Field work was conducted during the “Yakutia 2013” German–Russian expedition between 19 March and 14 April 2013 by the Alfred Wegener Institute Helmholtz Centre for Polar and Marine Research (AWI) and North-Eastern Federal State University in Yakutsk (NEFU). Vertical holes were drilled in the lake ice cover using a Jiffy ice auger with a diameter of 250 mm. Lake basin bathymetry was measured using a portable echo sounder. Ice cores were retrieved by drilling multiple holes around a central part. Water samples for hydrochemical analyses were collected prior to sediment coring using a UWITEC water sampler. Water samples were analysed in situ using a WTW Multilab 340i for pH, conductivity, and oxygen values at the day of retrieval during field work. A sub-sample of the original water was passed through a 0.45 µm cellulose–acetate filter, stored, and transported in 60 mL Nalgene polyethylene bottles for subsequent anion and cation analyses in AWI laboratories in autumn 2013. Cation samples were acidified during field work with HNO3, suprapure (65 %), to prevent microbial conversion processes and adsorptive accretion.

At 42 sites within the lake, short cores containing intact sediment surface material were retrieved using an UWITEC gravity corer. Water depth at sampling sites was measured using either a hand-held HONDEX PS-7 LCD digital sounder and/or the cord of the coring device when the lake ice cover disturbed the signal. The sediment was identified as clayish silt deposits with a predominantly dark (black) colour and a weak smell of hydrogen sulfide, a sticky and viscous mud mixed with plant and other organic residues. The uppermost ca. 2 cm at some sites had a dark red colouring indicating the redox boundary between oxygenated and anoxic sediments. We identified the uppermost 0.5 cm of short cores as surface sediments and subsampled these layers on site during fieldwork to avoid sediment mixture during transport. Sediment samples were transported in sterile Whirl-Pak bags and sediment cores were transported in plastic liners to the AWI laboratories in Potsdam, Germany, and stored at 4 ∘C in a dark room for further analyses and as back-up.

During this expedition long core material was also retrieved from multiple sites, including the northern and central parts of the lake, and is planned for publication in a separate paper.

According to the amount of the uppermost 0–0.5 cm layer in the short cores available, the sample size n for different sediment properties measured in this study varies.

Detrended correspondence analysis (DCA) with detrending by segments was performed on the chironomid and diatom data (rare taxa down-weighted) to determine the lengths of the sampled environmental gradients, from which we decided whether unimodal or linear statistical techniques would be the most appropriate for the data analysis (Birks, 1995). For diatom data the gradient lengths of the species scores were 2.07 and 1.49 standard deviation units (SDU) for DCA 1 and 2, respectively, suggesting that lineal numerical methods should be used. A principal component analysis (PCA) was used to explore the main taxonomic variation of the data (ter Braak and Prentice, 1988). The gradient lengths of chironomid species scores were 3.78 and 4.12 SDU, indicating that numerical methods based on a unimodal response model should be more appropriate to assess the variation structure of the chironomid assemblages (ter Braak, 1995). However, test PCA performed on chironomid data showed that the lineal method captures more variance of species data (ESM, Table 2); therefore, we further applied lineal methods for both chironomid and diatom data. In order to summarize the response of lacustrine biota to abiotic, physicochemical explanatory variables, a redundancy analysis (RDA) was performed on diatom and chironomid data in comparison to environmental variables (Figs. 2 and 3).

Initially, all environmental variables shown in this paper were tested in a RDA to assess the relationships between the distribution of bioindicator taxa and abiotic habitat parameters. Apart from the chemical and physical parameters of the lake and sediments (Fig. 5), we include in the analysis the presence/absence of the submerged vegetation and distances of the sampling stations from the shore and from the inflowing rivers. All explanatory variables were tested for normality prior to the analyses. Skewness reflects the degree of asymmetry of a distribution around its mean. Normal distributions produce a skewness statistic of about zero. Values that exceeded 2 standard errors of skewness were identified as significantly skewed (Sokal and Rohlf, 1995). Environmental variables with skewed distributions (gravel, grain-size EM2, smectite–chlorite, mica, K-feldspar) were log transformed and remaining parameters were left untransformed. To reveal intercorrelated parameters, we performed a variance inflation factor (VIF) analysis prior to ordination techniques to only retain non-correlated parameters in further multivariate analysis. Environmental variables with a VIF greater than 20 were eliminated, beginning with the variable with the largest inflation factor, until all remaining variables had values <20 (ter Braak and Šmilauer, 2012). A set of RDAs was performed on chironomid and diatom data with each environmental variable as the sole constraining variable. The percentage of the variance explained by each variable was calculated and the statistical significance of each variable was tested by a Monte Carlo permutation test with 999 unrestricted permutations. Significant variables (P≤0.05) were retained for further analysis. DCA, PCA, and RDA were performed using CANOCO 5.04 (ter Braak and Šmilauer, 2012).

Percentage abundances of the chironomid taxa that are absent or rare in the modern calibration data set were calculated at each sampling site in order to see the distribution of the taxa that could potentially hamper a T July reconstruction in the case of a palaeoclimatic study that could be done at each of the sampling sites. It is known that less reliability should be placed on the samples in which more than 5 % of the taxa are not represented in the modern calibration data or more than 5 % of the taxa are rare in the modern calibration data set (i.e. if the effective number of occurrences in the training set, Hill's N2, is less than five) (Heiri and Lotter, 2001; Hill, 1973; Self et al., 2011).

Species richness and the Simpson diversity on diatom and chironomid data were estimated after sample-size normalization using a rarefaction analysis of Hill numbers in the iNEXT package in R.

To assess the relative contribution of different sedimentary processes to the bulk sediment, such as fluvial or aeolian transport (Wang et al., 2015; Biskaborn et al., 2013b), a statistical end-member analysis on grain-size data was performed using the MATLAB modelling algorithm of Dietze et al. (2012). In this method, individual grain-size populations identified as end-member loadings (vol %, Fig. 4) as well as their contributions to the bulk composition identified as scores (%) were derived by eigenspace analysis, weight transformation, varimax rotations, and different scaling procedures.

