Holocene changes in sea-ice cover and polynya formation along the eastern North Greenland shelf: New insights from biomarker records

Understanding the processes controlling the natural variability of sea ice in the Arctic, one of the most dynamic components of the climate system, can help to constrain the effects of future climate change in this highly sensitive area. For the first time, a high-resolution biomarker study was carried out to reconstruct past sea-ice variability off eastern North Greenland. This area is strongly influenced by cold surface waters and drift ice transported via the East Greenland Current, meltwater pulses from the outlet glaciers of the Northeast Greenland Ice Stream and the build-up of landfast ice. The well-dated Holocene sedimentary section of Core PS93/025 provides insights into variations of the sea-ice conditions (regional and local sea-ice signal), oceanic and atmospheric circulation and the biotic response to these changes. These biomarker records show a reduced to variable sea-ice cover during the early Holocene between 10.2 and 9.3 ka, followed by a steady increase in sea-ice conditions during the mid Holocene. During the last 5e6 ka, sea-ice conditions remained more stable representing a seasonal to marginal sea-ice situation. Based on our biomarker records, stable sea-ice edge conditions, with a fully developed polynya situation occurred since the last 1 ka. © 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Background
Arctic sea ice is a key component in our climate system, regulating the global heat budget, influencing the atmospheric circulation (Serreze et al., 2016) and the global ocean circulation via the Atlantic Meridional Overturning Circulation (AMOC;S evellec et al., 2017). Alternations in Arctic sea-ice extent can result in major changes to deep-water formation in the Nordic Seas (Rudels and Quadfasel, 1991) and lower latitudinal climate of the North Atlantic region (Overland and Wang, 2005). Thus, the recent dramatic loss of Arctic sea ice ( Fig. 1) can affect global climate beyond the polar regions, changing the planetary albedo, surface heat and freshwater flux (Thomas and Dieckmann, 2008;S evellec et al., 2017). Sea ice is not only an important factor for biodiversity and -productivity in the Arctic, providing a habitat to ice-associated species including organisms such as sea ice-algae, fish, birds, and marine mammals (Rudels et al., 1996;Dieckmann and Hellmer, 2003), it also acts as insulating layer between the ocean and the atmosphere (Stroeve et al., 2012).
The particular importance of Arctic sea ice with respect to recent anthropogenic climate change is that it is one of the most dynamic components of the climate system. Due to a reduction in the iceocean albedo effect, declining Arctic sea ice is expected to amplify anthropogenic climate change in the polar regions (Manabe et al., 1992;Randall et al., 1998;Screen and Simmonds, 2010). Over the past three decades, summer sea-ice extent and thickness have declined by 50%, much faster than predicted by climate models (see review by Stroeve et al., 2012;Serreze et al., 2016;Walsh et al., 2017;Notz and Stroeve, 2018).
The oceanographic and sea-ice conditions on the eastern North Greenland shelf are influenced by the outflow of freshwater and sea ice (drift ice) from the central Arctic Ocean, as well as by rivers, fjords, islands and marine-terminating outlet glaciers from the Greenland coastline in the west (Arndt et al., 2017;Fig. 2b and c). The knowledge about regional and local drivers of the sea-ice history and oceanic and atmospheric interactions on the eastern North Greenland continental shelf during the Holocene, however, remains fragmentary due to its low accessibility. Over the last years reduced sea-ice conditions have enabled a number of research expeditions into this area, allowing the recovery of sediment cores for detailed studies of paleoclimate and paleoceanographic   (Birgel and Stein, 2004;supplemented). Kastenlot Core PS93/025 (red circle) location at 80 " N at the outer rim of the NEW Polynya (white dashed line, assumed area based on the TOC concentrations). Major surface ocean currents in the study area are the East Greenland Current (EGC, black arrows) and the West Spitsbergen Current (white arrows). Locations of further cores discussed in the text are marked as yellow circles. Northeast Greenland Ice Stream (NEGIS). b) Detailed map of the eastern North Greenland continental margin, including Core PS93/025 (black dot) taken from the upper edge of the Belgica Bank above the Westwind Trough, the southward flowing EGC (black arrow) and the northward flowing coastal Northeast Greenland Coastal Current (NEGCC, blue arrow). Norske Øer Ice Barrier is located in front of NEGIS. c) Satellite image (June 2019, credit: NASA Worldview: World Meteorological Organization, 1970) displaying outlet glaciers of NEGIS, landfast ice build-up along the coast, drift ice transported from the central Arctic Ocean and in between the seasonal formation of the NEW Polynya. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) conditions. Furthermore, proxy-based methods for past sea-ice reconstructions have been further developed and applied within the Fram Strait and Arctic Ocean regions (e.g., de Vernal et al., 2013;Pie nkowski et al., 2016;Belt, 2018;Belt et al., 2019;Limoges et al., 2018).
The Holocene (0e11.7 ka), generally more climatically stable than the previous glacial transition (Grootes et al., 1993), was punctuated by rapid changes in climate and ocean dynamics (Mayewski et al., 2004;Sundqvist et al., 2014). In the Northern Hemisphere, the Holocene is roughly divided into three main intervals: (1) the early Holocene (~11.7e9 ka); (2) the mid Holocene (~9e5 ka) and (3) the late Holocene (~5e0 ka) (Briner et al., 2016;Renssen et al., 2006;Wanner et al., 2011). Especially, the late Holocene near-boundary conditions witnessed short-term alternations (<500 yrs), e.g. the Medieval Warm Period, Roman Warm Period and Little Ice Age (Ljungqvist, 2010). A declining solar insolation, freshwater influence, volcanic activity, and changes in the AMOC have been suggested as drivers for these short-term climate events (Andersen et al., 2004a(Andersen et al., , 2004bLaskar et al., 2004;Rohling and P€ alike, 2005;Renssen et al., 2006;Wanner et al., 2011). The Arctic Oscillation (AO) and North Atlantic Oscillation (NAO) regulate sea level pressure and the outflow of drift ice through the Fram Strait (Rigor et al., 2002), and significantly influence the weather and climate in the Northern Hemisphere on interannualdecadal timescales (Darby et al., 2012). Positive or negative AO/ NAO modulations control the extensity of the Transpolar Drift system moving sea ice from Siberian shelves and the central Arctic Ocean in the direction of Greenland (Rigor et al., 2002;Bennike, 2004). So far, short-term variability of sea-ice extent is not sufficiently resolved in climate models, even though they have great relevance for predictions of the short-term variability in future climate (Loeb et al., 1997;Dieckmann and Hellmer, 2003;Funder et al., 2011).
This study utilizes biomarkers preserved in marine sediments from Core PS93/025, recovered from the outer eastern North Greenland shelf in the Northeast-Water Polynya (NEW Polynya, Fig. 1). This site is ideally suited to identify and disentangle the driving mechanisms of sea-ice distribution in the western Fram Strait. As proxies for the reconstruction of sea-ice cover we have used the sea-ice proxy IP 25 , a highly branched isoprenoid (HBI) monoene with 25 carbon atoms (Belt et al., 2007), in combination with specific open-water phytoplankton and terrestrial higher land plant biomarkers (Müller et al., 2009;Stein et al., 2012;Belt and Müller, 2013;Belt et al., 2015Belt et al., , 2019 for further details see chapter 4.1). Based on these high-resolution biomarker data we are able to reconstruct sea-ice variability, primary productivity, terrigenous input and seasonal formation of the NEW Polynya that evolved during the Holocene at the eastern North Greenland shelf. Our results are compared with the eastern Fram Strait sea-ice trends.

