Modern isotope hydrology and controls on δ D of plant leaf waxes at Lake El ’ gygytgyn , NE Russia

Stable isotope data from lipid biomarkers and diatom silica recovered from lake sediment cores hold great promise for paleoclimate and paleohydrological reconstructions. However, these records rely on accurate calibration with modern precipitation and hydrologic processes and only limited data exist on the controls on the δD values fornalkanoic acids from plant leaf waxes. Here we investigate the stable isotopic composition of modern precipitation, streams, lake water and ice cover, and use these data to constrain isotope systematics of the Lake El’gygytgyn Basin hydrology. Compound-specific hydrogen isotope ratios determined from n-alkanoic acids from modern vegetation are compared with modern precipitation and lake core top sediments. Multispecies net (apparent) fractionation values between source water (precipitation) and modern vegetation (e.g., εwax/precip mean value is−107± 12 ‰) agree with previous results and suggest a consistent offset between source waters and the δD values of alkanoic acids. We conclude that although there may be some bias towards a winter precipitation signal, overall δD values from leaf waxn-alkanoic acids record annual average precipitation within the El’gygytgyn Basin. A net fractionation calculated for 200-yr-integrated lake sediments yieldsε30/precip= −96± 8 ‰ and can provide robust net “apparent” fractionation to be used in future paleohydrological reconstructions.


Introduction
Hydrogen isotope ratios (δD) of lipid biomarkers are emerging as a new paleoclimatic and paleohydrological proxy (e.g., Sauer et al., 2001;Huang et al., 2002;Jacob et al., 2007;Feakins and Sessions, 2010;Castañeda and Schouten, 2011;Sachse et al., 2012 and references therein).Environmental water is the primary source of hydrogen for organic matter from photoautotrophic organisms; for terrestrial plants, this constitutes soil water that is ultimately derived from precipitation whereas aquatic organisms utilize lake water affected by lake hydrology.Multiple environmental variables affect the δD values of meteoric water such as temperature, elevation and water vapour source (Craig, 1961;Craig and Gordon, 1965;Dansgaard, 1964;Gonfiantini, 1986;Gat, 1996).The isotopic composition of meteoric water is also controlled by continental and "latitudinal effects" (Dansgaard, 1964;Craig, 1961;Gonfiantini, 1986;Gat, 1996;Bowen and Revenaugh, 2003;Kurita et al., 2004), which are reflected in the δD of plant leaf waxes (δD wax ) (Sachse et al., 2006;Smith and Freeman, 2006;Hou et al., 2008;Liu and Yang, 2008;Polissar and Freeman, 2010).Previous research has tested the relationships between key hydrological variables and leaf wax lipid δD values from modern vegetation along climate gradients, yielding strong linear relationships of δD wax with mean precipitation δD values (Sachse et al., 2004;Huang et al., 2004;Hou et al. 2008;Polissar and Freeman, 2010).Changes in sedimentary δD wax have been interpreted as shifts in precipitation δD (δD p ) (e.g., Liu and Yang, 2008;Tierney et al., 2008) or in Published by Copernicus Publications on behalf of the European Geosciences Union.
In general, abundance and composition of biomarkers can serve to identify past ecosystems and environmental parameters during their formation (Meyers, 2003).Fatty acids such as long-chain even carbon-numbered n-alkanoic acids (e.g., > 24 carbon chain length) typically originate from leaf waxes of terrestrial plants (Eglinton and Hamilton, 1967;Cranwell, 1974;Cranwell et al., 1987).In contrast, shortchain even carbon-numbered n-fatty acids (i.e., < 20 carbon atoms) are abundant as major constituents of cell membranes and waxes in many organisms, and their relative abundances have been attributed to phytoplanktonic and zooplanktonic sources (Cranwell, 1974;Cardoso et al., 1983;Cranwell et al., 1987;Meyers and Ishiwatari, 1993;Volkman et al., 1998).Aquatic photosynthetic organisms in lakes obtain hydrogen from lake water to produce their organic compounds, whereas terrestrial plants utilize hydrogen within the leaf water, which is subject to evapotranspiration (Leaney et al., 1985;Huang et al., 2004;Sachse et al., 2004;Zhang and Sachs, 2007).Thus, isotopically distinct source waters are used for the biosynthesis of aquatic versus terrestrial biomarkers (Sachse et al., 2006(Sachse et al., , 2012)).Lake water hydrogen isotopic variations are driven by δD p , lake hydrology, and the residence time of the lake water (i.e., Leng and Marshall, 2004;Jacob et al., 2007).Large, well-mixed lakes with long lake water residence times, such as modern Lake El'gygytgyn (∼ 120 yr;Fedorov et al., 2009Fedorov et al., , 2012)), will "average out" short-term seasonal variations in precipitation and stream input isotopic compositions, and will reflect mean annual precipitation (Leng and Marshall, 2004;Nolan and Brigham-Grette, 2007).Leaf wax lipids derived from terrestrial plants instead record the isotopic ratio of soil water, which is ultimately recharged by precipitation.Snowmelt is typically the dominant annual hydrological event in Arctic watersheds (Kane et al., 1992), and within the El'gygytgyn Basin soil water is annually recharged by snowmelt (Nolan and Brigham-Grette, 2007;Fedorov et al., 2012).
Proxy use of δD wax for paleoclimatic reconstructions depends on establishing a "net (or apparent) fractionation" between meteoric water and lipid δD values (ε lipid/w ).Previous research has focused on δD values of C 29 n-alkanes (see Sachse et al., 2012 for review and global data set).However only a few studies have investigated factors affecting leaf wax δD values from n-alkanoic acids (Huang et al., 2004;Chikaraishi and Naraoka, 2007;Hou et al., 2007a, b).Yet an increasing number of paleoclimate studies utilize δD values from alkanoic acids for paleoclimate reconstructions (Tierney et al., 2008;Sun et al., 2011;Thomas et al., 2012).
In addition to variable climatic and/or biosynthetic drivers affecting net fractionations, large interspecies variability has been found at the individual plant scale (e.g., Chikaraishi and Naraoka, 2003;Liu and Huang, 2005;Hou et al., 2008), particularly in highly arid sites (Feakins and Sessions, 2010).Several studies have linked this variability to life form (e.g., tree, shrub, grass) (Liu et al., 2006;Liu and Yang, 2008;Sachse et al., 2012), leading to concern for the potential to bias paleoclimatic reconstructions given inevitable vegetation shifts in response to climatic change.However, Feakins and Sessions (2010) find a strong correlation between δD of source water and the average δD wax value at a site suggesting catchment-scale average δD wax values reliably record regional climate despite a large spread in values between individuals.This is in good agreement with other calibration studies using core-top sediments (e.g., Huang et al., 2004;Sachse et al., 2004;Hou et al., 2008) that show much better correlations to climate than studies of individual plants due to temporal and spatial integration of plant inputs.
Here we establish local relationships between the δD of modern vegetation, sediment traps and the δD in modern El'gygytgyn lake sediments as the context for reconstruction of El'gygytgyn Basin paleohydrology.We investigate the controls on alkanoic acid δD values as an increasing number of paleoclimate studies make use of this compound class.These studies are supplemented by isotopic analysis of samples of modern meteoric water, including winter and summer precipitation, streams, lake water and ice cover which provide insight into the modern isotope hydrology.This approach enables us to (i) determine the net fractionation (ε wax/w ) between modern vegetation and source water within the El'gygytgyn Basin, (ii) assess the delivery of organic matter and target compounds for δD analysis, and (iii) establish a framework for separating variations in the isotopic composition of lake water from changes in the isotopic composition of precipitation in the δD wax paleohydrological proxy.

