A 2000-year record of lake ontogeny and climate variability from the north-eastern European Russian Arctic

Study area

The investigated lake (unnamed) is located in the lower reaches of the Moreyu River catchment, in north-eastern European Russia (68°17′N, 59°53′E). The region is dominated by lowland tundra vegetation and is characterized by discontinuous permafrost. The shallow lake occupies a former river branch of the meandering Moreyu River, with an original inflow in the south-east. The ~2-km-long lake still has floodplain or wetland stretches related to the former river system. The northern limit is separated from the active streambed of the Moreyu River by a narrow strip of sandbars and wetlands and is situated only 1 km away from the apex of the current Moreyu Delta to the north-west. The elevation of the lake is only 2–3 m above present sea level (a.p.s.l.), and field observations suggest that it could be affected by brackish water inflow at very high tides. Partly paludified tundra uplands with shallow active layer depths (<50 cm) are found to the west and east of the lake at 5–15 m a.p.s.l. (Figure 1).

OPEN IN VIEWER

Figure 1.

The investigated lake seen from the north-west. Inset: location of the study area in north-eastern European Russia.

The closest weather stations to the study area are Varandei, located on the Barents Coast ~95 km to the north-west, and Khoseda-Khard, corresponding to the current position of the arctic treeline ~140 km to the south. The mean July temperature (1971–2000) at Varandei is 9.3 ± 2.7°C (minimum–maximum range: 2.7–14.8°C) and at Khoseda-Khard 13.2 ± 2.3°C (minimum–maximum range: 8.1–17.8°C). By simple interpolation, mean July temperature for the study area can be estimated at ~10.5°C but displays high inter-annual variability.

Sediment sequence and radiocarbon dating

The following lake water parameters were measured during the fieldwork from lake water samples just below the water surface: conductivity 379 µS cm⁻¹, oxygen concentration 6.7 mg L⁻¹, turbidity 12.2 NTU and pH 7.5. The sediment sampling was performed in the northern part of the lake (water depth 2.3 m), from a rubber boat with a Russian peat corer in summer 2004. Logistical constraints related to the remoteness of the study site impeded the transport of heavier coring devices which are commonly used for this kind of lake sediment sampling. The sediment sequence consists of three separate cores that were connected via lithological horizons that represent a composite depth of ~86 cm. The bottom sediment was dark brown peat (~86–60 cm), with two dark grey brown clay intervals in between ~80 and 68 cm and a longer interval on top (~60–50 cm). A grey black clay layer (49–20 cm) occurred above it and was followed by a grey black clay layer with grey brown streaks (19–16 cm). A grey brown clay interval with black streaks (16–9 cm) was followed by a grey brown clay layer (7–0 cm) on the top sediment.

Seven samples were dated (83–86, 59–62, 43–45, 42–43, 33–34, 19–20 and 14–16 cm) using the accelerator mass spectrometry (AMS) ¹⁴C dating method. The dates from two bulk sediment samples and one sample of unknown macrofossil composition, corresponding to the upper lacustrine phase of the sequence and having reversed age estimate compared with the nearest dates, were rejected from the age–depth model (Figure 2). We ascribe these older and apparently erroneous ages to a hard-water effect in bulk lake sediments and aquatic plant macrofossils. Accepted dates are from bulk peat in the lower part of the core and selected terrestrial plant macrofossils for the top of the sequence. Compared with bulk dates from lake gyttja, radiocarbon dating of Holocene peat can be considered straightforward as this deposit consists almost entirely of locally grown plant remains (Blaauw et al., 2004). However, since there are obvious discrepancies in the chronology relative to the dated material, and there is a risk of significant changes in sedimentation rates and episodic redeposition and transport of older soils or sediments potentially as a result of flooding from the sea and the river, the results should be treated very cautiously, and the proposed chronology can only be considered as tentative. The dated samples, their material and age estimates are described in Table 1, and the relationship between sediment depth and estimated ages is illustrated in Figure 2. Radiocarbon dates were calibrated into calendar years using IntCal13 (Reimer et al., 2013) in the programme Calib (Stuiver and Reimer, 1993) Rev 7.0.4.

OPEN IN VIEWER

Figure 2.

Radiocarbon dates from the sediment sequence. The sample marked in white was rejected as an outlier together with the two dates from bulk lake sediments (not shown in this figure).

OPEN IN VIEWER

Table 1.