A Pearson correlation matrix of the main important variables (Fig. 5a) was calculated using the basic R core (R Core Team, 2012) and plotted using corrplot. To keep the false discovery rate below 5 %, a p-value adjustment was applied prior to assignment of colours using only values that revealed p<0.001 (Colquhoun, 2014). To identify the pattern, the correlation matrix was reordered according to the correlation coefficient. Exceptional sites within the heterogenic lake system lead to disturbance of good correlation coefficients within areas along natural borders, e.g. water depth isobaths. Spatial autocorrelation of variables was estimated using latitudes and longitudes recorded of each sample site and displayed as p values generated by Moran's autocorrelation coefficient (R package “ape”).

To guarantee the sustained availability of our research (Elger et al., 2016), the data used in this study are freely accessible at the PANGAEA data repository (10.1594/PANGAEA.906317).

Sampled surface waters of Bolshoe Toko (Table 1, ESM) were well saturated in O2 (101 %–113 %) with a pH value in the neutral range (6.8–7.2). Electrical conductivity was very low for all waters (35.1–39.1 µS cm-1), with slightly higher levels in the lagoon (67.8 µS cm-1). Traces of Al (mean 72 µg L-1), Fe (mean 46.6 µg L-1), and Sr (mean 37.1 µg L-1) were present, but there is no evidence of Pb, Cr, V, Co, Ni, or Cu. Mean sulfate concentrations (SO42-) were 2.35 mg L-1 on average, with lower values in the lagoon (0.51 mg L-1). The concentrations of nitrate (NO3-) were 0.76 mg L-1, but were lower in the lagoon (0.29 mg L-1). HCO3- was 37.5 mg L-1 in the lagoon and 14.9 mg L-1 on average in the remaining samples. There was no phosphorus in any sample. Overall the water can be characterized as water of the Ca–Mg–HCO3- type.

Surface waters were characterized by mean isotope values of -18.7 ‰, -140.2 ‰ and 9.5 ‰ for δ18O, δD and d excess, respectively (n=6). The isotopic composition was relatively uniform in the main lake basin (δ18O = -18.58±0.15 ‰, δD = -139±0.7 ‰), while the lagoon (PG2122) exhibited slightly lower δ18O (δD) values of -19.2 ‰ (-145 ‰) (Fig. 6).

In March 2013, isotope-depth profiles at PG2208 exhibited a slight isotopic enrichment trend from the surface to ∼5 m depth (∼+0.35 ‰ for δ18O), with a relatively uniform isotopic composition (δ18O = -18.2±0.2 ‰) below 10 m (Fig. 6a). These subtle variations likely reflect minor isotopic fractionation of surface waters during ice formation in spring and a well-mixed water column below. Conversely, the August 2012 depth profile at the western shoreline exhibited a gradual depleting isotope trend below ∼6 m depth, with marked variability that closely tracks water temperature changes (Fig. 6a). Meteorological data from the nearby weather station (Toko RS, 10 km northward) recorded heavy rainfall for August 2012 (25 mm above the long-term mean of 83 mm). Such precipitation events could cause temporary isotopic stratification or a variation in the isotopic signal throughout the water column. Due to ongoing mixing, these variations were then evened. In conclusion, variations in the isotopic composition throughout the August profile rather represent a temporal phenomenon and are not characteristic of Bolshoe Toko. In contrast, the lagoon showed a lighter isotope composition (δ18O = -18.9±0.2 ‰) than the main lake basin. All samples were positioned close to the Global Meteoric Water Line (GMWL, Fig. 6), indicating negligible evaporative effects on lake water isotope composition and a dominant influence of meteoric inputs both directly (i.e. precipitation) and indirectly (i.e. river inflows). The Local Meteoric Water Line for Yakutsk (dashed line; δD = 7.59⋅δ18O - 6.8), based on own data (monthly mean precipitation values between 1997 and 2006; n=106; from Kloss, 2008), is given for comparison and is indicative for more continental climate conditions.

The typical surficial lake bottom sediments consisted of either brown organic-enriched gyttja or sandy, organic-poor siliciclastic material. Sand contents ranged between 10.2 % and 96.2 % (mean 45.9 %, Fig. 7); silt contents ranged from 3.6 % to 83.3 % (mean 47.1 %); clay contents ranged from 0.2 % to 11.3 % (mean 5.8 %). Gravel was found only in four samples in the north-eastern near-shore areas, with contents of up to 13.1 %. The mean grain size ranged from 12 to 479 µm (mean 72 µm). The mean grain size generally correlated negatively with water depth (r -0.45). Mineral grains are composed mainly of quartz (32.7 %–76.2 %, mean 55.4 %), plagioclase (13.4 %–39.5 %, mean 26.2 %), K-feldspar (0.0 %–9.8 %, mean 5.6 %), and, to a smaller degree, pyrite (0.2 %–5.5 %, mean 3.3 %), hornblende (0.5 %–10.8 %, mean 3.1 %), mica (0.3 %–2.4 %, mean 1.1 %), and the clay minerals smectite, kaolinite, and chlorite (together 0.0 %–4.6 %, mean 2.0 %). The spatial distribution of minerals (Fig. 7) revealed a generally decreasing gradient of minerals relative to quartz starting from the Utuk River delta (proximal) towards the northern areas (distal).

The CLR-transformed XRF data (Fig. 8) revealed high proportions of Zr and intermediate to high Ti near the Utuk River inflow and at the northern and eastern shore proximal areas. Zr values decreased with increasing water depth towards the lake centre with the exception of the shallow lagoon, where low values were observed. Mn values were highest in the lake centre and at the very deep site at the western steep subaquatic slope, and intermediate at shallow areas close to the shore. A minimum in Mn was found in the lagoon. Fe tends to be highest in the southern part of the lake basin, in the very shallow site in the north, and in the lagoon. Br showed a variable distribution; however, high values were found at two sites within the eastern lagoon and correspond to high TOC contents.