Study area: location and climate
Lake El'gygytgyn is situated within a 3.58 × 0.04 Ma old meteorite crater (Layer, 2000) located in a previously unglaciated region of the far east Russian Arctic (67 • 30 N, 172 • 05 E; Glushkova and Smirnov, 2006; Fig. 1).The crater rim elevations vary between 600 and 930 m, while the lake surface lies at 492 m above sea level.The crater is approximately 18 km in diameter with a total watershed area of 293 km 2 (110 km 2 is lake surface), and the lake is 12 km wide and 175 m deep with a volume of 14.1 km 3 (Nolan and Brigham-Grette, 2007).Water temperatures in this cold monomictic, ultra oligotrophic lake are < 4 • C, and the lake is fed by 50 streams draining from within the crater rim (Cremer et al., 2005;Nolan and Brigham-Grette, 2007).Shallow lagoons with seasonal water temperatures > 6 • C are dammed by gravel berms surrounding the lake edge (Nolan and Brigham-Grette, 2007).Mean annual air temperature measured from 2002-2008 was −10.4 • C, and total precipitation rose from 350 mm a −1 to about 550 mm a −1 over this period with roughly equal rates between winter and summer precipitation (Nolan and Brigham-Grette, 2007;Nolan, 2012).The El'gygytgyn Basin is located in a zone of hypoarctic tundra (Yurtsev, 1973), and modern vegetation within the basin is dominated by lichen and discontinuous herbaceous taxa (Kohzevnikov, 1993;Minyuk, 2005;Lozhkin et al., 2007).Salix and Poaceae species comprise the dominant woody and non-woody taxa, respectively, and modern tree line is located ca. 150 km to the south and west of the lake (Lozhkin et al., 2006).

Water samples
Lake water and precipitation samples were collected during the 2003 El'gygytgyn field expedition in 30 ml Nalgene bottles and stabilized with one drop mercuric chloride (Juschus et al., 2005a).Water from incoming streams and the Enmyvaam outflow were likewise collected, and all water samples were stored at 4 • C at the Geoscience Dept., UMass Amherst, until analysis.Precipitation samples were also collected over the winter 2008-2009 (Table 1).

Modern vegetation samples
Herbarium specimens of typical modern plants growing close to the lake were sampled in summer 2003.Specimens were collected during summer 2003 for the purposes of establishing a standard pollen collection of the local vegetation at Lake El'gygytgyn (Minyuk, 2005).Subsamples (∼ 0.5 g) of seven representative specimens were removed from unfixed portions of the plants, rinsed with distilled water, dried at 40 • C, and homogenized with mortar and pestle.The total lipid extract (TLE) was separated into neutral and carboxylic acid fractions on a solid phase extraction (SPE) column containing 0.5 g of DSC-NH 2 stationary phase (Supelco).A neutral fraction was eluted with 4 mL hexane : dichloromethane DCM (1 : 1) and archived for future analysis.Fatty acids were eluted with 8 ml of 2 % formic acid in DCM and were subsequently derivatized as methyl esters (FAME) with BF 3 MeOH.

Sediment traps and sediment core tops
Sediment traps were deployed during the spring field expedition, 2003 (Fig. 2).Each sediment trap consisted of a base plate made of plastic with a steel rod to fix the trap in vertical position at the rope.Four plastic tubes with a diameter of 46 mm were screwed to the bottom of this plate.The rope was anchored at the lake bottom with a bag filled with pebbles and kept up in the water column by buoys fixed beneath the base of the lake ice (Juschus et al., 2005b).Unfortunately, sediment traps were only recovered at one location during the 2003 expedition yielding only six samples.
Samples of the sediment water interface were also collected during the spring 2003 field expedition.Coring was carried out from a floating platform with a gravity corer; attention was paid to horizontal preservation of the sediment water interface.Twelve of these samples were selected for biomarker analysis, of which four samples were selected for further preparation for compound-specific isotopic analysis.Additionally, a sample from the upper 0-2 cm of a 16-m-long sediment core (LZ1024) was also isolated and prepared for  analysis.These samples represent modern sedimentation in the lake.

Water isotopic analysis
The oxygen and hydrogen isotopic compositions of the water samples were analyzed at the Alfred Wegener Institute for Polar and Marine Research (AWI) Potsdam, Germany, with a Finnigan MAT Delta-S mass spectrometer equipped with two equilibration units for online determination following the methods of Meyer et al. (2000). in permil (Gonfiantini, 1978): The water samples show an analytical precision < 0.10 ‰ (1σ ) for δ 18 O and < 0.8 ‰ for δD, respectively (Meyer et al., 2000).

