The effect of species on lacustrine δ18Odiatom and its implications for palaeoenvironmental reconstructions

The oxygen isotope composition of diatom silica (δ18Odiatom) is increasingly being used to reconstruct climate from marine and lacustrine sedimentary archives. Although diatoms are assumed to precipitate their frustule in isotopic equilibrium with their surrounding water, it is unclear whether internal processes of a given species affect the fractionation of oxygen between the water and the diatom. We present δ18Odiatom data from two diatom size fractions (3–38 and >38 µm) characterized by different species in a sediment core from Heart Lake, Alaska. Differences in δ18Odiatom between the two size fractions varies from 0 to 1.2‰, with a mean offset of 0.01‰ (n = 20). Fourier transform infrared spectroscopy confirms our samples consist of pure biogenic silica (SiO2) and δ18Odiatom trends are not driven by contamination. The maximum offset is outside the range of error, but the mean is within analytical error of the technique (± 1.06‰), demonstrating no discernible species‐dependent fractionation in δ18Odiatom. We conclude that lacustrine δ18Odiatom measurements offer a reliable and valuable method for reconstructing δ18Owater. Considering the presence of small offsets in our two records, we advise interpreting shifts in δ18Odiatom only where the magnitude of change is greater than the combined analytical error.


Introduction
Diatoms are microscopic, unicellular algae with an external shell (frustule) composed of biogenic silica (opal, SiO 2 .nH 2 O) (Round et al., 2007). Distinct diatom communities exist due to variable ecological tolerances and optima (e.g. Moser et al., 1996;Battarbee et al., 2001), making them sensitive indicators of environmental change. Their siliceous frustules can be preserved over long timescales and are generally identifiable to the species level, enabling the reconstruction of past environments well beyond the instrumental record. The silica frustule comprises two layers: a tetrahedrally bonded silica layer (-Si-O-Si) which incorporates oxygen from the surrounding water during formation, and an outer hydrous layer (-Si-OH) which continues to exchange oxygen with the surrounding water following frustule formation . By isolating and extracting oxygen from the internal silica, the isotopic composition of the diatom (d 18 O diatom ) can be determined and used to reconstruct past changes in climate (Leng and Barker, 2006).
The d 18 O diatom value of a diatom frustule depends on the ambient water temperature and isotopic composition of the water (d 18 O water ) in which the diatom grows (Labeyrie, 1974;Juillet-Leclerc and Labeyrie, 1987). By assuming d 18 O diatom fractionates in isotopic equilibrium with this surrounding water, stratigraphic shifts in d 18 O diatom can be used to infer changes in climate over time. The aspect of climate captured by d 18 O diatom depends largely on local hydrology, climate, water residence time and seasonality of diatom growth. Previous studies have interpreted lacustrine d 18 O diatom records to reflect changes in the precipitation/evaporation balance (Rioual et al., 2001;Lamb et al., 2005), shifts in moisture source and regime (Shemesh et al., 2001a,b;Jones et al., 2004;Rosqvist et al., 2004;Leng et al., 2005;Schiff et al., 2009), changes in the amount and seasonality of precipitation (Barker et al., 2001;Morley et al., 2005;Mackay et al., 2013) and, to some extent, changes in temperature (Juillet-Leclerc and Labeyrie, 1987;Shemesh et al., 1995;Brandriss et al., 1998;Moschen et al., 2005). The oxygen isotope composition of marine diatoms has also been used to reconstruct palaeoceanographic conditions, whereby d 18 O diatom is interpreted to reflect changes in sea surface temperature and surface salinity (e.g. Shemesh et al., 1992;Sancetta et al., 1985), sea ice extent and meltwater events (e.g. Shemesh et al., 1995;Swann et al., 2013;Pike et al., 2013), as well as changes in climate-ocean dynamics and ocean productivity (e.g. Juillet-Leclerc and Schrader, 1987;Swann et al., 2006;Romero et al., 2011). d 18 O diatom is therefore increasingly being used as an indicator of palaeoclimate from both marine and terrestrial sedimentary archives (Leng and Barker, 2006;Swann and Leng, 2009).
Following the development of the technique in the 1970s and 1980s (Labeyrie, 1974;Labeyrie and Juillet, 1982), efforts over the last decade have significantly advanced our understanding of d 18 O diatom fractionation, as well as addressed problematic issues of sample contamination and silica maturation (for a review see Leng and Henderson, 2013). As diatoms are ubiquitous in most aquatic environments, d 18 O diatom data are particularly valuable in the absence of other available proxies for isotopic analysis, such as authigenic carbonates and carbonate-based microfossils (e.g. foraminifera, ostracods). Examples of such environments include oceans and lakes in high latitudes where carbonates are often limited or absent (e.g. Rosqvist et al., 1999Rosqvist et al., , 2004Hu and Shemesh, 2003;Schiff et al., 2009;Jones et al., 2004;Swann et al., 2010).

