A 250 ka oxygen isotope record from diatoms at Lake El’gygytgyn, far east Russian Arctic

In 2003 sediment core Lz1024 was drilled at Lake El’gygytgyn, far east Russian Arctic, in an area of the Northern Hemisphere which has not been glaciated for the last 3.6 Ma. Biogenic silica was used for analysing the oxygen isotope composition ( δOdiatom) in the upper 13 m long section dating back about 250 ka with samples dominated by one taxa in the< 10 μm fraction ( Cyclotella ocellata). Downcore variations inδ18O values show that glacial-interglacial cycles are present throughout the core and δOdiatom-values are mainly controlled byδOprecipitation. Changes reflect the Holocene Thermal Maximum, the Last Glacial Maximum (LGM) and the interglacial periods corresponding to MIS 5.5 and MIS 7 with a peak-to-peak amplitude between LGM and MIS 5.5 of118O = 5.3 ‰. This corresponds to a mean annual air temperature difference of about 9 C. Our record is the first continuousδOdiatom record from an Arctic lake sediment core directly responding to precipitation and dating back more than 250 ka and correlates well with the stacked marineδ18O LR04 (r = 0.58) andδD EPICA DomeC record ( r = 0.69). With δ18O results indicating strong links to both marine and ice-core records, records from Lake El’gygytgyn can be used to further investigate the sensitivity of the Arctic climate to both past and future global climatic changes.


The Arctic as an important area for paleoclimate reconstruction
The Arctic experienced the most rapid warming in the last 150 yr (Serreze et al., 2000;Kaufman et al., 2009;Bekryaev et al., 2010) with a continuation of this temperature trend above the global average projected for the 21st century (Serreze et al., 2000;Anisimov et al., 2007).Especially in climate-sensitive areas such as the Arctic (Serreze and Francis, 2006a, b;Miller et al., 2010;Serreze and Barry, 2011), longer, continuous terrestrial records for paleo-climate reconstructions are rare (Brigham-Grette et al., 2007;CCSP, 2009)  the Last Glacial Maximum (LGM) (Dyke and Prest, 1987;Lehman et al., 1991;Thiede, 2004) which destroyed sedimentary sequences in lakes especially in the western part of Siberia (Svendsen et al., 1999;Hubberten et al., 2004).The Beringia land bridge which connected Alaska (eastern Beringia) and northeast Russia (western Beringia) was exposed during the glacials (Hopkins, 1967(Hopkins, , 1982;;Brigham-Grette, 2001) linked to an eustatic sea-level drop due to emerging ice caps.However, due to its dry climate the Eastern Arctic experienced only valley glaciation (Glushkova, 2001) and the precipitation-controlled eastern margin of the Eurasian Ice Sheet never extended to Eastern Siberia (Svendsen et al., 2004;Melles et al., 2007).Lake sediments are known as excellent archives for the reconstruction of past climate and environmental changes which may provide insights as to how regions may respond to future change.The longest lacustrine palaeoenvironmental record in Siberia exists from Lake Baikal reaching back to 35 Ma before present (Tapponnier and Molnar, 1979).However, Lake Baikal is located south of the Arctic circle and the δ 18 O values of precipitation are influenced by south and southeast cyclones in July and August (Bezrukova et al., 2008).In an attempt to obtain long-term Quaternary lake records from the terrestrial, far east Russian Arctic drilling campaigns were conducted at Lake El'gygytgyn in an area of the Northern Hemisphere which has not been glaciated during at least the last five glacial/interglacial cycles and has the potential to be an archive of continuous and undisturbed sedimentation (Brigham-Grette et al., 2007;Melles et al., 2007).

The importance of δ 18 O (diatom) records
Oxygen isotope records from ice-cores (Greenland-Ice-Core-Project members, 1993; Grootes et al., 1993;Petit et al., 1999;EPICA members, 2004EPICA members, , 2006) ) or carbonate organisms (Zachos et al., 2001;Lisiecki and Raymo, 2005) are amongst the most important proxies for reconstructing paleoenvironmental conditions and paleoclimate.In regions of high latitude, where ice-records are not available or where carbonates are absent, as is the case for Beringia and Lake El'gygytgyn the oxygen isotope composition Figures from lacustrine biogenic silica, i.e. diatoms, often represents the best isotope proxy for reconstructing past climate changes (Leng and Marshall, 2004).By taking potential controls and effects into account, the analysis of oxygen isotopes in lacustrine biogenic silica provides one of the rare opportunities to gain a direct and continuous signal from paleo-precipitation beyond the LGM in the Eastern part of the Arctic.This can then be compared with other regional and global δ 18 O records to better understand the links between West Beringia and the wider global climate system.The preparation of the samples (e.g.Juillet-Leclerc and Labeyrie, 1987;Shemesh et al., 1995;Morley et al., 2004), contamination assessment (Lamb et al., 2007;Brewer et al., 2008;Mackay et al., 2011;Swann and Patwardhan, 2011) as well as the oxygen isotope analysis for diatoms are often challenging and time-consuming (Leng and Barker, 2006;Swann and Leng, 2009).Recently, improved protocols have been developed for the purification of small fractions and a more reliable contamination correction (Chapligin et al., 2012).In addition, a faster dehydration and analytical technique has also been introduced (Chapligin et al., 2010).However, when analysing 250 ka old material, diagenetic changes Introduction