Compound-specific isotopic analysis (CSIA)
The method for vegetation, sediment traps and all sediment samples is identical.Samples were freeze-dried, crushed and stored in combusted glassware.Soluble organic matter was extracted using an accelerated solvent extractor (ASE200, Dionex Corp., Sunnyvale, USA) with a dichloromethane/methanol mixture (9 : 1) at 100 • C and 1200 psi for three 15-min cycles.Given the relatively low amount of organic C in the samples (Melles et al., unpublished data), the maximum amount of sediment available was used (between ∼ 7 g-12 g dry sediment).The total lipid extract (TLE) was separated into neutral and carboxylic acid fractions on a solid phase extraction (SPE) column containing 0.5 g of DSC-NH 2 stationary phase (Supelco).A neutral fraction was eluted with 4 ml hexane : DCM (1 : 1) and archived for future analysis.Fatty acids were eluted with 8 ml of 2 % formic acid in DCM and were subsequently derivatized as methyl esters (FAME) with BF 3 MeOH.Compounds (i.e., FAMES) were identified and quantified using a Hewlett Packard 6890 series gas chromatograph-mass selective detector (GC-MSD) and a Hewlett Packard 6890 series gas chromatograph-flame ionization detector (GC-FID), respectively, and with comparisons to external standard mixtures.Both the GC-MSD and GC-FID were equipped with 5 % phenyl methyl siloxane columns (30 m × 0.25 mm ID, film thickness 0.25 µm).The GC oven temperature programs were also identical for both instruments: 60 • C (2 min), 15 • C min −1 to 315 • C (15 min) with helium as the carrier gas and the inlet temperature set at 320 • C. The δD of individual lipids was determined at the ESCSIS facility at Yale University using a Thermo Finnigan MAT 253 mass spectrometer with a gas chromatograph combustion (GCC) interface with a high-temperature pyrolysis furnace operated at 1440 • C (Hilkert et al., 1999) coupled with a Thermo Trace GC GC-MS equipped with a DB-5ms column (30 m, ID: 0.32 mm, film thickness: 0.5 µm, Agilent).The gas chromatography (GC) temperature program was 80 • C (2 min) to 120 • C at 12 • C min −1 and then to 320 • C at 6 • C min −1 (held 16 min).For δD analysis, each sample was analysed in triplicate and the H + 3 factor was determined daily prior to standard calibration and sample analysis.The results are reported using standard delta notation (i.e., δD values) in permil (‰) units.Isotope ratios were normalized relative to a co-injected standard mixture containing C 14 nalkane (δD 14 n-alkane = −71.7 ‰ ± 1.4 ‰) to condition the furnace and C 15 (δD 15 n-alkane = −88.4‰ ± 1.2 ‰) and C 38 n-alkanes (δD 14 n-alkane = −102.6‰ ± 1.3 ‰) as compounds of known isotopic values (whose δD values were determined by Arndt Schimmelman, Indiana University, via offline pyrolysis and dual inlet analysis).Instrument performance was also determined daily by measuring a suite of 5 FAMEs with a co-injected standard mixture (nC 14 , nC 15 , nC 38 ) of known isotopic values.The mean precision for replicate analyses of the standard compounds was typically better than 5 ‰ (1σ ).

Isotope mass balance correction
Fatty acid δD values were measured on methyl ester derivatives.A n-C 14 acid standard was derivatized with BF 3 MeOH (same lot number as used with samples) to calculate the δD contribution from the three H atoms added during the methylation reaction with BF 3 in MeOH.The bulk δD of this n-C 14 fatty acid standard (−240 ‰ ± 2 ‰; n = 5) was determined prior to derivatization offline by Thermal Conversion Elemental Analysis (TCEA, ESCSIS Yale University).Derivatized standard was measured in triplicate to yield an average δD value for the methyl contribution from the derivitization agent (−172 ‰ ± 3 ‰).This value was used to correct the measured δD values of fatty acid methyl esters for the added hydrogen atoms by mass balance.

Calculation of net "apparent" fractionations
Isotopic fractionations between two measured substrates, δD a and δD b , are reported as enrichment factors.Enrichment factors are reported in permil notation, implying a factor of 1000 (Cohen et al., 2007).Here we present the "net (or apparent) fractionation" (ε l/w ) between the δD of leaf wax (δD wax ) and the δD of source water (δD w ), commonly used in the plant leaf wax literature, where Although ε wax/w is typically calculated relative to δD of mean annual precipitation, this inherently assumes that the isotopic composition of the source water (δD w ) is identical to local δD values of precipitation (δD p ).In this study, the water available for plant uptake is largely spring snowmelt carried in streams and draining through the shallow active layer (see Sect. 4.1;Nolan, 2012).Therefore, we used the empirically determined average δD for stream inflow.We also calculate the net fractionation between plant leaf wax and annual average precipitation in the basin (ε wax/precip ) for comparison.

