Six individual sediment cores were collected along a north-south transect from McMurdo Sound coastal polynya to south of Terra Nova Bay in the southwestern Ross Sea in February 2014 and 2015 by the Korean Polar Research Institute vessel RV Araon during cruises ANA04C and ANA05B (Fig. 1, Table 1). Sea ice cover at the core sites is high (60 %–80 %, increasing to >80 % farther offshore) relative to the adjacent McMurdo Sound and Terra Nova Bay polynyas (Spreen et al., 2008; Matsuoka et al., 2021), reflecting northward advection of sea ice by katabatic winds and ocean currents, which accumulates against the Drygalski Ice Tongue (Arrigo and van Dijken, 2004; Brett et al., 2020). X-radiographic images of RS14-BC02, RS14-BC03, and RS14-BC04 were taken on 1 cm thick sediment slabs with Softex M-100W X-ray source and NTB EZ-320 digital X-ray scanner and were processed with iX-Pect EZ imaging software (Fig. A1). The three box cores (BC02, BC03, BC04) are mainly composed of olive grey diatomaceous mud, with evidence of bioturbation throughout (Figs. A1 and A2). Crude lamination is observed at 42–45 cm in BC03 and at 0–5 cm in BC04 (Figs. A1 and A2). Only the top 0–1 cm of the gravity cores (GC72, GC78, GC80) were subsampled and are composed of olive grey diatomaceous mud and no noticeable sediment loss or disturbance was observed during coring and recovery. Twenty-seven sediment samples were subsampled from these cores at various depth resolutions (see Table 1). The first subsamples in the cores are sampled from either 0–1 cm depth or 1–2 cm depth; these samples are referred to as “core top” throughout this study.

Age models were derived using 210Pb dating. 210Pb is produced in the atmosphere via decay of 222Rn, then deposited onto the ocean surface and scavenged from the water column by settling particles, accumulating in seafloor sediments where it decays with a half-life of 22.3 years (Appleby, 2002). Total 210Pb activity was determined indirectly by measuring its granddaughter 210Po via alpha spectrometry (Flynn, 1968) and assuming secular equilibrium between 210Pb and 210Po in sealed samples. This provides a proxy for total 210Pb activity, which was used to calculate excess (unsupported) 210Pb after subtraction of supported 210Pb, assumed to be in secular equilibrium with in situ 226Ra.

Sediment samples from cores BC02, BC03, and BC04 were analysed at the Institute of Environmental Science and Research Limited (ESR) in Christchurch, New Zealand for 210Po using alpha spectrometry (Appleby, 2002; Flynn, 1968). Sedimentation rates and sediment ages were calculated following Appleby (2002). Briefly, the sedimentation rate was calculated from the least-squares fit slope of excess 210Pb activity plotted on a logarithmic scale against depth. Raw 210Pb profiles showed evidence of a surface mixed layer in BC02 and BC04, consistent with bioturbation observed in the core logs. The uppermost two samples from BC02 and one sample from BC04 were therefore excluded from the least-squares regression. No mixed layer was identified in BC03 and all samples were retained. The constant flux and constant sedimentation rate (CFCS) model was applied, as sediment bulk density data were unavailable, precluding the use of mass-depth calculations required by alternative models such as CRS. Errors were estimated by Monte Carlo simulations using the R package “serac” (Bruel and Sabatier, 2020).

The CFCS model produced age estimates between 1856–2007 (±37 years at the base) for BC02 (Table 1, Fig. A2), with each centimetre of sediment representing approximately 4 years; sampling resolution was every 18–37 years. For BC03, age estimates spanned 1843–2008 (±5 years at the base), with ∼3.6 yr cm⁻¹ and a sampling resolution of ∼24 years. BC04 was dated to 1794–2007 (±21 years at the base), with ∼3.3 yr cm⁻¹ and sampling every 27–36 years.

