Highly branched isoprenoids for Southern Ocean semi-1 quantitative sea ice reconstructions : a pilot study from the 2 Western Antarctic Peninsula 3 4

4 Maria-Elena Vorrath1, Juliane Müller1,2,3, Oliver Esper1, Gesine Mollenhauer1,2, Christian 5 Haas1, Enno Schefuß2, Kirsten Fahl1 6 1Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Bremerhaven, Germany 7 2MARUM – Center for Marine Environmental Sciences, University of Bremen, Germany 8 3Department of Geosciences, University of Bremen, Germany 9 Correspondence to: Maria-Elena Vorrath, maria-elena.vorrath@awi.de 10 11 Abstract. Organic geochemical and micropaleontological analyses of surface sediments collected in the southern 12 Drake Passage and the Bransfield Strait, Antarctic Peninsula, enable a proxy-based reconstruction of recent sea 13 ice conditions in this climate sensitive area. The distribution of the sea ice biomarker IPSO25 supports earlier 14 suggestions that the source diatoms seem to be common in near-coastal environments characterized by an annually 15 recurring sea ice cover. We here propose and evaluate the combination of IPSO25 with a more unsaturated highly 16 branched isoprenoid alkene and phytosterols and introduce the PIPSO25 index as a potentially semi-quantitative 17 sea ice proxy. This organic geochemical approach is complemented with diatom data. PIPSO25 sea ice estimates 18 are used to discriminate between areas characterized by permanently ice-free conditions, seasonal sea ice cover 19 and extended sea ice cover. These trends are consistent with satellite sea ice data and winter sea ice concentrations 20 estimated by diatom transfer functions. Minor offsets between proxy-based and satellite-based sea ice data are 21 attributed to the different time intervals recorded within the sediments and the instrumental records from the study 22 area, which experienced rapid environmental changes during the past 100 years. 23 24


Introduction
In the last century, the Western Antarctic Peninsula (WAP) has undergone a rapid warming of the atmosphere of 3.7 ± 1° C, which exceeds several times the average global warming (Vaughan et al., 2003).Simultaneously, a waters, this has been strengthened by Smik et al. (2016a).Known source organisms of C25:3 HBI trienes (Fig. 1 shows molecular structures of both the E-and Z-isomer) are, for example, Rhizosolenia and Pleurosigma diatom species (Belt et al., 2000(Belt et al., , 2017)).In the subpolar North Atlantic, the HBI Z-triene has been used to further modify the so-called PIP25 index (Smik et al., 2016b) -an approach for semi-quantitative sea ice estimates.Initially, PIP25 was based on the employment of phytoplankton-derived sterols, such as brassicasterol and dinosterol (Kanazawa et al., 1971;Volkman, 2003), to serve as open-water counterparts, while IP25 reflects the occurrence of a former sea ice cover (Belt et al., 2007;Müller et al., 2009Müller et al., , 2011)).Consideration of these different types of biomarkers helps to discriminate between ice-free and permanently ice-covered ocean conditions, both resulting in a lack of IP25 and IPSO25, respectively (for further details see Belt, 2018;Belt and Müller, 2013a).Uncertainties in the source-specificity of brassicasterol (Volkman, 1986) and its identification in Arctic sea ice samples (Belt et al., 2013), however, require caution when pairing this sterol with a sea ice biomarker lipid for Arctic sea ice reconstructions.In this context, we note that Belt et al. (2018) reported that brassicasterol is not evident in the IPSO25 producing sea ice diatom Berkeleya adeliensis.While the applicability of HBIs (and sterols) to reconstruct past sea ice conditions has been thoroughly investigated in the Arctic Ocean (Belt, 2018;Stein et al., 2012;Xiao et al., 2015), only two studies document the distribution of HBIs in Southern Ocean surface sediments (Belt et al., 2016;Massé et al., 2011).The circum-Antarctic data set published by Belt et al. (2016), however, does not report HBI triene abundances.Significantly more studies so far focused on the use of IPSO25 and the HBI Z-triene for paleo sea ice reconstructions and these records are commonly compared to micropaleontologial diatom analyses (e.g., Barbara et al., 2013;Collins et al., 2013;Denis et al., 2010).
Here, we provide a first overview of the distribution of IPSO25, HBI trienes, brassicasterol and dinosterol in surface sediments from the northern part of the WAP (southern Drake Passage and Bransfield Strait).These biomarker data are completed by diatom analyses and remote sensing sea ice data.We further introduce and discuss the socalled PIPSO25 index (phytoplankton-IPSO25 index), which, following the PIP25 approach in the Arctic Ocean (Müller et al., 2011), may serve as a semi-quantitative indicator of past Southern Ocean sea ice cover.These biomarker-based sea ice estimates are compared to sea ice concentrations derived from diatom transfer functions and satellite-derived data on the recent sea ice conditions in the study area.