Vital effects on d 18 O diatom fractionation
Diatoms are generally assumed to precipitate their frustules in isotopic equilibrium with the surrounding water (d 18 O water ), and any biological fractionation is presumed to be constant. Furthermore, there is a growing consensus that fractionation between d 18 O diatom and d 18 O water is in the order of c. À0.2‰˚C (e.g. Brandriss et al., 1998;Moschen et al., 2005;Dodd and Sharp, 2010), enabling changes in d 18 O diatom to be quantified over time. However, the d 18 O diatom composition of a diatom frustule often differs from that predicted by thermodynamics, an offset often referred to as a 'vital effect' (or disequilibrium effect) (Leng and Barker, 2006). Such vital effects are well documented and better understood in biogenic carbonates such as foraminifera in the marine environment (Shackleton et al., 1973;Erez, 1978) and ostracods in the lacustrine environment (Xia et al., 1997;Wetterich et al., 2008). In these instances, vital effects may even serve as a valuable palaeoenvironmental indicator, for example to reconstruct ocean thermal stratification (e.g. Fairbanks and Wiebe, 1980;Ravelo and Fairbanks, 1992;Mulitza et al., 1997) or vertical temperature gradients (Tedesco et al., 2007;Steph et al., 2009) due to the varying depths and temperatures occupied by different species. However, the d 18 O of foraminiferal calcite is not only controlled by the physical and chemical properties of the water column (e.g. d 18 O water , temperature, depth), but also by 'species-specific' vital effects: internal processes such as calcification rate (Ortiz et al., 1996), symbiont photosynthesis (Spero and Lea, 1993) and respiration (Wolf-Gladrow et al., 1999) of a given species, which are known to affect isotope fractionation. Some studies have also demonstrated that offsets can be found between the d 18 O of ostracod valves at different stages of growth (e.g. juvenile vs. adult) (Von Grafenstein et al., 1999;Keatings et al., 2002).
Similarly, d 18 O vital effects in diatoms could result from species-specific fractionation, with a varying fractionation factor due to species-dependent d 18 O uptake in equilibrium conditions, or a shortage or excess supply in non-equilibrium conditions (Chapligin et al., 2012). The internal processes within the frustule could theoretically alter d 18 O diatom relative to the isotopic signal of the surrounding water, by an amount that could vary from species to species. However, unlike biogenic carbonates, these effects are poorly understood and much harder to observe, their small size (2-200 mm) making it difficult to isolate individual diatom species for isotope analysis. Isotope measurements are typically made on samples containing many tens of different diatom species. The possibility that different diatom species fractionate oxygen differently is therefore problematic, especially as the abundance of different species changes through time downcore, and it is not possible to analyse the same species in every sample. A number of studies have found no evidence for vital effects between individual diatom species (e.g. Sancetta et al., 1985;Schmidt et al., 2001;Moschen et al., 2005). When small (0.2-0.6‰) offsets have been observed between different species, they are usually within the range of analytical reproducibility (e.g. Shemesh et al., 1995;Brandriss et al., 1998). More recently, however, an offset of c. 3.0-3.5‰ was suggested for two size fractions of marine diatoms in the north-west Pacific (Swann et al., 2007(Swann et al., , 2008. A potential species-driven offset has considerable implications for the applicability of d 18 O diatom as a palaeoclimatic proxy in the marine environment, particularly as this magnitude of variation equates to a 15-16˚C temperature difference. In the north-west Pacific, these offsets were attributed to size-related effects associated with larger diatoms (>75 mm; Swann et al., 2007), although the results were largely inconclusive.
Similar to biogenic carbonates, the habitat in which a diatom grows (e.g. benthic vs. planktonic; water depth) may also affect its isotope composition. For example, the d 18 O of lake water may become more negative with increasing depth if lighter isotopes are preferentially evaporated at the surface (e.g. Von Grafenstein et al., 1999). In addition, water temperature decreases with depth and may also influence the oxygen isotope fractionation of diatom silica (Juillet-Leclerc and Labeyrie, 1987;Shemesh et al., 1992;Brandriss et al., 1998;Moschen et al., 2005), which must be considered when interpreting records of d 18 O diatom . An offset between diatom species could also derive from the timing of the diatom bloom in the water column and whether the d 18 O diatom signal captured has a seasonal component (e.g. spring vs. autumn; temperature and precipitation). The interpretation of d 18 O diatom from any sedimentary record therefore requires an understanding of the external processes affecting d 18 O diatom at a particular site, such as climate, water inflow and outflow, lake residence time and diatom ecology.
Here, we examine the issue of species-dependent vital effects in d 18 O diatom using lacustrine sediment material from Heart Lake on Adak Island, part of the Aleutian Island chain in south-west Alaska. By analysing the d 18 O diatom value of two size fractions containing different diatom species, we investigate whether species-specific fractionation of oxygen is evident in a lacustrine environment.