Approximately 10–15 mg of dried, homogenised, and inorganic carbon-removed sample was analysed for total organic carbon (TOC) and nitrogen (TON) via dynamic flash combustion (The FlashSmart Elemental Analyzer Thermo Fisher Scientific). CO2 and N2 gases are detected by a Thermal Conductivity Detector, and the results are calibrated against three certified standards, resulting in an error of 5 % for carbon and 8 % for nitrogen, based on replicate measurements of soil certified standards.

Organic geochemical extraction and analysis were performed in the joint Earth Sciences New Zealand/Te Herenga Waka-Victoria University of Wellington Organic Geochemistry Laboratory following Naeher et al. (2012) and Verret et al. (2025). Briefly, approximately 7 g of freeze-dried and homogenised sediment were extracted by ultrasonication using dichloromethane(DCM)/methanol(MeOH) (3 : 1, v : v). An internal standard containing 5a-cholestane, n-C19 alcohol, and n-C19:0 fatty acid was added to the total lipid extract (TLE) to calculate relative concentrations. The TLE was saponified with 6 % KOH in MeOH and the fatty acids were separated from the neutral compounds and derivatised by boron trifluoride in MeOH to fatty acid methyl esters (FAMEs). To determine the number and position of double bonds, the FAME fractions were derivatised with 2-amino-2-methyl-1-propanol (AMP) to form 2-alkenyl-4, 4-dimethoxyloxazline (DMOX) derivatives. The neutral compounds were separated into three fractions by silica gel column chromatography using n-hexane, n-hexane : DCM (1 : 2, v : v), and DCM : MeOH (1 : 1, v : v) to obtain apolar fraction containing HBIs (F1), the ketone fraction (F2), and the polar fraction (F3). The F3 fraction was derivatised prior to analysis using pyridine and N, O-Bis(trimethylsilyl)trifluoroacetamide. Procedural blanks were used for data quality control and to investigate any potential laboratory contamination.

Samples were analysed on a gas chromatography mass spectrometer at Earth Sciences New Zealand (GC-MS; Agilent Technologies, model: 7890A and 5975C) with a 60 m-long capillary column (Agilent J&W DB-5MS, inner diameter of 250 µm, and film thickness of 0.25 µm), and helium as the carrier gas (1 mLmin-1). Fatty acids and apolar fractions were injected on the GC-MS splitless at an inlet temperature of 300 °C, and the MS was operated in full scan (m/z 50–700). The temperature program of the oven was 70 °C, 70–150 °C at 20 °Cmin-1, 150–320 °C at 4 °Cmin-1 and then kept isothermal at 320 °C for 15 min. The derivatised polar fraction and DMOX fatty acid derivatives were injected and analysed as above, except the GC-MS was kept isothermal at 320 °C for up to 25 min.

Lipids were identified by retention time (time taken for a compound to elute from the GC column), total ion current (TIC) mass spectra output, and fragmentation patterns. Biomarker concentrations were quantified using the TIC integrated peak area of each compound and known concentrations of an internal standard, then normalised to the TOC content of the sediment. Biomarker proxies and indices were calculated to identify biomarker sources and study environmental changes.

The carbon preference index for fatty acids (CPI), which indicates the amount of microbial activity or amount of modification (Kalpana et al., 2021), was calculated for each sample using Eq. (1). 1CPI=∑C12-28(even)∑C13-29(odd), Average chain length (ACL), which indicates the dominant plant source, e.g. terrestrial plant material/ocean algae, was calculated using the concentrations of saturated fatty acids following Eq. (2): 2ACL14-32=∑(Cn×n)∑Cn, The C16:1ω7/C16:0 ratio was calculated to identify diatom sources (Nichols et al., 1993, 1986; Skerratt et al., 1995, 1998; Smik et al., 2016b; Smith et al., 1986). The C15 fatty acid index indicates levels of microbial activity and was calculated following Eq. (3): 3C15fatty acid index=(iC15+aC15)C15, IPSO₂₅ was quantified using an internal standard and normalised to TOC. The phytoplankton IPSO₂₅ index (PIPSO₂₅) was calculated using brassicasterol (P_BIPSO₂₅) and dinosterol (P_DIPSO₂₅) following nomenclature and Eq. (4) by Vorrath et al. (2019): 4PIPSO25=IPSO25IPSO25+(c×phytoplankton marker), The balance factor is applied to account for concentration differences between IPSO₂₅ and the phytoplankton biomarker (c = mean IPSO25/mean phytoplankton biomarker) (Belt and Müller, 2013; Smik et al., 2016a; Vorrath et al., 2019). C factors were calculated individually for each core site.

