Reconstructing spring sea ice concentration in the Chukchi Sea over recent centuries: insights into the application of the PIP25 index

In this study, we aimed to reconstruct spring (April–June) sea ice changes in the western Arctic Ocean over recent centuries (ca. the last 250 years) by measuring biomarker distributions in a multicore (ARA01B-03MUC) retrieved from the Chukchi Shelf region and to evaluate outcomes against known or modelled estimates of sea ice conditions. Specifically, we analyzed for the Arctic sea ice proxy IP25 and assessed the suitability of a further highly branched isoprenoid (HBI) lipid (HBI III), epi-brassicasterol, and dinosterol as complementary biomarkers for use with the so-called phytoplankton marker-IP25 index (PIP25; PIIIIP25, PBIP25, and PDIP25, respectively). The presence of IP25 throughout core ARA01B-03MUC confirms the occurrence of seasonal sea ice at the study site over recent centuries. From a semi-quantitative perspective, all three PIP25 indices gave different trends, with some dependence on the balance factor c, a term used in the calculation of the PIP25 index. PIIIIP25-derived spring sea ice concentration (SpSIC) estimates using a c value of 0.63, determined previously from analysis of Barents Sea surface sediments, were likely most reliable, since SpSIC values were high throughout the record (SpSIC > 78%), consistent with the modern context for the Chukchi Sea and the mean SpSIC record of the 41 CMIP5 climate models over recent centuries. PBIP25-based SpSIC estimates were also high (SpSIC 108%−127%), albeit somewhat over-estimated, when using a c value of 0.023 obtained from a pan-Arctic distribution of surface sediments. In contrast, PDIP25 values using a pan-Arctic c value of 0.11, and PIP25 data based on the mean biomarker concentrations from ARA01B-03MUC, largely underestimated sea ice conditions (SpSIC as low as 13%), and exhibited poor agreement with instrumental records or model outputs. On the other hand, PBIP25 values using a c factor based on mean IP25 and epi-brassicasterol concentrations exhibited a decline towards the core top, which resembled recent decreasing changes in summer sea ice conditions for the Chukchi Sea; however, further work is needed to test the broader spatial generality of this observation.