Model parameters
In order to assess the sensitivity of the lake basin to changes in the lake water residence time and changing δD p and δ 18 O p over time, we constructed an isotope mass balance model of the lake (Fig. 3): where δ lake(i) represents the initial δ value of the lake water, δ in the isotopic composition of input or precipitation, τ and t the lake residence time and time step (here defined as 1 due to the annual resolution of the tree-ring time series), respectively.Initial lake conditions were assigned based on the present-day mean δD lake (−155 ‰) and are similar to reconstructed input approximations.As long-term records of past δD p are lacking in this region, we used δ 18 O data from tree ring cellulose in northern Siberia (Sidorova et al., 2008) to reconstruct the isotopic ratios of past precipitation for input to our model.In order to derive isotope ratios of precipitation from δ 18 O cellulose , we used the relationship between source water (here assumed to be precipitation) and δ 18 O cellulose described in Anderson et al. (2002).We assigned a midrange value of 0.6 for the fraction of leaf water not subject to evaporation (f ) (e.g., Saurer et al., 1997)    where n is the number of data points and R 2 is the coefficient of determination (Fig. 4).Streams, lake ice and  4).These results are consistent with previous observations within the El'gygytgyn Basin (Schwamborn et al., 2006) and in northern Siberia (Meyer et al., 2002;Sugimoto et al., 2003;Kurita et al., 2004;Opel et al., 2011).
The isotopic compositions of precipitation and temperature are strongly correlated at mid-and high latitudes (Dansgaard, 1964), occurring because the degree of rain-out from the atmosphere through condensation is highly correlated with the condensation temperature (see Jouzel et al., 1997 for review).Figure 5 1).Relatively isotopically enriched values cluster predominantly along the south facing and western slopes (Fig. 2).Most negative δ 18 O values and δD values were found along the eastern side of the crater, in streams with relatively high topographic gradients (Fig. 2).Decreasing δ 18 O and δD values broadly correlate with higher relief (R 2 = 0.44), with the exception of the shortest stream (Stream 37).There is a general trend toward increasing δ 18 O and δD values over the course of the summer sampling interval (R 2 = 0.33).Sampled profiles of the lake water show small (δ 18 O: Chapligin et al., 2012; δD: ∼ 11 ‰, Table 4) seasonal variations in the upper 2 m surface waters.These variations are likely due to spring (May) input from streams carrying isotopically depleted snowmelt and evaporative enrichment during summer (August) (Chapligin et al., 2012).Below 2 m water depth, the lake is well mixed with an average δD value of −155.1 ± 0.3 ‰ (δ 18 O: −19.82 ‰; Chapligin et al., 2012).

Modern vegetation samples
We report abundances and hydrogen isotope ratios for C 20 , C 24 , C 26 , C 28 , and C 30 n-alkanoic acids from representa-tive samples of modern vegetation from within the Lake El'gygytgyn Basin (Fig. 6).Despite the small sample set (n = 7), we observe a large interspecies variation in the distribution of compounds with some species producing relatively high concentrations of short-chain compounds typically associated with aquatic organisms (i.e., nC 20 and nC 22 acids; Cranwell et al., 1987;Ficken et al., 2000).The average chain length (ACL = (C n •n)/ C n , where n = 20, 22,24,26,28,30) for modern vegetation varied between species, ranging from 24 to 27 (Table 2).Concentrations are normalized per g of total lipid extract (TLE).The highest concentrations of C 30 n-alkanoic acid were found in the Douglasia ochotensis and Rumex sp.samples, whereas lowest concentrations were observed in Artemisia and Bryophyta sp.High concentrations of nC 28 acid were found in the Bryophyta sp., but concentrations of all other n-alkanoic acids in the Bryophyta sp. were very low or below detection limit.
For all plants yielding measurable n-alkanoic acids, δD values are reported in Table 3.We observe differences in nalkanoic acid chain length abundances for different species as well as isotopic offsets between different chain lengths in some species.Thus, a mass-weighted mean leaf wax δD value (δD wax ) was calculated to provide a common basis for comparison of the δD of different species regardless of chain length abundances (Table 3).Within El'gygytgyn lake sediments, only nC 30 acid can be attributed solely to a terrestrial source (Holland et al., 2012).Therefore, we also compared δD wax values with the δD values of nC 30 acid (i.e., δD 30 ).The hydrogen isotope compositions of C 30 n-alkanoic acid (δD 30 ) strongly correlate with the concentration-weighted δD wax values (r 2 = 0.97) confirming δD 30 values are representative of integrated δD values for terrestrial leaf waxes from plants within the Lake El'gygytgyn watershed (ranging between −227 ‰ and −260 ‰).Epsilon values between δD wax and stream water (ε wax/streams ) and δD wax and annual average precipitation mean (ε wax/precip ) were calculated and are similar to net fractionations calculated solely with nC 30 -acid (stream water: ε 30/streams and annual average precipitation: ε 30/precip ) (Table 3).

Sediment traps and lake sediments
Concentrations of even chain n-alkanoic acids (nC 20 , C 22 , C 24 , C 26 , C 28 , and C 30 ) varied widely between sediment traps and samples of the upper 0-2 cm of lake sediments (Fig. 7).Sediment traps show variable distributions of compounds and concentrations with water depth with highest concentrations of all compounds at 90 m water depth (Fig. 7).Individual compounds were normalized by total yield of nalkanoic acids to show changes in the relative proportions of compounds (normalized to total yield of n-alkanoic acids) with water depth (Fig. 8  High concentrations of all compounds were observed in the upper 0-2 cm of lake sediments, increasing in concentrations with increasing lake water depth (note log scale, Fig. 7).The distribution of all compounds is consistent across all samples as indicated by calculated values of average chain length (ACL; Table 2).The hydrogen isotope compositions of sedimentary n-alkanoic acids are also reasonably consistent across samples (i.e., sedimentary δD 30 values varied by ∼ 10 ‰; within 2σ analytical error).Net fractionation values calculated between terrestrially sourced sedimentary nC 30 acid and different source waters (stream water; ε 30/streams or annual average precipitation; ε 30/precip ) are similar but show slightly greater fractionation between nC 30 acid and stream water (mean ε 30/streams = −114 ± 13 ‰; mean ε 30/precip = −105 ± 13 ‰; Table 3).regimes, while higher slope values are from sites with maritime climates (Lacelle, 2011).East Siberian Arctic winter precipitation exhibits a wide range of stable isotopic compositions (e.g., Kurita et al., 2004), and although the values presented here are not weighted means, the range of variation (e.g., δ 18 O values = 17 ‰ and δD values = 19 ‰) is similar to that previously reported within the El'gygytgyn Basin (Schwamborn et al., 2006).Modern winter precipitation at Lake El'gygytgyn closely follows the δD and δ 18 O values of precipitation from other Arctic sites in the GNIP database.

