Alkenone Distributions and Hydrogen Isotope Ratios Show Changes in Haptophyte Species and Source Water in the Holocene Baltic Sea

The Baltic Sea, a dynamic, marginal marine basin, experienced a number of large changes in salinity during the Holocene as a result of fluctuations in global and local sea level related to melting of glacial ice sheets and subsequent isostatic rebound. These changes likely had pronounced effects on the species composition of haptophytes, a common phytoplankton group found in the Baltic Sea. This dynamic environment provides the ideal setting to study how species change impacts distribution and hydrogen isotope ratios of long‐chain alkenones (δ2HC37), haptophyte‐specific biomarkers. Here we analyzed the aforementioned parameters in Holocene sediments covering the contrasting hydrological phases of the Baltic Sea. Alkenone distributions changed with different Baltic Sea salinity phases, suggesting that species shifts coincide with salinity change. δ2HC37 values show two major shifts: one in the middle of the freshwater Ancylus Lake phase (10.6 to 7.7 ka) and a second at the transition from the brackish Littorina Sea phase (7.2 to 3 ka) into the fresher Modern Baltic (3 ka to the present). The first shift represents a significant enrichment of 50‰, which cannot be explained by salinity or species changes only. At this time, the isotopically depleted ice sheets had melted, and only the relatively enriched freshwater source remained. The second shift, coincident with a change in distribution, is likely caused by a change in species composition alone. These findings show that hydrogen isotope ratios of long‐chain alkenones, combined with their relative distribution, can be used to reconstruct changes in source water.