Regional setting
The Fram Strait is the only deep-water connection between the Arctic and global oceans. Arctic Deep Water, filling the Fram Strait at depths greater than 800 m is overlain by Atlantic Intermediate Water (AIW) (Aagaard and Coachman, 1968a;Rudels and Quadfasel, 1991). The upper water layer is zonally divided by local surface ocean currents. On the eastern side of the Fram Strait relatively warm, saline Atlantic surface waters enter the Arctic Ocean via the West Spitsbergen Current (WSC, Fig. 2a). In contrast, the western Fram Strait is dominated by the outflow of cold Arctic surface waters of the East Greenland Current (EGC) (Aagaard andCoachman, 1968a, 1968b;Rudels et al., 1999). The low-saline EGC flows south along the eastern North Greenland shelf (Fig. 2a). These regional ocean currents result in a strong zonal surface temperature gradient. This zonal difference in sea-surface temperature causes a large disparity in sea-ice cover between the western and eastern Fram Strait. Consequently, west Spitsbergen is characterized by the northernmost ice-free water conditions during winter, while the sea-ice cover along the continental shelf off eastern North Greenland stretches much further south, due to regional sea-ice build-up and transport from the Arctic Ocean (Aagaard, 1982;Vinje, 1977). Sediments, drift ice and low-saline water masses formed on the Eurasian shelfs are transported via the Transpolar Drift across the central Arctic Ocean and eventually through the Fram Strait (Aagaard andCoachman, 1968a, 1968b;Eicken et al., 1995;Reimnitz et al., 1995;Schlichtholz and Houssais, 1999;Smedsrud et al., 2017).
Sea-ice conditions over the wide continental eastern North Greenland shelf in the western Fram Strait between 79 and 80 " are additionally influenced by the waxing and waning of the~700 km long Northeast Greenland Ice Stream (NEGIS), the largest ice stream in eastern North Greenland (Aagaard and Coachman, 1968b, Fig. 2b). This fast-flowing ice stream (20 m/year) drains 20% of the Greenland Ice Sheet and seasonally supplies large amounts of ice and freshwater to the eastern North Greenland shelf (Rignot and Kanagaratnam, 2006;Joughin et al., 2010;Rignot et al., 2011).
The coastal area north of NEGIS at~80 " N is characterized by the highly variable NEW Polynya (Fig. 1). Coastal polynyas are mostly ice-free areas surrounded by a denser sea-ice cover (Smith et al., 1990;WMO, 1985;Stirling, 1980). Northerly winds in combination with local ocean currents remove ice from the coastal region while grounding of icebergs and ice islands can prevent ice advection from the Arctic into the polynya region (Barber and Massom, 2007).
When the NEW Polynya opens up in April/May, reaching its final extension in summer, an anticyclonic circulation pattern appears over almost the entire~300 km wide and shallow Belgica Bank with water depths ranging from 40 to 300 m (Bourke and Garrett, 1987;Kuklinski and Bader, 2007, Figs. 1 and 2c). A combined effect of a fast-ice barrier (the Norske Øer Ice Barrier), katabatic winds and a northward flowing North East Greenland Coastal Current (NEGCC) form the NEW Polynya (Fig. 2b, Schneider and Bud eus, 1997). The water column in the NEW Polynya is characterized by AIW at the bottom overlain by Polar Water (PW) that circulates anticyclonic at the surface (Bourke and Garrett, 1987). Additionally, the surface waters in the polynya region are affected by melting of sea ice along the marginal ice zone and freshwater input from ice melting on land. The freshwater input from these processes stratifies the upper water column and fosters intense biological production until nutrients become depleted (Hirche et al., 1991;Schneider and Bud eus, 1994;Bud eus and Schneider, 1995). A significant ecosystem for primary producers (e.g., diatom blooms), marine mammals and seabirds is supported by the seasonally reduced ice cover of the recurring seasonal NEW Polynya (Stirling, 1980;Wadhams, 1981;Barber and Massom, 2007). Today, sea-ice cover varies from a dense sea-ice cover in winter to stable-ice-edge conditions during late spring/summer. This is also supported by elevated Total Organic Carbon (TOC) concentrations in sediments here, relative to surrounding areas ( Fig. 2a; Birgel and Stein, 2004). However, understanding of the evolution and processes leading to polynya formation remain inconclusive.