Quantitative diatom slides were prepared following Crosta et al. (2020) at Environnements et Paléoenvironnements Océaniques et Continentaux (EPOC), University of Bordeaux, to investigate diatom composition and abundances. Briefly, a subsample of dried sediment was immersed in a hydrogen peroxide and anhydrous tetra sodium pyrophosphate solution on a hotplate until reaction with organic matter was complete. The samples were centrifuged, diluted, and one drop was transferred onto a coverslip immersed in a petri dish. The water was decanted, and the dry coverslip was then glued onto a slide using NOA61. Diatom analysis was performed on an Olympus BH2 phase contrast photomicroscope at a magnification of ×1000. Two slides were made per sample, and at least 300 diatom valves were counted per sample across the two slides following the method outlined by Crosta and Koç (2007). Diatoms were identified to species level where possible, or species group level, following the taxonomic criteria detailed in Al-Handal and Wulff (2008), Almandoz et al. (2008), Andreoli et al. (1995), Baldauf and Barron (1991), Saggiomo et al. (2021), and Warnock and Scherer (2015). Absolute diatom abundances (ADA) were calculated by scaling counted valves to the total processed sample volume and normalising to sediment dry weight, yielding concentrations expressed as valves per gram of dry sediment (Crosta et al., 2008). The relative abundance of each species/group was determined as the proportion of the species/group to the whole diatom assemblage in each sample.

To assess the sources of variance and relationships among biomarker and diatom variables, a Principal Component Analysis (PCA) was performed on the combined dataset containing core top and downcore samples (n=27) in R (version 4.4.1, R Core Team, 2024) using prcomp, with variables log-transformed (log⁡(x+1)). Variable contributions to the first two dimensions were calculated and visualised using the factoextra package (version 1.0.7; Kassambara and Mundt, 2016), with results displayed as a colour gradient on the correlation circle (Fig. 8).

To differentiate between phytoplankton and microbial sources in the southwestern Ross Sea, we applied a multi-proxy biomarker approach incorporating fatty acid profiles, HBIs, sterols, and diatom assemblages. This framework enables us to distinguish between pelagic and sympagic diatom communities, sea ice-associated taxa, P. antarctica, and bacterial inputs. While saturated fatty acids are produced by most organisms, variations in chain length and relative dominance can indicate different sources. A predominance of C16:0 is commonly associated with diatoms (Skerratt et al., 1995; Nichols et al., 1986; Smik et al., 2016b), while C18:0 and unsaturated C18 fatty acids such as C18:1ω9 are more characteristic of P. antarctica as shown by natural blooms and pure culture studies (Skerratt et al., 1995, 1998; Nichols et al., 1986; Hamm et al., 2001). However, saturated fatty acids are less diagnostic as they have multiple sources and can be derived from the degradation of unsaturated compounds. Diatom dominated assemblages also exhibit elevated concentrations of C16:1ω7 and polyunsaturated fatty acids (PUFAs) such as C20:5ω3 and C20:4ω6, as well as high C16:1ω7/C16:0 ratios, as shown by natural diatom blooms across Antarctica (Skerratt et al., 1995; Nichols et al., 1986; Smik et al., 2016b). Sea ice dwelling diatom communities from cultures and Ross Sea samples are further distinguished by long-chain monounsaturated fatty acids, including C24:1ω11, C24:1ω9, C26:1ω11, and C26:1ω9 (Nichols et al., 1993, 1986). Although diatom cultures contain C20:1ω9 (Skerratt et al., 1998), this compound is also associated with zooplankton (copepod) (Yang et al., 2016; Zhang et al., 2026). Bacterial contributions are inferred from branched-chain fatty acids, particularly iso- and anteiso-C15:0 and C17:0, which occur in pairs (Skerratt et al., 1995; Kaneda, 1991). In addition, C18:1ω7 is produced by several bacterial taxa (Wilson et al., 2010). These patterns, summarised in Table 2, provide a basis for interpreting fatty acid distributions in terms of source inputs.