Introduction
Haptophyte algae are important primary producers and believed to be one of the most substantial producers of calcium carbonate globally (Laguna et al., 2001;Paasche, 2002) inhabiting both lacustrine and marine environments (Theroux et al., 2010). Haptophytes are extremely diverse, with over 300 documented species (Jordan & Chamberlain, 1997). Certain species of the order Isochrysidales are the only known producers of long-chain alkenones and polyunsaturated methyl and ethyl ketones with chain lengths between 35 and 42 carbon atoms (de Leeuw et al., 1980;Volkman et al., 1980;Theroux et al., 2010;Araie et al., 2018). These long-chain alkenones are often well preserved in sediments and provide highly useful proxies to understand past changes in global climate, in particular the U K′ 37 index, a proxy for sea surface temperature (Brassell et al., 1986;Müller et al., 1998;Prahl & Wakeham, 1987;Tierney & Tingley, 2018). Furthermore, the hydrogen isotope composition of long-chain alkenones (δ 2 H C37 ) has been shown to strongly relate to salinity and the δ 2 H values of water (e.g., Chivall et al., 2014a;M'Boule et al., 2014;Sachs et al., 2016;Schouten et al., 2006;Weiss et al., 2017), which has led to the use of δ 2 H C37 values to reconstruct paleosalinity in a number of different marine environments (Pahnke et al., 2007;van der Meer et al., 2007van der Meer et al., , 2008; Leduc et al., 2013;Kasper et al., 2014Kasper et al., , 2015Petrick et al., 2015;Simon et al., 2015;Weiss, de Bar, et al., 2019). However, other factors such as growth rate (Sachs & Kawka, 2015), growth phase (Chivall et al., 2014a;Wolhowe et al., 2009), and light intensity Weiss et al., 2017) are also known to affect the δ 2 H values of alkenones. While the latter factors exhibit significant controls on δ 2 H values in culture, environmental studies do not always reflect these effects and often salinity variations appear to be the strongest signals , and references therein).
One particular factor that is not well constrained yet is the impact of species composition on the δ 2 H values of alkenones. It has been noted that species fractionate hydrogen differently relative to each other, resulting in 2 H enrichment of alkenones produced by marginal marine species relative to δ 2 H values of alkenones from marine species (Schwab & Sachs, 2011;M'Boule et al., 2014;Nelson & Sachs, 2014;. Alkenone-producing haptophytes are separated into three distinct groups based on 18S ribosomal RNA (Theroux et al., 2010), and each group seems to prefer niches with a distinct salinity range. Group I haptophytes are presently uncultured but have been found in freshwater, lacustrine environments in the Northern Hemisphere (Theroux et al., 2010;Longo et al., 2016Longo et al., , 2018Plancq, Cavazzin, et al., 2018;Plancq, McColl, et al., 2018). Group II haptophytes are the most cosmopolitan, found in both oligohaline and hypersaline lakes Theroux et al., 2010), and cultured representatives include Isochrysis galbana and Ruttnera lamellosa. Marine species fall under Group III haptophytes, which include Emiliania huxleyi and Gephyrocapsa oceanica (Theroux et al., 2010). Only recently, analytical advances have allowed for better constraint on the contributions of each of these three groups to the pool of sedimentary alkenones. Groups can now be distinguished because of improved separation that allows for detection of tri-unsaturated alkenone positional isomers with different double bond positions at Δ 14,21,28 instead of the common Δ 7,14,21 configuration Longo et al., 2013Longo et al., , 2016. These so-called "tri-unsaturated isomers" have been detected alongside the traditional C 37:3 methyl, C 38:3 ethyl and methyl, and C 39:3 ethyl alkenones and allow for better distinction between haptophyte groups. Each haptophyte group appears to have different alkenone distributions: Group I is characterized by the presence of the novel tri-unsaturated isomers of C 37 -C 39 , in addition to the regular di-, tri-and tetra-unsaturated C 37 -C 39 alkenones; Group II does not synthesize the novel tri-unsaturated isomers and rarely produces C 38 methyl alkenones but instead synthesizes the standard suite of C 37 -C 39 and is also thought to produce the shorter chain C 35 and C 36 alkenones. The latter alkenones have been detected in marginal marine environments generally characterized by having low salinities compared to the open ocean (Rontani et al., 2001;Xu et al., 2001;Fujine et al., 2006;van Soelen et al., 2014;Warden et al., 2016;. These shorter chain alkenones have also appeared in an E. huxleyi culture after being kept in cultivation for a few years (Prahl et al., 2006). In contrast to the C 37 -C 40 alkenones, the C 35 and C 36 are thought to be synthesized from the C 37:2 and C 38:2 , respectively, via β-oxidation and produced by mutants of Group II species in the natural environment . Finally, Group III produce the standard suite of C 37 methyl and C 38 ethyl and methyl alkenones, with low abundance of the C 39 ethyl alkenones in some cases (Longo et al., 2013). All three groups are known to live in the Baltic Sea (Kaiser et al., 2019) providing the perfect opportunity to determine how distribution changes can enhance our understanding of shifts in alkenone producers when environments experience large fluctuations in salinity.
Besides different alkenone distributions, the hydrogen isotope ratios of long-chain alkenones also differ by group; more depleted δ 2 H C37 values are noted for Group III haptophytes compared to Group II when both are grown at the same salinity (Chivall et al., 2014a;M'Boule et al., 2014). Based on distributions and hydrogen isotope ratios of alkenones, mixed contributions of Groups I and II as well as Groups II and III have been noted in some environments (Schwab & Sachs, 2011;Nelson & Sachs, 2014;, but a mixing of all three groups based on distributions has only been observed for surface sediments of the Baltic Sea (Kaiser et al., 2019). In that case, a transect from the Kattegat into the Baltic Sea basin showed the presence of all three haptophyte groups reflected in changing alkenone distributions. This was confirmed by 18S rDNA preserved in surface sediments (Kaiser et al., 2019). Additionally, the relative percent contribution of the C 37:4 alkenone (%C 37:4 ) increases for Group II species (Chivall et al., 2014b) and has been proposed as a salinity proxy in culture and modern Baltic Sea sediments (Kaiser et al., 2017).
The findings of Kaiser et al. (2019) have potentially large consequences for interpreting the sedimentary record of the Baltic Sea with respect to δ 2 H C37 composition, as the Baltic Sea experienced varying salinities throughout the Holocene. The Baltic Sea began as the freshwater Baltic Ice Lake at around 13 ka BP and transitioned into a brackish basin known as the Yoldia Sea (from 11.6 to 10.6 ka BP; Moros et al., 2002) as deglaciation led to rising sea levels (Mörner, 1995). Continental uplift as a result of the isostatic adjustment to the melting of the Scandinavian Ice Sheet closed off the Baltic Sea from the North Sea, giving rise to the Ancylus Lake phase noted for having surface waters that were fresh (from 10.6 until approximately 7.7 ka BP; Moros et al., 2002). This freshwater phase ended when a large marine transgression brought North Sea water into the basin, shifting the Baltic Sea into the brackish Littorina Sea phase (Moros et al., 2002). The Littorina Sea lasted for a few thousand years, from 7.2 to around 3 ka BP with a gradual freshening into the Modern Baltic phase as a consequence of continued continental uplift, which was mainly noted in the Baltic proper (Andrén et al., 2000;Björck, 1995). Initial work on a Holocene sediment core has shown that hydrogen isotope ratios of long-chain alkenones from the Baltic Sea might reflect changes in species composition (Warden et al., 2016), but techniques used in that study did not allow for detection of tri-unsaturated isomers and alkenones of lower abundance; thus, no detailed reconstructions of species changes could be made.
Here we generated a comprehensive down-core record of δ 2 H C37 values and alkenone distributions from the Arkona Basin in the Baltic Sea analyzed on an RTX-200 60 m GC column, allowing for the separation and detection of individual alkenones and isomers. Detection of these isomers allowed for changes in species composition to be determined based on alkenone distributions which were related to the different Baltic Sea phase shifts. The Arkona Basin also provided an ideal setting for testing how well hydrogen isotope ratios of long-chain alkenones are recording salinity changes in marginal marine environments and how isotopic change relates to species shifts.