Sediment material
Kastenlot Core PS93/025e2 was taken during RV Polarstern Cruise PS93.1 (2015) on the continental shelf off eastern North Greenland (80 " 28.90 0 N, 8 " 29.40 0 W) in 290.2 m water depth ( Fig. 2a and b, Stein, 2016). The coring location was selected based on detailed Hydrosweep and Parasound surveys. These surveys were used to identify small depressions with thick fillings of deglacial to Holocene sediments on the eastern North Greenland shelf, an area that is generally characterized by lag deposits due to strong bottom current activity. The recovered sediments consist of dark grayish brown, homogenous silty clay with relatively low magnetic susceptibility and wet bulk density values (Stein, 2016). Because of coring loss in the upper 25 cm of giant Kastenlot Core PS93/025e2, we correlate the uppermost part of the Kastenlot Core with a Box Core (GKG) PS93/025e1 covering the upper 43 cm of near surface sediments, based on the TOC content and spliced the records together at 27 cm to obtain a continuous and complete sedimentary record (named as Core PS93/025). From the Kastenlot and GKG cores, subcores were taken for multidisciplinary studies. One set of subcores devoted to biomarker analysis, was stored at # 20 " C to prevent biomarker degradation.

Chronology
The age model is based on 19 accelerator mass spectrometry (AMS) 14 C ages measured on macro-and microfossil species, including benthic foraminifera (mixed species), planktic foraminifera (Neogloboquadrina pachyderma), bivalves, and scaphopods (not further identified) from GKG Core PS93/025e1 and Kastenlot Core PS93/025e2 (Table 1). AMS 14 C dating was carried out at the Alfred Wegener Institute Bremerhaven using the "Mini Carbon Dating System (MICADAS) and at the Leibniz-Laboratory in Kiel (Graphite Target). Radiocarbon ages were converted to calibrated calendar years before present (cal. years BP) using the calibration software CALIB 7.1 and the Marine13 calibration curve (Stuiver et al., 2019, Table 1). Five AMS 14 C ages at depths of 93.5 cm, 110.5 cm, 130.5 cm, 238 cm and 264 cm (Table 1, grey shaded areas) turned out to be reworked material within the sedimentary section and were therefore excluded. At a depth of 93.5 cm two AMS 14 C datings were performed on two different organism groups (bivalve and benthic foraminifera), but only one was used for the age model (Table 1).
Based on the radiocarbon ages we present two age models (Fig. 3). The first approach uses a 2nd order polynomial regression through the radiocarbon ages (see Zehnich et al., 2019). The second chronology is based on modeling by using the Bayesian R package Bchron (Parnell et al., 2008) and the marine 13 calibration curve (Reimer et al., 2013). With the Bchron model, two outliers at 62.5 and 125.5 cm would lead to extreme values up to 100 cm/kyr in the calculated sedimentation rate for which no evidence is found in the core data. We therefore excluded these two results from the Bayesian model. In the present study we have used the Bchron age model for three reasons: Firstly, the Bayesian chronology remains within the error margin for the individual radiocarbon measurements; secondly, the 2nd order polynomial can only curve once, which may not fully reflect the variable nature of sediment input or flux under changing climatic conditions; thirdly, the Bayesian approach provides uncertainties and increases robustness of the age model. 10,000 age-depth realizations were obtained to estimate the median age and 95% confidence intervals at 1 cm resolution (Fig. 3). We acknowledge that application of the Bchron age model results in a higher variability of calculated sedimentation rates and that a possible vertical displacement (from reworking or Table 1 AMS 14 C radio carbon ages for Kastenlot Core PS93/025 were determined on benthic and planktic foraminifers, bivalves and scaphopods. Bold CALIB median numbers were used for both age models (red ones were excluded in the Bchron age model).

Lab-ID
Depth ( bioturbation) of some of the dated materials cannot be ruled out. Nevertheless, the overall trends are very similar for both age models as shown, for example, in the IP 25 accumulation rate records (Fig. 3). The sedimentation rates from the Bchron age model range between 25 and 30 cm/kyr over the past 10.2 ka, except for three intervals of high sedimentation rates (up to~40e45 cm/kyr) between 0.8 and 1.1 ka, 5.7e6 ka and 9e10 ka (Fig. 3).