Clim
The values of the slope (7.35) and the y-intercept (−9.62) are lower than the GMWL but very close to the GNIP-based circum-polar regression, indicating a nearly unaltered precipitation signal from the original moisture source during winter months.In contrast, summer precipitation shows a clear kinetic fractionation of isotopes with lower slope and d-excess values.This may be related to enhanced evaporation occurring during summer months in typical tundra settings due to the water surplus on the land surface and continuous 24-h net radiation (Ohmura, 1982;Boike, 1997Boike, , 1998)).Reprecipitated moisture derived from recycled water masses from the land surface may also be included (Sugimoto et al., 2003;Kurita et al., 2004;Schwamborn et al., 2006).The seasonal contrast in modern precipitation may also be explained by the strong seasonality in the dominant weather patterns.Modern synoptic weather patterns identified from National Centers for Environmental Prediction (NCEP) global reanalysis data show strong low pressure systems over the Aleutians during winter and broad high pressure systems to the south and east with weaker lows to the north during the summer (Nolan et al., 2012).Strong Aleutian lows during winter tend to bring cold Arctic air from the east and north to the lake, whereas summer synoptic weather patterns bring warm continental air from the south and west (Nolan et al., 2012).Thus, these dominant weather patterns affect both the seasonal temperature change and source of moisture delivery to Lake El'gygytgyn.Seasonal air temperature changes affect the saturation vapour pressure of air masses and subsequently the degree of rainout, yielding decreasing precipitation isotope ratios with decreasing temperatures and along trajectories of atmospheric vapour transport.These trajectories deliver winter precipitation from a proximal north and eastern moisture source (i.e., E. Siberian, Chukchi and Bering seas and possibly North Pacific) and recycled moisture from the Siberian land surface from the south and west during the summer months.High d-excess values of winter precipitation in the El'gygytgyn Basin support a proximal moisture source with minimal alteration of the original isotopic composition while low d-excess values during summer months are consistent with recycled continental moisture from the south and west.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.
Input to the lake via streams draining from the crater rim is predominantly driven by winter snowmelt and shows spatial variations in isotopic composition according to stream gradient and slope aspect (Fig. 2).Stream samples are more 18 O and D depleted than overall summer precipitation isotope ratios with values closer to mean annual precipitation (δD p ).This suggests that stream waters carry a large winter precipitation signal offset by some input of isotopically enriched summer precipitation and/or evaporative enrichment of the soil water.Water tracks, characteristic of per-mafrost terrain and immature drainage systems, are common on El'gygytgyn slopes with rapid drainage of moisture through the active layer in early spring/ summer (Nolan and Brigham-Grette, 2007;Federov et al., 2012).The residence time of the water in the soil typically produces a dampening of the seasonal variations observed in precipitation (Buhay and Edwards, 1995;Anderson et al., 2002).However, this residence time is exceedingly short within the El'gygytgyn Basin.Therefore, stream water effectively represents the soil moisture in the basin, draining rapidly and recharged annually by snowmelt.Soil moisture data collected from different depths within the El'gygytgyn Basin (to base of active layer; [2002][2003][2004][2005][2006][2007][2008] show peak soil moisture in early spring prior to peak soil temperatures and with individual summer precipitation events having only a minor effect on soil moisture content (Federov et al., 2012).Deep permafrost in this region (∼ 500 m; Yershov, 1998;Schwamborn et al., 2006) precludes deep groundwater input, and streams are therefore representative of soil moisture recharged by winter precipitation draining through the shallow active layer during the short summer season.
Streams with south facing and western slopes display relatively enriched isotope compositions.Lower gradients for these slopes, longer transit times and seasonally greater net incoming solar radiation likely resulted in 18 O and D enrichment of these stream waters due to evaporation.These streams were also sampled late in the summer season, well after peak discharge and when flow in many of the smaller streams was reduced (< 1 m 3 s −1 ) and is often just a trickle (Nolan and Brigham-Grette, 2007).Thus, despite summer sampling during maximum expected evaporative enrichment, the isotopic ratios of stream samples appear biased towards more negative winter precipitation isotope ratios though offset by some minor evaporative enrichment.
Small seasonal changes are also observable in the isotopic composition of the lake water.Profiles of the lake water column show seasonal enrichment of the surface waters in summer and depleted surface waters in early spring.Although thermally stratified in winter due to ice cover, the lake is rapidly mixed following snowmelt and ice breakup in late May (Nolan and Brigham-Grette, 2007) with little variation below the uppermost surface waters.Isotopically depleted lake surface waters (2 m depth, Table 4) collected in late May are likely the combined result of melting lake surface snow and ice and input from streams carrying isotopically depleted snowmelt.Evaporative enrichment of the lake surface waters in summer (August) is small and likely very brief given vertical mixing due to strong winds (Nolan and Brigham-Grette, 2007) and has little to no influence on the overall isotopic composition of the lake (Chapligin et al., 2012).
The mean hydrogen isotope composition of the lake water (δD lake ) is largely controlled by the hydrogen isotope composition of precipitation (δD p ), duration of ice cover and the residence time of the lake.Input to the lake is a combination of direct precipitation on the lake surface and stream www.clim-past.net/9/335/2013/Clim.Past, 9, 335-352, 2013 discharge.Streams account for 85 % of water input to the lake (Fedorov et al., 2012), and prolonged or perennial ice cover prevents evaporative enrichment of surface waters.Accordingly, the average isotopic ratio of modern lake water is slightly more depleted in 18 O and D when compared to mean annual precipitation values, reflecting either greater input from isotopically depleted snowmelt and/or a lag in response based on lake residence time to past changes in δD p .
To assess the impact of this potential lag, lake water isotopic response was modelled by isotope mass balance (Fig. 3).Variations in δD p are reflected in the lake water isotopic response within decades (≥ 10 yr), but the magnitude of the response is strongly attenuated by the residence time.Modern residence times for Lake El'gygytgyn are ∼ 120 yr (Federov et al., 2009(Federov et al., , 2012) ) but could have been greater during ice-covered, glacial conditions.Although the modelled δD lake does not record the full magnitude of minor fluctuations of δD p , these results suggest δD lake does still respond to large magnitude changes in input on long glacial-interglacial timescales (see also Chapligin et al., 2012).