Oceanographic setting
The study area includes the southern Drake Passage and the Bransfield Strait located between the South Shetland Islands and the northern tip of the WAP (Fig. 2a and b).While the oceanographic setting in the Drake Passage is dominated by the Antarctic Circumpolar Current (ACC), a complex current system prevails in the Bransfield Strait.
According to Sangrá et al. (2011) a branch of the ACC enters the Bransfield Strait in the west as the Bransfield Current, carrying transitional waters under the influence of the Bellingshausen Sea (Transitional Bellingshausen Sea Water, TBW).The TBW is characterized by a well-stratified, fresh and warm water mass with summer sea surface temperatures (SST) above 0° C. Below the shallow TBW, a narrow tongue of circumpolar deep water (CDW) flows along the slope of the South Shetland Islands (Sangrà et al., 2011).In the eastern part, transitional water from the Weddell Sea (Transitional Weddell Sea Water, TWW) enters the Bransfield Strait through the Antarctic Sound and from the Antarctic Peninsula (AP).This water mass corresponds to the Antarctic Coastal Current (Collares et al., 2018;Thompson et al., 2009).The TWW is significantly colder (summer SST < 0° C) and saltier due to extended sea ice formation in the Weddell Sea Gyre.The two water masses are separated at the sea surface by the Peninsula Front characterized by a TBW anticyclonic eddy system (Sangrà et al., 2011).While the TWW occupies the deep water column of the Bransfield Strait (Sangrà et al., 2011), it joins the surface TBW in the southwestern Bransfield Strait (Collares et al., 2018).
Due to high concentrations of dissolved iron on the shelf (Klunder et al., 2014), the area around the WAP is characterized by a high primary production with high vertical export fluxes during early summer associated with the formation of fast sinking mineral aggregates and fecal pellets (Kim et al., 2004;Wefer et al., 1988).The Peninsula Front divides the Bransfield Strait into two biogeographic regimes of high chlorophyll and diatom abundance in the TBW and low chlorophyll values and a pre-dominance of nanoplankton in the TWW (Gonçalves-Araujo et al., 2015), which is also reflected in the geochemistry of surface sediments (Cárdenas et al., 2018).Materials and Methods

Sediment Samples and radiocarbon dating
In total, 26 surface sediment samples obtained by multicorers and box corers during the RV Polarstern cruise PS97 (Lamy, 2016) were analyzed (Fig. 2, Table 1).All samples were stored frozen and in glass vials.The composition of the sediments ranges from foraminiferal mud in the Drake Passage to diatomaceous mud with varying amounts of ice rafted debris in the Bransfield Strait (Lamy, 2016).
14 C radiocarbon dating of two samples from the PS97 cruise and one from the Polarstern cruise ANT-VI/2 (Fütterer, 1988) was conducted using the mini carbon dating system (MICADAS) at the Alfred Wegener Institute following the method of Wacker et al. (2010).The 14 C ages were calibrated to calendar years before present (cal BP) using the Calib 7.1 software (Stuiver et al., 2019) with an estimated reservoir age of 1178 years, derived from the six closest reference points listed in the Marine Reservoir Correction Database (www.calib.org).

Organic geochemical analyses
For biomarker analyses, sediments were freeze-dried and homogenized using an agate mortar.Also after freezedrying, samples were stored frozen to avoid degradation.The extraction, purification and quantification of HBIs and sterols follow the analytical protocol applied by the international community of researchers performing HBI and sterol-based sea ice reconstructions (Belt et al., 2014;Stein et al., 2012).Prior to extraction, internal standards 7-hexylnonadecane (7-HND) and 5α-androstan-3β-ol were added to the sediments.For the ultrasonic extraction (15 min), a mixture of CH2Cl2:MeOH (v/v 2:1; 6ml) was added to the sediment.After centrifugation (2500 rpm for 1 min), the organic solvent layer was decanted.The ultrasonic extraction step was repeated twice.From the combined total organic extract, apolar hydrocarbons were separated via open column chromatography (SiO2) using hexane (5 ml  Belt (2018), the identification of IPSO25 and HBI trienes is based on comparison of their mass spectra with published mass spectra (Belt et al., 2000).Regarding the potential sulfurization of IPSO25, we examined the GC-MS chromatogram and mass spectra of each sample for the occurrence of the HBI C25 sulfide (Sinninghe Damsté et al., 2007).The C25 HBI thiane was absent in all samples.For the quantification, manually integrated peak areas of the molecular ions of the HBIs in relation to those of 7-HND were used.An external calibration for HBI diene and trienes was applied using a sample with known HBI concentrations from the Lancaster Sound, Canada, to account for the different response factors of the HBI molecular ions (m/z 346; m/z 348) and the fragment ion of 7-HND (m/z 266).The identification of sterols was based on comparison of their retention times and mass spectra with those of reference compounds run on the same instrument.Comparison of peak areas of individual analytes and the internal standard was used for sterol quantification.The error determined by duplicate GC-MS measurements was below 0.7%.The detection limit for HBIs and sterols was 0.5 ng/g sediment.Absolute concentrations of HBIs and sterols were normalized to total organic carbon contents (for TOC data see Cárdenas et al., 2018).