Conclusions References
Tables Figures

Back Close
Full cold-monomictic and ultra-oligotrophic (Nolan and Brigham-Grette, 2007).50 streams drain into the lake, the Enmyvaam River is the only outflow in the southeast of the Lake and the total catchment area is 293 km 2 (Nolan and Brigham-Grette, 2007).
Several cores were drilled in the last decade (Fig. 1b) including the 16.5 m core Lz1024 drilled in 2003 (67 • 30.13 N; 172 • 06.46 E) examined in this study.A gravity corer (UWITEC Corp., Austria) was used for the water-sediment transition and changed to a percussion piston corer further down the sediment (Melles et al., 2005).In total, 96 samples spread over the first 13 m were chosen for analysis, covering the last three glacial-interglacial cycles.From 2 cm sections sub-samples with average dry weight of 1.9 g were taken.As the first test series of 10 samples distributed equidistantly downcore did not provide enough material after purification, two subsequent samples were merged and the average age and depth determined.The first age model for core Lz1024 was proposed by Juschus et al. ( 2007) using isochronous sediment layers and susceptibility variations to correlate core Lz1024 to the previously dated 13.5 m core PG1351 with the final age model and revised after tuning the magnetic susceptibility to the insolation at 70 • N (Nowaczyk et al., 2002(Nowaczyk et al., , 2007(Nowaczyk et al., , 2012)).The core base of Lz1024 was dated to a maximum age of 350 ka (Frank et al., 2012;Nowaczyk et al., 2012).However, the lower sequence is disturbed (V.Wennrich, personal communication, 2012) which is why the focus of this study is on the 250 ka record.

Conclusions References
Tables Figures

Back Close
Full material for oxygen isotope analyses in most core sections.Consequently, in an attempt to obtain an isotope record dominated by a single taxa, the <10 µm fraction was purified according to an 8-step method developed by Chapligin et al. (2012) which is similar to the 4-step method (Morley et al., 2004).After freeze-drying the samples were treated with H 2 O 2 (32 %) and HCl (10 %) to remove organic matter and carbonates and wet sieved into <10 µm and >10 µm fractions.Four multiple heavy liquid separation (HLS) steps with varying densities (from 2.25 to 2.15 g cm −3 ) were then applied using a sodium polytungstate (SPT) solution before being exposed to a mixture of HClO 4 (65 %) and HNO 3 (65 %) for removing any remaining micro-organics.

Isotope analysis
Oxygen isotope analysis was performed with a PDZ Europa 2020 mass spectrometer (MS-2020; now supplied by Sercon Ltd., UK) following the method of Chapligin et al. (2010).Prior to isotope analyses, hydrous groups (H 2 O, -OH) were removed from an initial sample weight of 1.5 to 2 mg using ramp degassing under inert gas flow conditions (inert Gas Flow Dehydration = iGFD technique) in a horizontal ceramic tube furnace flushed with helium (Chapligin et al., 2010).The dehydrated sample was then reacted by laser fluorination under BrF 5 atmosphere to liberate O 2 based on Sharp (1990).The oxygen was separated of by-products and directed to the mass spectrometer for isotope analysis.Every fifth sample was a biogenic working standard to verify the quality of the analyses.For this, the biogenic standard BFC (Leng et al., 2001;Leng and Sloane, 2008;Chapligin et al., 2010) calibrated within an inter-laboratory comparison was used (Chapligin et al., 2011).With a δ 18 O values of +29.0 ± 0.3 ‰ BFC is the closest diatom working standard to the Lake El'gygytgyn samples (δ 18 O values range between +19 and +26 ‰) available.Additionally, the international quartz standard NBS28 (δ 18 O = +9.57‰; Hut, 1987;Gr öning et al., 2007) was measured.Long-term standard measurements of δ 18 O values in silica show standard deviations of 1σ = 0.15 % for quartz and 1 σ ≤ 0.25 % for biogenic silica working standards such as BFC (Chapligin et al., 2010).Replicate isotope analyses were performed Introduction