Location and Climate History
A 12 m piston core (64PE410-S7) was recovered from the Arkona Basin of the Baltic Sea (54°50.910′N, 13°2 1.412′E) with RV Pelagia in May 2016. The Arkona Basin is located in the western edge of the Baltic Sea next to the open connection with the North Sea via the Kattegat and Skagerrak (Figure 1). Situated at the western edge of the Baltic Sea, the Arkona Basin is perfectly suited for finding evidence of major marine transgressions and regressions occurring throughout the Baltic Sea climate history. The Arkona Basin has slightly higher salinities at present (~8 g kg −1 ) compared to the Baltic proper as a result of an open connection with the Kattegat. The phases in the development of the Baltic Sea during the Holocene are relatively well established (Björck, 1995), but the timing of these events in the Arkona Basin differs somewhat from timing reported for the Baltic proper, specifically the timing of the marine transgression that ended the Ancylus Lake phase (Warden et al., 2016). Around~7.7 ka, a large marine transgression occurred, bringing saline water into the basin and permanently restoring the connection to the North Sea (Moros et al., 2002;Warden et al., 2016).

Long-Chain Alkenones
Samples for long-chain alkenone analyses were taken every~100 cm in the top section of the core and then every 10 cm between 930 cm and the bottom (1,216 cm). In total, 41 sediment samples were freeze dried, and organics were extracted using a Dionex Accelerated Solvent Extractor. Following Accelerated Solvent Extractor, total lipid extracts were collected and separated into three fractions using small aluminum oxide columns. An apolar fraction was separated using hexane:dichloromethane 9:1 (v:v), a ketone fraction using hexane:dichloromethane 1:1 (v:v), and a polar fraction using dichloromethane: methanol 1:1 (v:v). Following separation over an aluminum oxide column, alkenones ended up in the 10.1029/2019GC008751 Geochemistry, Geophysics, Geosystems ketone fraction and were cleaned over a silver nitrate impregnated silica gel column using 100% dichloromethane to remove coeluting compounds, and alkenones were eluted using dichloromethane: ethyl acetate 1:1 (v:v). Alkenone concentrations were determined by using a gas chromatograph with a flame ionization detector equipped with an RTX-200 column (Restek, 60 m × 0.32 mm × 0.5 μm) and quantified based on a known concentration of a C 19 ketone internal standard (supporting information Table S1). Alkenones were identified using a gas chromatograph coupled to a mass spectrometer equipped with the same 60 m RTX-200 column. The GC temperature program was as follows: 70 to 250°C at 18°C/min, 250-320°C at 1.5°C/min, and then kept at 320°C for 25 min at a flow rate of 1.5 ml/min. Two indices have been proposed to trace the presence of Group I haptophytes : In these equations, "a" represents the traditional C 37:3 and C 38:3 alkenones, and "b" represents the novel triunsaturated isomers. These indices, along with the binary mixing model from Longo et al. (2018) %Group II ¼ RIK 37 −0:56 0:44 ; were used to determine the percentage of Group I and Group II haptophytes in samples containing triunsaturated isomers. When Group I species were present, temperature was calculated using R3b index: with % C37:4 being the concentration of C 37:4 as a percentage of total C 37 alkenones. Furthermore, salinity was estimated from the RIK 37 salinity calibration from Kaiser et al. (2019):

Hydrogen Isotope Analyses
Hydrogen isotope ratios of alkenones were measured using gas chromatography coupled to an isotope ratio mass spectrometer via a high temperature conversion reactor. The GC was also equipped with an RTX-200 60 m column using the temperature program described in . At the start of each day, the H 3 + factor was calculated and corrected for, which ranged from 2.5 to 2.6 ppm nA −1 , 3.1 to 3.3 ppm nA −1 , and 6.8 to 6.9 ppm nA −1 , with a day-to-day variation of <0.1 ppm nA −1 . Tuning of the ion source was responsible for the observed differences in H 3 + factor over the entire sample measuring period. Prior to measuring samples, a n-alkane mixture (Mix B, supplied by A. Schimmelmann, Indiana University), which covers a δ 2 H range from approximately −50‰ to −250‰, was measured. Sample analyses were only conducted when the average difference and standard deviation between online and offline values was less than 5‰.
Squalane and a C 30 n-alkane were coinjected with each sample run for further control on the stability of the system and ranged from −169 ± 3‰ and −78 ± 8‰, respectively, fitting well with their offline established values of −170 ± 4‰ (squalane) and −79 ± 5‰ (C 30 ). High standard deviation for the C 30 n-alkane is attributed to coeluting compounds in the sample fractions. Samples were run in duplicate, and error bars represent the standard errors of duplicate isotope ratio analyses. The RTX-200 60 m column allows for baseline separation of alkenones with different degrees of unsaturation and different chain length. The tri-unsaturated isomers were not baseline separated on the IRMS, but they were well resolved on the GC-FID. Hydrogen isotope ratios presented here are the result of manual peak integrations of all the alkenones for a given chain length (e.g., δ 2 H C37 = C 37:4 , C 37:3 , C 37:3 *, and C 37:2 ).