The HBI biomarker IPSO₂₅ diene is used across the Southern Ocean to reconstruct paleo Antarctic sea ice (Belt, 2018; Johnson et al., 2021; Tesi et al., 2020; Etourneau et al., 2013; Schmidt et al., 2018; Lamping et al., 2020, 2021; Sadatzki et al., 2023; Vorrath et al., 2023, 2020), and has also been used in conjunction with two open-water biomarkers, two sterols, brassicasterol (24-methylcholesta-5,22E-dien-3β-ol) and dinosterol (4α,23,24trimethyl-5α-cholest-22E-en-3β-ol) (Smik et al., 2016a). In the Southern Ocean, the index has been identified as PIPSO₂₅ by Vorrath et al. (2019) and has been used around the Bransfield Strait (Lamping et al., 2021; Vorrath et al., 2023, 2020), the Amundsen Sea (Lamping et al., 2020), and near Wilkes Land (Sadatzki et al., 2023). In Antarctic seawater and sediment, high concentrations of brassicasterol are typically associated with P. antarctica blooms (Skerratt et al., 1995; Villinski et al., 2008), consistent with the high concentrations of brassicasterol found in Phaeocystis cultures (Nichols et al., 1991). Brassicasterol was present in 37 % of diatom cultures studied by Rampen et al. (2010), most notably pennate diatoms, with species such as Stauroneis constricta, Phaeodactylum tricornutum, and Dickieia ulvacea containing particularly high proportions (>60 % of total sterols). However, none of the species noted in Rampen et al. (2010), were observed in our diatom assemblages. Brassicasterol has been found to correlate with sea ice diatom abundance and spring sea ice extent (Guo et al., 2024). Dinosterol, although occasionally detected in diatom blooms (Skerratt et al., 1995), is primarily produced by dinoflagellates and is considered a source-specific indicator of this group (Skerratt et al., 1995; Villinski et al., 2008; Thomson et al., 2004; Volkman, 2006; Wisnieski et al., 2014). However, the southwestern Ross Sea is dominated by diatoms rather than dinoflagellates (Arrigo et al., 2000, 2015; McMinn et al., 2010; Noble et al., 2013) which may lead to a larger diatom source of dinosterol. We use these biomarkers to investigate the sources and environmental drivers of phytoplankton-derived compounds in our surface samples, as well as to interpret downcore changes in sea ice extent using HBI concentrations and sterol-based ratios.

To complement biomarker analysis, we assessed diatom assemblages, and diatom taxa were grouped according to their ecological affinities following previous studies (Leventer and Dunbar, 1988; Leventer et al., 1993; Crosta et al., 2004; Armand et al., 2005; Crosta et al., 2005; Campagne et al., 2016). These groupings are further supported by the PCA performed on the three box cores analysed in the present study (Fig. 8). The Sea Ice group comprises Entomoneis spp., Nitzschia stellata, Synedropsis fragilis, Synedropsis recta, and Synedropsis hyperboreoides. These taxa form part of the bottom-ice community (Horner, 1985), inhabiting platelet ice beneath fast ice (Riaux-Gobin et al., 2000), and are also found in polynyas following seasonal ice break-up (Riaux-Gobin et al., 2003). Owing to their lightly silicified frustules, these species are rarely preserved in sediments. When present, they indicate high ice-associated primary production under favourable light conditions and exceptional preservation (McMinn and Ryan, 2000). Their elevated abundance in sediments is generally interpreted as evidence of fast-ice melting (Leventer and Dunbar, 1988).