Additive log ratios (ALR) of Mn/Fe were variable, with intermediate values found at sites surrounding the Utuk River inflow and low values within the lagoon and at basin central sites. High values were located at the deepest lake site as well as in the shallow north-eastern region. Both Sr/Rb and Zr/Rb ratios showed high values directly in front of the Utuk River inflow and decreased with distance toward the basin centre. Both Sr/Rb and Zr/Rb exhibited intermediate to high values in the north-eastern lake region and lower values in the lagoon. Si/Ti ratio values demonstrated an increasing trend from the southern lake region and lagoon to the northern lake region.

The contents of total organic carbon (TOC, Fig. 9) range from 0.1 % to 12.3 % (mean 4.9 %). Maximum values occurred in the eastern area, intermediate values in the central basin, and lowest values in the northern shallow water areas. The difference between TOC and total carbon is within the error of the devices, and hence no inorganic carbon was detected. TOC contents and the TOC/TN ratios were highest near the Utuk River inflow in the southern part of the lake, in the lagoon, and in proximity to the eastern shoreline. δ13C was measured in 15 samples and showed maximum values at the eastern shore (-25.7 ‰) and minimum values elsewhere (-27.8 ‰).

Radiocarbon dating of a surface sample at site PG2139 (0–0.5 cm) indicated an age of 720±30 14C yr BP (Lab-ID: Poz-105350, NaOH-SOL), while PG2207 (0–0.5 cm) suggested 1790±130 14C yr BP (Lab-ID: Poz-105355, NaOH-SOL). Considering that the carbon concentration dissolved in sample PG2207 was too low (0.03 mgC), we use sample PG2139 as an estimated reservoir effect on the lake caused by the input of old carbon. Given that a hypothetical sediment surface is just a momentum only collectable as a range of past surfaces and there was more time available for radioactive decay at 0.5 cm depth than at 0 cm, the actual reservoir effect will be a little bit lower and should be confirmed by 210Pb and 137Cs measurements of downcore material before establishing an age–depth model for sediment cores.

The Bolshoe Toko diatom assemblages were characterized by boreal and arcto-alpine types and exhibited distinct spatial variations across the lake. In total, 142 different diatom taxa were found at 23 sites, dominated by planktonic species Pliocaenicus bolshetokoensis (Genkal et al., 2018) (0.0 %–27.9 %, mean 14.7 %), Cyclotella comensis (0.0 %–23.1 %, mean 10.9 %), and benthic species Achnanthidium minutissimum (0.0 %–38.0 %, mean 11.8 %). The relative content of planktonic species (Fig. 9) was 2.0 %–73.7 % (mean 54.2 %), epiphytic species 19.2 %–83.9 % (mean 36.4 %), and epibenthic species 2.6 %–23.0 % (mean 9.3 %). The spatial distributions of the main taxa are presented in Fig. 10. Small benthic fragilarioid species were represented by 0.0 %–27.6 % (mean 7.4 %), Naviculoid species ranged from 3.3 % to 12.9 % (mean 7.2 %), and Aulacoseira species ranged from 0.0 % to 10.8 % (mean 4.5 %). Pliocaenicus bolshetokoensis maximum abundance occurred in areas of deepest water such as the southern part of the lake and in the eastern lagoon. Cyclotella species were more abundant in the central lake and were not as strictly bound to water depth as Pliocaenicus. Aulacoseira species displayed no clear spatial pattern, but were less abundant in the northern shallow water areas. Tabellaria species were more abundant in shallow near-shore areas than in central and deep-water areas.

Achnanthoid (monoraphid) species were most abundant in near-shore areas, especially near the eastern lake terrace. Fragilarioid (araphid) species were common in the southernmost part near the inflow, as well as the lagoon. Other benthic species, i.e. Navicula, Cymbella, and Eunotia, were generally more abundant in shallow near-shore areas than in deeper water areas.

In pelagic areas planktonic diatoms were generally more abundant than epiphytic and epibenthic species. Epiphytic species, however, predominated in some shallow areas in the northern and eastern parts of the lake. Epibenthic species occurred in smaller abundances in shallow lake littorals. Together with an increased amount of non-planktonic species, the Simpson diatom species diversity was higher in the northern and eastern parts of the lake. The chrysophyte index was high near the river inflow in the south and along the river-like bathymetrical structure, as well as the lagoon where another small river inflowed into the lake. The Mallomonas index, reported for high nutrients and low pH (Smol et al., 1984), was highest near the inflow and in the central part and lowest in near-shore areas in the north and east. The maximum f-index value, representing the highest valve preservation, was found in the near-shore areas, whereas lower values were found at the shallow bathymetrical structure in the central part of the lake. Maximum valve concentrations were observed in the central and northern lake basin.

The initial RDA with all environmental variables indicated that axes 1 and 2 explain 39.6 % of variance in diatom species data. After deleting all intercorrelated variables, 13 parameters with VIFs <20 were left for manual selection with Monte Carlo test. The analysis revealed eight statistically significant (p≤0.05) explanatory variables: TOC/TN, TOC, water depth, distance from river, distance from the shore, presence of vegetation, sand, and EM3, (ESM diatoms, Fig. 2). Eigenvalues for RDA axes 1 and 2 constrained by eight significant environmental variables constitute 81 % and 59 %, respectively, of the initial RDA, suggesting that the selected significant variables explain the major variance in the diatoms data. The RDA biplots of the species scores and sample scores (Fig. 2) show that diatom species and sites are grouped according to the main environmental forcing responsible for their spatial distribution. The clearest environmental signals in the RDA are related to water depth, habitat preferences and river influence. The upper left quarter of the biplot is strongly influenced by water depth, grain size (EM3), and the ratio between TOC and TN. The species found next to water depth are planktonic Cyclotella taxa, whereas Aulacoseira is closer to TOC/TN and the total carbon content. In the lower right quarter epiphytic and benthic taxa prevail, i.e. achnanthoid, naviculoid, and cymbelloid taxa, associated with the presence of vegetation and coarser (sand) substrate conditions. The distances to river and to shore are crossing the lower left quarter and are associated with different planktonic Cyclotella and achnanthoid taxa, while in the opposite direction, with increasing Utuk River influence, fragilarioid taxa, Eunotia, Tabellaria, and Gomophonema prevail, next to the high influence of TOC/TN.