Conclusions References
Tables Figures

Back Close
Full for 49 samples from the core Lz1024 which show a mean standard deviation of 0.13 ‰ (1 σ, ranging between 0.00-0.41‰)
where the percentage of purity was 100 % subtracted by the percentage of remaining contamination (% cont.).Determining % cont.by counting contamination particles/diatoms under a light microscope underestimates this parameter in the <10 µm fraction due to the presence of small, diffuse clay and silt particles (Chapligin et al., 2012).Thus, % cont.was calculated by dividing the sample percentage of Al 2 O 3 by the Al 2 O 3 percentage from a 100 % contamination end member according to Brewer et al. (2008), Swann and Leng (2008) and Mackay et al. (2011).Chapligin et al. (2012) tested different contamination assessment techniques to Lake El'gygytgyn sub-surface samples and found no significant difference between this method and calculating

Species assessment
The diatom assemblage in the <10 µm fraction of 40 purified samples equally distributed throughout the core was assessed by counting under light microscope (LM) (Chapligin et al., 2012, based on Morley et al., 2004, and Swann et al., 2008).Five random fields of view were selected along a transect in the middle of the slide and at least 200 valves as well as fragments were counted whenever possible.The biovolume was calculated by defining average area-sizes (Chapligin et al., 2012) for each species as it better reflects the actual proportion of each taxa when analysing for oxygen isotopes than using the number of valves.

FTIR analysis for studying diagenetic effects
The question of potential diagenetic changes and their impact on the oxygen isotope composition in the amorphous structure was addressed by performing diffuse reflectance Fourier Transform Infrared Spectroscopy (FTIRS) at Ume å University, Sweden.This method determines individual compounds in biogenic silica by recording the different characteristic vibration patterns (absorption peaks) of the chemical bonds present in the structure (siloxane bonds: Si-O-Si; silanol bonds: Si-OH; H 2 O) over a certain wavelength window (Schmidt et al., 2001;Gendron-Badou et al., 2003;Moschen et al., 2006;Leng et al., 2009;Swann and Patwardhan, 2011).An FTIR spectrometer (Bruker IFS 66v/S) equipped with a diffuse reflectance accessory (Harrick Inc.) was used with 11 mg of sample material mixed with 500 mg KBr for the analysis under vacuum (4 mbar) conditions (Vogel et al., 2008;Ros én et al., 2010).16 purified diatom samples from different core depths (first sample at 0.09 m, last sample at 12.35 m) were selected and scanned (64 scans per sample) with measurements taken every 4 cm −1 for the spectral region between 3750 and 400 cm −1 .All wavelength absorption peaks >1500 cm −1 were removed as these peaks are not related to biogenic silica or linked to molecular vibrations of hydroxide (-OH stretching).
Focusing on wavelengths <1500 cm −1 allows only the silica bonds from the diatom Introduction

Conclusions References
Tables Figures

Back Close
Full frustules to be examined.Common peak centers for amorphous silica are 450 cm −1 , 800 cm −1 and 1100 cm −1 for various modes of siloxane bonds and 945 cm −1 for silanol bonds (Gendron-Badou et al., 2003;Swann and Patwardhan, 2011).In our study the data was baseline-corrected and normalised, so that the area under each curve adds up to 100 %, with individual peaks separated by applying a Gaussian peak fitting (Fig. 4a) (Swann and Patwardhan, 2011) using the software Fytik, Version 0.9.3 (http://fityk.nieto.pl/).Peak types were then assigned (Si-OH and Si-O-Si) and the percentages calculated from the respective areas.However, when using FTIR each Si-O-Si band has its own "sensitivity", so absolute unweighted calculations are not suitable for comparing individual samples.Therefore, the relative change in Si-OH bond determined by a single peak, was calculated.The highest peak area (top sample at 0.09 m, 1.3 ka) was set at 100 % (using the size of the un-normalised -Si-OH peak following peak fitting).

Diatom assemblage
Five different types of diatoms and particles were counted for 40 samples downcore (Cyclotella ocellata, Pliocaenicus seczkinae, Surirella fragments, contamination particles and other diatom species) with an average of 236 ± 57 counts.The diatom assemblage in the analysed <10 µm fraction was almost mono-specific (Fig. 3