Study site
Heart Lake, located on Adak Island in the north-west Pacific (51.87˚N, 176.63˚W; Fig. 1), is a small (<0.25 km 2 ) freshwater, through-flow lake. Separating the Bering Sea to the north and the Pacific Ocean to the south, the island experiences a cool, wet and windy maritime climate (Rodionov et al., 2005). Mean precipitation in January and July is 150.4 and 71.6 mm, respectively   (NOAA, 2013

Methodology
Sample material A 5.5-m-long sediment core (AS-10-1D) was recovered from Heart Lake in July 2010. A GPS-enabled sonar depth recorder guided bathymetric profiling, and the core site was selected adjacent to the deepest part of the basin at 7.6 m depth. The core was recovered using a percussion corer operated from a floating platform. The core was split, photographed and described (Krawiec, 2013), with 20 samples taken downcore for d 18 O diatom and diatom species analysis. The samples range in age from 8.0 to 1.0k cal a BP based on a best-fit age-depth model constructed using accelerator mass spectrometry 14 C ages and correlated tephra ages from nearby Andrew Lake .

Sample preparation
Samples for d 18 O diatom were prepared using a stepwise process of chemical digestion, differential settling, sieving and heavy liquid separation based on Morley et al. (2004 reactions ceased (to remove organic material), before using 5% HCl to eliminate any carbonates. Following differential settling, all samples were centrifuged in sodium polytungstate (3Na 2 WO 4 9WO 3 .H 2 O) (SPT) heavy liquid at 2500 r.p.m. for 20 min, resulting in the separation and suspension of diatoms from the heavier detrital residue. This process was repeated three times using specific gravities of 2.50, 2.30 and 2.25 g mL À1 . After the final SPT separation, samples were washed five times at 1500 r.p.m. for 5 min to remove any remaining SPT. Purified diatom samples were then sieved at 38 and 3 mm, resulting in two size fractions of 3-38 and >38 mm for d 18 O diatom analysis. The <3 mm fraction was discarded as it was too small ($1 mg) to be analysed. The remaining fractions (3-38 and >38 mm) of the sample were then treated with an additional stage of 30% H 2 O 2 at 70˚C for 1 week and centrifuge-washed to ensure no traces of organic matter remained.
Sub-samples of the purified diatom material were retained and mounted on cover slips using Naphrax 1 , Brunel Microscopes Ltd. The samples were visually inspected for contamination before diatom assemblage analysis using light microscopy (300 frustules per sample). Standard diatom preparation and analysis (Battarbee et al., 2001) was also performed on the unprocessed lake sediment to identify the diatom assemblages present and to determine whether any species were lost during preparation for diatom isotope analysis.