The F. curta group includes Fragilariopsis curta, F. cylindrus, and F. vanheurckii. These taxa occur within the phytoplankton assemblage associated with or beneath sea ice (Leventer and Dunbar, 1987; Gersonde and Zielinski, 2000). High abundances of F. curta and F. cylindrus characterise environments with extensive sea ice cover (Burckle et al., 1987) and strongly stratified surface waters (Beans et al., 2008). Although some lightly silicified F. cylindrus valves may be lost during settling (Burckle et al., 1987), their distribution in surface sediments exhibits a pronounced increase towards regions of high sea ice concentration, indicating that sedimentary abundances largely reflect surface production (Armand et al., 2005; Esper and Gersonde, 2014). Furthermore, high abundances of the F. curta group occur within the early spring laminae off Adélie Land (Denis et al., 2006; Maddison et al., 2012). Collectively, these taxa are considered indicative of persistent fast or pack ice that melts relatively late in the seasonal cycle (Armand et al., 2005; Campagne et al., 2016).

The F. cryophilic group consists of Fragilariopsis obliquecostata, F. sublinearis, and F. ritscherii. These species are associated with cold waters and extensive sea ice cover (Armand et al., 2005), but generally bloom later in the growing season, as indicated by their occurrence within late spring laminae (Denis et al., 2006; Maddison et al., 2012). Their heavily silicified frustules are well preserved in sediments. In particular, the distribution of F. obliquecostata in offshore sediments has been used to reconstruct the past position of the summer sea ice margin (Gersonde and Zielinski, 2000).

The E. antarctica group comprises terminal and intercalary valves of Eucampia antarctica var. recta, the only variety identified in western Ross Sea sediments. High abundances of this taxon have been reported from both the water column (Fryxell and Prasad, 1990) and modern sediments of the southern Weddell Sea (Zielinski and Gersonde, 1997; Armand et al., 2005), as well as deglacial sediments from the Antarctic Peninsula (Peck et al., 2015; Barbara et al., 2016). The heavily silicified valves are generally well preserved in sediments, and elevated abundances are widely interpreted as indicators of glacial or sea ice meltwater influence (Burckle, 1984; Cunningham et al., 1999; Peck et al., 2015).

The Centric Cold Water group comprises Actinocyclus actinochilus, Shionodiscus ritscheri, Stellarima microtrias, and Thalassiosira tumida. These taxa develop in turbulent, cold surface waters during the summer ice-free season (Fryxell, 1989), potentially following inoculation from sea ice (Garrison et al., 1987). Their heavily silicified frustules are generally well preserved in sediments (Cunningham and Leventer, 1998; Stickley et al., 2006). The Open Ocean group comprises several taxa present at relatively low abundances in the studied box cores, including Actinocyclus curvatulus, Asteromphalus hookeri, A. hyalinus, A. parvulus, Corethron pennatum, Coscinodiscus oculus-iridis, Fragilariopsis kerguelensis, Shionodiscus gracilis, S. trifulta, Thalassiosira lentiginosa, T. maculata, T. oliverana, the cold variety of T. antarctica, Thalassiothrix spp., and Trichotoxon reinboldii. Most of these species characteristically inhabit open-ocean environments (Hasle, 1969; Fenner et al., 1976; Krebs et al., 1987; Froneman et al., 1995; Grigorov et al., 2014; Rigual-Hernández et al., 2015). Their heavily silicified frustules favour excellent preservation in sediments (Crosta et al., 2005; Rigual-Hernández et al., 2016). An exception is C. pennatum, a tubular species that is generally preserved only following mass-bloom events (Torricella et al., 2026). Taxa belonging to both the Centric Cold Water and Open Ocean groups are typically associated with summer laminae (Denis et al., 2006; Maddison et al., 2012) and indicate relatively warm conditions and prolonged ice-free periods promoted by incursions of warm water masses (Cunningham and Leventer, 1998; Maddison et al., 2006).