Radiocarbon dates.

Laboratory no	Depth (cm)	Dated material	14C age BP	95% range cal. BP	Remarks
Poz-49712	14–16	Terrestrial plant macrofossils	730 ± 110	907–533	Very small sample
Hela-1168	19–20	Bulk lake sediments	5660 ± 50	6565–6312	Rejected
Hela-1169	33–34	Bulk lake sediments	8615 ± 65	9761–9482	Rejected
Hela-1101	42–43	Unidentified plant remains	1805 ± 65	1874–1569	Rejected
Poz-19667	43–45	Terrestrial plant macrofossils	1110 ± 30	1071–937
Hela-1054	59–62	Bulk peat	1720 ± 60	1815–1525
Poz-16803	83–86	Bulk peat	1960 ± 30	1990–1829

Palaeoecological analyses

Subfossil chironomid analysis was performed using standard methods (Brooks et al., 2007). Subsamples of 1–5 g wet weight were treated with 10% KOH and sieved through a 100-µm mesh. The sieved samples were transferred to a Petri dish, and the chironomid head capsules were hand-picked with fine forceps under a stereo microscope at 25× magnification. The head capsules were mounted in Euparal ventral side upwards, on microscope slides. A minimum of 50 head capsules was set as a target counting sum for each sample (Heiri and Lotter, 2001; Larocque, 2001; Quinlan and Smol, 2001). The head capsules were identified using a light microscope at 400× magnification. Identifications were based on Wiederholm (1983) and Brooks et al. (2007). In addition to chironomids, other subfossil insect remains, and oribatid mite exoskeletons were enumerated from the sediment subsamples and identified according to, for example, Heiri and Lotter (2007) and Luoto (2009a).

Diatom slide preparation followed standard methods (Battarbee et al., 2001) using the water-bath technique (Renberg, 1990). Slides were mounted using Naphrax^®. The diatom accumulation rate was estimated using microsphere markers (Battarbee and Kneen, 1982). Diatom nomenclature followed Krammer and Lange-Bertalot (1986–1991).

Statistical analyses

Detrended correspondence analysis (DCA) was used as an ordination method to examine temporal changes in chironomid and diatom assemblages since the lengths of the gradient in the data sets were above 2 standard deviation (SD), thus requiring the use of a unimodal method. DCAs were carried out using CANOCO 4.5 (ter Braak and Šmilauer, 2002). The DCAs were run with square-root transformed relative taxa abundances with down-weighting of rare taxa in the programme CANOCO 4.5 (ter Braak and Šmilauer, 2002).

Chironomid-based mean July air temperature reconstruction was performed using the high-latitude Eurasian training set and a two-component weighted averaging-partial least squares (WA-PLS) model (Self et al., 2011). The model with 81 calibration sites has a bootstrapped coefficient of determination (R²) of 0.92, and a root mean squared error of prediction (RMSEP) of 1.0°C with mean and maximum biases of −0.03°C and 1.38°C, respectively. The training set can be considered suitable for the core site since its limnological features and climate characteristics are within the training set gradients. Most importantly, the temperature changes expected at the downcore site are covered by the temperature gradient of the training set (8.8–19.0°C). Furthermore, the location of the downcore site is geographically inside the training set area. All data (training set and core) were square-root transformed. Sample-specific error estimates were performed using the modern analogue technique (MAT) with 999 bootstrap iterations.

Salinity (measured as conductivity) was reconstructed using the European Diatom Database (EDDI, available online) using the combined salinity data set, which provided the best representation of the subfossil diatom assemblages. In EDDI, we used the WA inverse model, which was the most suitable in terms of its predictive performance (R² = 0.76, RMSEP = 2.28 µS cm⁻¹, maximum bias = 0.94 µS cm⁻¹). pH was reconstructed using the AL:PE training set and WA inverse model (R² = 0.78, RMSEP = 0.35, maximum bias = 0.62). The reconstructions were performed with the Ernie software (Environmental Reconstruction using the EDDI diatom database) v. 1.2 (Juggins, 2001). The diatom training sets can be considered suitable for the core sites based on their limnological features. The AL:PE training set includes sites from northern European Russia (Kola Peninsula) and has a pH range of 4.5–8.0. Although the combined salinity training set does not include sites from Russia, its conductivity gradient (40–400,000 µS cm⁻¹) well covers the expected variability at the downcore site.