XRF Core Scanning
The inorganic geochemical composition of Core 64PE410-S7 was obtained with an Avaatech X-ray fluorescence (XRF) core scanner at the Royal Netherlands Institute for Sea Research. This XRF core scanner contains a 100 W rhodium X-ray tube and a Rayspec cubed SiriusSD silicon drift detector. Prior to the XRF-core-scan measurements, the sediment surfaces of the core sections were flattened and covered with a 4 μm thin Ultralene foil. For calcium (Ca) and titanium (Ti), the used settings were 10 kV, 0.4 mA, and 10 s, without a primary beam filter. For bromine (Br), the used settings were 30 kV, 0.15 mA, and 10 s, with a thin palladium primary beam filter. The core sections were measured continuously with a 1 cm resolution, excluding some gaps that were caused by gas expansion during core retrieval. The irradiated surface area of the X-Ray beam was set at 1 cm ("down core") by 1.2 cm ("cross core") using the slits of the scanner. All spectral data were processed using the bAxil spectrum analysis software to calculate the element intensities (in counts). The standard tablet SARM-4 was measured before and after each core section, showing that precision was high (relative standard deviation <10%) for all elements targeted here.

Chronological Framework
The age model for Core 64PE410-S7 is based on a combination of 14 C dating and correlation of Ca/Ti and Br records to previously dated, similar records close by (318310: 54°50.34′N; 13°32.03′E, and 318340: 54°5 5.77′N; 13°41.44′E) (Warden et al., 2016; Figure S1). Bromine is considered to be a good indicator of marine organic carbon content (Ziegler et al., 2008), which can thus be linked to the total organic carbon content obtained through loss on ignition in close-by Core 318310 (Warden et al., 2016). Five important transitions are present within sediments obtained from the Arkona basin (see Warden et al., 2016), which are clearly distinguishable in the carbonate (Ca/Ti) and organic matter (Br and loss on ignition) contents of the 10.1029/2019GC008751 Geochemistry, Geophysics, Geosystems sediment cores (numbered 1 to 5 in Figure S1): first, the transition between the Yoldia Sea and the Ancylus Lake phase at~10.6 ka; second, the Ancylus Lake regression pinpointed by a peak in organic matter at~10.2 ka; third, the start of the Ancylus Lake/Littorina Sea transitional phase defined by the increase in organic matter contents at~7.7 ka; fourth, the transgression maximum indicated by the carbonate maximum at 7.2 ka; and fifth, the occurrence of a peak in organic matter during the Medieval Climate Anomaly at 0.9 ka. The first is not present in the core studied here; however, the dip in Ca/Ti content at~10.5 ka in Core 318340 (Warden et al., 2016) can be linked to that in Core 64PE410-S7 (marked orange in Figure  S1). Additionally, we obtained three 14 C dates on mollusk shells to improve age control for the interval younger than 7 ka ( Figure S1). Shells were radiocarbon dated by Beta Analytic, corrected for the local reservoir effects (reservoir age = 376 years; Lougheed et al., 2013), and calibrated using CALIB (Stuiver et al., 2018). The final age model for all depths was created using the Bacon software package in R (Blaauw & Christen, 2011). In Bacon, the error estimates for 14 C ages were also applied to the XRF tie points (as these are based on correlation to the nearby 14 C dated cores), and Student's t test values were set to 33 and 34 to allow for more narrow error estimates as suggested in the Bacon manual. The resulting age model suggested that sedimentation rates varied between 55 and 250 cm/Kyr over the studied interval. The large variation in sedimentation rates for Core 64PE410-S7 and the nearby cores from Warden et al. (2016) is likely related to the shallow water depth in the Arkona Basin (~45 m). The direction of sediment delivery must have varied over time, with a more northeastward source of sedimentation prior to 7.5 Kyr and a shift to a westward source around 7.5 Kyr. An increase in sedimentation rates around 7.5 Kyr is likely instigated by the marine transgression brought from the North Sea and thus an extra indication of the start of the Littorina Sea phase.

