum to a shorter time interval between 6.8 and 5.9 ka based on diatom assemblage analyses of a sediment core in the western Bransfield Strait. We propose that this temporal offset may relate to regionally different responses, glacial retreat patterns impacting oceanic pathways and the position of frontal systems controlling primary productivity within Bransfield Strait. The generally decreasing WSI and variable PIPSO₂₅ values further depict different trends compared to PIPSO₂₅ values determined for the JPC10 in Palmer Deep (Fig. 4d; Etourneau et al., 2013), which suggest an overall increase in spring sea ice along the WAP until 4.2 ka. Though minima in spring sea ice at 7.5, 6.5 and 5.4 ka at core site PS97/072-1 may be related to PIPSO₂₅ minima observed for JPC10, the lack of Middle-Holocene-aged tie points in our core from the Bransfield Strait prevents us from making conclusions about a common driver for these sea-ice reductions.

Regarding ocean temperatures, we observe a sustained warming in RI-OH' SOT, punctuated by a cooling at 5.5 ka (Fig. 5c), while TEX${}_{\mathrm{86}}^{\mathrm{L}}$ temperatures depict a subtle cooling of ca. 0.5 ^∘C between 8.2 and 7 ka followed by a warm reversal until 6 ka and a further cooling until 4.2 ka (Fig. 5e). This Middle Holocene slight cooling trend is also observed in the TEX${}_{\mathrm{86}}^{\mathrm{L}}$ records from the core sites in Palmer Deep at the WAP (Fig. 5f and g; Etourneau et al., 2013; Shevenell et al., 2011). The similarity between these records encourages us to assume that these TEX₈₆-derived temperatures from along the WAP and NAP are driven by spring insolation rather than being a reflection of annual mean ocean temperature conditions.

4.4 Late Holocene and Neoglacial from 4.2 ka until today

The Late Holocene, covering the past 4.2 kyr, is characterized by highly variable winter sea ice and decreasing spring sea-ice cover at core site PS97/072-1, as indicated in the MAT-derived WSI and a decline in PIPSO₂₅ values over the past 2 kyr (Fig. 4a and b). Rather constant biogenic opal and TOC contents (Fig. 3e and f), however, suggest that primary productivity remained relatively unaffected by this reduction in spring sea-ice cover. While decreasing IPSO₂₅ concentrations between 2.5 ka and the core top (Fig. 3d) suggest a reduced productivity of the sea-ice diatom species synthesizing this molecule, no significant changes are observed in the sea-ice diatom assemblage (Fig. 3c), which supports the assumption that only a restricted group of diatoms – at least Berkeleya adeliensis – produce IPSO₂₅ (Belt et al., 2016). The warm open-ocean diatom assemblage follows an overall declining trend throughout the Late Holocene, which is not reflected in the highly variable and slightly increasing HBI triene concentrations (Fig. 3a and b), and a prominent decrease in HBI triene concentrations occurs only at 1 ka. While the observation of cooler sea surface temperatures and a diminished spring sea-ice cover indicated by the joint decrease in the warm open-ocean diatom assemblage and PIPSO₂₅ values since 2 ka may seem counterintuitive, Milliken et al. (2009) report a similar development in Maxwell Bay since 2.6 ka. Interestingly, records of diatom and radiolarian assemblages of a sediment core (Gebra-2) collected in close vicinity to PS97/072-1 document an overall increase in sea-ice taxa over the past 3 kyr, with distinct Neoglacial events characterized by higher (denser and longer) sea-ice cover (Bárcena et al., 1998). The lower sampling resolution and missing age control for the past 3 kyr in PS97/072-1, however, hamper a more detailed comparison of diatom species in our core with those investigated for Gebra-2. The Neoglacial increase in spring sea-ice cover is also indicated by a prominent rise of PIPSO₂₅ values determined for JPC10 in Palmer Deep (Fig. 4d; Etourneau et al., 2013). Similarly, deposition of ssNa in the EDML ice core (Fischer et al., 2007) has increased since 2 ka.

Minimum PS97/072-1 PIPSO₂₅ values at 0.5 ka result from notably reduced IPSO₂₅ and HBI triene concentrations (Fig. 3b and d). While this pattern of minimum HBI triene and IPSO₂₅ concentrations is similar to the period between 13.8 and 13.5 ka, which was characterized by cold conditions and a pronounced – potentially perennial – ice cover, the elevated TOC and biogenic opal values, as well as the presence of diatoms associated with warm open-ocean conditions at 0.5 ka, point to favourable ocean conditions. We hence relate this drop in HBI concentrations to a shift in the diatom community rather than to an abrupt readvance of ice cover.