Organic bulk sediment parameters
Total organic carbon (TOC), total carbon (TC) and total nitrogen (TN) were measured using freeze-dried and homogenized bulk sediment samples in intervals of 1 cm. TOC measurements were performed with a CarboneSulfur Analyser (CS-125, Leco) after removing any carbonate by adding hydrochloric acid to a 0.1 g subsample. TC and TN contents were measured on sediment samples of 20 mg sediment with a CarboneNitrogeneSulfur Analyser (Elementar III, Vario). Subsequently, the carbonate content (CaCO 3 ) was calculated from TOC and TC (CaCO 3 ¼(TC-TOC) x 8.333), assuming that carbonate predominantly occurs in the form of calcite, where 8.333 is the stoichiometric calculation factor. The TOC vs. total nitrogen (C/N) ratio may give some first-order information about the organic-carbon source, with higher/lower values more indicative for a terrigenous/marine origin (for details and limitations of this approach see Stein and Macdonald, 2004).
Component assignment was based on comparison of GC retention times with those of reference compounds and published mass spectra (sterols: Boon et al., 1979;Volkman, 1986;HBIs: Belt et al., 2007;Brown and Belt, 2016). The concentration of each biomarker was calculated by setting its individual GC-MS ion responses in relation to those of respective internal standards. For the quantification of the sterols (quantified as trimethylsilyl ethers), the molecular ions m/z 470 for brassicasterol, m/z 472 for campesterol, m/z 486 for b-sitosterol, and m/z 500 for dinosterol were used in relation to the molecular ion m/z 348 of the internal standard androstanol. For the quantification of IP 25 , HBI II and HBI III, their molecular ions (m/z 350 for IP 25 , m/z 348 for HBI II, and m/z 346 for HBI III) were compared to the fragment ion m/z 266 of the internal standard 7-HND. The different responses of these ions were balanced by an external calibration. For further details we refer to Fahl and Stein (2012), Brown et al. (2014) and Belt (2018).
The calculated Kovats Index for IP 25 is 2085, for HBI II 2084 and 2046 for HBI III. All biomarkers of this study were normalized against the amount of sediment and TOC content.
We calculated the PIP 25 indices according to the equation by Müller et al. (2011), where c is a concentration balance factor (to account for concentration differences between IP 25 and phytoplankton biomarkers (P), i.e., brassicasterol, dinosterol or HBI III: On average, concentrations of brassicasterol are about five times higher, dinosterol about ten times higher and HBI III about two times lower than the IP 25 concentrations in Core PS93/025, resulting in balance factors of 0.23, 0.11 and 2.44, respectively.
Biomarker concentrations were converted into accumulation rates by using the following equations (e.g. Stein and Macdonald, 2004):

Statistical analyses
Significant shifts in biomarker records of Core PS93/025 (Fig. 4) were identified using the changepoint analysis from the R package ECP with a significance level of 99% (James and Matteson, 2013;R Core Team, 2018). The variability of average HBI and sterol concentrations of the five intervals defined in the Holocene record of Core PS93/025 is shown in Fig. 5.
For spectral analysis, IP 25 , brassicasterol and HBI III time series were resampled at 0.03 kyr resolution prior to spectral analysis. Then spectral analysis was carried out using the multitaper method (MTM) with three 2p tapers (Thomson, 1982) and LOWSPEC noise estimation as implemented in the R package "astrochron" (Meyers, 2012). Confidence levels were calculated by applying the LOWESSbased (Cleveland, 1979) procedure of Ruckstuhl et al. (2001).

Organic/inorganic bulk parameters
In the following, data are presented versus age. Single records of the bulk parameters versus depth are available in the supplementary material S1. The TOC record of Core PS93/025 can be divided into three intervals with low values of 0.5e0.6% between 10.2 and 7.9 ka, constant values of~0.6% from 7.9 to 5.1 ka and increasing values from 5.1 ka towards the top, reaching 0.8e0.9% in the last 1 ka (Fig. 4). High C/N ratios of 7.5e9 occur in the lower part of the core between 10.2 and 8.8 ka whereas lower values between about 6 and 7.5 are typical for the interval from 8.8 ka onwards (Fig. 4). Maximum inorganic carbon (IC) values of 0.8e1.0% (i.e., 7e8.5% CaCO 3 ) were determined in the lower part between 10.2 and 7.9 ka. After 7.9 ka, IC values sharply decrease, and values of 0.55e0.65% (i.e., 4.5e5.5% CaCO 3 ) are typical for the interval between 7.9 and 5.5 ka. In the upper part of the core between 5.5 ka and the top, IC values vary between 0.35 and 0.65% (i.e., 3e5.5% CaCO 3 ), with higher values of 0.6e0.65% (i.e., 5e5.5% CaCO 3 ) most dominant in the interval between 1.3 and 0.6 ka (Fig. 4).