Contamination assessment and correction
Out of 96 samples, 80 contained sufficient material after all purification steps for both oxygen isotope measurements and contamination analysis (>5 mg).Only a low correlation exists between the final mass of the sample after purification and Al 2 O 3 percentage (r 2 = 0.14).However, all samples above 2.5 % Al 2 O 3 had final masses below 10 mg which indicates that in horizons with low biogenic productivity (BSi <10 %) a purification resulting in enough cleaned sample material can be rather difficult.
Measurements reveal an increase for the next 20 ka with a peak of +24.4 ‰ (5.85 m, 127.2 ka) before a progressive decrease down to +20.7 ‰ (7.35 m, 162.0 ka).This peak at 5.85 m (127.2 ka) is the absolute maximum of the record.Between 7.35 m and 9.56 m (162.2 ka and 187.6 ka) no measurements could be performed as the two samples within this period did not provide sufficient material.The δ 18 O value of +20.1 ‰ at 9.56 m (187.6 ka) supports the decreasing trend.From the short-term minimum downcore a rapid increase until 197.6 ka was observed, plateauing until 209.3 ka with an average δ 18 O = +23.2‰.Further downcore, a slight decline of 1.5 ‰ was detected until values increased again to a maximum of +23.6 ‰ (243.9 ka).

Diagenetic effects
In order to test the samples for diagenetic effects which could potentially alter δ 18 O values with time FTIRS measurements were performed.The determined absolute percentage of Si-OH bonds ranges between 5.6 and 9.0 %.As absolute unweighted calculations are not suitable for comparing individual samples the change in Si-OH bond percentages relative to the highest peak area (top sample at 0.09 m, 1.3 ka set at 100 %) was calculated resulting in maximum values in the upper core section between 0-30 ka (average 88 %), a local minimum at 90 ka (64 %) and relatively constant values further downcore (average for the period between 90 ka and 230 ka = 68 %).A relative decrease of 2.9 % per 10 ka is observed in the first 90 ka (total decrease of 26 %) with a correlation between Si-OH percentage and age of r 2 = 0.63).For the relatively constant part of the Si-OH percentage record from 90 ka to 231 ka only a weak correlation exists (r 2 = 0.30).These numbers are not absolute and should only be used qualitatively to separate larger from smaller Si-OH layers.However, different calculations by peak height ratio (945 cm −1 , SiOH, vs. 800 cm −1 , inter-tetrahedral Si-O-Si bending vibrations) or peak area ratio (990-860 cm −1 vs. 860-750 cm −1 ) confirmed this trend.

CPD Introduction Conclusions References
Tables Figures

Back Close
Full In a deep, well-mixed lake such as Lake El'gygytgyn with little evaporation and a defined small catchment area δ 18 O lakewater is mainly determined by the oxygen isotope composition of precipitation (δ 18 O p ). Precipitation is either directly going into the lake or is supplied via one of the small streams (Chapligin et al., 2012;Wilkie et al., 2012).The residence time is about 100 yr (Fedorov, 2012)  Assuming a winter to summer precipitation ratio of 60 % to 40 % (Nolan and Brigham-Grette, 2007) et al., 1993).Due to the relatively stable lake water temperatures and the negligible effect from ∆T lake we assume the effect of air temperature changes as predominant.Nolan et al. (2012) identified modern synoptic weather patterns and climate trends affecting air temperatures from National Centers for Environmental Prediction (NCEP) global reanalysis data.Strong Aleutian lows and strong high pressure centres over the Arctic Ocean prevail in the winter bringing cold Arctic air and moisture from the East and North to Lake El'gygytgyn.Summers are characterized by weak low pressure systems over Siberia and broad high pressure systems to the south and east providing warm continental air and recycled moisture to the lake.
For this area and the lake system, the general warming or cooling trends for air temperature are more important in contrast to changes in storm tracks being negligible (Nolan et al., 2012).Furthermore, Nolan et al. (2012) suggest that these weather patterns have been relatively stable with time and are likely representative of this and other interglacial periods.
The modern day and interglacial mean annual air temperature is −10.3 • C (Nolan and Brigham-Grette, 2007) while mean annual air temperatures for glacials were assumed to be about −25 • C using the coldest value from modern day NCEP daily reanalysis Introduction

Conclusions References
Tables Figures

Back Close
Full  (Swann et al., 2010).Using the Dansgaard relationship of +0.6 ‰ per • C this temperature difference of about 15 • C is equivalent to a long-term amplitude in δ 18 O p by 9 ‰.This should be significantly higher when using MIS 5e interglacial temperatures instead of modern day temperatures.
If only condensation temperature controlled this peak-to-peak amplitude of ∆ 18 O = 5.3 ‰, this would correspond to a mean annual air temperature change of about 9 • C between LGM and MIS 5e interglacial (according to Dansgaard, 1964).
Potentially, the change in the isotope signal due to air temperatures could be amplified by changes in continentality patterns.Lower sea levels and extended sea-ice in glacial periods increase the continentality of the location and the effect of the Rayleigh distillation and; thus, further decrease the δ 18 O p .In contrast, there are three possibilities to explain a reduced amplitude.3. The global isotope-temperature relationship (Dansgaard et al., 1964) is based on recent precipitation from different regions.There is no detailed information if this δ 18 O p -MAAT relationship is valid for the region, nor if it might have changed through time.Additionally, every single δ 18 O diatom sample refers to centennial time scales, thus longer than the period considered by Dansgaard (1964).However, the reasonable order of magnitude of ∆ 18 O in the Lake El'gygytgyn isotope record indicates that air temperatures should be the main driver.