Late Holocene ocean temperature reconstructions for core PS97/072-1 display different patterns. Generally increasing diatom-derived SSSTs are only punctuated by a cooling event at 3.1 ka, while RI-OH' SOT remains relatively constant with a very subtle cooling of ca. 0.2 ^∘C between 1.5 ka and the present, which could be linked to the slight decrease in summer insolation (Fig. 5a–c). The decrease in TEX${}_{\mathrm{86}}^{\mathrm{L}}$ SOT by about 1 ^∘C between 4 and 3.3 ka in eastern Bransfield Strait is also depicted in the TEX${}_{\mathrm{86}}^{\mathrm{L}}$ data from the Palmer Deep core JPC10 (Fig. 5e and f; Etourneau et al., 2013). The following warming reflected in PS97/072-1 TEX${}_{\mathrm{86}}^{\mathrm{L}}$ SOT until ca. 2 ka may relate to the establishment of open marine conditions fostering primary productivity at the Perseverance Drift north of Joinville Island (northern tip of the AP) as a result of warm-water intrusions (Kyrmanidou et al., 2018). This warming is reversed by another cooling at about 2 ka – coincident with an abrupt temperature increase of ca. 4 ^∘C depicted in the ODP1089 TEX₈₆ SOT record in Palmer Deep (Fig. 5g; Shevenell et al., 2011). The latter warming is not displayed in the TEX${}_{\mathrm{86}}^{\mathrm{L}}$ data of the nearby JPC10, and we relate this contrast to the different approaches used to determine SOT (i.e. TEX₈₆ vs. TEX${}_{\mathrm{86}}^{\mathrm{L}}$ omitting the crenarchaeol regio-isomer, which seems to be less important for membrane adaptation in polar waters; Kim et al., 2010).

Evidently, temperature trends at the AP in the Late Holocene are highly variable between different areas (Allen et al., 2010; Barbara et al., 2016; Bárcena et al., 1998; Bentley et al., 2009; Etourneau et al., 2013; Mulvaney et al., 2012; Shevenell et al., 2011), and this is likely associated with the complex oceanographic and atmospheric settings. This heterogeneous pattern, however, contrasts with the currently observed large-scale ocean warming along the AP driven by intrusions of ACC-derived warm CDW onto the continental shelf of the WAP (Couto et al., 2017) and the NAP (Ruiz Barlett et al., 2018), as well as the overall loss of sea ice (Parkinson and Cavalieri, 2012), which supports the assumption that the recent changes impacting the AP already exceed natural variability.

5 Conclusions

We reconstructed the sea ice and climate development at the NAP since the last Deglacial using the sediment core PS97/072-1 from the eastern Bransfield Strait. Pursuing a multi-proxy approach that focuses on organic geochemical bulk and biomarker analyses, diatom assemblage studies and transfer functions, and IRD data, we identified different Deglacial and Holocene environmental conditions impacted by sea ice and ocean temperature changes. Our results reveal the retreat of a perennial ice cover after the ACR and an overall sea-ice reduction and warming summer ocean temperatures during the Holocene. The late Deglacial, from 13.8 to 11.7 ka, was a highly dynamic period: until 13.4 ka, primary productivity was low due to a permanent ice cover during the ACR. The ACR terminated with a shift to slight warming conditions at 13 ka along with a reduction in the length of the sea-ice season, which permitted phytoplankton productivity, at least during summer. The Early Holocene, from 11.7 to 8.2 ka, was characterized by increasing summer ocean temperatures, further decreasing sea-ice cover in terms of duration and/or sea-ice concentration, and highly variable winter sea-ice cover. In the Middle Holocene, from 8.2 to 4.2 ka, increased advection of BSW and CDW led to a shortened sea-ice season confined to winter and spring and also to rising summer ocean temperatures which fostered primary production, indicating the Middle Holocene climatic optimum. During the Late Holocene, the core site experienced distinct fluctuations in WSI, with concentrations shifting between 90 % and 60 %, while PIPSO₂₅ values declined continuously, suggesting a less intensive or shorter spring sea-ice cover. We note that GDGT-based TEX${}_{\mathrm{86}}^{\mathrm{L}}$ and RI-OH' SOTs correspond to spring and summer insolation, respectively, which may explain the divergent trends displayed by both SOT proxies. Clearly, while this observation may help with the interpretation of other Southern Ocean GDGT-based temperature estimates and the reconstruction of seasonal SOT variability, more investigations into the mechanisms driving GDGT synthesis in polar waters are needed.