The herein presented phytoplankton-IPSO25 index (PIPSO25) is calculated using the same formula as for the PIP25 index following Müller et al. (2011): The balance factor c (c = mean IPSO25 / mean phytoplankton biomarker) is applied to account for the high offsets in the magnitude of IPSO25 and sterol concentrations.
Stable carbon isotope composition of IPSO25, requiring a minimum of 50 ng carbon, was successfully determined on five samples using GC-irm-MS.The ThermoFisher Scientific Trace GC was equipped with a 30m Restek Rxi-5 ms column (0.25 mm diameter, 0.25 µm film thickness) and coupled to a Finnigan MAT 252 isotope ratio mass spectrometer via a modified GC/C interface.Combustion of compounds was done under continuous flow in ceramic tubes filled with Ni wires at 1000° C under an oxygen trickle flow.The same GC program as for the HBI identification was used.The calibration was done by comparison to a CO2 reference gas.The values of d 13 C are expressed in per mill (‰) against Vienna PeeDee Belemnite (VPDB) and the mean standard deviation was <0.9 ‰.An external standard mixture was measured every six runs, achieving a long-term mean standard deviation of 0.2‰ and an average accuracy of <0.1 ‰.

Diatoms
Details of the standard technique of diatom sample preparation were developed in the micropaleontological laboratory at the Alfred Wegener Institute (AWI) in Bremerhaven.The preparation included a treatment of the sediment samples with hydrogen peroxide and concentrated hydrochloric acid to remove organic and calcareous remains.After washing the samples several times with purified water, the water was removed and the diatoms were embedded on permanent mounts for counting (see detailed description by Gersonde and Zielinski, 2000).
The respective diatom counting was carried out according to Schrader and Gersonde (1978).On average, 400 to 600 diatom valves were counted in each slide using a Zeiss Axioplan 2 at x1000 magnification.In general preservation state of the diatom assemblages was moderate to good in the Bransfield Strait and decreased towards the Drake Passage where it is moderate to poor.
Diatoms were identified to species or species group level and if possible to forma or variety level.The taxonomy follows primarily Hasle and Syvertsen (1996), Zielinski andGersonde (1997), andArmand andZielinski (2001).
capitulata, two varieties with gradual transition of features between them and no significantly different ecological response.The species Fragilariopsis curta and Fragilariopsis cylindrus were combined as F. curta group taking into account their similar relationship to sea ice and temperature (Armand et al., 2005;Zielinski and Gersonde, 1997).Furthermore, the Thalassiosira gracilis group comprises T. gracilis var.gracilis and T. gracilis var.expecta because the characteristic patterns in these varieties are often transitional, which hampers distinct identification.
Although the two varieties Eucampia antarctica var.recta and E. antarctica var.antarctica display different biogeographical distribution (Fryxell and Prasad, 1990), they were combined to the E. antarctica group.This group was not included in the transfer function (TF) as it shows no relationship to either sea ice or temperature variation (Esper and Gersonde, 2014a, b).Besides the E. antarctica group, we also discarded diatoms assembled as Chaetoceros spp. group from the TF-based re-constructions, following Zielinski et al. (1998) and Esper and Gersonde (2014a).This group combines mainly resting spores of a diatom genus with a ubiquitous distribution pattern that cannot be identified to species level due to the lack of morphological features during light microscopic inspection.Therefore, different ecological demands of individual taxa cannot be distinguished.