Modern vegetation
Modern vegetation within the El'gygytgyn Basin produced a range of n-alkanoic acids with variable concentrations and δD values.Reconstructions of the δD of meteoric water rely on a constant "net fractionation" between lipid and meteoric water.Differences between plant species and vegetation type may affect the net fractionation (e.g., Chikaraishi and Naraoka, 2003;Smith and Freeman 2006;Hou et al., 2007b).However, within watersheds these differences are integrated and a representative multi-species net fractionation factor may be appropriate for paleohydrologic reconstructions (Feakins and Sessions, 2010).Modern vegetation within the El'gygytgyn Basin includes 249 identified plant species and approximately an additional 100 rare species (Kohzenikov, 1993;Minyuk, 2005).
In order to evaluate how representative the seven sampled species are of all vegetation within the basin, we calculated an unweighted average distribution of compounds (Fig. 9).This unweighted average was then used to produce a representative assemblage of n-alkanoic acids within the El'gygytgyn Basin (Fig. 9).This modern vegetation distribution pattern was then compared with an average distribution of compounds from modern sediments calculated from all four sediment core tops.The two patterns are similar, although, as expected, there are discrepancies given the relatively small sample set and the modern floristic diversity within the basin.The modern vegetation distribution pattern appears particularly sensitive to Bryophyta weighting.Reducing the contribution from Bryophyta from the overall modern vegetation pattern results in a distribution similar to that observed in modern lake sediments.This may be be- cause Bryophyta may not be a significant portion of overall plant biomass in the basin or alternatively may not be major contributor of terrestrially sourced n-alkanoic acids to the lake.Additionally, Salix arctica and Artemisia spp.are abundant on El'gygytgyn slopes and modification of their proportional contributions produces a nearly identical distribution of compounds to modern lake sediments (Fig. 9).Minor differences may certainly be attributed to and accounted for with the further inclusion and proportional representation of the remaining vegetation types and species.This preliminary survey suggests that although the proportional contributions of each species to the overall pattern may change, these six species (excluding Bryophyta) are generally representative of the overall production and input of even-chain n-alkanoic acids from modern vegetation to the lake basin.
The hydrogen isotope composition observed in sampled vegetation may also be representative basin-wide.The range of interspecies δD variability observed around Lake El'gygytgyn is within other published ranges (e.g., Chikaraishi and Naraoka, 2003;Liu and Huang, 2005;Sachse et al., 2006;Hou et al., 2007;Feakins and Sessions, 2010), although these authors note that there is significant scatter at the level of individual plants across a wide range of climatic conditions.Some studies have linked δD values to life form (e.g., tree, shrub, and grass; Smith and Freeman, 2006;Liu et al., 2006;Liu and Yang, 2008).However, no clear linkage between δD values and life form was found in our study.Yet, despite a large spread between individuals, catchment scale average δD wax values generally do seem to reliably record the climate conditions of the region (e.g., Huang et al., 2004;Sachse et al., 2004;Feakins and Sessions, 2010) thereby allowing for a constant offset to be used for paleoenvironmental reconstructions.
The climate signals preserved in alkanoic acids from modern vegetation within the El'gygytgyn Basin closely reflect the isotopic composition of mean annual precipitation, with a possible bias towards winter precipitation due to the nature of water availability in an arid Arctic context.Within arid Arctic regions, and particularly at this site, snowmelt is the dominant annual hydrological feature (see Sect. 5.1) (Nolan and Brigham-Grette 2007;Federov et al., 2012).Given the presence and depth of extensive permafrost within the basin, this precludes deeper groundwater input and the majority of water available for plant uptake occurs as rapid drainage of snowmelt (i.e., winter precipitation) through the shallow active layer.The short growing season in this region -coinciding with warmer temperatures and enhanced snowmeltalso contributes to growth and production of biomass utilizing soil moisture dominated by stream water (active layer runoff) carrying a large winter precipitation signal offset by some input of isotopically enriched summer precipitation.Soil moisture data in this basin also show individual summer precipitation events have only a minor effect on soil moisture content (Federov et al., 2012).Thus, water availability may be the primary mechanism by which the isotopic composition of mean annual precipitation is recorded by the leaf wax proxy in the El'gygytgyn Basin.
Our calculated ε wax/w values (Table 3) display an overall average net fractionation between modern vegetation and stream water (ε wax/streams ) of −116 ± 12 ‰.However, this assumes that the source water available for plant uptake is directly represented by streams in the El'gygytgyn Basin (i.e., δ w = δ streams ).Potential changes in the seasonality of precipitation and Arctic summer productivity (i.e., timing of biosynthesis with respect to peak snowmelt and runoff) suggests caution in application of ε wax/streams in paleohydrological reconstructions.These values could also be affected by inclusion of more plant samples and species given the large range of known δD diversity between individuals (Hou et al., 2007b).Average net fractionation between modern vegetation and annual average precipitation (ε wax/precip ) in the El'gygytgyn Basin is slightly more positive (−107 ± 12 ‰).These values are very similar to average net fractionations of −99 ± 8 ‰ reported from a North American transect of C 28 n-alkanoic acids (Hou et al., 2008) suggesting consistent offset between source waters and the δD values of alkanoic acids.
Net fractionation between long chain (C 27 , C 29 and C 31 ) n-alkanes and source water in arid to semi-arid ecosystems also display similar values (−94 ± 21 ‰; Feakins and Sessions, 2010) although n-alkanoic acids are expected to be enriched in D compared to n-alkanes from the same source by about 25 ± 16 ‰ based on their biosynthetic origins (Chikaraishi and Naraoka, 2007).A broad trend to less negative net fractionation values between C 29 alkanes and mean annual precipitation was found globally for drier regions (characterized by relative humidity < 0.7 and evapotranspiration < 1000 mm yr −1 ; Sachse et al., 2012).Our data suggest that net fractionation values for alkanoic acids may not follow the same trend, but larger global data sets for net fractionations for terrestrial n-alkanoic acids are needed.
In a global compilation, Sachse et al. (2012) report significant differences in leaf wax n-alkane δD values among different plant life forms.Liu and Yang (2008) report a high latitude effect with smaller n-alkane ε wax/w values for woody plants versus grasses at high latitudes.Within the El'gygytgyn Basin, Salix and Poaceae comprise the dominant woody and non-woody taxa respectively (Minyuk, 2005;Lozhkin et al., 2007).Grasses (Poaceae), which are dominated by C 3 graminoids at this latitude, produce leaf waxes that are more D-depleted relative to source water likely due to physiological differences in leaf architecture as well as in location and timing of wax synthesis (Hou et al., 2007;Sachse et al., 2012).Although limited, our data also show less negative n-alkanoic acid ε wax/w values for woody plants versus grasses (Table 3).These results suggest that vegetation changes, especially during major climatic shifts, have the potential to affect the overall δD wax signal in catchment sediments.Thus, additional data from other paleoclimate proxies (e.g., pollen assemblages) should be used in combination with molecular paleohydrological interpretations from δD values of terrestrial n-alkanoic acids.