Contamination assessment
Fourier transform infrared spectroscopy (FTIR) was used as a rapid, non-destructive means to assess the chemical composition of biogenic silica within our samples , and ultimately sample purity (see Swann and Patwardhan, 2011). Sixteen purified diatom samples, together with the within-run laboratory diatom standard (BFC mod ), were analysed using FTIR (Fig. 3). FTIR produces energy absorption spectra of a sample over a range of wavelengths. Individual absorption peaks correspond to specific chemical bonds and compounds, with pure silica displaying peaks in two distinct regions at >2500 cm À1 (hydroxyl) and <1300 cm À1 (silica) (Fidalgo and Ilharco, 2001;Patwardhan et al., 2006;Swann and Patwardhan, 2011). FTIR analyses of all purified diatom isotope samples indicate peaks corresponding to the BFC mod standard, known to represent clean, fossillized diatomite (Fig. 3). Spectral deviation from the standard would indicate additional compounds within the sample and contamination by non-diatom components (Swann and Patwardhan, 2011). Peaks centred at $450, $800 and $1100 cm À1 indicate pure silica and suggest the samples comprise purely diatoms. Additionally, scanning electron microscopy (SEM) was used to check sample purity before oxygen isotope analysis (Fig. 4).

Oxygen isotope analysis
Purified diatom samples were analysed for d 18 O diatom at the NERC Isotope Geosciences Laboratory (UK) using a stepwise fluorination method outlined in Leng and Barker (2006). The outer hydrous layer of the diatom, known to freely exchange isotopically with water (e.g. Juillet-Leclerc and Labeyrie, 1987), was removed in a pre-fluorination stage using a BrF 5 reagent at low temperature. This was followed by a full reaction at high temperature to liberate oxygen that was converted to CO 2 (Clayton and Mayeda, 1963) and measured for d 18 O diatom using a MAT 253 dual-inlet mass spectrometer. All d 18 O values were converted to the VSMOW scale using the within-run laboratory standard BFC mod , and are reported here in per mil (‰).

Diatom assemblages
Diatom frustules are well preserved in all 60 samples as indicated by SEM images that show no evidence of valve dissolution (Fig. 4). The smaller size fraction (3-38 mm) is taxonomically diverse, with 120 diatom species identified. The flora is dominated by seven species (Psammothidium levanderi, Rossithidium pusillum, Cyclotella rossii, Pseudostaurosira brevistriata, Staurosira construens, Stauroforma exiguiformis and Staurosirella pinnata) with a combined relative abundance of 76% across all samples (Fig. 4B). Subtle shifts  in species assemblages are evident among the samples, with the most notable transition at c 320 cm depth, when the abundance of P. levanderi and C. rossii decline markedly and the abundance of the small Staurosira/Staurosirella species increasingly dominate. In the larger size fraction (>38 mm), 28 different diatom species were identified. The assemblages are dominated by six species (Didymosphenia geminata, Surirella robusta, Surirella splendida, Pinnularia turnerae, Rhopalodia gibba and Campylodiscus hibernicus), with a combined relative abundance of 85% across all samples (Fig. 4A). The diatom assemblage in this larger fraction is distinctly different from that of the smaller size fraction; they are not merely larger specimens of the same species. The transition at $320 cm depth is also evident in the larger size fraction, with a marked increase in D. geminata, P. turnerae and R. gibba, which replace C. hibernicus and Surirella species. The unprocessed bulk sediment samples are composed of the same diatom assemblages found in both the purified d 18 O diatom size fractions (Fig. 4C). However, the larger >38-mm diatoms only represent $1% of the relative abundance across all bulk sediment samples.
The d 18 O diatom values from Heart Lake range from þ28.8 to þ33.4‰ (Fig. 5). Comparing the two size fractions, d 18 O diatom values differ by 0-1.2‰ (r 2 ¼ 0.75, P < 0.05), with a mean difference of À0.01‰. Duplicate analyses of d 18 O diatom indicate an analytical reproducibility (1s) of AE 0.19‰ for the smaller (3-38 mm) fraction, AE 0.49‰ for the larger (>38 mm) fraction and AE 0.31‰ for the BFC mod laboratory diatom standard. All 20 pairs have d 18 O diatom values within the combined analytical uncertainty of AE 1.06‰ (2s) for the two size fractions. Neither fraction is consistently more isotopically positive or negative relative to the other, and the two values are similar down core, aside from two samples at 147.5 and 161.5 cm depth where they diverge.
The relationships between d 18 O diatom and diatom assemblages were evaluated using principal components analysis (PCA) (Fig. 5B, C) (ter Braak and Prentice, 1988). A PCA was applied to a correlation matrix based on the dominant diatom species in all 20 samples, in both size fractions. The stratigraphic changes are captured in the first and second PCA axes, which account for 49.9 and 29.9% of variance in   (Fig. 5B), and 55.3 and 17.0% in the large fraction (Fig. 5C).