Biomarker
Results of all biomarker records are plotted in mg g # 1 TOC and as accumulation rates in mg cm # 2 ky # 1 versus age (Fig. 4). In addition, each biomarker is plotted in mg g # 1 Sed and mg g # 1 TOC versus depth, presented in the supplementary material S2. Due to the good correlation of campesterol and b-sitosterol, both records were combined into one record of terrigenous biomarkers that is used in the discussion. The sterols (brassicasterol, dinosterol, campesterol and b-sitosterol) of Core PS93/025 vary more or less synchronously over the entire Holocene record, which can be subdivided into three main intervals (Fig. 4). High concentrations of the phytoplankton biomarkers brassicasterol (4e6 mg g # 1 TOC) and dinosterol (~8e10 mg g # 1 TOC) and the terrigenous biomarkers campesterol and b-sitosterol (both~13 mg g # 1 TOC) occur between 10.2 and 9.3 ka. Relatively stable, but low values of brassicasterol (~2 mg g # 1 TOC), dinosterol (~4 mg g # 1 TOC), campesterol and bsitosterol (~5 mg g # 1 TOC), are obvious between 9.3 and 1 ka. During the last 1 ka, all sterols reach maximum values with brassicasterol (up to 8 mg g # 1 TOC), dinosterol (up to 15.5 mg g # 1 TOC), campesterol and b-sitosterol (both up to 22.5 mg g # 1 TOC) (Fig. 4).
The IP 25 record can also be subdivided into three intervals. Low IP 25 values of 0.4 mg g # 1 TOC characterize the interval between 10.2 and 7.9 ka, while slightly higher values of~0.6 mg g # 1 TOC were found between 7.9 and 4.8 ka (Fig. 4). From 4.8 ka onwards, the IP 25 concentrations increase continuously, reaching maximum values of 1.4 mg g # 1 TOC during the last 1 ka (Fig. 4). The HBI III record was divided into five intervals. High HBI III concentrations occur from 10.2 to 9.3 ka (values up to~0.6 mg g # 1 TOC), between 7.9 and 5.5 ka (up to 0.4 mg g # 1 TOC) and during the last 1 ka (up to 0.8 mg g # 1 TOC), intercalated by intervals of lower concentrations between 9.3 and 7.9 and 5.5 and 1 ka with values below 0.1 mg g # 1 TOC (Fig. 4) Overall, accumulation rates of all biomarkers reflect similar trends during the Holocene.