For estimating winter sea ice (WSI) concentrations we applied the marine diatom TF MAT-D274/28/4an, comprising 274 reference samples from surface sediments in the western Indian, the Atlantic and the Pacific sectors of the Southern Ocean, with 28 diatom taxa and taxa groups, and an average of 4 analogs (Esper and Gersonde, 2014a).The WSI estimates refer to September sea-ice concentrations averaged over a time period from 1981 to 2010 at each surface sediment site (National Oceanic and Atmospheric Administration, NOAA; Reynolds et al., 2002Reynolds et al., , 2007)).The reference data set is suitable for our approach as it uses a 1° by 1° grid, representing a higher resolution than previously used and results in a root mean squared error of prediction (RMSEP) of 5.52% (Esper and Gersonde, 2014a).We defined 15% concentration as threshold for maximum sea-ice expansion following the approach of Zwally et al. (2002) for the presence or absence of sea ice, and 40% concentration representing the average sea-ice edge (Gersonde et al., 2005;Gloersen et al., 1993).MAT calculations were carried out with the statistical computing software R (R Core Team, 2012) using the additional packages Vegan (Oksanen et al., 2012) and Analogue (Simpson and Oksanen, 2012).Further enhancement of the sea-ice reconstruction was obtained by consideration of the abundance pattern of the diatom sea-ice indicators allowing for qualitative estimate of sea-ice occurrence, as proposed by Gersonde and Zielinski (Gersonde and Zielinski, 2000).

Sea ice data
The mean monthly satellite sea ice concentration was derived from Nimbus-7 SMMR and DMSP SSM/I-SSMIS passive microwave data and downloaded from the National Snow and Ice Data Center (NSIDC; Cavalieri et al., 1996).An interval from 1980 to 2015 was used to generate an average sea ice distribution for each season, spring (SON), summer (DJF), autumn (MAM) and winter (JJA).
The distribution of IPSO25 in the study area shows a clear northwest-southeast gradient (Fig. 3a) with concentrations increasing from the continental slope and around the South Shetland Islands towards the continental shelf.Maximum IPSO25 concentrations are observed in the Bransfield Strait.According to Belt et al. (2016), deposition of IPSO25 is highest in area covered by landfast sea ice and platelet ice during early spring and summer.
We suppose that core sites PS97/068 to PS97/073 in the central and eastern Bransfield Strait are located too distal to be covered by fast ice and suggest that peak IPSO25 concentrations at these sites may refer to the frequent drift and melt of sea ice exported from the Weddell Sea into the Bransfield Strait.The vertical export of biogenic material from sea ice towards the seafloor may be accelerated significantly by the formation of organic-mineral aggregates, fecal pellets or by (cryogenic) gypsum ballasting, which promotes a rapid burial and sedimentation of organic matter in polar settings (De La Rocha and Passow, 2007;Wefer et al., 1988;Wollenburg et al., 2018).
Lateral subsurface advection of organic matter (incl.biomarkers) through the TWW, however, may also contribute to elevated IPSO25 concentrations at these sites.
IPSO25 was not detected in sediments from the permanently ice-free areas in the Drake Passage.Highest concentrations of both HBI trienes are found in the eastern Drake Passage and along the continental slope, while their concentrations in the Bransfield Strait are rather low (Fig. 3b and c) suggesting unfavorable environmental conditions (ocean temperature, sea ice cover) for their source diatoms.
We observe higher concentrations of brassicasterol and dinosterol in the eastern part of the Drake Passage and, in contrast to the observation made for HBI trienes, also in the eastern and central Bransfield Strait (Fig. 3d and e).
Dinosterol and, in particular, brassicasterol are known to have different source organisms including diatoms, dinoflagellates, prymnesiophycean algae and cyanobacteria (Volkman, 1986) and we assume that this diversity may account for the higher concentration of these lipids in Bransfield Strait sediments, while concentrations of HBI trienes, mainly derived from diatoms, are significantly lower.Sediments collected along the Hero Fracture Zone in the western Drake Passage (Fig. 2) contain only minor amounts of biomarkers except for elevated brassicasterol concentrations observed at stations PS97/048-1 and 049-2 (Fig. 3d).This part of the Drake Passage is mainly barren of fine-grained sediments and dominated by sands (Lamy, 2016), which may point to intensive winnowing by ocean currents impacting the deposition and burial of organic matter.We consider that also degradation of HBIs and sterols may affect their distribution within surface sediments.Rontani et al. (2014) report a higher sensitivity of tri-unsaturated HBIs to oxidation but also note that oxidation conditions in pelagic environments (i.e.their source organisms' habitat) are not as significant as those within sea ice.A more recent study by Rontani et al. (2019) on surface sediments shows that IPSO25 may potentially be affected by autoxidative and bacterial degradation but oxidation products are found in only minor proportions.In general, further investigations into degradation processes affecting both HBIs and sterols within sediments would address an important knowledge gap regarding in-situ biochemical modifications of the biomarker signal.