δ 18 O diatom Holocene study
Apart from this study Swann et al. (2010) examined the oxygen isotope composition of diatoms for the first 23 ka at Lake El'gygytgyn in relatively high resolution (Fig. 6).Differences exist between the two studies and a relatively constant offset of 3.4 ± 1.0 ‰, n = 16 was observed.The studies are compared and simplified in the following section to study1 (Swann et al., 2010) and study2 (our data).While for study1 the 5-75 µm fraction was used (comprised of Cyclotella ocellata and Pliocaenicus costatus) to provide enough material for isotope analysis, in study2 the <10 µm fraction was used in order to obtain single-species samples of Cyclotella ocellata.Differences in diatom assemblages (and different living conditions between both species) between the two studies could have caused the different δ 18 O values, although numerous studies have found no evidence of a species-dependent isotope fractionation in diatoms (Schmidt et al., 1997(Schmidt et al., , 2001;;Brandriss et al., 1998;Moschen et al., 2005).Chapligin et al. (2012) analysed the Pliocaenicus-dominated >10 µm fraction vs. the Cyclotella-dominated <10 µm fraction for sub-surface samples from Lake El'gygytgyn and detected no significant difference in δ 18 O diatom values between the two size fractions after contamination correction.
The purity of diatoms in a sample could cause differences in δ 18 O values between both studies.The % cont. in study1 was assessed by counting under a light microscope, resulting in a mean sample purity of 95.9 ± 1.9 % (1 σ), while study2 showed a Introduction

Conclusions References
Tables Figures

Back Close
Full purity of 94.6 ± 3.4 %.Study1 did not correct the measured values for contamination where as study2 did.However, higher contamination should cause lower δ 18 O values for study1 which is not the case.Different dehydration techniques were applied in both laboratories (step-wise fluorination = SWF; inert Gas Flow Dehydration = iGFD), but an inter-laboratory comparison showed for both techniques results within standard deviation on various biogenic working standards (Chapligin et al., 2011) between the two laboratories.The removal of Si-OH groups in Nickel cylinders by using SWF as applied in study1 could potentially remove oxygen from clay structures whereas iGFD removes just H 2 O and -OH groups.However, study2 corrected for contamination, so a significant difference due to different dehydration techniques can also be excluded.Both studies show different analytical reproducibility (mean SD (1 σ) = 0.34 ‰ for study1 vs. 0.13 ‰ for study2) but a similar trend can be observed.In both records, δ 18 O values increase until 8-10 ka corresponding to the HTM (study1: +27.0 to +27.4 ‰, increase between 3-10.7 ka; study2: +23.4 ‰ peak at 8.4 ka) and decrease until 20-22 ka corresponding to the LGM (study1: +22.4 to +22.5 ‰ at 21.4-21.6ka; study2: +19.1 to +20.3 ‰ at 21.2-23.1 ka).Hence, no obvious reason for the offset in δ 18 O values between the two studies could be observed, but both records show similar trends indicating that conclusions on palaeoclimatic changes can be drawn based on relative δ 18 O changes in the individual studies.