Introduction
Over the last four decades, the Arctic Ocean has experienced a persistent loss of sea ice, which is one of its main characteristics (e.g. Stroeve et al 2007, Serreze andStroeve 2015). Such trends have been based mainly on satellite passive-microwave records spanning the last 40 years (e.g. de Vernal et al 2013), although some historical records extend back to ca. 1850 AD (e.g. Walsh et al 2016). Such historical records have revealed seasonal variability, with the most profound sea ice reduction occurring in late summer (September), especially evident in the Chukchi Sea (western Arctic Ocean), even though winter (i.e. March) sea ice extent has remained largely unchanged in this region (Walsh et al 2016). Deciphering longer-term changes in sea ice on a seasonal and spatial basis remains a key aim in paleoclimatology.
A number of biogenic or geochemical proxies have been developed and applied to reconstruct sea ice conditions in the past (de Vernal et al 2013 and references therein). Amongst these is a mono-unsaturated highly branched isoprenoid (HBI) alkene (IP 25 -Ice Proxy with 25 carbon atoms; Belt et al 2007), which is biosynthesized by certain Arctic sea ice dwelling diatoms during the spring bloom (Brown et al 2014) and, on sea ice melt, is deposited in underlying sediments (Belt et al 2007). Since its initial discovery, IP 25 has become an established northern hemisphere proxy for the qualitative occurrence of seasonal sea ice in various paleo records (e.g. Belt andMüller 2013, Belt 2018).
The absence of IP 25 in marine sediments has previously been considered to indicate either ice-free conditions or permanent ice cover (e.g. Belt et al 2007), although the validity of this simple two end-member interpretation still requires further investigation (Belt 2018). In any case, the co-measurement of IP 25 and various phytoplankton biomarkers was first proposed to differentiate between these end-members, since absent IP 25 but elevated phytoplankton biomarker concentrations would likely reflect ice-free conditions, while the absence (or very low abundance) of both biomarker types would be more consistent with intervals of permanent (or near permanent) ice cover (Müller et al 2009). Thereafter, Müller et al (2011) suggested combining IP 25 and open-water phytoplankton lipid biomarker concentrations in the form of a so-called phytoplankton marker-IP 25 index (PIP 25 ; see equation (3)). Importantly, the employment of this biomarker-based ratio method necessitates the use of a balance factor c (see equation (4)) to take account of the significant imbalance that often exists between IP 25 and phytoplankton biomarker concentrations (Müller et al 2011). The balance factor c is derived either from mean IP 25 and phytoplankton biomarker concentrations measured in an individual down-core sediment profile (e.g. Müller et al 2012 or is given as a fixed value, which can be obtained from a regional or pan-Arctic study of surface sediments (e.g. Xiao et al 2015, Smik et al 2016. However, the suitability of both approaches is in need of further investigation.
To date, three different PIP 25 have been used as indices for semi-quantitative proxy measures of sea ice change in some paleo sea ice reconstructions in the Arctic Ocean. These employ epi-brassicasterol (P B IP 25 , e.g. Müller et al 2011), dinosterol (P D IP 25 , e.g. Stoynova et al 2013), or a tri-unsaturated HBI (HBI III) (P III IP 25 , e.g. Belt et al 2015, Smik et al 2016 as the phytoplankton-derived biomarkers. According to Müller et al (2011), PIP 25 values greater than ca. 0.75 are indicative of extended sea ice cover, while values between 0.1 and 0.5 suggest variable or less ice cover. In the Barents Sea, surface sediment data exhibit a strong linear relationship between P III IP 25 and spring sea ice concentration (SpSIC), with outcomes far less dependent on the balance factor c than for P B IP 25 obtained from the same sediments (Smik et al 2016). This relationship, which uses a fixed, regional c value of 0.63, has been applied subsequently to obtain semiquantitative estimates of SpSIC for the Barents Sea over Holocene and longer timeframes (Cabedo-Sanz and , Berben et al 2017, Köseoğlu et al 2018. Notably, virtually all previous investigations into the use of the PIP 25 index have been based either on surface sediments, which are generally attributed to the most recent (years-decades) deposition, or downcore records spanning hundreds to millions of years (see Belt 2018 for a review). In contrast, relatively few studies have focused on IP 25 and other biomarker data in short cores representing recent decades-centuries, for which documented and high-resolution modelled sea ice records could be used for comparison, testing, and calibration purposes (Alonso-García et al 2013, Weckström et al 2013, Cormier et al 2016, Pieńkowski et al 2017. Further, only one such short core has previously been studied from a location (northern Barents Sea; Vare et al 2010, Köseoğlu et al 2018) with essentially uniform (and extensive) spring sea ice conditions against which corresponding consistency in biomarker data could be evaluated.
In the current study, we therefore measured IP 25 , HBI III, epi-brassicasterol, and dinosterol concentrations in a multicore collected from a site in the Chukchi Sea and compared the corresponding PIP 25 records with other published sea ice records (Walsh et al 2016) and modelled sea ice properties in historical simulations from the Coupled Model Intercomparison Project (CMIP5; Taylor et al 2012). Since the core site is one that has experienced relatively constant spring sea ice conditions but declining summer sea ice extent over the most recent centuries, it provided a good opportunity to identify the most suitable biomarker-based proxy method for reconstructing such sea ice trends in the relatively recent past, but beyond that of surface sediments. In doing so, we further show the potential of the P III IP 25 approach for subsequent application to paleo sea ice reconstruction in longer-term records.

Material and methods
Sample collection Sediment core ARA01B-03MUC (23 cm long) was recovered from the shelf of the Chukchi Sea (73.52°N, 168.94°W, 72.5 m water depth) during the 2010 R/V ARAON Expedition (ARA01B) using a MUC 8 multicorer developed by Oktopus GmbH ( figure 1(A)). The study site is influenced by Pacific waters flowing northwards across the Chukchi Shelf along three principal pathways associated with major topographic depressions in the western, eastern, and central Chukchi Sea: more saline (>32.5), nutrient-rich Anadyr Water, fresher (<31.8), more nutrient-limited Alaska Coastal Water, and intermediate salinity  )). According to recent satellite observations, sea ice covers most of the Chukchi Sea from December to May, and retreats during summer with a minimum value of less than 0.6×10 6 km 2 in September. In recent decades, sea ice concentration has especially decreased from July to November (Onarheim et al 2018). The sediments of core ARA01B-03MUC consist of olive-gray clayey silt. The sediment core was sectioned at 1 cm intervals (and a further section at 0.5 cm) on board and stored at -20°C until further treatment.