The d 13 C values of IPSO25 are between -10.3‰ and -14.7‰ which is the commonly observed range for IPSO25 in surface sediments and sea ice derived organic matter (Massé et al., 2011, Belt et al., 2016), and contrasts the low d 13 C values of marine phytoplankton lipids in Antarctic sediments (-38‰ to -41‰ after Massé et al., 2011).
Contrary to the finding of elevated Z-triene concentrations in surface waters along an ice-edge (Smik et al., 2016a) and earlier suggestions that this biomarker may be used as a proxy for MIZ conditions (Belt et al., 2015;Collins et al., 2013), we observe highest concentrations of the Z-and E-triene at the permanently ice-free northernmost stations in the eastern Drake Passage.This is also apparent for brassicasterol and dinosterol supporting an open marine (pelagic) source for these sterols.Moderate concentrations of HBI trienes at the continental slope along the WAP and in the Bransfield Strait likely refer to primary production at the sea ice margin during spring and summer indicating seasonally ice-free waters in high production coastal areas influenced by upwelling (Gonçalves-Araujo et al., 2015).The similarity in the distribution of the Z-and the E-triene in our surface sediments -the latter of which so far is not often considered for Southern Ocean paleoenvironmental studies -supports the assumption of a common diatom source for these HBIs (Belt et al., 2000(Belt et al., , 2017)).Since brassicasterol and dinosterol are highly abundant in both seasonally ice-covered Bransfield Strait sediments as well as in permanently ice-free Drake Passage sediments, their use as an indicator of fully open-marine conditions is questionable.Elevated concentrations of both sterols in the Bransfield Strait could either point to an additional input of these lipids from melting sea ice (Belt et al., 2013) or a better adaptation of some of their source organisms to cooler and/or icedominated ocean conditions.Production and accumulation of these lipids in (late) summer (i.e. after the sea ice season) may be considered as well.This observation highlights the need for a better understanding of the source organisms and the mechanisms involved in the synthesis of these sterols.

A novel sea ice index for the Southern Ocean: PIPSO25
The main concept of combining the sea ice proxy with an indicator of an ice-free ocean environment (i.e. a phytoplankton biomarker, Müller et al., 2011), aims at a semi-quantitative assessment of the sea ice conditions.
By reducing the light penetration through the ice, a thick and perennial sea ice cover limits the productivity of bottom sea ice algae (Hancke et al., 2018), which results in the absence of both sea ice and pelagic phytoplankton biomarker lipids in the underlying sediments.Vice versa, sediments from permanently ice-free ocean areas only lack the sea ice biomarker but contain variable concentrations of phytoplankton biomarkers (Müller et al., 2011).
The co-occurrence of both biomarkers in a sediment sample suggests seasonal sea ice coverage promoting algal production indicative of sea ice as well as open ocean environments (Müller et al., 2011).
Following the PIP25-approach applied in the Arctic Ocean (Müller et al., 2011;Belt and Müller, 2013;Xiao et al., 2015), we used IPSO25, HBI triene and sterol data to calculate the PIPSO25 index.Depending on the biomarker reflecting pelagic (open ocean) conditions, we define PZIPSO25 (using the Z-triene), PEIPSO25 (using the E-triene), PBIPSO25 (using brassicasterol), and PDIPSO25 (using dinosterol).Since the concentrations of IPSO25 and both HBI trienes are in the same range, the application of the c-factor is not needed here.For the calculation of PBIPSO25 the c-factor is 0.0048, for PDIPSO25 it is 0.0137.Front described by Sangrà et al. (2011).This front is reported to act as a barrier for phytoplankton communities (Gonçalves-Araujo et al., 2015) and is associated with the encounter between TWW carrying Weddell Sea sea ice through the Antarctic Sound and the TBW.The high PIPSO25 values suggesting an extended sea ice cover west of the Peninsula Front (station PS97/054 and PS97/056) result from minimum concentrations of pelagic biomarkers and slightly higher concentrations of IPSO25.