The chironomid-based temperature reconstruction was compared with chironomid DCA axis 1 scores (main direction of community variance) to verify that the changes in assemblages were driven by temperature changes (Bigler et al., 2003). This was done by examining the Pearson product–moment correlation coefficient (r) and level of statistical significance (p) between the inferred temperatures and DCA scores. Furthermore, the nearest modern analogue of the fossil samples in the temperature training set was calculated using MAT with chi-squared distance and training set and core data square-root transformed to test the suitability of the training set for the fossil data.

Lake ontogeny

The radiocarbon dating results from the peat samples indicated that the sediment record covered the past ~2000 years. The lithology of the lower part of the core consisted of dark sediment material with peat layers. This part of the sediment profile, accumulated between ~2000 and 1200 cal. yr BP, was characterized by terrestrial and semi-terrestrial animal remains, such as oribatid mites, mosquitoes and dark-winged fungus gnats (Figure 3), thus indicating a limno-telmatic phase in the lake ontogeny. Subfossil oribatids, for example, are typically found in sediment samples collected from the limno-telmatic contact, while they are absent or rare from the sediments collected from the actual lake basin (Luoto, 2012). In addition, the relative proportion of semi-terrestrial (or terrestrial) chironomids was high (Figure 3) and, for example, the semi-terrestrial Chaetocladius dentiforceps type (Paasivirta, 2012) reached its maximum of 80% in the lowermost part of the sediment sequence (Figure 4).

The limno-telmatic phase had a low abundance of diatoms, and their diversity was also low (Figure 5). The diatom community consisted of circumneutral mesotrophic taxa, most of which are common for the northern Ural region (Solovieva et al., 2005, 2008). Pinnularia lundii, which tends to occur in relatively high-conductivity waters (Roberts and McMinn, 1996; WFD-UKTAG, 2014), dominated the phase (Figure 5). In addition, two other taxa which had their maximum abundances in the limno-telmatic phase, Navicula pusilla and Diploneis interrupta, are typically found in coastal brackish waters (Denis, 1994; Snoeijs and Potapova, 1995) that suggest marine influence at the time.

The upper part of the sediment core, from ~1200 cal. yr BP to the present, consisted of lacustrine clay sediments and was characterized by decreased abundance in terrestrial and semi-terrestrial mites and insects, and, for example, the terrestrial dark-winged fungus gnats disappeared permanently (Figure 3). Following the decreased number of terrestrial and semi-terrestrial taxa, the invertebrate community of the lacustrine phase changed into having a strongly increased relative abundance of aquatic chironomids (Figure 3). These aquatic chironomids included typical taxa for northern Russia with intermediate temperature optima, such as Psectrocladius (P.) sordidellus type, C. anthracinus type and T. lugens type (Self et al., 2011). The above taxa are also common in oligohumic lakes of northern Finnish Lapland (Luoto et al., 2016) indicating the diminished influence of surrounding bogs on the limnology. In addition to lacustrine chironomids, riverine chironomids were common in this phase until ~700 cal. yr BP (Figure 3). Of the stream chironomids, Derotanypus was the most abundant, and it appeared to the stratigraphy already in the mid-part of the limno-telmatic phase and began to increase towards the lacustrine phase, where it reached its maximum before 700 cal. yr BP (Figure 4). Derotanypus is rare in lake sediments (Brooks et al., 2007), but it is often found in Arctic tundra streams (Medeiros et al., 2011). Therefore, it is very likely that the lake basin had river inundations between the time period of ~1800–700 cal. yr BP as a result of increased regional precipitation causing stronger stream flow events in the Moreyu River and hence river floods transporting aquatic insects and their remains into the lake basins of the catchment. Increased effective moisture is also suggested by the basin ontogeny, as it apparently turned from a wetland into a lake (Figure 3). The hydrological dynamics of rivers emptying into the Arctic Ocean are often a spring phenomenon associated with snow melting in the catchment (Yang et al., 2004), and the subsequent flooding results in an infusion of suspended materials and nutrients that also likely affected the current study lake. The increase in the abundance of Procladius, which is common in eutrophic lakes (Brooks et al., 2001; Luoto, 2011), from ~1200 cal. yr BP onwards may be related to a gradual flood-derived nutrient enrichment of the basin. However, Procladius is also known to succeed in situations where the environmental conditions continuously change (Heiri and Lotter, 2003), such as during frequent flooding events suggesting that the changes may be more related to the dynamic environment.