Mean surface sediment δ18Odiatom was 22.8 ‰ (minimum 21.9 ‰, maximum 23.6 ‰, n=9, Fig. 9) with a standard deviation of ±0.6 ‰ (1σ). The spatial distribution indicated higher values ∼23.3 ‰ in the deeper south-western part of lake (PG2113, 2115, 2005) and lower values ∼22.3 ‰ in the shallower northern lake basin (PG2139, 2140, 2147, 2209). The two samples from the lagoon exhibited values of 22.2 ‰ in the shallower northern area and 23.6 ‰ in the deeper part. Four samples from the southern part could not be purified well enough and had contamination corrections >2 ‰.

A total of 79 different chironomid taxa were present in the surface sediment samples, of which 48 belong to the subfamily Orthocladiinae, 25 to Chironominae (15 from the triba Tanitarsini and 10 from the triba Chironomini), 4 to subfamily Diamesinae, and 2 to Tanypodinae.

The initial RDA with all environmental variables shows that axes 1 and 2 explain 46.7 % of variance in the taxon data. Most of the environmental parameters were intercorrelated, and following sequential deletion of all redundant variables, eight parameters with VIFs <20 remained for the further analysis. The manual Monte Carlo test selection demonstrates four statistically significant (p≤0.05) explanatory variables: TOC/N, water depth (WD), distance from river, and presence of vegetation (Table 2). Distance from the river and presence of vegetation showed lower than TOC/N and WD level of significance. However, we still use these parameters for interpretation of the chironomid data, as there was a clear gap between the four chosen parameters (p=0.001 to 0.059) and much higher p values (>0.25) of the following tested parameters (TC, distance to the shore, silt, clay). Eigenvalues for RDA axes 1 and 2 constrained by four significant environmental variables were 0.200 and 0.150, respectively, and constituted 70 % and 85 % of the RDA performed on all environmental variables (0.289 and 0.177, respectively). This minor difference suggests that the four selected variables sufficiently explain the major gradients in the chironomid data.

The RDA biplot of the sample scores shows that sites are grouped by their location in relation to the major environmental variables (Fig. 11), and distribution of chironomid taxa along the RDA axes reflects their ecological spectra. Figure 11 and Table 6 in the ESM show median values of eco-taxonomical chironomid groups and their relation to environmental parameters.

Sites most strongly influenced by the inflowing rivers grouped in the lower left quadrant of the biplot, as the vector in the upper right quarter shows increasing distance from the river mouth. In total 64 chironomid taxa were found in this group of sites, and of these 33 were only found here. Chironomid fauna were chiefly represented by phytophilic littoral taxa from the Orthocladiinae genera Cricotopus, Orthocladius, Eukiefferiella, and Parakiefferiella etc. (Fig. 11). Another important feature is the presence of a relatively high amount of lotic environmental taxa, among which are several Diamesa taxa, Rheocricotopus effusus-type, Synorthocladius, Brillia, and for lotic–lentic environments Parakiefferiella bathophila-type, P. triguetra-type, Nanocladius rectinervis-type, N. branchicolus-type, several Eukiefferiella taxa, and Stictochironomus.

The group in the opposite upper right quadrant represents the northern part of the lake situated far from the inflowing rivers. Here, mainly profundal taxa prevail, i.e. Procladius, Polypedilum nubeculosum-type, Cryptochironomus (eurytopic), and Heterotrissocladius maeaeri-type 1 (acidophilic).

The lower right group of sites represent eastern shallow littoral with presence of macrophytes. Species richness and proportion of semiterrestrial and littoral taxa in this group is generally low. Littoral taxa were generally phytophilic: Cricotopus intersectus-type, C. cylindraceus-type, Dicrotendipes nervosus-type (mesotrophic), and Cladotanytarsus mancus-type and Psectrocladius sordidellus-type (acid-tolerant mesotrophic). Most abundant profundal taxa here belong to the acid-tolerant Heterotrissocladius genera represented by H. macridus-type, H. maeaeri-types 1 and 2, H. grimschawi-type (acidophilic), and to the subfamily Tanypodinae represented by Procladius. The sites grouped in the opposing upper left quadrant represent lotic and lotic–lentic taxa (Diamesinae, Thenimaniella clavicornis-type, Eukiefferiella claripennis-type, Eukiefferiella fittkaui-type, several Orthocladius taxa).

Sediment-geochemical and physical properties of the uppermost surface of the sediment basin in Bolshoe Toko are spatially variable. Physical properties of particles within the surface sediments depend chiefly on transportation processes and the characteristics and availability of clastic compounds in the lake catchment. The main catchment comprises the Stanovoy mountain range in the south channelled through the Utuk River into Bolshoe Toko. Accordingly, the lake experiences annual input of suspended material through a single source at the Utuk River mouth that likely is at its maximum during spring snowmelt (Bouchard et al., 2013). The grain-size data and their end-members (Figs. 4 and 7) indicate that the relative proportions of sand, silt, and clay are somewhat constant in proximity to the Utuk River inflow but change towards the north and at the lake shoreline. Whereas in the central northern lake basin the amount of silt increases, the proportions of sand increase along the northern shoreline on top of the drowned moraine. Gravel is only present in samples near the lake terraces in the east. The constant distribution in the southern–central lake basin reflects the river input. Decreasing river influence and hence decreasing water transport energy with increasing distance from the river mouth leads to the observed predominance of finer grain-size (silt dominated) samples in the northern central parts of the lake. Sandy samples along the shoreline reflect direct input from the moraines around the northern part of the lake. Other relevant within-lake sedimentary processes include shore erosion and inwash and winnowing of fine sediment grains by surface currents as well as alluvial processes and debris flows which continue basin ward as subaquatic flows. The restriction of gravel at the eastern shore can be attributed to the availability of source material and suitable transport pathways of coarser clasts from the third moraine. In consequence to the described lateral transport trajectories and local control factors within the lake, there is only weak negative correlation between mean grain size and water depth (r -0.45, Figs. 7 and 12).