PIPSO25 values based on the E-triene are about 0.2 higher compared to PZIPSO25, due to the generally lower concentrations of the E-triene (Table 1).The sterol-based PIPSO25 values display a generally similar pattern as PZIPSO25and PEIPSO25, respectively, and we note a high comparability between the PEIPSO25 and PBIPSO25 values (r 2 = 0.73).Some differences, however, are observed in the southwestern part of the Bransfield Strait (station PS97/056) where PBIPSO25 indicates a lower sea ice cover and in the central Bransfield Strait (stations PS97/068 and PS97/069) where PBIPSO25 and PDIPSO25 reflect only MIZ conditions.The HBI-triene based PIPSO25 indices hence seem to reflect the oceanographic conditions within the Bransfield Strait more satisfactorily.We note that the brassicasterol or dinosterol-based PIPSO25 index links environmental information derived from biomarker lipids belonging to different compound classes (i.e.HBIs and sterols), which have fundamentally different chemical properties.This requires special attention as, for example, selective degradation of one of the compounds may affect the sedimentary concentration of the respective lipids.Previous studies linking HBI and sterol-based sea ice reconstructions with satellite-derived or, with respect to downcore paleo studies, paleoclimatic data, however, demonstrate that the climatic/environmental conditions controlling the production of HBIs and sterols seem to exceed the influence of a potential preferential degradation of these biomarkers within the sediments (e.g., Berben et al., 2014;Cabedo-Sanz et al., 2013;Müller et al., 2009Müller et al., , 2012;;Müller and Stein, 2014;Stein et al., 2017;Xiao et al., 2015).A comparison of PIP25 records determined using brassicasterol and the HBI Z-triene for three sediment cores from the Arctic realm covering the past up to 14.000 years BP (Belt et al., 2015) reveals very similar trends for both versions of the PIP25 index in each core, which may point to, at least, a similar degree of degradation of HBI trienes and sterols through time.More such studies are needed to evaluate the preservation potential of HBIs and sterols in Southern Ocean sediments.

Comparison of satellite-derived modern sea ice conditions and biomarker data
Satellite-derived sea ice data were averaged over the time period from 1980 to 2015 for all four seasons (Table 2) and are considered to reflect the modern mean state of sea ice coverage around the WAP.The sea ice concentration is expressed to range from 0 to 100 % and, although the error can be up to 15 %, concentrations below 15 % still suggest the occurrence of sea ice.These low sea ice concentrations are usually neglected for the determination of the sea ice extent, which is defined as the ocean area with a sea ice cover of at least 15 %.The spring and winter sea ice concentrations are shown in Figure 4 e-f.Winter sea ice is estimated to not extend north of 61° S (Fig. 4 f) and varies between 1 % and 50 % in the study area, while sea ice is reduced to less than 20 % in spring (Fig. 4e, Table 2).
Sea ice concentrations of up to 50 % are common in winter between the South Shetland Islands and north of the Antarctic Sound where the influence of TWW is highest.Permanent sea ice cover is uncommon in the Bransfield Strait and around the WAP and this area is mainly characterized by a high sea ice seasonality and drift ice from the Weddell Sea (Collares et al., 2018).Comparisons of individual biomarker concentrations with satellite sea ice data reveal a weak and positive correlation between IPSO25 concentrations and winter sea ice concentrations (r 2 = 0.5), while no correlation is found between sea ice and pelagic biomarker concentrations (r 2 < 0.1 for all relations).
Correlations of PIPSO25 values with satellite-derived sea ice concentrations (for spring, summer, autumn and winter) contrast earlier observations made for the PIP25 index in the Arctic Ocean, where the closest linear relationship is found mainly with the spring sea ice coverage (i.e. the blooming season of sea ice algae; Müller et al., 2011;Xiao et al., 2015).
Esper and Gersonde (2014a) studied the response of diatom species to changes in environmental conditions and their response to the non-linear behaviour of sea ice dynamics (Zwally et al., 2002).In contrast to ice free areas or areas of permanent sea ice cover, areas characterized by the transition from consolidated to unconsolidated sea ice show rapid changes in satellite derived sea ice concentrations (ranging from 90 % to 15 %) and exhibit a large variability in species composition.To reflect this curve in sea ice we hence chose a cubic polynomial regression (polynomial of third degree) to determine the relation between PIPSO25 values and satellite data depicting sea ice concentrations of more than 20 %.