Discussion
All 20 pairs of diatom size fractions have differences in d 18 O diatom within the 2s uncertainty range of this technique. Furthermore, the mean difference between the two data sets is close to zero (m ¼ À0.01‰) and indicates the two d 18 O diatom records are not statistically different (Fig. 6).
The diatom assemblages from the two size fractions are composed of entirely different species, and the relative abundance of each species varies through time (Fig. 4). The main growing season of diatoms identified in Heart Lake occurs in spring following winter snow melt, when sediment and nutrient input to the lake is high and temperatures begin to increase. While some species such as D. geminata are known to reside in freshwaters year-round, the main bloom occurs in late spring and summer (Whitton et al., 2009). C. hibernicus, which is dominant in the lower section of the core, also blooms in both spring and autumn (Griffiths, 1923;Ramrath et al., 1999). Autumn diatom blooms are typically caused by the breakdown of summer stratification and entrainment of nutrients while there are still sufficient light levels for growth (Round et al., 2007). Temperatures in Heart Lake are consistently low year-round and the lake water is well mixed, meaning autumn blooms are unlikely to occur. As a result, both size fractions of the diatom sample capture the spring/summer d 18 O diatom signal when their silica frustule is formed (Moschen et al., 2005), and this rules out possible 'seasonal-effects' on d 18 O diatom values.
Aside from the planktonic/tychoplanktonic species C. rossii, all the dominant diatom species in Heart Lake are generally benthic and occupy the same habitat and pool of water (d 18 O water ). Heart Lake is 7.6 m deep and is likely to be well mixed. Given the similarity between the d 18 O value of precipitation and lake water at Heart Lake, isotopic enrichment due to evaporation is insignificant. Any so-called water column effect is probably only applicable to deeper lakes than Heart Lake, or within the marine environment where there may be variations associated with different water masses. We can therefore assume all pairs of diatom fractions analysed for d 18 O diatom formed their silica frustules under the same environmental conditions (i.e. depth, temperature, d 18 O water ). Although there are subtle differences between species habit, with some being solitary (e.g. C. hibernicus), some colonial (e.g. D. geminata), others motile (i.e. C. rossii), attached to substrata (e.g. R. gibba) or a combination of the above, these attributes appear insignificant given there is no discernible difference in d 18 O diatom .
Evidence of a size-related species effect on d 18 O diatom has previously been documented in the marine environment, although these results are rather inconclusive. Swann et al. (2007) report more positive values of d 18 O diatom in smaller diatoms compared with larger ones, but further research suggested the opposite (Swann et al., 2008). Diatom size is also inherently linked to growth rate, with most diatoms exhibiting a gradual reduction in size/growth with increasing maturity and successive cell division (Round et al., 2007). We cannot quantify the growth rate of specific diatoms within our sediment record, but we note in each size fraction the diameter of any given species does not vary visibly (Fig. 4). While it would be incorrect to assume growth rates are consistent across all species, there is no evidence for a relationship between diatom size and the amount of fractionation in our samples, with no one size fraction consistently more positive or negative in d 18 O diatom relative to the other (and within analytical error).
Visual inspection of all samples by light microscopy and SEM revealed no obvious sign of contamination (e.g. SPT, minerals, tephra), which is further confirmed by FTIR analysis. As the fluorination process will liberate oxygen from any oxygen-bearing mineral in the sample (Brewer et al., 2008), having ensured the diatoms are clean and free from contaminant, we consider the d 18 O diatom data to be reliable.