The modelled end-member loadings of the observed grain-size classes (Figs. 4 and 7) indicate an EM1 major peak in fine silt that represents fluvial sediment input. EM2 has peak values in fine to medium sandy grain-size fractions and in the northern part of the lake indicative of depositional processes associated with the erosion of moraines distal from the river inflow, where the hydrological dynamics in the lake basin are weak. The weak positive correlation between EM3 and the concentration of diatom valves (r 0.44) likely represents both in situ diatom valves that could not be removed from allochthonous sediment particles during sample processing, and possibly redistributed ice-rafted debris (Wang et al., 2015).

Intermediate concentrations of TOC and high ratios of TOC/TN in the south as compared to the north suggest differences in catchment characteristics, i.e. a considerable allochthonous contribution of terrestrial plant material from the Utuk River. This assumption is supported by previous findings that show non-vascular plants, i.e. phytoplankton and other algae, with TOC/TN ratios between ca. 5 and 10, while organic matter from vascular land plants has higher values of about 20 (Meyers and Teranes, 2002). High values of TOC/TN in lake sediment surfaces at river inflows have also been observed in other studies (Vogel et al., 2010). δ13C is generally low on average (-26.8 ‰) and only slightly higher at the eastern shore (-25.7 ‰), suggesting a strong overall dominance of C3 plants and phytoplankton in the bulk organic matter fraction (Meyers, 2003). It remains unclear as to the degree of old and reworked organic carbon, e.g. from charcoal deposits, transported to the lake.

The distribution of elements from the XRF scanning data suggests strong abiotic relationships with grain-size and mineral distributions. We focus on heavier elements because lighter elements, even though commonly in higher concentrations, show potential contribution from multiple sources. Sr/Rb ratios and Zr are negatively correlated with kaolinite and chlorite (r -0.73 and -0.85, respectively). As described in Kalugin et al. (2007), Rb substitutes for K in clay minerals. The Sr/Rb ratios do not however show a significant correlation with grain-size parameters, as found in other studies (Biskaborn et al., 2013b). We assume therefore that Sr, as a substituent for Ca, is influenced by multiple minerals represented in different grain-size fractions, i.e. K-feldspar (r 0.45) and Hornblende (r 0.24). Associated with high metamorphic grades in the Stanovoy mountains, Sr is preferentially taken into the K-feldspar phase (Virgo, 1968). Conversely, the Zr/Rb ratio correlates well with the sand fraction (r 0.50) and with the mean grain size (r 0.49), but negatively with silt (r -0.54) and clay (r -0.39). We account for this effect by a higher diversity of minerals in the input of the Utuk River supplying the lake basin with mafic Ca-rich metamorphic rocks from the Stanovoy mountains. The strong influence of the Utuk River in the spatial distribution of physicochemical sediment components is further demonstrated by the decreasing gradient of minerals relative to quartz starting from the Utuk River towards the northern lake basin (Fig. 7). The most representative indicator of grain-size variations in surface sediments is given by CLR-transformed values of Ti, which correlate well with the sand fraction (r 0.74) and the mean grain size (r 0.88).

Si/Ti ratios have traditionally been used as a proxy for the biogenic silica content of sediments (Melles et al., 2012). This stems from the fact that Ti is generally attributed to detrital influx and Si to both detrital and biogenic (diatom) origins. At Bolshoe Toko positive correlations between Si/Ti ratios, diatom valve concentrations (r 0.36) and the ratio of planktonic to benthic diatoms (r 0.42) suggests that Si/Ti may be useful to trace the relative portion of diatom valves in intermediate grain-size fractions. Moreover, the Si/Ti ratio correlates significantly with silt (r 0.81).

Mn/Fe ratios have been ascribed to redox dynamics associated with bottom water oxygenation processes (Naeher et al., 2013). In Bolshoe Toko, however, the detrital input of ferrous minerals, i.e. pyrite, suggests that Mn/Fe ratios cannot be directly attributed to redox processes in the surface sediments. This is supported by the correlation of Fe with the sand fraction (r 0.6) and grain size (r 0.59). Accordingly, we found no significant correlations between Mn/Fe and other abiotic or biotic proxies.

Lastly, there is an uncertainty in the spatial distribution of elements measured by XRF techniques. We attribute this lack of clear patterns to (1) methodological hurdles to apply XRF techniques to surface sediments commonly rich in water and organic material, and (2) multiple sources of the same elements coming from minerogenic input, grain-size differences in individual samples and different intensities of redox processes at different habitat settings. The high variance of elements are therefore representative of the high complexity of this lake system, rather than unequivocal validations or falsifications of the applicability of XRF scanner data as an environmental proxy at Bolshoe Toko.