Radioisotope analyses
Sub-samples (1 cm intervals) were indirectly measured for 210 Pb activity (t 1/2 =22.23±0.12 year) using the 210 Po method (Robbins andEdgington 1975, Nittrouer et al 1979) with an SSB alpha spectrometer (Canberra Inc., PIPS) at the Korea Basic Science Institute (South Korea). The analytical error is on average 2.3±0.8 mBq g −1 (figure 2). Excess 210 Pb activities ( 210 Pb ex ) were calculated by subtraction of supported level values ( 210 Pb sup ) from total activity ( 210 Pb total ). The apparent sedimentation rate (cm yr −1 ) was calculated from 210 Pb ex using the constant flux and constant sediment accumulation rate model based on a slope of the logarithmic regression line (figure 2) as follows: b Apparent sedimentation rate cm yr , 1 where λ is the radioisotope decay constant ( 210 Pb, 0.031 14 per year) and b is the slope of the regression line.

Bulk geochemical analyses
The analytical error was less than ±0.1 wt% for TOC content.

Lipid biomarker analyses
Lipid analyses were conducted at the University of Plymouth (UK) according to Belt et al (2012). Briefly, freeze-dried sediments (ca. 1-2 g) were extracted by sonication (dichloromethane (DCM):methanol (MeOH); 2:1 v:v, 3×3 ml), and partially purified to remove polar components and elemental sulphur using tetrabutylammonium (TBA) sulfite reagent. Internal standards (9-octylheptadec-8-ene: 9-OHD and 5-α-androstan-3β-ol; 0.01 μg) were added prior to extraction for the quantification of HBIs and sterols, respectively. The total extracts were separated into apolar (HBIs) and polar (sterols) fractions using hexane and DCM:MeOH (1:1, v:v), respectively. HBIs and sterols were analyzed using gas chromatographymass spectrometry (GC-MS) with conditions described elsewhere (Belt et al 2012). The polar fractions were derivatized using N,O-bis(trimethylsilyl)trifluoroacetamide (BSTFA) prior to the GC-MS analysis. Individual compounds were identified by total ion current (TIC; m/z 50-500 mass range) chromatograms, while selective ion monitoring chromatograms were used to quantify the abundances of HBI III (m/z 346), IP 25 (m/z 350), epi-brassicasterol (m/z 470), and dinosterol (m/z 500). The different response factors (RFs) were applied to account for the differences in mass spectral responses of selected compounds and the internal standards. However, dinosterol concentrations were obtained by only correcting the selected ion contributions to the total ion counts of 5-α-androstan-3β-ol and dinosterol identified in a reference sample.

Climate model simulations
Sea ice concentrations and sea ice thicknesses in the CMIP5 historical runs by 41 climate models for the period from AD 1862 to AD 2004 were used for comparing modelled sea ice properties with the proxy records. The CMIP5 historical runs were performed by coupled climate models to simulate observed climate changes during the 20th century, with forcing of observed atmospheric composite changes from the nineteenth century to near present (Taylor et al 2012).

210
Pb geochronology The depth profile of excess 210 Pb is presented in figure 2. Measured 210 Pb total activities ranged from 20 to 87 mBq g −1 of dry sediment weight, and 210 Pb sup between 12 and 23 cm had an average value of 22.5±1.6 mBq g −1 (figure 2). The concentration of excess 210 Pb decreased nearly exponentially with sediment depth. The mean sedimentation rate for core ARA01B-3MUC was 0.09 cm yr −1 (R 2 =0.92), corresponding to a ca. 250 year record.

Model data
The ensemble mean of annual sea ice concentrations of the CMIP5 models averaged over the region of the Chukchi Sea (70-80°N and 150°E-150°W) showed a gradual decrease from around AD 1980 ( figure 5). Notably, the winter (January-March) and spring (April-June) mean sea ice concentrations remained high (>96% and >90%, respectively), whilst, after around AD 1980, the sea ice concentrations decreased noticeably in summer (July-September) and autumn (October-December).