Discussion
4.1. Proxies used for past sea-ice reconstructions on the eastern North Greenland shelf: some general aspects In this study past sea-ice reconstructions on the eastern North Greenland shelf were carried out by using the sea-ice proxy IP 25 (Belt et al., 2007) in combination with the tri-unsaturated HBI (HBI III, phytoplankton biomarker) and specific sterols (phytoplankton and terrigenous biomarkers). IP 25 is produced by diatoms that inhabit the sea ice with highest concentrations during spring algae blooms (Belt et al., 2007;Brown et al., 2014Brown et al., , 2016. Thus, IP 25 is a reliable and direct indicator for past seasonal sea-ice presence. Sediment traps studies evidence the release of ice algal organic carbon (OC) containing the IP 25 signal, that sinks to the ocean surface during the ice melt (Belt et al., 2008;Fahl and Stein, 2012). Over the past years, IP 25 has been successfully used to reconstruct sea-ice variability across the Arctic (for reviews see Stein et al., 2012;Belt and Müller, 2013;Belt, 2018) even in regions with a near-permanent sea-ice cover (Xiao et al., 2015). However, this proxy has a few limitations (Belt, 2018). Sea-ice algae are absent in ice-free areas but also absent or only occur in very minor amount in areas characterized by a more closed and thick sea-ice cover that prevents significant algae productivity. Thus, IP 25 is not found or only found in very minor amount in the underlying sediments. Additionally, several other factors may hamper the IP 25 production and accumulation, for example lower salinities in fjords or nearcoastal surface waters due to large river input or meltwater discharge (Xiao et al., 2013;H€ orner et al., 2016;Ribeiro et al., 2017), or brine channels within the sea ice, or reduced light availability through an overlying snow cover (Belt and Müller, 2013). Lately, the discussion about degradation of IP 25 in terms of ice-algal carbon as foodstock (grazing) within the water column, and oxidation of IP 25 within the sea ice, has received greater awareness (for details see Belt, 2018). Another substitutional sea-ice proxy (biosynthesized by Arctic diatoms) and often very well correlated with IP 25 , might be the HBI II diene (Brown et al., 2014), with concentrations often somewhat higher than those of the IP 25 (for details see Belt, 2018).
IP 25 in combination with open-water phytoplankton biomarkers allows a more detailed classification of sea-ice conditions. As openwater phytoplankton biomarkers Müller et al. (2009Müller et al. ( , 2011 originally used specific sterols (brassicasterol, dinosterol), and proposed a combined phytoplankton-IP 25 proxy (PIP 25 ) that allows a more semi-quantitative description of spring sea-ice conditions (for PIP 25 calculations see subchapter 2.2.3; for general information about methods and interpretation of the approach see Müller et al., 2011;Belt et al., 2015;Smik et al., 2016). According to Müller et al. (2011), sea-ice conditions were classified into ice-free conditions (PIP 25 ¼ 0), a low to variable sea-ice cover (PIP 25 > 0.1) with low IP 25 and high phytoplankton biomarker values, the marginal ice zone (PIP 25 > 0.5) with intermediate values and extended sea-ice conditions (PIP 25 > 0.75) with high IP 25 and low phytoplankton biomarker values. A perennial closed sea-ice cover would result in IP 25 and phytoplankton biomarker concentrations near zero, i.e. PIP 25 would become indeterminable and thus often artificially set to 1. These extreme values, however, have to be interpreted with caution and the single biomarker concentrations should be considered as well (cf., Stein et al., 2017;Belt, 2018). Furthermore, the degree of accuracy of PIP 25 may depend on the combination of specific phytoplankton biomarkers (and their pelagic origin), the c balance factor (c.f. subchapter 2.2.3), the algorithmic relationship between the PIP 25 and the different sea-ice conditions, as well as the temporal shifts within the downcore records (Belt, 2018).
More recently, the HBI III triene (e and z triene) has been used as a complimentary proxy to IP 25 . HBI III is probably an indicator for pelagic diatom productivity of the species Rhizosolenia setigera in the photic zone of the water column, and is strongly associated in (sub-) polar regions with the Marginal Ice Zone (MIZ) (Collins et al., 2013;Smik et al., 2016;Belt, 2018). The consideration of iceassociated IP 25 and HBI II alongside with the pelagic HBI III triene (e and z triene) might help to distinguish between different sea-ice scenarios (see Belt, 2018). Positively correlated IP 25 and HBI III profiles were assumed to reflect rapidly changing (advancing/ retreating) sea-ice margins or offshore polynyas (cf., Belt, 2018) and negatively correlated profiles are attributed to temporal shifts in the winter/summer sea-ice margin (Belt et al., 2015;Berben et al., 2017). Furthermore, different interpretations of the proxy data might be possible when using this approach in different regions.  For example, based on data of surface sediments from the Barents Sea and around Svalbard, Smik et al. (2016) concluded that the HBI III triene in combination with IP 25 (P III IP 25 ) is a promising indicator for winter sea-ice extent and/or the MIZ. Findings from eastern North Greenland fjord surface sediment samples seem to support this statement since HBI III concentrations were also found to be most abundant in locations proximal to the ice edge (Ribeiro et al., 2017). In contrast to the Barents Sea area, however, the HBI maximum concentrations correlate with the mid-July MIZ in the Greenland fjord area that is characterized by much heavier sea-ice conditions in general.
To identify terrigenous organic carbon, we have used specific sterols (b-sitosterol and campesterol), C/N ratios and TOC values. For further details of the identification of organic carbon sources see Stein and Macdonald (2004); Stein (1999, 2012) and Belt (2018).
In summary, multivariate analysis of individual open-water phytoplankton and terrigenous biomarkers in combination with IP 25 and PIP 25 records may help to identify different sea-ice scenarios and organic carbon source areas (marine vs. freshwater) (see Fig. 6 and suppl. mat. S3). During the late Holocene between 10.2 and 9.3 ka, low IP 25 and HBI II fluxes reflect reduced to moderate sea-ice algae productivity, while considerably high phytoplankton biomarker concentrations and fluxes of brassicasterol, dinosterol and HBI III (Fig. 4) indicate prolonged open-water conditions with enhanced biological productivity at the outer eastern North Greenland continental shelf. Such a reduced sea-ice situation is also clearly reflected in the related minimum P B IP 25 , P D IP 25 and P III IP 25 indices of <0.3 (Fig. 7) and the scatter plot of IP 25 indices against the phytoplankton biomarker brassicasterol and HBI III (Fig. 6). Considering the generally reduced sea-ice conditions and the distinct HBI III maximum, this may represent a winter MIZ situation (Fig. 4, interval 1) as proposed by Belt et al. (2015) for the modern Barents Sea. Strongly reduced sea-ice conditions were also reconstructed for the inner East Greenland shelf (Core PS2641-4) and the eastern Fram Strait (MSM5/5-712-2 and MSM5/5-723-2) during the early Holocene based on relatively high phytoplankton biomarker values, high CaCO 3 and lowered IP 25 and IRD contents . The eastern Fram Strait sea-ice conditions during the early Holocene most likely correspond to the present-day summer situation (cf., Fig. 1c).