Biogenic silica, TiO 2 , magnetic susceptibility
For a longer time-series, we compared measurements of biogenic silica (BSi [%]), TiO 2 [%] and magnetic susceptibility analysed on the same core (Lz1024; Frank et al., 2012) to our δ 18 O record (Fig. 7).BSi can be generally used as a proxy for nutrient availability and bioproductivity or primary production (Colman et al., 1995).TiO 2 is commonly used as a proxy for clastic sediment supply from fluvial (Haug et al., 2001) or eolian (Yancheva et al., 2007) transport and often related to weathering (Minyuk et al., 2011).
To compare the δ Frank et al., 2012) we interpolated between sample depths to get comparable data.Both, BSi and TiO 2 records do not correlate well with the δ 18 O record (δ 18 O vs. BSi, r = −0.14;δ 18 O vs. TiO 2 , r = −0.33).The anti-correlation of the overall record is mainly due to the first three meters of the core and the high peak of BSi at the time interval corresponding to the LGM.This can be explained by enhanced weathering (maximum in TiO 2 percentage) and a subsequent nutrient availability (maximum in BSi percentage) accompanied by a minimum in δ 18 O values.Magnetic susceptibility in lake sediments reflects the concentration of magnetic mineral phases which is controlled by preservation in Lake El'gygytgyn sediments (Nowaczyk et al., 2007).Anoxic bottom and pore waters dissolve the respective mineral phases while keeping preserved under oxic conditions.Thus, high and low magnetic susceptibility values can represent interglacials and glacials but a high organic content during interglacials can also induce anoxic bottom water conditions.
None of the compared proxies are directly related to temperature and precipitation and hence, only weak correlations exist between these proxies and the diatom δ henceforth simplified to EPICA) was used for comparison further downcore as well as the mean monthly (July) insolation at 67 • N (Laskar et al., 2004) being indicative for the main diatom blooming period (Fig. 8).
For comparing the EPICA, the LR04 record and the mean monthly (July) insolation at 67 • N with our δ 18 O record we interpolated between sample ages.The correlation of the diatom δ 18 O record and the stacked δ 18 O LR04 record (Fig. 9; r = +0.58)and δD EPICA record (Fig. 9; r = +0.69)are significant (p < 0.05) indicating that a clear precipitation driven climate signal is preserved in the δ 18 O record from diatoms at Lake El'gygytgyn.
The correlation between the diatom δ 18 O record and the mean monthly (July) insolation at 67 • N is r = +0.56.When correlating this record to the different Milankovitch cycles, the 41 ka cycle caused by the obliquity of the ecliptic is clearly the best fit (r = +0.62;eccentricity: r = +0.29;precession: r = −0.38).Hence, solar forcing, especially the 41 ka obliquity cycle, seems to be the main forcing factor on longer timescales for δ 18 O in diatoms at Lake El'gygytgyn.On the one hand, this further supports the proposed age model, on the other hand this shows the potential of δ 18 O in diatoms at Lake El'gygytgyn as a long-term climate archive back to 3.6 Ma.

Paleoclimate reconstruction
Glacial-interglacial cycles are clearly visible in the presented δ 18 O diatom record at Lake El'gygytgyn dating back to 250 ka.This includes prominent minima and maxima such However, the LGM minimum in the Lake El'gygytgyn δ 18 O record is earlier than those in LR04 and EPICA (20-18 ka), whereas the insolation minimum at 67 Again, this maximum correlates well with the maxima of the EPICA and the insolation record.However, the increase towards higher δ 18 O values occurs earlier than in the other records and already at ∼133 ka the values are close to the peak maximum.Further maxima are present in the time interval corresponding to MIS 7.1 and 7.3 (plateau ∼+23.0 ‰, ∼209-203 ka) as well as MIS 7.5 (+23.6 ‰, 243.9 ka).The intermediate period of MIS 7.2 could not be detected despite the analysis of four samples within this time.Hence, the δ 18 O record shows a plateau which is more similar to the weakly developed minima of the EPICA and LR04 records than to the distinct minimum of the insolation record at ∼206 ka.The δ 18 O minimum corresponding to MIS 7.4 (+21.5 ‰, ∼231 ka) correlates well with the EPICA, LR04 and insolation minima.The time corresponding to MIS 7.5 is reflected by the local δ 18 O maximum (+23.6 ‰, 243.9 ka) which occurs slightly earlier than in the EPICA, LR04 and insolation record (∼240 ka).
According to the presented δ 18 O diatom record of Lake El'gygytgyn similar temperatures occurred in the time of the HTM, MIS 3 and MIS 7, where as the MIS 5e interglacial was significantly warmer than the Holocene.In the global marine LR04 record abrupt warming and slow cooling trends were observed whereas in our δ of condensation in the atmosphere) with downcore variations indicating that glacialinterglacial cycles are present throughout the whole core.The Holocene Thermal Maximum (HTM; δ 18 O∼+23 ‰; 8.9 ka), the Last Glacial Maximum (LGM; +19.1 ‰, 23.1 ka) as well as the MIS 5e interglacial (+24.4 ‰, 127.2 ka) and the interval corresponding to MIS 7.1, 7.3 (plateau around +23.0 ‰, ∼209-203 ka) and MIS 7.5 (+23.6 ‰, 243.9 ka) are clearly visible.A peak-to-peak amplitude of 5.3 ‰ between the MIS 5e absolute maximum and the LGM absolute minimum was detected.This shift would be equal to a change in mean annual air temperature of about 9 • C when assuming that changes in due to the coverage of large areas by glaciers, e.g. during Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