Diatom communities in Yakutia respond rapidly to environmental changes including hydrochemical parameters, water depth, nutrients, and catchment vegetation type (Pestryakova et al., 2018). Planktonic diatom species are ubiquitous across Bolshoe Toko, with a distinct tendency of the ratio between planktonic and benthic species to greater water depths (r 0.77, Figs. 5 and 12), due to the limited availability of light for benthic species (Gushulak et al., 2017; Raposeiro et al., 2018). Especially Aulacoseira species were never abundant along the shallower northern and eastern shorelines. The primary difference between the two most abundant genera in the lake is that Pliocaenicus exhibits the highest abundances proximal to the inflow and in the south-eastern lagoon, whereas Cyclotella are more abundant in the lake centre and absent in the lagoon. Little is as yet known about the new species Pliocaenicus bolshetokoensis (Genkal et al., 2018). Our findings suggest factors other than water depth (r 0.39), such as proximity (e.g. nutrient supply) to the Utuk River and small streams, as controlling parameters for bloom intensities of this species. Cyclotella, however, are restricted to stratification of the water column and hence are more abundant at distance from the river mouth, where incoming water causes turbulence (Rühland et al., 2003; Smol et al., 2005). Cyclotella are therefore also believed to benefit from recent air temperature warming trends and will likely increase in abundance (Paul et al., 2010). Aulacoseira is a dense, rapidly sinking tychoplanktonic group of species requiring water turbulence to remain in the photic zone (Rühland et al., 2008, 2015), which explains the lower abundances in the northern and hydrologically less dynamic zones within the lake. Lightly silicified Tabellaria species are known to occur in zigzag planktonic colonies, yet they also appear as short-valved populations in the benthos (Lange-Bertalot et al., 2011; Biskaborn et al., 2013a; Krammer and Lange-Bertalot, 1986–1991). In Bolshoe Toko, the spatial distribution of Tabellaria indicates benthic habitats are more favourable than planktonic.

The most common non-planktonic species in Bolshoe Toko belong to achnanthoid (monoraphid) genera, of which most species are epiphytic. Epiphytic species exhibit a stronger negative correlation with water depth (r -0.68) than epibenthic species (r -0.4), indicating that aquatic plants, in turn controlled by water transparency, pH, water depth and nutrient status (Valiranta et al., 2011), have an important function in the lake ecosystem (Fig. 12). The highest abundance of achnanthoid and cymbelloid valves occurs at 400 m distance to the northern shore at a water depth of 0.5 m.

Fragilarioid species are adapted to rapidly changing environments and are thus good indicators of ecosystem variability (Wischnewski et al., 2011). The peak occurrences of Staurosira species, which are pioneering small benthic fragilarioids (Biskaborn et al., 2012), therefore indicates the formation of a new ecosystem habitat type in the lagoon at the south-eastern lake basin. We assume this basin is successively separated from the main basin and will eventually form a small isolated remnant lake, similar to “Banya” lake (Fig. 1). High productivity of epiphytic species and low detrital input suggested by elemental and grain-size data, together with higher organic content (high TOC and Br), indicate a calm sedimentological regime with high bioproductivity. Similar neutral pH values measured in water samples from the central basin and the lagoon (Table 1) questions pH as a main driving factor of the Eunotia peak in the lagoon. However, Barinova et al. (2011) suggest a 5.0–5.8 pH range for the identified Eunotia species, which rather indicates that the pH values obtained during April in 2013 are not representative for the annual average and the specific catchment of the lagoon, which likely will differ from this point measurement. The ice break-up during spring and transport of water from the catchment restricted to the lagoon likely leads to milieu differences in the lagoon relative to the main basin.

High autocorrelation coefficients (Moran's I p values) for species richness and valve concentration indicate strong local influence of biotic processes, i.e. reproduction, leading to spatial autocorrelation (Legendre et al., 2005). The lowest observed autocorrelation for the diatom planktonic / benthic ratio confirms the strong relationship between diatom species assemblage composition and water depth. A strong relationship between diatom diversity and water depth is supported by a study comparing morphological count data and phylogenetic species data gained by next-generation sequencing DNA analysis (Stoof-Leichsenring et al., 2019).

The RDA biplot of diatoms (Fig. 2) suggests that both water depth and distance to river are important lake attributes accounting for the species distributions across the lake. Especially Eunotia, fragilarioids, Tabellaria, and also Aulacoseira subarctica appear more frequently at sites that are close to the Utuk River mouth (e.g. PG2113, PG2115, PG2117, and PG2118). The high TOC/TN ratios in these samples illustrates the strong riverine input of allochthonous material. In the biplots, high water depth is primarily associated with Cyclotella species (and Aulacoseria), while Aulacoseira species tend to be additionally influenced by incoming rivers and also thrive closer to the shorelines. Areas close to river mouths are usually dominated by river taxa and species that prefer higher nutrient content related to river input and associated early ice-cover melting (Kienel and Kumke, 2002). Accordingly, the influx of diatoms from wetlands in the lake catchment is an important additional factor influencing the spatial diatom distribution (Earle et al., 1988). Compared to direct conductivity, water depth and nutrient controls, the link between temperature and diatom species is poorly understood in Yakutian lake systems (Pestryakova et al., 2018) and should be avoided.

Our RDA also shows that a high diversity of benthic, and particularly epiphytic diatom species, i.e. several achnanthoid species and some naviculoid taxa, plot in the opposite direction from water depth together with vegetation and the coarse grain-size fraction. Kingston et al. (1983) revealed spatial diatom variability in the Laurentian Great Lakes, where the stability of diatom assemblages increased with water depth. In shallower marginal waters of the Great Lakes, the availability of diverse habitats, including benthic and periphytic niches, leads to high species diversity. According to our data in Bolshoe Toko, the Simpson diversity index suggests higher effective numbers of dominant species associated with increased habitat complexity (Kovalenko et al., 2012), i.e. availability of water plants and benthic substrates in shallower depths along the eastern and northern shores. Thus, higher diversity in this area is facilitated by differential catchment preferences. However, it can be assumed that due to lesser water supply rates from the small northern part of the catchment (Fig. 1), a single location at the north-eastern lake margin will likely not receive significantly higher loadings of nutrients as compared to the Utuk River coming from the igneous mountain range. Nevertheless, moraine deposits typically contain high amounts of silt and clay which can more easily be weathered and altered to fertilizing substances that are transported into the calm and shallower northern part of the basin.

The indices of chrysophyte cysts and Mallomonas relative to diatom cells exhibit indistinct patterns in spatial distributions but a slight tendency towards proximity to river input and high water depths. Although chrysophyte cysts commonly represent planktonic algae (Smol, 1988b), periphytic taxa are also common in boreal regions (Douglas and Smol, 1995) with cool and oligotrophic conditions (Gavin et al., 2011). Mallomonas was reported as an indicator of lake eutrophication and acidification (Smol et al., 1984).