Sea-ice variability off eastern North Greenland during the
The low sea-ice coverage and relatively warm surface water conditions due to the intensified inflow of warmer recirculating Atlantic Water from the eastern towards the western Fram Strait (Koç et al., 1993;cf. Carstens et al., 1997;Volkmann, 2000;Pados and Spielhagen, 2014), favored planktic blooms at our core site, as reflected by elevated numbers of planktic and benthic foraminifers and highest (biogenic) CaCO 3 content and flux (Fig. 4). Warmer seasurface conditions were also reconstructed for the Nordic Seas and the East Greenland shelf based on foraminifers and diatoms (Bauch et al., 2001;Andersen et al., 2004aAndersen et al., , 2004b. The appearance of Armeria scabra and the sub-arctic mussel Mytilus edulis, both species are well-known for their occurrence in eastern North Greenland, further support warmer summer temperatures than today during the early Holocene between 10 and 8.5 ka (Hjort and Funder, 1974;Bennike and Weidick, 2001).
During this warmer early Holocene period, elevated fluxes of terrigenous organic matter indicated by high b-sitosterol and campesterol values and high C/N ratios (Fig. 4), and the decrease of IRD deposition (Telesi nski et al., 2014;Zehnich et al., 2019) may have resulted from melting by local glaciers from eastern North Greenland and related meltwater discharge (cf., Blaschek and Fig. 6. a) Scatter plot (mg g # 1 TOC) of IP 25 versus brassicasterol with different sea-ice scenarios after Müller et al. (2011). b) scatter plot (mg g # 1 TOC) IP 25 versus HBI III of Core PS93/ 025-2 highlighting the early Holocene warm phase (orange shaded area) and the assumed seasonal formation of the NEW polynya during the last 1 ka (blue shaded area). Data from the five Holocene intervals are indicated by different colors. The yellow circles mark data from the uppermost 1-6 cm of Core PS93/025, probably representing modern-type conditions. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) Renssen, 2013). This may have provided nutrient-rich freshwater to the eastern North Greenland shelf causing stratification of the surface layer and intensified phytoplankton growth (Waniek et al., 2005). This explanation would also explain the high positive correlation between the phytoplankton (i.e., brassicasterol and dinosterol) and the terrigenous biomarkers (b-sitosterol and campesterol) (suppl. material S3). In summary, our data of Core PS93/025 and multiproxy records (e.g., IRD, d 18 O, d 13 C, benthic and planktic foraminifera abundances) from other nearby cores (Telesi nski et al., 2014) relate the phase between 10.2 and 9.3 ka to be linked to the Holocene Thermal Maximum (HTM), where increased mean local summer insolation led to summer temperatures 4e5 " C warmer than today in northern Greenland (Kaufman et al., 2004;Laskar et al., 2004;Reeh, 2004;Lecavalier et al., 2017). Around 9.3 ka a prominent change in environmental conditions occurred, as reflected in a significant shift from high to low concentrations and accumulation rates of all biomarkers (Figs. 4 and 5) suggesting reduced sea-ice and open-water algal productivity. This change may be related to an extended sea-ice cover as shown in the elevated PIP 25 indices (Fig. 7). Superimposed to this change towards an increased sea-ice cover, three distinct maxima and minima in the higher-resolution carbonate concentration and planktic foraminifer are obvious, implying short-term disturbances in seasurface conditions (Fig. 4). A reasonable explanation for these disturbances might be short-lived small cooling events, similar to the 8.2 Event (Kobashi et al., 2007;Rasmussen et al., 2007;Thomas et al., 2007), which was probably triggered by freshwater outbursts from the Laurentide Ice Sheet-dammed lakes into the North Atlantic Ocean and had a considerable impact on the thermohaline circulation (Barber et al., 1999;Yu et al., 2010). Thus, the minima in our records (Fig. 4) between 7.9 and 9.3 ka may reflect increased sea-ice conditions and strongly reduced phytoplankton productivity due to abrupt freshwater outbursts.

Mid Holocene (7.9e4.9 ka)
Elevated IP 25 and HBI II values, but overall low and stable (with minor fluctuations) phytoplankton biomarker (brassicasterol and dinosterol) and terrigenous biomarker concentrations and accumulation rates revealed during the mid Holocene (Fig. 4). Rising PIP 25 indices indicate a gradual transition from reduced to seasonal sea-ice conditions along the outer eastern North Greenland continental shelf (Fig. 8b). Contemporaneously, decreasing CaCO 3 concentrations and relatively low abundances of foraminifers (Zehnich et al., 2019) occur during the same interval. In fact, a decreasing warm Atlantic Water advection may have caused a gradual cooling evidenced by lowered SSTs in the Fram Strait (Bauch et al., 2001;Jennings et al., 2002;Müller et al., 2012) and constantly reduced phytoplankton productivity at Core PS93/025. Meanwhile, increased abundances of Arctic diatom taxa point to a strengthened EGC between 6.8 and 5.5 ka (Ran et al., 2006) that would have transported more sea ice (drift ice) from the Arctic along the eastern North and East Greenland shelf southward towards Iceland (Andrews et al., 2009;. Maximum concentration and flux of HBI III (Fig. 4) might indicate the build-up of a stable MIZ in the interval between about 7 and 6 ka. Following Ribeiro et al. (2017) and their study of surface sediments from the eastern North Greenland shelf, this probably represents a (spring) summer MIZ situation (Fig. 4, interval 3).

The late Holocene (5e1 ka)
During the late Holocene, the three PIP 25 indices do not show any trend, but fluctuate around 0.6 (Fig. 7). These values represent typical seasonal to marginal sea-ice conditions. The ice-associated IP 25 and HBI II records only display a minor increase between 5 and 1 ka (Figs. 4 and 5). This is in contrast to several IP 25 records from other Arctic regions, including the Yermak Plateau (Müller et al., 2009), north of Iceland (Cabedo-Sanz and  and Spitsbergen . Continuously rising P B IP 25 and P D IP 25 in the eastern Fram Strait  point to extended sea-ice conditions following the Neoglacial Cooling trend (Fig. 7). This difference in sea-ice conditions between the eastern and western Fram Strait is also supported by model simulations indicating an increase of sea ice by 30% at 6 ka at the eastern Fram Strait and around Spitsbergen and only a minor increase in sea-ice concentrations at the western Fram Strait (Müller et al., 2011). The increase in sea-ice cover in the eastern Fram Strait may be explained by a declining inflow of warm Atlantic Water via the WSC (c.f., Werner et al., 2013). The western Fram Strait, on the other hand, is continuously influenced by the cold EGC and drift-ice export from the Arctic Ocean.