The Chaetoceros resting spore (CRS) group comprises vegetative cells and resting spores of Chaetoceros subgenus Hyalochaete, with resting spores predominating. This group may dominate diatom assemblages in Antarctic coastal waters (Leventer, 1991; Annett et al., 2010) and underlying sediments (Armand et al., 2005), reflecting periods of elevated primary productivity in stratified surface waters that ultimately lead to nutrient depletion (Annett et al., 2010). Sustained production nevertheless requires periodic nutrient replenishment through wind-driven mixing, as these opportunistic bloom-forming taxa are otherwise outcompeted by species adapted to low-nutrient conditions (Annett et al., 2010; Campagne et al., 2016). High abundances of CRS have been reported from both spring and summer laminae off Adélie Land (Denis et al., 2006; Maddison et al., 2012), indicating production throughout the growing season under favourable environmental conditions. Similarly, abundant CRS in deglacial Antarctic shelf sediments (Cunningham et al., 1999; Leventer et al., 2006) have been linked to enhanced surface-water stratification associated with retreating ice sheets and increasingly prolonged ice-free summers (Gilmer et al., 2025). The heavily silicified, unornamented resting spores exhibit excellent preservation potential.

Antarctic shelf waters are among the most productive marine environments (Arrigo et al., 2008). Water depths are generally less than 1000 m, while sedimentation is rapid and episodic (Schloss et al., 1999; Kim et al., 2015). Consequently, in this region diatom preservation is typically excellent, and sediments are characterised by exceptionally high biogenic silica contents (Ledford-Hoffman et al., 1986; Dunbar et al., 1998; Presti et al., 2003; Denis et al., 2009; Chiarini et al., 2019; Gilmer et al., 2025). Together, these assemblages were used alongside biomarker evidence to reconstruct spatial patterns in phytoplankton community structure and infer recent environmental changes in the southwestern Ross Sea.

Fifty-eight species or groups of species were identified in the sediment samples. The complete list of species and data is available at PANGAEA (de Jong et al., 2026). The diatom assemblage in all the sediment samples is characterised by high abundances of Fragilariopsis curta, Chaetoceros resting spores (CRS), Thalassiosira antarctica cold, and Fragilariopsis cylindrus (Fig. 7). However, the F. curta group dominates the diatom assemblage across the record and ranges from 42 % to 70 % of all diatoms. Average proportions of CRS ranges between 14 % and 22 %, while Thalassiosira antarctica cold represents on average ∼8 % of all diatoms counted. The absolute diatom abundance (ADA) is expressed as valves per gram dry weight (vgdw-1), and ranges from 97×106 to 488×106 vgdw-1 across the sediment samples.

Diatom abundances and species vary spatially. Diatom abundance decreased northward, with the highest ADA observed in the core top of BC02 (303 ± 76×106 vgdw-1), followed by BC03 (262 ± 30×106 vgdw-1), and the lowest ADA seen in BC04 (199 ± 91×106 vgdw-1). To assess sea ice and phytoplankton variability, spatial variation in “centric cold-water diatoms”, CRS, and F. curta were compared to the average proportion of sea ice cover across the southwestern Ross Sea from 2002–2011 (Spreen et al., 2008), covering the time represented by the core tops (Fig. 6). Relative proportions of “centric cold-water diatoms” decreased northward away from the McMurdo polynya, whereas proportions of F. curta were highest towards the north. CRS, similar to the “centric cold-water diatoms”, decreased on a south-to-north gradient, away from the polynya.