Sediment traps
The abundances and distribution of compounds produced by vegetation directly affect the integrated signal captured in lake sediments, and, thus, any sedimentary isotopic signature could be complicated by the inclusion of compounds from multiple sources with differing isotopic composition.Parallel compound-specific δ 13 C analysis shows that, within El'gygytgyn sediments, nC 30 acid derives from a strictly terrestrial source while other even long chain n-alkanoic acids derive from a mixed terrestrial and aquatic source (Wilkie et al., 2013;Holland et al., 2013).This same study further shows that nC 20 and nC 22 acids likely derive from an aquatic source.However, production of these compounds in modern vegetation (this study; see above) initially suggests some contribution of a terrestrial source of these compounds to lake sediments.This explicitly requires intact delivery of these terrestrially sourced compounds to the lake bottom.Our results show the abundances of these compounds vary with depth in the water column, suggesting multiple processes in the water column influencing the ultimately integrated signal captured in lake sediments.
Even short chain n-alkanoic acids (e.g., nC 14 -nC 22 ) typically produced by phytoplankton and/or bacteria (Cranwell et al., 1987;Gong and Hollander, 1997) are more labile and subject to rapid turnover in the water column (Lee et al., 2004;Jones et al., 2008).Thus, the observed trend in nC 20 acid is consistent with aquatic production in the upper www.clim-past.net/9/335/2013/Clim.Past, 9, 335-352, 2013 surface waters and degradation with depth.Combined terrestrial inputs and aquatic production of even chain nC 20 -C 26 acids likely contribute a higher proportion of these compounds to surface waters, while the sharp decline by 50 m depth could reflect early degradation of the more labile shorter chain nC 22 acid.Although nC 24 and nC 26 acids are typically attributed to terrestrial sources, some aquatic production (i.e., subaquatic/floating macrophytes and filamentous algae similar to those observed in El'gygytgyn streams and lagoons) of these mid-chain fatty acids has been reported (Cranwell et al 1987;Ficken et al., 2000).
Relative abundances of nC 24 and nC 26 acids increase in the deepest water column sample, exceeding values observed in upper surface waters.This suggests either additional input at depth (e.g., delivery of terrestrial material from dense hyperpycnal flows from warming moat and lagoon waters), autochthonous production at intermediate to lower depths in the water column (> 90 m water depth), and/or advantageous capture of sinking organic debris entrained in and subsequently released by melting snow and lake ice.However, delivery of terrestrially sourced C 24 and C 26 acids by hyperpycnal flows is unlikely as this would also result in a concomitant increase in concentrations and relative proportions of nC 28 and nC 30 acids.While we cannot rule out autochthonous preferential production of nC 24 and nC 26 acids below 90 m water depth, capture of previously ice entrained sinking organic debris seems likely during the period of sediment trap deployment (late May to mid-July).Substantial lake ice persisted for most of the trap deployment, with extensive ice melting, development of leads and moat formation occurring throughout sample collection (Fig. 8).Ice-rafted fatty acids (IRFAs), possibly entrained during the previous fall freeze and accumulated over winter through aeolian deposition on lake snow and ice, would be released to the water column sporadically during lake ice melt and breakup.These settling plumes of IRFAs could explain the increase in total concentrations of all fatty acids at 90 m water depth, while shifting proportions with depth could reflect variable melt and release times of the ice over the site.The isotopic ratios of these IR-FAs would still reflect their original biological sources regardless of transport and ice residence times though the signal from compounds of both aquatic and terrestrial sources would be obscured.Further deployment and study of timeseries sediment traps is required to fully assess modern water column processes and contributions to lake floor sediments.

Lake sediments
Lake sediments represent an integrated signal from the watershed, and the carboxylic acids within them typically originate from multiple sources (Meyers, 2003).Concentrations of fatty acids in core top sediments increase exponentially towards the deepest part of the basin consistent with greater accommodation space and lower sediment supply resulting in less dilution of the organic signal.An age model devel-oped on the deepest core from the central part of the lake (LZ1024, 170 m water depth) yields an age of 200 years for the upper 0-1 cm.This sampling resolution for the lake surface sediments implies that seasonal to interannual variations are lost and the "modern" sediments are an integrated average of climate and deposition for the last 200 yr.The consistent distribution of compounds across surface sediment samples regardless of water depth confirms a catchmentintegrated organic signal reaching the lake bottom.Thus, the climatic signal from leaf waxes in lake surface sediments is a combined average for all vegetation changes in the catchment in the last 200 yr (i.e., molecular abundances).Additionally, the sedimentary δD wax also represents an integrated signal of precipitation for the last 200 yr.Longterm δD p records are lacking at this site.However, the reconstructed estimates of δD p from the northern Yakutia δ 18 O cellulose record (Sidorova et al., 2008) yields a 97-yraverage reconstructed δD p of approximately −147 ‰.Net fractionations between core top sediments and either streams (ε C30/streams = −103 ± 5 ‰) or precipitation (ε C30/precip = −95 ± 5 ‰) closely agree with ε wax/w values from modern vegetation (streams: ε wax/streams = −116 ± 12 ‰; precipitation: ε wax/precip = −107 ± 12 ‰).Calculation of these values inherently applies present-day input to a time-integrated signal, yet recalculation with the derived time-averaged δD p still yields near identical net fractionation values (∼ −100 ‰).