Sea-ice distribution and NEW polynya formation during the last 1 ka
During the last 1 ka, maximum concentrations and accumulation rates of IP 25 and open-water phytoplankton and terrigenous biomarkers, as well as TOC were recorded at Core PS93/025 (Figs. 4 and 5). These significant changes are related to the seasonal formation of the NEW Polynya with suitable environmental conditions for the generation of these biomarkers. Sea ice was pushed towards the open ocean by strong katabatic winds coming from the Greenland Ice Sheet and causing the NEW Polynya to open up (Fig. 8c). Such open-water conditions in its centre and sea-ice melting at the margins caused increased fluxes of terrigenous suspension and nutrients to the surface waters within the polynya and favored primary productivity, indicated by highest brassicasterol, dinosterol and HBI III values (Fig. 4) in the NEW Polynya. The build-up of a stationary (summer) MIZ is especially reflected in the elevated HBI III concentrations (light blue field in Fig. 6b; cf., Belt et al., 2015;Ribeiro et al., 2017). Low carbonate fossil deposition (M. Zehnich, pers. comm. 2019) and low CaCO 3 accumulation during this interval of high productivity (Fig. 4) may result from carbonate dissolution, a process that is commonly observed in areas of high productivity and seasonal sea-ice formation (Steinsund and Hald, 1994;Knies et al., 2003). Similar findings were recorded for the recent NEW Polynya situation, where nutrients are constantly supplied by the NEGCC and favor primary productivity (Schneider and Bud eus, 1997). Dissolved organic carbon occurs in a larger amount within the NEW Polynya compared to the surrounding surface waters and is most likely released by melting sea ice and ice algae (Skoog et al., 2001). Unfortunately, foraminifer assemblages from Core PS93/025 do not give a hint on carbonate dissolution during the late Holocene (Zehnich et al., 2019). Core top biomarker concentrations (uppermost 1-6 cm) in our record match very well with modern observations of sea-ice cover and TOC values in the NEW Polynya ( Fig. 2a; Birgel and Stein, 2004). Knies et al. (2017) described Arctic polynyas as sea-ice factories and connected the increase in seasonal sea-ice cover with a strengthened outflow of drift ice from the Arctic Ocean along the Greenland shelves during the last 0.5 ka. Ice algae benefit from the light and nutrients availability and are able to grow within the melting sea ice (Belt et al., 2007), which would explain the elevated number of sympagic algae produced IP 25 and HBI III (Figs. 5 and 4,interval 5). Several other studies based on foraminifer and diamicton records from East Greenland and Southeast Greenland and other marine and terrestrial proxies support a strengthening of the EGC during the last 0.6e0.7 ka (Jennings and Weiner, 1996;Nesje et al., 2000;Sepp€ a et al., 2002;Andersson et al., 2003;Moros et al., 2006;Bendle and Rosell-Mel e, 2007;Sejrup et al., 2010;Mernild et al., 2011;.

Short-term cyclic variability in Holocene environmental conditions
Spectral analysis reveals significant power at frequencies exceeding the 90% confidence level and corresponding to periodicities of 110e130, 170, 220 and 500 years in the proxy records of IP 25 , HBI III and brassicasterol throughout the Holocene (Fig. 9). Such short-term cycles of 110e130 year periods have been observed in the oxygen isotope record of the NGRIP (NGRIP-Members, 2004) and are likely linked to variations in solar activity (Vonmoos et al., 2006). A spectral power centered around a period of 200 years (brassicasterol) may correspond to the De Vries/ Suess cycle of solar activity. North Atlantic variability at this timescale has been previously found in the 512 year D 14 C cycle and is connected to variabilities in the North Atlantic Deep Water Formation (NADW) (Stuiver and Braziunas, 1993). While the exact period of spectral peaks may be biased by uncertainties in the age model, the correspondence between our and other oceanic and terrestrial variations in paleoclimate proxies with cyclic signals in solar activity (Stuiver et al., 1995;Ito and Yu, 1999;Chapman and Shackleton, 2000;Hu et al., 2003;Zhao et al., 2010) suggests a possible key role of the amount of total solar irradiance as a driver of oceanic and atmospheric change in the Arctic region. Thus, the atmospheric response to reduced solar irradiance could lead to increasing Arctic sea-ice drift and cooling of both the ocean surface and high northern latitude continents and trigger the negative state of AO/NAO (Xu et al., 2014).

Conclusions
A new high-resolution biomarker record from Core PS93/025 provides new insights into the Arctic sea-ice evolution and for the first time documents the formation of the NEW Polynya on the eastern North Greenland continental shelf during the late Holocene. The presence of IP 25 throughout the Core PS93/025 confirms that there has been seasonal sea ice in the area during the entire record. Our biomarker proxies indicate relatively rapid changes in sea-ice conditions at~9 ka and~1 ka, i.e., sea-ice conditions progressed through three major stages over the course of the Holocene. During the early Holocene, we recorded a reduced but variable sea-ice cover. Between about 9.3 and 5.5 ka, sea-ice coverage increased towards seasonal conditions. Based on terrigenous biomarkers and IRD we assume a stronger regional than local sea-ice signal at core site PS93/025, due to the high influence of drift ice transported from the central Arctic Ocean along the eastern North Greenland shelf. During the late Holocene, especially during the last 1 ka, our records reflect the seasonal formation of the NEW Polynya leading to stable sea-ice edge conditions and a fully developed polynya situation. Probably, cyclic changes in the solar activity acted as trigger for the short-term variability in sea-ice cover during Holocene times.