Downcore, ADA decreased in BC04, while BC02 and BC03 have a relatively stable diatom abundance downcore (Fig. 7). The F. curta group decreased downcore in BC04 (Fig. 7). CRS shows the opposite trend and increased downcore in BC04 and BC02 (Fig. 7). No downcore trend is observed in relative abundances of Thalassiosira antarctica. Similarly, the F. cryophilic group, the open ocean group, the sea ice group, the centric cold water group, and E. antarctica do not demonstrate a strong increasing or decreasing trend (Fig. 7).

A PCA of the biomarker and diatom dataset (Fig. 8) resolved 57.3 % of total variance across the first two dimensions (Dim1=37.8 %, Dim2=19.5 %). Three groupings emerge. First, the F. curta group, C24:0, and the sterols (dinosterol, brassicasterol) plot together, with C24:0 showing a strong positive correlation with F. curta (r=0.80) and a strong negative correlation with Chaetoceros resting spores (r=-0.67), while dinosterol and brassicasterol show more moderate correlations with the same variables (r=0.58 and -0.40; r=0.47 and -0.33, respectively). Second, Chaetoceros resting spores, the open ocean group, the centric cold water group, the sea ice diatom group, the F. cryophilic group, and T. antarctica plot in the opposing quadrant, consistent with a negative correlation with the first group. Third, the fatty acids, and HBIs cluster together and show negligible correlation with either diatom grouping (r=-0.19 to 0.11), indicating these variables are statistically independent of the ordination axis separating the two diatom groupings, with the notable exception of C24:0 which does not align with the rest of the fatty acid group.

The sediment archive preserved in BC02, BC03, and BC04 provides a ∼200-year paleorecord of diatom and biomarker variability in the southwestern Ross Sea. These records show both shared patterns and site-specific differences. Total diatom abundance is stable downcore in BC02 and BC03, and decreases in BC04, while most fatty acid biomarkers show a general decline in the upper ∼20 cm across all three cores, consistent with early post-depositional degradation. BC03 and BC04 also show episodic laminations (Figs. A1 and A2), which may reflect reduced bioturbation under more persistent sea ice cover at these sites. In contrast, several diatom groups (e.g., F. curta group, Chaetoceros RS) and the sea ice biomarker IPSO₂₅ show more gradual, longer-term downcore trends.

The relative abundance of F. curta, a well-established sea ice diatom proxy particularly associated with seasonal ice zones (Armand et al., 2005; Gersonde and Zielinski, 2000; Leventer and Dunbar, 1988), decreases downcore in BC02 and BC04, while it remains relatively stable throughout the record in BC03 (Fig. 9). Together, these trends suggest an overall increase in sea ice concentration in the southwestern Ross Sea from 1900 CE to present, providing an independent line of evidence to support biomarker-based reconstructions. Similarly, IPSO₂₅ shows a decreasing trend downcore in all three cores, indicating an increasing signal of fast and platelet ice toward the present (Belt, 2018). The PIPSO₂₅ index, which integrates IPSO₂₅ with the open-water sterols dinosterol and brassicasterol, is considered a more robust measure of sea ice influence than IPSO₂₅ alone (Lamping et al., 2021; Vorrath et al., 2023, 2020). Despite the low resolution, the PIPSO₂₅ values in all three cores show an increasing trend since ∼1900 CE (Fig. 9). Collectively, diatom and biomarker records suggest a 200-year increase in sea ice duration in the southwestern Ross Sea.

Our findings align closely with other Ross Sea records indicating increasing sea ice conditions over the past two centuries. Diatom-based sediment records from the southwestern Ross Sea show a similar long term trend in sea ice presence (Leventer and Dunbar, 1988; Leventer et al., 1993). Yang et al. (2021) reconstructed the northern limit of sea ice extent in the Ross Sea and observed relatively stable conditions pre-1900, followed by an increasing trend since ∼1950 CE (Fig. 9). Similarly, Dalaiden et al. (2021) used ice core records and data assimilation to reconstruct Ross Sea sea ice extent, also finding an increase beginning around 1950 after a stable pre-1900 period (Fig. 9). A coastal ice core record from the Ross Sea shows a stable sea ice area from the 1880s to the 1950s, a reduction from the 1950s to the 1990s and an increase from 1993 (Sinclair et al., 2014). Although these studies quantify sea ice extent and area at the regional scale, and our records reflect coastal sea ice duration at three coastal-proximal sites, most reconstructions point toward enhanced sea ice conditions in the Ross Sea since the mid-20th century (Fig. 9).