Alkenone Distributions
Sediments from the Arkona Basin covering the last 10.6 Kyr were analyzed for long-chain alkenones. Nineteen different alkenones were detected, with chain lengths ranging from 35 to 40 carbon atoms (Table S1). Tri-unsaturated isomers of C 37:3 methyl and C 38:3 ethyl alkenones were detected in addition to the conventional C 37:3 methyl and C 38:3 ethyl alkenones in some samples. These tri-unsaturated isomers have previously only been reported for lacustrine sediments (Longo et al., 2013Dillon et al., 2016;Plancq, Cavazzin, et al., 2018;Plancq, McColl, et al., 2018) and surface sediments in the Baltic Sea basin (Kaiser et al., 2019). Changes in alkenone distribution patterns align with Baltic Sea phase boundaries ( Figure 2). Alkenone distributions for the Yoldia Sea phase are typical for Marine Group III, that is, lacking tri-unsaturated isomers as well as C 39 and C 40 alkenones (cf. Longo et al., 2013). The first emergence of triunsaturated isomers and C 39 ethyl and methyl alkenones occurs at the Yoldia Sea to Ancylus Lake boundary (Figure 2), signifying the first occurrence of Group I species in this record. The Ancylus Lake distribution is characterized by the presence of tri-unsaturated isomers and C 39 ethyl and methyl and C 40 ethyl alkenones (Figure 2). This pattern points to a mixing of Groups I and II haptophytes throughout the Ancylus Lake phase, and distributions are stable across the entire phase. Two samples in the early Ancylus Lake phase show a more marine distribution pattern although still containing the tri-unsaturated isomers indicative of freshwater species. During the Yoldia Sea phase, the southern Baltic experienced short transgressions related to higher sea levels and less continental uplift (Björck, 1995). It appears that the increase in sea level rise overshadowing isostatic rebound also may have continued into the early Ancylus Lake phase, causing some input of saline waters at the onset of this phase.
To constrain the species mixing of Groups I and II haptophytes, the percentage of Groups I and II producers was determined using equation (3), which is based on alkenone distributions in Northern Hemisphere lacustrine sediments and suspended particulate matter (Longo et al., 2018). Group I is indeed dominant throughout the Ancylus Lake phase (57% to 75%), with Group II only contributing substantially at the beginning and the end of the phase (60% to 75%). This is supported by the RIK 38E index (equation (2) and Table 1), another indicator of Groups II and I haptophytes , as it remains below a value of 0.75 which is characteristic for Group I haptophytes.
The transition between the Ancylus Lake and the brackish Littorina Sea phase lasts for~0.5 Kyr and is characterized by a Group III signature, with only conventional C 37 methyl and C 38 ethyl alkenones (Figure 2). However, in the latter half of the transition, the C 35:2 and C 36:2 alkenones appear for the first time 10.1029/2019GC008751 Geochemistry, Geophysics, Geosystems ( Figure 2). The Littorina Sea phase is distinguished from other phases by the presence of the alkenones of shorter chain-length as well as the absence of tri-unsaturated isomers and C 39 and C 40 alkenones. The presence of the shorter chain alkenones is an indicator for the presence of Group II haptophytes . The final shift in alkenone distribution corresponds to the change from the Littorina Sea into the Modern Baltic phase and represents a mixed Groups II and III signal indicated by the presence of the C 35:2 , C 36:2 , and C 39 ethyl alkenones (Figure 2).

Temperature and Salinity Estimates 3.2.1. Temperature
The U K 37 and U K′ 37 indices are well-established sea surface temperature proxies (Brassell et al., 1986;Müller et al., 1998;Prahl & Wakeham, 1987;Rosell-Melé et al., 1995). The latter index excludes the tetraunsaturated C 37 alkenone in the equation in response to low concentrations of C 37:4 in marine sediments. Application of the temperature calibration for the U K 37 using Atlantic surface sediments (Rosell-Melé et al., 1995) gives negative temperature estimates for this record, while the culture calibration from Prahl and Wakeham (1987) gives a temperature range between 0.3 and 2.3°C, much lower than present-day spring mean SST of 8°C . These results are not surprising because the calibrations are based on Group III alkenone distributions characterized by low abundance of the C 37:4 . Groups I and II are known to produce higher concentrations of C 37:4 (Chivall et al., 2014b;Schulz et al., 2000) which would lead to different calibrations. When excluding the tetra-unsaturated C 37 alkenones from the equation (U K′ 37 index) and using the global temperature calibration for the U K′ 37 index (Müller et al., 1998), temperature estimates range from 2.2 to 8.5°C across the record (Figure 3a and Table 1). Haptophytes in the present-day Baltic Sea are known to bloom during the months of April to June (Blanz et al., 2005;Hällfors, 2004), so it is reasonable to expect that alkenones should be reflecting spring temperature values. Modern spring mean temperature is 8°C , and the most recent sample suggests a temperature of 8.5°C using the global U K ′ 37 calibration, which is in good agreement. Therefore, the U K′ 37 index seems to be the more reliable of the two indices to reconstruct temperatures for this record.
Nevertheless, the presence of different Groups I and II alkenone producers is known to complicate the use of these well-established U K 37 and U K′ 37 indices in lacustrine and brackish environments . New indices have been proposed based on Groups I and II distributions that include novel tri-unsaturated isomers in the temperature ratios (Longo et al., , 2018. Since Group I haptophytes are the most abundant in the Ancylus Lake phase (10.6 to 7.7 Kyr), the R3b index seems more applicable for temperature reconstructions during this time period. Using the R3b temperature calibration from Longo et al. (2018), temperature estimates range from 4.0 to 13.5°C throughout the Ancylus Lake phase, capturing more variation than estimates using the Müller et al. (1998) global calibration for the U K′ 37 index, which range from 4.7 to 7.7°C (Figure 3a and Table 1).