The contour lines in Figure 4 a-d show the observed extent of 15 %, 30 %, 40 % and 50 % winter sea ice compared to the PIPSO25 values.In the northeastern part of the study area, the HBI triene based PIPSO25 indices align well with the contour lines of winter sea ice concentrations and depict the gradient from the marginally ice-covered southern Drake Passage towards the intensively ice-covered Weddell Sea.In the southwestern part of the Bransfield Strait, all PIPSO25 indices suggest a higher sea ice cover than it is reflected in the satellite data.This may be explained by the transport (and melt) of drift ice through the TWW, joining the TBW at the southwestern Peninsula Front.In contrast, also the absence of IPSO25 at stations PS97/052 and PS97/053, off the continental slope, is in conflict with the satellite data depicting an average winter sea ice cover of 23 %.Earlier documentations that the IPSO25 producing sea ice diatom Berkeleya adeliensis favors land-fast ice communities in East Antarctica (Riaux-Gobin and Poulin, 2004) and platelet ice occurring mainly in near coastal areas (Belt et al., 2016) could explain this mismatch between biomarker and satellite data, which further strengthens the hypothesis that the application of IPSO25 seems to be confined to continental shelf or near-coastal and meltwater affected environments (Belt et al., 2016).Alternatively, strong ocean currents (i.e. the ACC) could impact the deposition of IPSO25 in this region.
As photosynthesis is not possible and a release of sea ice diatoms from melting sea ice is highly reduced during the Antarctic winter, the observation of a stronger correlation between recent winter sea ice concentrations and PIPSO25 sea ice estimates is unexpected.We hence suggest that this offset may be related to the fact that the sediment samples integrate a longer time interval than is covered by satellite observations.Radiocarbon dating of selected samples that contain calcareous material reveals an age of 100 years BP in the vicinity of the South Shetland Islands (station PS97/059-2) and 142 years BP at the Antarctic Sound (station PS1546-2, Table 3).A significantly older age was determined for a sample of N. pachyderma from station PS97/044-1 (4830 years BP) which likely denotes the winnowing and/or very low sedimentation rates in the Drake Passage.Bioturbation effects and uncertainties in reservoir ages potentially mask the ages of the near-coastal samples.
Nevertheless, since also other published ages of surface sediments within the Bransfield Strait (Barbara et al., 2013;Barnard et al., 2014;Etourneau et al., 2013;Heroy et al., 2008) are in the range of 0-270 years, we consider that our surface samples likely reflect the paleoenvironmental conditions that prevailed during the last two centuries (and not just the last 35 years covered by satellite observations).In the context of the rapid warming during the last century (Vaughan et al., 2003) and the decrease of sea ice at the WAP (King, 2014;King and Harangozo, 1998), we suggest that the biomarker data of the surface sediments relate to a spring sea ice cover, which must have been enhanced compared to the recent (past 35 years) spring sea ice recorded via remote sensing.
Presumably, the average spring sea ice conditions over the past 200 years might have been similar to the modern (past 35 years) winter conditions, which would explain the stronger correlation between PIPSO25 sea ice estimates and winter sea ice concentrations.

Comparison of diatom-based sea ice estimates and biomarker data
The diatoms preserved in sediments from the study area (Table 4) can be associated with open ocean and sea ice conditions.North of the South Shetland Islands, the strong influence of the ACC is reflected in the high abundance of open ocean diatom species such as Fragilariopsis kerguelensis and Thalassiosira lentiginosa (Esper et al., 2010).The two diatom species Fragilariopsis curta and Fragilariopsis cylindrus -known to not produce HBI (Belt et al., 2016;Sinninghe Damsté et al., 2004) mark the vicinity to sea ice (Buffen et al., 2007;Pike et al., 2008) and indicate fast and melting ice, a stable sea ice margin and stratification due to melting processes and the occurrence of seasonal sea ice.The high abundance of these species in our samples is in good agreement with high and moderate IPSO25 concentrations and PIPSO25 values in the Bransfield Strait and around the South Shetland Islands, respectively.The only HBI source diatom identified is the HBI Z-triene producing Rhizosolenia hebetata (Belt et al., 2017), which is present in four samples in rather small amounts and does not show a relation to the measured Z-triene concentrations (Table 1 and 4).The source diatom of IPSO25 Berkeleya adeliensis was not observed (or preserved) in the samples, and we assume that other, hitherto unknown, producers may exist.
We applied the transfer function of Esper and Gersonde (2014a) with four analogs (4an, Table 4) to our samples to compare the different estimates of sea ice cover based on biomarkers and diatoms.A positive correlation of the linear relationship is found between WSI concentrations derived from diatoms and the PIPSO25 indices based on HBI trienes (PZIPSO25 with r 2 = 0.76; PEIPSO25 with r 2 = 0.77, Fig. 6a).The correlations of sterol-based PIPSO25 values with WSI are slightly lower but in the same range (PBIPSO25 with r 2 = 0.74; PDIPSO25 with r 2 = 0.69, Fig. 6b).A slightly weaker correlation is noted for diatom-and satellite-based winter sea ice concentrations (r 2 = 0.63; Fig. 6c).Overall, the diatom approach indicates higher sea ice concentrations than the satellite data with an offset of up to 65 %.This may be due to different sources of satellite reference data used for the transfer function or also due to the fact that the sediment samples integrate a longer time period with a higher sea ice cover than the satellite data (see discussion in section 4.3).