Differences among studies likely reflect a combination of factors. First, sediment cores generally have lower temporal resolution than ice cores and may not capture short term variations, such as the stable conditions before 1950 reported by Dalaiden et al. (2021) and Yang et al. (2021). Chronological uncertainty is another factor, as the CFCS lead dating model assumes constant sedimentation rates, a constant Pb flux, and no mixing or diffusion of Pb in the sediment (Appleby, 2002; Bruel and Sabatier, 2020). The dating uncertainty could introduce offsets in the timing of the observed trends. Second, spatial variability across the Ross Sea is well documented, with satellite observations from 1997 to 2013 showing an overall increase in sea ice extent (Yuan et al., 2017), but with significant anomalies (Dale et al., 2017; Krauzig et al., 2024). It is plausible that the southwestern Ross Sea experienced an earlier or stronger increase in sea ice duration compared with more northern or offshore regions. Sea ice variability in the southwestern Ross Sea is influenced by local processes such as fast ice formation and breakup, coastal polynya activity, and storm driven winds (Radlwimmer et al., 2026). These processes differ from those governing regional scale sea ice variability across the wider Ross Sea, where larger scale oceanic and atmospheric processes influence sea ice at the offshore margins (Dalaiden et al., 2025, 2021). Consequently, regional reconstructions may not fully capture the sea ice variability recorded at our coastal core sites. Finally, methodological and spatial differences may also contribute. Our records capture sea ice duration and concentration near the coast (Armand et al., 2005), while Dalaiden et al. (2021) and Yang et al. (2021) reconstructed sea ice extent at the regional scale. Although these metrics are related, changes in sea ice extent do not necessarily correspond to changes in local sea ice concentration and duration. Thus, differences between regional sea ice extent and the conditions recorded at our coastal sites may reflect both the spatial scale variability in sea ice across the Ross Sea and the distinct processes influencing coastal versus offshore sea ice. Despite these differences in metric and spatial coverage, all records presented in Fig. 9 show broadly consistent increases in sea ice conditions since ∼1950, underscoring the robustness of this trend across multiple archives and proxies.

Fatty acids in the surface sediments reflect inputs from a mixture of phytoplankton and bacterial sources, including diatoms and P. antarctica. The PCA analysis (Fig. 8) groups the fatty acids and HBIs together, suggesting shared post-depositional processes such as degradation. In contrast, the sterols do not follow the same grouping as the fatty acids and HBIs, consistent with their comparatively low reactivity and limited downcore decrease (Vorrath et al., 2019; Rontani et al., 2019, 2012). Downcore profiles of fatty acids show a marked decrease in concentrations within the upper ∼20 cm (corresponding to ∼76–95 years ± 2.4–22.2 years), consistent with early post depositional degradation. Below this interval, fatty acid concentrations remained relatively stable. This pattern suggests that post-depositional alteration is an important driver of fatty acid concentrations; however, it is unlikely to be the sole factor influencing downcore variability. Changes in primary production, organic matter flux, and sedimentation processes also contribute to the observed trends. Beyond the ∼20 cm fatty acid decrease, no consistent downcore trends were observed, which may suggest relatively little change in phytoplankton community composition over the past ∼200 years, despite increasing sea ice influence (Yuan et al., 2017; Dalaiden et al., 2021). However, interpretation is limited by the low sampling resolution (4–8 cm), which may obscure finer-scale variability. Higher-resolution records would allow more detailed reconstruction of short-term variability and potential responses of phytoplankton communities to sea ice changes.