Salinity
Salinities were reconstructed using %C 37:4 for the entire record and the RIK 37 equation for the Ancylus Lake phase since it was the only phase to contain C 37 tri-unsaturated isomers (equations (6) and (7), respectively; Kaiser et al., 2017Kaiser et al., , 2019. Values ranged from 4.1 to 8.7 g kg −1 (%C 37:4 ) and 2.0 to 5.2 g kg −1 (RIK 37 ), with the %C 37:4 yielding slightly higher values (Table 1). The reconstructed salinities based on %C 37:4 started to rise at the end of the Ancylus Lake phase leading into the Littorina Sea phase, and both %C 37:4 and RIK 37 calibrations showed an increase in salinities around 7.7 ka during the marine transgression (Figure 3b). In general, %C 37:4 suggests a salinity value closest to the modern-day spring mean of 8.1 g kg −1  and hence may be the most appropriate for reconstructing salinity of latter part of this record. However, due to the presence of tri-unsaturated isomers during the Ancylus Lake phase, the RIK 37 salinity calibration might provide better estimates during this phase, since it includes the tri-unsaturated isomers. Indeed, using the RIK 37 calibration yielded an average salinity of 3 ± 0.4 g kg −1 , which seems more logical than an average of 6 ± 0.5 g kg −1 (obtained from the %C 37:4 calibration) for this relatively fresh phase in the Baltic Sea history (Jensen et al., 1999). The Ancylus Lake phase in the Arkona Basin was dominated by lacustrine macrofossils, diatoms, and sedimentary deposits (Jensen et al., 1999), reinforcing a low salinity to even freshwater environment. Furthermore, the %C 37:4 shows more fluctuations in salinity, whereas the RIK 37 suggests a more steady decrease in salinity, in line with a gradual shift from a brackish into a lacustrine, freshwater environment during the Early Ancylus Lake phase. In general, when tri-unsaturated isomers are present, indices accounting for these isomers seem more appropriate for reconstructing temperature and salinity. The above observations confirm the necessity of obtaining optimal separation of individual alkenones to enable detection of these informative isomers.
Salinities reconstructed using the %C 37:4 equation from Kaiser et al. (2017) align well with modern-day salinity values for the Arkona Basin. That being said, reconstructed values are relatively stable at around 8 g kg −1 across the Littorina Sea phase, a period of higher salinity than either the Ancylus Lake or Modern 10.1029/2019GC008751

Geochemistry, Geophysics, Geosystems
Baltic phases as the result of increased North Sea inflow (Emeis et al., 2003). This strong marine influence is not obvious in the %C 37:4 based record. Haptophyte species produce different relative amounts of the C 37:4 alkenone, and these differences are even noted between species from the same group-for example, I. galbana was shown to have a relative abundance of C 37:4 around 5% to 10% and R. lamellosa around 35% to 40% (Chivall et al., 2014b). The changing haptophyte community composition affects the relative abundance of these alkenones potentially leading to such stable salinity estimates across the higher salinity Littorina Sea phase into the fresher Modern Baltic phase.  (Longo et al., 2018). The R3b index includes the tri-unsaturated C 37 and C 38 isomers and was therefore only applied to samples containing said isomers. (b) Reconstructed salinities using the Baltic Sea %C 37:4 calibration (Kaiser et al., 2017) and the RIK 37 calibration (Kaiser et al., 2019), both from Baltic Sea surface sediments. Similar to the R3b temperature index, the RIK 37 index includes the tri-unsaturated C 37 isomer, thus was only applied to samples containing the tri-unsaturated C 37 isomer. Present-day spring mean temperature (8°C) and salinity (8.1 g kg −1 ) from the World Ocean Atlas 2013 Zweng et al., 2013) are denoted by the dotted line, and abbreviations for corresponding Baltic Sea phases are at the top.