Conclusions
This study provides a detailed assessment of the stable isotopic composition of modern hydrology within the El'gygytgyn Basin and the controls on the δD signature from plant leaf waxes within the El'gygytgyn Basin.Our results suggest different source regions for winter and summer precipitation, characterized by an unaltered isotopic signature from a northern and eastern moisture source in winter and a south-southwest continental moisture source including recycled moisture from the land surface.LMWL determined within the El'gygytgyn Basin is consistent with other circumpolar regressions (Kurita et al., 2004;Lacelle, 2011), though the local spatial δ-T relationships are lower than previously reported.However, evaporative enrichment is not a strong control on δD lake and the lake responds relatively quickly to large magnitude changes in δD p .Changes in the length of the residence time are reflected in the degree of dampening of the δD p -driven signal, with longer residence times averaging out shorter-term (e.g., decadal) changes.Inflow from streams in the basin represents soil water (δ soil water δ streams ) that is annually recharged by snowmelt, drained rapidly and relatively unaffected by summer precipitation events.Enhanced evaporative enrichment on southfacing slopes causes δ streams to more closely reflect annual average δD p .Plants utilize this water within the active layer so δD wax should reflect δ streams .However, large interspecies variability and potential changes in seasonality of precipitation suggest caution in applying ε wax/streams in paleohydrological reconstructions.Therefore, it may be more prudent to use ε wax/precip for paleohydrological reconstructions.We observed an offset between woody vs. grass samples though given the small sample set here this might be inconclusive.These results suggest that vegetation changes have the potential to affect the overall δD wax signal in catchment sediments.Larger sample sets including more species and multiple individuals within each genus are needed.In modern lake sediments, ε C30/precip = −95 ± 5 ‰, which is determined only from terrestrially sourced nC 30 acid and represents catchment-wide integrated signal (200 yr of all basin vegetation and δD p ). Utilizing reconstructed (∼ 100 yr average) δD p cellulose to calculate net fractionations for 200yr-integrated lake sediments yields ε 30/precip = −96 ± 8 ‰.This time-averaged ε 30/precip value provides a robust net "apparent" fractionation to be used in future paleohydrological reconstructions.

Fig. 2 .
Fig. 2. Location of the sediment traps deployed in Lake El'gygytgyn in 2003, streams and the δD of sampled stream water.View to the NE with 5x vertical exaggeration.Coloured dots indicate measured δD of stream waters collected in summer 2003.Red line denotes watershed; streams are indicated in light blue lines.Contours indicate lake bathymetry.

Fig. 3 .
Fig. 3. Isotopic response of the lake water to changes in residence time and variable δD p of input (precipitation).Pink lines denotes reconstructed (a) δD and (b) δ 18 O of precipitation from δ 18 O cellulose(Sidorova et al., 2008) based onAnderson et al. (2002).Blue curves show modelled response of δD lake and δ 18 O lake with various residence times (10, 50, 100 and 300 yr).
shows the δD-T and δ 18 O-T relationships for the El'gygytgyn Basin for both winter and summer precipitation (n = 26).Positive spatial relationships (δ 18 O = 0.43 T -18.5, R 2 = 0.68; δD = 3.1 T -143.7,R 2 = 0.64) are observed between the isotopic compositions of precipitation and air temperature.Temperatures used are "precipitation temperature" (i.e., surface air temperature during a precipitation event) as isotopic values are discrete events, not weighted means.The slope coefficients of the δ-T relationships presented here are similar to the spatial δ 18 Omean air temperature (MAT) relation for Taymyr and Lena data (δ 18 O = 0.59 MAT −11.47,R 2 = 0.91;Boike et al., 1997) and data from 47 European IAEA/WMO stations to the west (δ 18 O = 0.59MAT −14.35;Rozanski et al., 1992).However, the slope coefficient for δD-T relationship at

Fig. 5 .
Fig. 5. Relationship between δ 18 O and δD of precipitation with temperature within the El'gygytgyn Basin.

Fig. 6 .
Fig. 6.Concentrations and δD values of n-alkanoic acids measured in modern vegetation samples from within the El'gygytgyn Basin.
Fig. 8. (a) Concentrations of n-alkanoic acids normalized to total yield of n-alkanoic acids by water column depth in Lake El'gygytgyn.(b) Large moat formed by ice melt 500 m to the north of the southern shore on 17 June 2003 -approximately 1/3 way through sediment trap deployment (from Melles et al., 2005).Note extent of snow and lake ice cover during typical summer month.(c) Sediment traps deployed 31 May 2003 and recovered 19 July 2003 (from Melles et al., 2005).

Fig. 9 .
Fig. 9. Average distribution of n-alkanoic acids from modern vegetation within the El'gygytgyn Basin (far left) and in core top sediments (far right).Modelled distributions are shown with varying proportions of vegetation species.Compounds are normalized to total yield of n-alkanoic acids.

Table 1 .
Mean values (δ 18 O, δD, d-excess) for H 2 O sample sets.95 % confidence interval of model values determined from the Online Isotopes in Precipitation Calculator (OIPC), version 2.2 (Bowen and Revenaugh, 2003).b Arithmetic mean of OIPC modelled monthly values for Siberian winter months (November-April).c Arithmetic mean of OIPC modelled monthly values for Siberian summer months (June-August). a The stable isotope ratios r O = 18 O/ 16 O and r H = D/ 1 H are measured for both sample and reference, and the final 18 O and δD values are calculated relative to Vienna Standard Mean Ocean Water (VSMOW)

Table 2 .
Concentrations of selected n-alkanoic acids from vegetation, sediment traps from various water depths, and lake bottom sediment from various depths with the El'gygytgyn Basin.Concentration (µg g −1 TLE)

Table 4 .
δ 18 O and δD values of lake water from Lake El'gygytgyn at various depths(May and August, 2003; November 2008).