Taphonomic effects on the preservation of subfossil assemblages are generally influenced by clastic transport mechanisms depending on the lake morphology (Raposeiro et al., 2018). The preservation of diatom valves in Bolshoe Toko is found to be lowest in samples from a plateau-like feature in the central part of the lake bottom, which indicates increased re-working associated with bottom currents and/or increased dissolution of diatom valves due to lesser accumulation rates and/or increased grazing activity of herbivorous organisms (Flower and Ryves, 2009; Ryves et al., 2001).

The spatial distribution of δ18Odiatom from the sediment surface indicates higher δ18Odiatom values at the deeper, south-western part of the lake with a difference of approximately 1 ‰ compared to lower δ18Odiatom values in the shallower northern part. This could reflect a combination of spatial δ18Owater variations, water temperatures, and/or a potential species-driven fractionation effect. However, existing studies demonstrate no apparent species composition effects on lacustrine δ18Odiatom (Bailey et al., 2014; Chapligin et al., 2012b). Additionally, the sieving step reduces the assemblage before the isotope analysis to a small size interval, resulting in a similar species composition. Furthermore, dissolution effects in nature and during sample preparation could have an impact on δ18Odiatom. However, we suppose differential dissolution to have a minor influence on the spatial variability of δ18Odiatom at BT samples tackled in our study as these are (1) of similar age, (2) have been treated with wet chemistry at low temperatures, and (3) after preparation do not show any microscopical signs of dissolution effects, i.e. a low diatom dissolution index (Smith et al., 2016).

Regarding δ18Owater variability, waters sampled at the same time in different parts of the lake show a uniform isotopic composition (within ±0.15 ‰) and indicate an isotopically well-mixed lake. Considering this is a one-time recording, slight seasonal variation between shallower and deeper parts (for example due to evaporation) cannot be excluded and could account for some differences in 18O. However, lake surface evaporation would result in isotopic enrichment and overall higher δ18Odiatom values.

Alternatively, the lake temperature in which the diatoms grow has an impact of ca. -0.2 ‰ ∘C-1 on δ18Odiatom (Brandriss et al., 1998; Dodd et al., 2012; Moschen et al., 2005). Shallower areas heat up faster, especially in the photic zone. The temperature profile near to the western shoreline taken in August 2012 (Fig. 6) shows 12 ∘C at the surface, with an average of approximately 10 ∘C in the first 15 m of the water column decreasing to approximately 6 ∘C in 30 m depth. Although a spatial difference of 5 ∘C in the photic zone for causing a 1 ‰ shift is rather unlikely, this could account for part of the variation in surface δ18Odiatom.

The chironomid RDA indicates that spatial variations are primarily influenced by the distribution of tributary rivers. For example, high species diversity is found adjacent to the Utuk River inflow (2117) and in the south-eastern lagoon fed from a small inflowing stream (PG2122). Semiterrestrial taxa, like Smittia-Parasmittia, Pseudosmittia, and Limnophies-Paralimnophies, have been found only here, with the highest abundances of 6 % and 3.2 % at the sites opposite of the inflowing rivers (PG2117 and PG2122), suggesting these taxa were transported from marshy river deltas.

Species at lentic sites with no tributary influence are primarily controlled by water depth. Deep profundal sites of the lake have much lower taxonomic richness in chironomid communities. Higher taxonomic richness at site PG2118 can be explained by an enriching riverine influence. High proportions of lotic and lotic–lentic taxa lead to a high taxonomic similarity of this profundal site to littoral sites in the south and south-east. Similarly, in relation to temperature, sublittoral and profundal sites both have much higher representation of the taxa characteristic of semi-warm conditions and lower abundances of the taxa preferring warm and cold conditions. However, high depths of the sublittoral and profundal sites lead to the development of a poor chironomid fauna at these sites. High distance from the shore and presumably only weak transportation of chironomid remains of littoral fauna to the profundal zone could be another limiting factor for diversity of chironomid communities in the profundal zone.

Eastern relatively shallow littorals are inhabited by more diverse, phytophilic, mesotrophic, and partly acidophilic fauna with absence of lotic taxa, related to a less disturbed and turbulent environment and presence of macrophytes. This fauna has higher abundance of the semi-warm and warm taxa. The presence of mesotrophic to eutrophic and acidophilic taxa can be attributed to paludification of the shore zone and decomposition of macrophytes and submerged vegetation in the shallow littoral (Nazarova et al., 2017b).

It is still debated how spatial and local environmental processes influence the distribution of chironomids at a small spatial scale in a lake (Luoto and Ojala, 2018; Yang et al., 2017). It is known that within one water body the concentration of chironomid head capsules can vary from zero to several thousand per 1 cm3 of sediments (Kalinkina and Belkina, 2018; Walker et al., 1997) depending on factors such as water depth, rate of sediment accumulation, the hydrological conditions, or anthropogenic influence. Water depth in particular is a major driving factor of chironomid assemblages (Ali et al., 2002; Luoto, 2012; Vemeaux and Aleya, 1998), with depth optima of several species consistent across broad spatial scales (Nazarova et al., 2011). Chironomid remains from the deepest zones of Bolshoe Toko represent an assemblage of elements of profundal necrocenosis (Hofmann, 1971) mixed with secondary components of littoral fauna transported with in-lake hydrological and sedimentary processes into the profundal zone from outside. Thus, the re-deposition of littoral taxa into the profundal zone is an important factor that affects the final composition and abundance of subfossil assemblages. While in small lakes, subfossil assemblages from the profundal zone quite adequately reflect the fauna of the entire water body (Brooks and Birks, 2001; Walker and Mathewes, 1990), our findings support the hypothesis that in large lakes the taphonomy of chironomid communities seems to be more complex (Yang et al., 2017; Árva et al., 2015).

Compared to small lowland lakes of central and northern Yakutia, sedimentary processes are quite different in Bolshoe Toko. One reason is the lack of thaw slumps, subsidence, and other permafrost-related phenomena (Biskaborn et al., 2013b) that are typical of shallow thermokarst lake settings across northern permafrost regions (Biskaborn et al., 2016, 2012, 2013a, b; Bouchard et al., 2016; Schleusner et al., 2015; Subetto et al., 2017).