Hydrogen Isotope Ratios
Hydrogen isotope ratios of integrated C 37 and C 38 alkenones, and for the C 36:2 alkenones when present, were measured (Table 2, Figure 4). Isotope ratios for all C 37 and C 38 were integrated. δ 2 H values of C 37 and C 38 showed two major shifts: a substantial enrichment of~50‰ starting at approximately 10 ka (mid-Ancylus Lake phase) and a depletion of~25‰ at around 3 ka (Littorina Sea to Modern Baltic phase boundary). The C 36:2 first occurred at the start of the Littorina Sea phase, and δ 2 H values were similar to those of the C 37 alkenones. The δ 2 H values of the C 36:2 became enriched across the transition from the Littorina Sea into the Modern Baltic phase by around 15‰, deviating from the C 37 and C 38 alkenones which became increasingly depleted by as much as 40‰ in the most modern sample at 1.37 ka (Figure 4). It has been hypothesized that the C 36:2 ethyl alkenone is synthesized from the C 38:2 ethyl alkenones by β-oxidation . During the Littorina Sea phase, there was relative isotopic enrichment of the C 38 compared to the C 36 , but this reversed in the Modern Baltic phase. It might be the case that when C 36:2 is synthesized from C 38:2 , the loss of four hydrogens via β-oxidation causes the C 36 to become relatively 2 H depleted. In the more modern phase, there might be a greater amount of C 38 produced by a different species (Group III), causing relative depletion of the C 38 pool which overwrites this degradation effect.
The two isotopic shifts observed for the C 37 and C 38 alkenones (at approximately 3 and 10 ka) reveal two separate effects that do not coincide with large shifts in reconstructed salinity values. Reconstructed salinities were stable from the Littorina Sea into the Modern Baltic phase at the Arkona Basin. The shift toward more negative values at 3 ka aligns with the phase boundary between the Littorina Sea and the Modern Baltic phase and coincides with a shift in the alkenone producing population from Group II dominance to a mixed Groups II and III signal noted by changes in distributions ( Figure 2). As explained above, the C 37 and C 38 alkenones differ from the C 36:2 by up to 40‰ in the most modern sample. Our isotope data from the Modern Baltic phase cannot confirm the synthesis of C 36:2 from C 38:2 via β-oxidation because values do not align with ratios proposed by Zheng, Dillon, et al. (2016; Figure S2) and instead suggest mixing of different haptophyte groups contributing alkenones to the sedimentary alkenone pool. Each group has a characteristic isotopic range measured from culture experiments (Schouten et al., 2006;Chivall et al., 2014aChivall et al., , 2014bM'Boule et al., 2014;Sachs et al., 2016;Weiss et al., 2017). Group II spans a range of −189‰ to −97‰ over a salinity gradient from 10 to 35, and Group III values span from −263‰ to −173‰ over a salinity gradient from 24 to 42. As salinity becomes lower, more depleted isotope ratios for C 37 and C 38 alkenones are expected due to a lowering of the δ 2 H of the water as well as increased isotopic fractionation (Schouten et al., 2006;Chivall et al., 2014aChivall et al., , 2014bM'Boule et al., 2014;Sachs et al., 2016;Weiss et al., 2017). Compiled Group III culture data (M'Boule et al., 2014;Sachs et al., 2016;Schouten et al., 2006;Weiss et al., 2017) together with the sea water δ 2 H values suggest that salinities of 7 to 8 g kg −1 result in isotope ratios of around −218‰, close to measured values for this interval. Group II species, on the other hand, tend to have more enriched isotope ratios at these salinities (around −186‰; based on results from Chivall et al., 2014a, 2014b, and M'Boule et al., 2014. The presence of the C 36:2 implies Group II contribution to the sedimentary alkenone pool. While the C 36:2 isotope ratios have not been reported from culture experiments, the similarity between C 36:2 with both the C 37 and C 38 during the Littorina Sea phase suggests that when produced by Group II species, the three do not differ by more than~15‰. Collectively, these observations imply that the isotope shift at 3 ka is caused by a change in species composition.
The large isotope enrichment of~50‰ starting at 10 ka occurs in the middle of the Ancylus lake phase and does not coincide with a change in alkenone distributions, and hence, the alkenone distributionbased salinity estimates show no large fluctuations (Figures 2 and 3). Group II culture regressions suggest δ 2 H values should be around −210‰ for a salinity of zero (Chivall et al., 2014a(Chivall et al., , 2014bM'Boule et al., 2014). This value is more enriched than values measured for the early Ancylus Lake phase; thus, an alternative mechanism must be invoked to explain the depleted values measured for this freshwater phase in Baltic Sea history. Source water isotope ratios also exhibit a strong control on δ 2 H values (Englebrecht & Sachs, 2005;Paul, 2002). During this time period, the Scandinavian Ice Sheet was rapidly decreasing in size, as a result of an increase in high latitude summer insolation and changes in ablation rates (Cuzzone et al., 2016). The melting Scandinavian Ice Sheet acted as an additional water source for the Baltic Sea. The Scandinavian Ice Sheet formed during the last glacial and was significantly depleted in hydrogen isotope ratios with a δ 2 H value of around −310‰ suggested by modeling data (de Boer et al.,