The diatoms indicated increased salinity and pH for the lacustrine phase, although the conditions remained relatively stable for the whole period. While the used diatom-based training set was not specifically designed for our study region and is hence not an ideal tool for the current salinity or conductivity reconstruction, the constant presence of high-conductivity diatom Amphora lybica and several brackish Nitzschia taxa, such as N. levidensis (Sylvestre et al., 2004), clearly indicates continuous marine influence on the lake. The increase in salinity was probably related to periodic inflows of brackish water from the nearby sea coast, and the elevated pH was most likely caused by the basin ontogeny from an acidic wetland into a circumneutral lake. The presence of planktonic diatoms (Aulacoseira taxa) is typical for the mesotrophic lakes of northern Urals (Solovieva et al., 2005, 2008) suggesting increased trophic conditions.

A gradually cooling climate from ~1200 until ~700 cal. yr BP is indicated by the increases in cold-adapted chironomids (Heiri et al., 2011; Luoto, 2009b; Self et al., 2011), such as Sergentia coracina type, Protanypus and Corynocera oliveri type (Figure 4). After ~700 cal. yr BP, these taxa decreased, while more warm preferring C. anthracinus type and Procladius increased suggesting warmer climate conditions towards the present. However, from 10 cm onwards, there is a clear decrease in vegetation-associated chironomids (Brodersen et al., 2001; Brooks et al., 2007; Luoto, 2010), such as Psectrocladius sordidellus type, Cricotopus intersectus type and Paratanytarsus (Figure 4) that would suggest a decrease in aquatic macrophytes related to either colder water temperatures or increased lake level. In fact, there is a slight increase in the diatom-inferred salinity at this time that could imply flooding of brackish water into the lake basin, which could influence the lake level and survival of freshwater plants. However, it is noteworthy that the abundance of all the vegetation-associated chironomids remained very low until the present, indicating that these taxa have not recovered from the environmental stress that caused their decline. Salt water intrusions can also have direct influence on chironomids (Thienpont et al., 2015), but the present results do not show any clear evidence for the increase in salt-tolerant taxa or decline of freshwater taxa during the period of increased diatom-inferred salinity. Only the abundance of Paratanytarsus, which can tolerate salt water but is also a typical freshwater taxon (Dickson et al., 2014), increases concurrently with the diatom-based salinity reconstruction (Figures 4 and 5).

In the most recent sediment layers, there was a marked increase in chironomids that have preference to eutrophic lakes (Brooks et al., 2001; Luoto, 2011), such as Procladius, Stempellinella and Einfeldia pagana type (Figure 4). In addition to the recent changes in chironomid assemblages, there was a noticeable change in the diatom assemblages as Stephanodiscus parvus, S. minutulus and Diploneis modica increased in the surface sediment layer. Brooks et al. (2001) described a similar increase in the chironomids Procladius and Einfeldia and the diatom S. parvus along an eutrophication process in a lake in the United Kingdom that would suggest that also this study lake has experienced a recent nutrient enrichment. It is possible that the appearance of Stephanodiscus species in the surface sediment sample indicating recent enrichment of the lake is a consequence of climate warming trend in the region (Solovieva et al., 2005, 2008). However, the diatom-inferred pH indicates that the lake has become slightly more acidic during the recent times that would be against the interpretation of eutrophication, but diatoms further indicate a subtle decrease in salinity in the most recent sediment samples. Variations in temperature and salinity also affect algal blooms, for which an important condition is that temperature increases and salinity decreases. Therefore, the recent changes in the aquatic communities might result from the cumulative effects of temperature and limnology. Moreover, in freshwater environments, the effects of climate change are very often similar to those observed under eutrophication (Guilizzoni et al., 2012; Luoto et al., 2015; Schindler, 2001; Visconti et al., 2008), and hence, also in the present record, the chironomid and diatom communities appear to have experienced a eutrophication-like response to recent climate warming.