Application of PIPSO25 as a semi-quantitative sea ice index
Precise and, in particular, quantitative reconstructions of past sea ice coverage are crucial for a robust assessment of feedback mechanisms in the ice-ocean-atmosphere system.While diatom transfer functions provide a valuable tool, additional information on sea ice conditions in coastal ice-shelf proximal areas, which are often affected by opal dissolution, are essential.The PIPSO25 approach seems to be a promising step into this direction, though our data obtained for the WAP are not yet sufficient for a full calibration.PIPSO25, diatom and satellite sea ice data, however, reveal positive correlations (Figs. 5 and 6) and depict similar gradients in sea ice cover.The observed offset between satellite data and biomarker-and diatom-based sea ice estimates likely relates to the fact that the instrumental records cover a significantly shorter or more recent time interval than the studied sediments.The recent rapid warming along the WAP (Vaughan et al., 2003) hence complicates attempts to calibrate these proxy data against observational data.Regarding the interpretation of PIPSO25 in terms of sea ice coverage in the study area, lower PIPSO25 values (<0.15 for PZIPSO25; <0.31 for PEIPSO25; <0.22 for PBIPSO25 and PDIPSO25) roughly seem to reflect unconsolidated, drifting winter sea ice and a nearly ice-free spring season.Higher values (>0.71 for PZIPSO25; >0.9 PEIPSO25; >0.6 for PBIPSO25 and PDIPSO25) would refer to an extended winter sea ice cover (up to 91 % in some years) with ice floes remaining until summer.

Conclusion
The distribution of the sea ice biomarker IPSO25 and related HBI trienes and sterols as well as diatoms in a suite of surface sediments from the southern Drake Passage and the WAP reflects recent sea ice conditions reasonably well.The herein proposed sea ice index PIPSO25 indicates seasonal sea ice cover along the coast of the WAP and in the Bransfield Strait, whereas mainly ice-free conditions prevail in the Drake Passage.In general, this pattern is consistent with satellite-derived sea ice data and diatom-based sea ice estimates and we note that the PIPSO25 index seems a promising approach towards semi-quantitative sea ice reconstructions in the Southern Ocean.The recent rapid warming in the study area, however, affects the comparability of proxy and satellite data.The fact that the surface sediments integrate a significantly longer time interval than the remote sensing data thwarts attempts to calibrate PIPSO25 values against observed sea ice concentrations.Additional data from other circum-Antarctic coastal (and distal) environments and investigations into potential calibration methods are needed to further develop this index as a quantitative sea ice proxy.Table 1:

The
PIPSO25 values are 0 in the Drake Passage and increase to intermediate values at the South Shetland Islands and the continental slope and reach highest values in the Bransfield Strait (Fig.4a-d).Minimum PIPSO25 values are supposed to refer to a predominantly ice-free oceanic environment in the Drake Passage, while moderate PIPSO25 values mark the transition towards a marginal sea ice coverage at the continental slope and around the South Shetland Islands.Elevated PIPSO25 values in samples from the northeastern Bransfield Strait suggest an increased sea ice cover (probably sustained through the drift of sea ice originating in the Weddell Sea).This pattern reflects the oceanographic conditions of a permanently ice-free ocean north of the South Shetland Islands and a seasonal sea ice zone at the WAP influenced by the Weddell Sea as described byCárdenas et al. (2018).Both HBI triene-based PIPSO25 indices show constantly high values at the coast of the WAP of >0.7 (PZIPSO25) and >0.8 (PEIPSO25), respectively, in the southern Bransfield Strait paralleling the southwest-northeast oriented Peninsula
Coordinates of sample stations with water depth, concentrations of IPSO25, Z-and E-trienes, brassicasterol and dinosterol normalized to TOC, d 13 C values for IPSO25, and values of sea ice indices PIPSO25 based on the Z-and E-triene, brassicasterol and dinosterol.Concentrations below the detection limit are expressed as 0. The PIPSO25 could not be calculated where IPSO25 and the phytoplankton marker is absent (blank fields).