The Bolshoe Toko mineral composition is primarily influenced by the Utuk River, and only samples in extremely shallow areas are influenced by direct shoreline input. The grain-size signal is influenced by dissolution effects associated with organic matter and in situ growth of diatom valves. Conversely, the coarser fractions parallel minerogenic compositions and water depth. Accordingly, the grain-size distribution originated from multiple processes and should only be considered an environmental proxy when combined with biotic indicators.

Diatoms are spatially distributed according to their preferred habitat. Aside from the spatial habitat conditions associated with basin morphology, an additional consideration is the annual duration and thickness of lake ice cover (Keatley et al., 2008; Smol, 1988a). For instance, planktonic communities in Lake Baikal, including Aulacoseira species, are found to grow under the ice if the surface snow properties (i.e. thickness, density) allow sufficient light penetration (Jewson et al., 2009; Mackay et al., 2005). Generally, planktonic and benthic diatom species have strategies to survive in ice-covered lakes by growing in benthic mode, forming resting spores, or attaching to the ice-cover substrate (D'souza, 2012). Hence, the duration and presence of ice cover can significantly impact both changes in assemblage composition and spatial distribution, particularly including the ratio of planktonic to benthic diatoms (Wang et al., 2012a; Bailey et al., 2018).

The applicability of chironomids for temperature reconstructions reveals clear spatial constraints; 22 % of the taxa in sites with riverine influence are absent or rare from the FE mean July chironomid-based temperature inference model (Nazarova et al., 2015), whereas fewer of these rare/absent taxa occur in the central and northern littoral, sublittoral, and profundal parts of the lake (Fig. 4). However, low taxonomic richness of the profundal zone also hampers palaeoclimatic inferences. Also, the number of chironomid head capsules are generally lower here relative to littoral sites. Maximum taxonomic diversity in areas influenced by lake tributaries can be explained by both a taxonomic enrichment from the lake catchment as well by more favourable oxygen and nutrient conditions.

The applicability of δ18Odiatom as a proxy of past hydroclimate conditions at Bolshoe Toko is facilitated by the main controls influencing δ18Odiatom, which are here found to be (1) lake water temperature (Tlake) and (2) lake water isotope composition (δ18Olake) (Dodd and Sharp, 2010; Leng and Barker, 2006; Labeyrie, 1974; Leclerc and Labeyrie, 1987). The fractionation between lake water and biogenic opal can be calculated when comparing δ18Olake (mean: -18.7 ‰) with recent surface sediments of Lake Bolshoe Toko and their respective mean δ18Odiatom (of +22.8 ‰) using this isotope fractionation correlation between sedimentary diatom silica and water as determined by Leclerc and Labeyrie (1987). The mean Tlake can be estimated to ca. 6 ∘C for the photic zone/diatom bloom. This estimate is at the lower end of summer temperatures between 4.8 and 12 ∘C. The corresponding derived mean isotope fractionation factor for the system diatom silica–water α=1.0424 matches the fractionation factor for sediments proposed by Leclerc and Labeyrie (1987) well (α(silica–water)=1.0432).

Additionally, as lacustrine δ18Odiatom also reflects the isotopic composition of the water where the diatoms grow (δ18Olake), δ18Odiatom typically reflects meteoric inputs associated with precipitation and riverine inflows (Fig. 6b). For example, existing studies have used lacustrine δ18Odiatom to reconstruct past changes in precipitation amount and seasonality, the precipitation–evaporation balance, spring snowmelt inputs, and synoptic-scale shifts in atmospheric circulation (Bailey et al., 2015, 2018; Meyer et al., 2015; Kostrova et al., 2013; Mackay et al., 2013). It is envisaged that changes in δ18Odiatom through time at a single site in Bolshoe Toko will yield insights into the long-term air temperature and palaeohydrological history of the region.

Positive feedback mechanisms between benthic algae and chironomid larvae in benthic ecosystems are well documented (Herren et al., 2017). Chironomids in Bolshoe Toko, however, showed less significant correlations with benthic diatom species but weak correlations with planktonic species and lake attributes associated with benthic habitats and water depth, highlighting the potential of chironomids for independent water depth and temperature reconstruction in future sediment core studies (Nazarova et al., 2011).

High correlation coefficients between organic carbon and Pliocaenicus bolshetokoensis (0.66) and silt (0.65) suggest that the accumulation of organic matter and intermediate grain-size fraction is, to a certain degree, controlled by the productivity of siliceous microalgae (Biskaborn et al., 2012). A strong contribution of plankton indicates that TOC/TN ratios can provide insights into the relative influx between land and water plants (Meyers and Teranes, 2002). The relatively weak correlation between TOC/TN ratios and water depth (r 0.51) demonstrates the accuracy limits of TOC/TN as a proxy for relative lake-level changes. This is caused by transport and accumulation of allochthonous organic matter in proximity to the Utuk River. Furthermore, correlations between TOC/TN and TOC, as well as negative correlations with grain-size indicators, suggest diagenetic alteration (i.e. loss) of nitrogen in the surface sediments (Galman et al., 2008).

The distinct difference between two samples along the subaquatic slope near the western shore (diatoms, minerals, organics) indicates redistribution of sediment. Downslope transport of surface layers over the time could lead to redistribution of old material into the deepest parts of the basin. Due to higher accumulation rates, a sediment core from the deepest part of the basin would potentially provide a higher temporal resolution but also a higher risk of repositioned sediment layers. On top of redistribution processes, hump-shaped relations between lake depth and species diversity observed in other studies suggest that the total subfossil species assemblages are better represented at intermediate depths than at the maximum depth (Raposeiro et al., 2018). A coring site at intermediate depth in the shallow northern and sedimentologically calm sector of the basin would enable the tracking of different river and glacial influences and offers greater chances of undisturbed successions of bioindicator time series.