Species-specific effects
In the large size fraction (>38 mm), the diatom species C. hibernicus and R. gibba show the strongest correlation with downcore variation in d 18 O diatom , in positive and negative associations (i.e. with more positive d 18 O diatom , the abundance of C. hibernicus increases and the abundance of R. gibba decreases) (Fig. 5C). In the small (3-38 mm) diatom fraction, C. rossii and S. construens are most closely related with d 18 O diatom , in positive and negative associations (Fig. 5B). The remaining dominant diatom species, in both size fractions, appear unrelated to the d 18 O diatom vector in the PCA, with several species being orthogonal to the d 18 O diatom gradient (D. geminata, S. splendida, P. turnerae, P. brevistriata, R. pusillum; Fig. 5B, C). Given only two different species drive $50% of the variance in each size fraction, and there is no discernible difference in the d 18 O diatom signal from these two fractions, we therefore find no evidence to suggest there is a species-driven effect controlling d 18 O diatom . C. hibernicus disappears from Heart Lake above 254.5 cm, but there is no evidence of a concurrent shift in the d 18 O diatom record at this time. The data therefore suggest stratigraphic shifts in diatom assemblages are ecological responses to climatic and environmental changes, as well as in the d 18 O diatom record, rather than driving the isotopic signal thorough differences in species-specific fractionation. Determining the precise environmental and ecological factors driving species assemblages and changes to d 18 O diatom is, however, beyond the scope of this paper.
The diatom composition of the two size fractions analysed here represents different species assemblages and are considered independent of each other. If species-dependent vital effects were present, we would expect the d 18 O diatom data for each size fraction to deviate and be consistently offset from one another. While the data do not establish whether diatoms precipitate their silica in isotopic equilibrium with lake water, they demonstrate different species of diatoms fractionate oxygen isotopes at a similar magnitude.
Conclusions d 18 O from diatom silica is generally presumed to precipitate in isotopic equilibrium with the surrounding water, but the presence of species-dependent vital effects on fractionation has, until now, been unclear. Our d 18 O diatom data from Heart Lake reveal only small differences (0-1.2‰, n ¼ 20) between two size fractions containing different diatom species assemblages. Given all differences are within the combined analytical error of the technique, it suggests there is no species-or size-related effects controlling fractionation of d 18 O diatom and bulk d 18 O diatom samples are suitable for investigating palaeoenvironmental change at Heart Lake.
Diatom species analyses of both the raw sediment and the purified diatom samples reveal some diatom species were lost during preparation for diatom oxygen isotopes. However, considering these species account for <1% of overall abundance across all samples, we conclude the purified samples analysed for d 18 O diatom are representative of the species found within the raw sediment. It is therefore advised that samples for d 18 O diatom follow the same rigorous preparation and analytical procedures employed here.
d 18 O diatom measurements from lacustrine diatom silica are a reliable and valuable method for reconstructing past d 18 O water . As diatoms are found in nearly all aquatic environments, d 18 O diatom records offer an important source of information in regions devoid of other proxies available for isotopic analysis (e.g. carbonates), such as in the high-latitude regions. Considering the presence of small offsets in our two records, we advise interpreting shifts in d 18 O diatom only where the magnitude of change is greater than the combined analytical error for those samples.