10.1029/2019GC008751
Geochemistry, Geophysics, Geosystems 2014). Holocene Baltic Sea precipitation has a value of −68.1 ± 0.7‰ (Fröhlich et al., 1988), significantly more enriched compared to the melting Scandinavian Ice Sheet in the Early Holocene. The difference between these two sources is large enough that a relatively modest, yet substantial, contribution of ice melt-derived source water would be reflected in δ 2 H C37 values, resulting in significant isotopic depletion (Figure 4). The steep decline in δ 2 H C37 values seems to indicate a rather rapid phase of melting of the Scandinavian Ice Sheet, which disappeared completely at about 9.5 ka ( Figure 5). The largest part of the Scandinavian ice sheet melted before this time. This major melting phase seems to be recorded in the hydrogen isotopic composition of only the C 38 alkenones. Prior to the 50‰ isotope shift, the water in the Arkona Basin was derived from a combination of rain water and melt water from the ice sheet, which explains the relatively depleted values during this period. Only at the end of the melting phase does the relative contribution of melt water drop significantly, reflected as an enrichment in the hydrogen isotopic composition of the Arkona Basin water and in turn recorded by alkenone hydrogen isotopes.
To constrain this, the δ 2 H values of source water can be reconstructed using calibration data from haptophyte cultures and a simple end-member mixing model with meltwater and precipitation together contributing to the Baltic Sea water. The offset between δ 2 H values of water and the δ 2 H values of alkenones is not constant and is affected by biological fractionation (fractionation factor α) occurring during alkenone synthesis (  2017). The fractionation factor α from two Group II species I. galbana and R. lamellosa at a salinity of around 3 g kg −1 (value obtained from the alkenone estimates; Figure 3) is 0.841 and 0.832, respectively (Chivall et al., 2014a(Chivall et al., , 2014bM'Boule et al., 2014). It should be noted that the presence of Group I alkenone distributions indicates freshwater conditions, thus salinities between 2 and 5 may not be accurate, but these estimates will be used for roughly determining the δ 2 H values of source water. Assuming this fractionation value and using the measured δ 2 H values of alkenones across the~50‰ shift, we calculated that the δ 2 H values of the water should change from −99‰ to −40‰ using I. galbana α values and −89‰ to −30‰ using R. lamellosa α values. Following this, we could estimate the percent contribution of both ice melt-derived source water and precipitation using −310‰ (ice) and −68‰ (precipitation) as end-members. As expected, it showed that there was a decrease in the percent contribution from ice melt-derived source water starting at approximately 10 ka ( Figure 5). However, reconstructed contributions exceed 100% (precipitation) or go into negative values (ice) at some points, suggesting that a freshwater source with a heavier δ 2 H value than that of modern precipitation must have been present. Indeed, a rise in water level during this time period coincides with the creation of the Dana River, suggesting another freshwater source between 9.5 and 9.0 ka (Björck, 1995). Thus, a more 2 H-enriched freshwater end-member might not be unrealistic. Still, these values cannot fully explain the observed range in δ 2 H H2O values. This would either require a heavier precipitation source, which is not likely in view of the reconstructed temperatures, or admixing of some sea water. The sea water either somehow entered the basin (although it was closed off from the North Sea during this time), or some sea water . Hydrogen isotopes ratios of C 36:2 and integrated C 37 and C 38 alkenones over the last 10.6 Kyr in the Arkona Basin. Two major isotope shifts occur: The C 37 and C 38 alkenones become significantly enriched in the middle of the Ancylus Lake phase, and the C 36:2 deviates from the C 37 and C 38 at the Littorina Sea to Modern Baltic phase shift. Error bars represent reproducibility of duplicate hydrogen isotope analyses.

Figure 5.
Reconstructed hydrogen isotope ratios of water plotted alongside the corresponding ice melt contribution during the Ancylus Lake phase. Ice melt contribution was calculated using a simple two end-member mixing model, with modeled Scandinavian Ice Sheet values and modern precipitation ratios as the two end-members. We accounted for fractionation values using results from cultures of two different Group II species Isochrysis galbana and Ruttnera lamellosa. Percent contribution of ice becomes negative, suggesting that the Scandinavian Ice Sheet no longer contributed water to the Baltic Sea after 9.5 ka and was replaced by a much lighter source.

10.1029/2019GC008751
Geochemistry, Geophysics, Geosystems remained from an earlier, more saline phase in the basin (i.e., Yoldia Sea). Regardless of the δ 2 H values of both (or three when including sea water) end-members, it can be estimated that the freshwater input directly from the melting Scandinavian Ice Sheet ceased to contribute to the freshwater budget around 9.5 ka. The R3b index indicated a spike in temperature coincident with the shift in source water, suggesting that temperatures might have locally increased in response to the ice sheet being fully melted, which directly affected the local albedo.

Conclusions
A large number of alkenone isomers were detected in Holocene age sediments from the Arkona Basin using improved gas chromatography separation, allowing for better determination of the haptophyte population in the Holocene Baltic Sea. Tri-unsaturated isomers, indicative of Group I alkenone producers, were prominent during the Ancylus Lake freshwater phase from 10.6 to 7.7 ka. For this reason, temperature and salinity calibrations based on indices including tri-unsaturated isomers provide more reliable estimates during this time period, circumventing issues associated with traditional alkenone-based temperature indices (i.e., anomalously high or low temperatures). Changes in distribution show mixing between Groups III and II in the Yoldia Sea, Littorina Sea, and Modern Baltic phases.
Hydrogen isotope ratios of long-chain alkenones inform about changes in salinity and species composition but also provided insight into shifts in the dominant freshwater sources at the Arkona Basin. Results presented here show that analysis of well-separated alkenone distributions, in conjunction with hydrogen isotope ratios, provide a more in-depth understanding of species, salinity, and hydrological changes in paleoenvironmental settings, especially in marginal marine environments.