Discussion
The sedimentation rate (0.09 cm yr −1 ) obtained at our core site falls within the range reported for other sediment cores from the Chukchi Shelf (0.03−0.37 cm yr −1 ; . The nearly exponential decrease of the excess 210 Pb suggests that mixing by bioturbation and/or physical processes was minor at our core site. Nonetheless, it should be noted that the Chukchi Shelf can have a significant degree of bioturbation due to high benthic macrofaunal populations and biomass. For example, Kuzyk et al (2013) reported surface mixed layers of 10-30 cm on the Chukchi Shelf (<200 m water depth). However, previously published data from a sediment core (73.36°N, 175.62°W) geographically closest to our core site (73.52°N, 168.94°W), has similar sedimentation rates of 0.06 and 0.08 cm yr −1 using 210 Pb and 137 Cs, respectively (Cooper and Grebmeier 2018). Accordingly, it appears that our core site is located far enough north that the more southerly observations of bioturbation on the Chukchi Shelf may not be so significant.
The presence of IP 25 throughout core ARA01B-03MUC (figure 3) provides proxy evidence for seasonal sea ice occurrence at the core site over recent centuries. The IP 25 concentration of the core top sample (1.9 μg g −1 TOC) was at the lower end of values (3.6±1.9 μg g −1 TOC) found previously in surface sediments from the Chukchi Plateau near to our Chukchi Shelf core site (Xiao et al 2015). The phytoplankton biomarker concentrations of the core top sample (14.9 and 22.1 μg g −1 TOC for epi-brassicasterol and dinosterol, respectively) were also of the same order of magnitude to those from the Chukchi Plateau (77±34 and 37±12 μg g −1 TOC for epibrassicasterol and dinosterol, respectively), even though the epi-brassicasterol concentrations were slightly lower (Xiao et al 2015). The HBI III concentrations were not reported in the previous analyses of Chukchi Sea sediments (Stoynova et al 2013, Xiao et al 2015, Polyak et al 2016, Stein et al 2017, although values in ARA01B-03MUC (1−4 ng g −1 ) were within the range of those (0.1−30 μg g −1 ) reported in surface sediments from the Barents Sea The general trends of the three PIP 25 records calculated using the c values based on the current sediment core data (i.e. using equation (4)) were dependent on the specific phytoplankton-derived biomarkers used ( figure 4). Thus, while both P III IP 25 and P D IP 25 records showed an increasing sea ice trend towards the core-top (ca. recent decades), the P B IP 25 values decreased over the same interval. However, in all three cases, only relatively few PIP 25 values were above the proposed threshold (0.75; Müller et al 2011) indicative of the extensive sea ice cover that characterizes the core site, with many substantially lower (i.e. below 0.5; figure 4). In contrast, P III IP 25 and P B IP 25 values were all greater than 0.75 when using c values of 0.63 and 0.023 obtained from the Barents Sea (Smik et al 2016) and pan-Arctic (Xiao et al 2015) databases, respectively, indicative of extensive sea ice cover (Müller et al 2011), a well-known feature of the core site. Use of the pan-Arctic c value of 0.11 for P D IP 25 , however, gave consistently low values (P D IP 25 ca. 0.2-0.5; figure 4), implying variable or low sea ice extent (Müller et al 2011) and, therefore, under-estimates of sea ice conditions at the core site. On the basis of these outcomes, the most reliable measures of spring sea ice concentration for the core site were derived from PIP 25 values using IP 25 and HBI III or epi-brassicasterol (but not dinosterol), and using fixed values for c. This conclusion is supported further through conversion of PIP 25 data to estimates of SpSIC using the calibrations of Smik et al (2016) and Xiao et al (2015). Thus, P III IP 25 -derived SpSIC estimates were all >78% in close agreement with the modelled values (>90%) and those from satellite data (https://nsidc.org/data/nsidc-0051) in the modern era (figure 6). The corresponding P B IP 25 -derived SpSIC was, however, slightly over-estimated (108% −127%), while the P D IP 25 -derived values were mainly well below 50% (figure 6). As such, the closest agreement between biomarker-derived and modelled SpSIC over recent centuries was obtained using P III IP 25 data based on IP 25 and HBI III using a c value obtained from Barents Sea surface sediments (i.e. c=0.63). Further refinement of the optimal absolute value of c for the Chukchi Sea might potentially be obtained through future and more accurate quantitative analysis of surface sediments from the study region.
Finally, we note that the only biomarker record that mimics the well-known pronounced reduction in late summer sea ice cover in the Chukchi Sea in recent decades (figure 4(D); Walsh et al 2016) was the P B IP 25 profile using a value of c based on the core data (i.e. Figure 4(B); c=0.17). Since some recently reported sediment trap data from the Chukchi Sea showed that the epi-brassicasterol flux was still relatively high in late summer, while the HBI III flux was reduced in summer compared to spring values (Bai et al 2019), it follows that certain sterols potentially integrate, to some degree, both spring and summer conditions. Accordingly, the major production of IP 25 in sea ice and HBI III in open waters along the sea ice edge during the spring season appear to provide the most reliable biomarker pair for estimating PIP 25 -derived SpSIC in the Chukchi Sea, while the additional use of P B IP 25 may potentially provide complementary insights into subsequent summer sea ice trends. (2015M1A5A1037243, PN19090), and the University of Plymouth.

Data availability statement
Any data that support the findings of this study are included within the article.