%
cont. by multiplying the sample's Al 2 O 3 percentage with a clay assemblage factor which depends on the Al 2 O 3 percentage of the average clay types present in the core.Hence, the Al 2 O 3 percentage for each sample was analysed by Energy-Dispersive Xray Spectroscopy (EDS) (3-5 repetitions, diameter of excited-area size: 100-120 µm) and the % cont.and δ 18 O cont.end-members determined from a heavy fraction after the first heavy liquid separation containing no diatoms (verified under scanning-electron microscope, SEM) resulting in Al 2 O 3 < 10 µm = 16.6 % and δ 18 O cont.= +6.5 ‰ Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | , biovolume [%] with Cyclotella ocellata constituting >90 % of all diatom biovolume in 34 out of 40 samples, and 70-90 % in 4 samples).Only at 10.15 m and 10.75 m depth (203 and 208 ka) did the usually >10 µm sized species Pliocaenicus seczkinae predominate the <10 µm fraction (smaller frustules) with biovolume abundances of 52 % and 58 % respectively.Discussion Paper | Discussion Paper | Discussion Paper | 18 O of +19.1 ‰ at 1.61 m/23.1 ka.The δ 18 O values continuously increase until +22.4 ‰ (1.98 m, 34.3 ka) with subsequent fluctuations within about 1.0 ‰ for the next 70 ka and prominent minima of +19.6 ‰ (3.36 m, 70.5 ka) and +20.0 ‰ (4.80 m, 107.3 ka) Discussion Paper | Discussion Paper | Discussion Paper | temperature, blooming period and δ 18 O waterVarious factors can influence the δ18 O values of lacustrine, biogenic silica (Fig.5).The lake temperature in which the diatoms grow has an effect on the fractionation between δ 18 O water and δ 18 O diatom of about −0.2 ‰ per • C(Brandriss et al., 1998;Moschen et al., 2005;Dodd and Sharp, 2010; Dong and JingTai, 2010).Lake El'gygytgyn lacks a true thermocline as in winter the lake is thermally stratified, but only with a difference of 2.3• C from the top to bottom of the water column (0.7• C at 3 m water depth, 3• C at 170 m depth).As soon as the ice melt begins, the first 30 m are mixed (at 1.4• C, beginning of June) and in mid-July the lake is completely mixed with temperatures from July to Mid-October ranging between 2.2 • C to 3.8 • C (observations from 2002, Nolan and Brigham-Grette, 2007).Diatoms are phototrophic organisms and the major blooming period was assessed by Cremer et al. (2005) to the time from early spring (under-ice) through open-water conditions until the re-establishment of ice coverage in autumn with the highest increase of Cyclotella ocellata in the first meters of the photic zone between 1 June and 20 July.Thus, due to this short major blooming period variations in lake temperature should not exceed ±1 • C, equal to water temperature induced changes in δ 18 O values of ±0.2 ‰ which is within the instruments error for biogenic silica (1σ = 0.25 ‰, Chapligin et al., 2010).Chapligin et al. (2012) has further shown that no significant isotopic differences exist in the water column throughout various times of the year, so only minor effects from seasonality are expected to be reflected in δ 18 O diatom values.With a deep photic zone of about 34 m (Dehnert and Juschus, 2008), good mixing conditions and a location in the cold Arctic environment, no significant impact on the δ 18 O diatom values from varying lake water temperature can be expected from different ages, samples or from taxa blooming in different seasons.Evaporation trends were not detected when analysing the δ 18 O lake water profile at three different times of the year (May, August, November;Chapligin et al., 2012).Despite the fact that these observations account mainly for the recent system, the reconstructed differences in lake level (terraces: +35-40 m, middle Pleistocene; +9-11 m, late Pleistocene; +2-3 m, Holocene;Glushkova and Smirnov, 2007) and subsequent differences in evaporation rates can potentially change the hydrological conditions (i.e. the residence time) and cause a shift in δ 18 O lakewater .However, changes in the extension of the photic zone should be considerably low throughout time.Hence δ 18 O diatom is basically influenced by δ 18 O lakewater .
causing no major delays in the transfer of the δ 18 O p signal to δ 18 O lakewater for the last 100 yr as shown in the isotope mass balance model of the lake (Wilkie et al., 2012).The model additionally shows that this residence time buffers short-term changes and reduces the amplitude in δ 18 O diatom to approximately one third on a decadal scale (Wilkie et al., 2012).However, the low sedimentation rate only allowed for sampling on a centennial scale and for gaining average δ 18 O diatom values.Compared to this sample-related time resolution and the relatively low resolution from the δ 18 O diatom record (74 samples/250 ka), the residence time can have reduced the amplitude by "cutting off local maxima or minima" of the δ 18 O diatom record to a certain extent but considerably less than in the 100 yr model.Modern lake water isotope composition of δ 18 O AVG = −19.8‰ (Chapligin et al., 2012) matches the oxygen isotope composition of precipitation ranging between −14 and −34 ‰ for snow samples of single winter events (mean −23.2 ‰) and rain δ 18 O values ranging from −15.9 ‰ to −12.4 ‰ (mean −14.3 ‰) (Wilkie et al., 2012).Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

1.
The different lake temperature (about −0.2 ‰ per • C) and air temperature (about +0.6 ‰ per• C) factors cannot be mixed up and merged, as air temperature varies between glacial and interglacials from −10.3 • C to −25 • C while water temperatures at these latitudes in well mixed lakes with a depth of 170 m must have been always around 4• C with only small variations (modern and interglacial lake temperatures about 2-4• C; see Sect.4.1.1).Hence, the effect from lake temperatures (e.g.potentially higher lake temperatures during the MIS 5e interglacial or the Greenland interstadial 1) would reduce the δ 18 O diatom values and, thus, the amplitude slightly (e.g. by 1 ‰ when assuming a maximum temperature difference of 4 to 5 • C). 2. The amplitude reduction from δ 18 O p to δ 18 O water could also originate from the lake residence time acting as a buffer for extreme variations in δ 18 O p over time (see Sect. 4.1.2).This effect could increase with longer residence time in cold phases generally characterized by lower temperature and humidity and thus potentially less moisture reaching the lake setting.Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 18 O record with both data sets (BSi record, Ros én et al., 2010; TiO 2 Discussion Paper | Discussion Paper | Discussion Paper | is remarkable though, that δ 18 O peaks often occur earlier than TiO 2 minima or BSi maxima percentages (cf.δ 18 O at 5.50-6.22m depth vs. TiO 2 and BSi at 5.30-5.80m; δ 18 O, 2.40-2.70m vs. TiO 2 and BSi, 2.40-2.00m) which could indicate a more direct atmospheric signal responsible for δ 18 O while there is a delayed reaction in the more indirect proxy records.4.3Comparison with other regional and global climate recordsThe diatom δ 18 O record was compared with prominent climate curves such as the global marine δ 18 O benthic stack LR04 (Lisiecki and Raymo, 2005; henceforth simplified to LR04) and the glacial δ 18 O record from the North Greenland Ice Core Project (NGRIP; North-Greenland-Ice-Core-Project members, 2004; henceforth simplified to NGRIP).As this latter record ends at ∼123 ka the δD Dome-C record from the European Project for Ice Coring in Antarctica (EPICA; EPICA members, 2004, 2006; Discussion Paper | Discussion Paper | Discussion Paper | as the Holocene Thermal Maximum (HTM, δ 18 O ∼ +23 ‰, 8.9 ka).A post-HTM cooling trend can be observed, while in the pre-HTM time the Younger Dryas stadial (Greenland Stadial 1) has not been identified in the δ 18 O record which might be due to the relatively low resolution.However, in the higher resolved δ 18 O record of Swann et al. (2010) a local minimum in the Younger Dryas exists, but is not clearly visible within the general trend.The LGM (δ 18 O = +19.1 ‰, 23.1 ka) reflects the absolute minimum of the record and corresponds to minima in the δ 18 O NGRIP, δD EPICA Dome-C and stacked δ 18 O LR04 records (Fig. 8) as well as in the July insolation record from 67 • N. Discussion Paper | Discussion Paper | Discussion Paper | 18 O record both warming and cooling processes seem to occur with the same speed.However, Discussion Paper | Discussion Paper | Discussion Paper |
Chapligin et al. (2012)showed that the geochemical mass-balancing correction technique is reasonable for samples containing up to 2.5 % Al 2 O 3 or <15 % contamination.From the 80 analysed samples 76 samples had an average Al 2 O 3 percentage of 0.94 ± 0.16 % (Fig.3, SiO 2 and Al 2 O 3 [%], ∆ 18 O corr−measured ).For four samples the Al 2 O 3 percentage was >2.5 % and for one more sample the Al 2 O 3 percentage of 2.3 % resulted in a high δ 18 O value correction of >2.5 ‰.These samples were discarded from further interpretations.
the annual average of δ 18 O values would be −19.3‰.Stream δ 18 O values ranged from −24.2 ‰ to −16.7 ‰ (Wilkie et al., 2012).Hence, the oxygen isotope composition of the diatom samples reflects δ 18 O lakewater which is mainly controlled by δ 18 O p (pathway via streams or directly on/into the lake).
perature, and/or continentality.According toDansgaard (1964)the mean annual air temperature (MAAT) at surface is directly related to the mean annual δ 18 O p signal (shift of +0.7 ‰ per • C; in high latitudes/continental regions +0.6 ‰ per • C, Rozanski data points, occurs at ∼70 ka and matches the NGRIP and the insolation minimum.After the LGM minimum it is the second lowest minimum reflecting a cold time period.The MIS 4 is preceeded by a characteristic maximum corresponding to the MIS 5e interglacial (+24.4 ‰, 127.2 ka) being the absolute maximum of the record.