Temperature reconstruction

The chironomid-based mean July air temperature reconstruction indicated high temperatures at ~2000 cal. yr BP followed by a decreasing trend until ~1700 cal. yr BP, after which the temperatures again increased towards the end of the limno-telmatic phase. However, as the sediment section between ~2000 and 1200 cal. yr BP appears to resemble more of a wetland than a lake, the chironomid-based temperature reconstruction may not be reliable. Furthermore, owing to the large proportion of semi-terrestrial chironomids (Figure 3), these samples had poor taxa representativeness (absent or rare) in the temperature training set and, consequently, poor modern analogues (Figure 6). In addition, most of these samples contained less head capsules than the minimum set for reliable quantitative analysis, and therefore, the temperature reconstruction for the whole limno-telmatic phase is potentially unreliable. In fact, the chironomid DCA axis 1 scores for the entire sediment profile showed major changes for the limno-telmatic phase where it corresponded closely with the chironomid-inferred temperatures, whereas the changes in the lacustrine phase were very small and the relationship between the DCA scores and inferred temperatures broke apart (Figure 6). However, when the DCA was performed for the lacustrine section only, the scores again showed good correlation with the inferred temperatures, suggesting that the chironomids were responding to changes in temperature. A potential problem for the temperature reconstruction also lays in the fact that riverine chironomids occurred in the samples between ~1700 and 700 cal. yr BP (Figure 3). However, previous studies from north-eastern Europe have shown that the influence of streams and river inundations on chironomid-based temperature reconstruction is not very significant (Engels et al., 2008; Luoto, 2010; Luoto and Nevalainen, 2015). Furthermore, the most abundant stream taxon, Derotanypus, was not present in the temperature training set and, therefore, does not have any direct influence on the inferred temperatures. Nonetheless, Derotanypus is associated with very cold environments (Nazarova et al., 2015) that fit well with the current chironomid-based reconstruction, which suggests coldest temperatures during the period when Derotanypus was present (Figures 4 and 6).

A major factor influencing chironomid communities is lake-level changes (Engels and Cwynar, 2011) that could potentially compromise the reliability of the reconstruction. However, a study on boreal lakes has shown that since the inlake water depth gradient is usually directly linked with the water temperature profile, it has the same control on chironomid distribution as the air temperature gradient at large spatial gradients (Luoto et al., 2014). In addition to hydrological conditions, dissolved organic carbon (DOC) and lake humic state have significant influence on chironomid communities (Larocque et al., 2006; Luoto, 2013). In this study lake, the chironomid taxa suggesting dystrophic conditions are the same as the semi-terrestrial taxa in the lower part of the profile, whereas the chironomids of the lacustrine phase are all common in oligohumic lakes. This provides evidence that there can be serious problems with the temperature reconstruction of the semi-terrestrial phase, but the lacustrine phase is probably more reliable. The influence of salinity changes has likely had no significant effect on the temperature reconstruction since true salt water chironomids (Dickson et al., 2014) are absent from the fossil assemblages (Figure 4).

Although the lacustrine phase of the sediment profile appears to provide a valid temperature reconstruction, the comparison of the trends with other climate records is difficult owing to the uncertainties in the chronology. A tentative time frame was constructed based on the four radiocarbon dates we consider reliable (Figure 2), but we cannot exclude the possibility of changes in sedimentation rates between them or the risk of redeposition of material following flooding events from the river and the sea. Therefore, the current chronology should be regarded only as tentative.

The temperature reconstruction in the lacustrine phase showed a gradual decreasing trend towards ~700 cal. yr BP (Figure 6). This temperature decline, associated with the so-called Neoglacial cooling, has also been identified from a previous pollen-based reconstruction from Lake Kharinei and Lake Tumbulovaty in north-eastern European Russia (Salonen et al., 2011). Salonen et al. (2011) showed that this cooling began already at ~3500 cal. yr BP. Because the current temperature reconstruction resembles closely this trend, the warmer chironomid-inferred temperatures in the limno-telmatic phase of the present core data may be correct in their general trends but not in their values, since the temperature variability of 6°C within this section does not appear very realistic. The late-Holocene cooling period caused a rapid retreat of Picea treeline and triggered an abrupt initiation and activation of permafrost processes and, consequently, a major expansion of the northern Russia tundra (Salonen et al., 2011). In the present record, the gradual cooling trend was interrupted by a rapid short-lived warming episode at ~700 cal. yr BP when the temperatures increased by 2.3°C (Figure 6). This period was also reflected in the sediment lithology, where the grey brown clay was accompanied by grey brown streaks (19–16 cm). A warm phase at the same time (~600–700 cal. yr BP) has also been detected in a 1000-year-long tree-ring-based reconstruction from the eastern side of northern Urals (Briffa, 2000; Briffa et al., 1995). However, we found no clear evidence of the presence of the Medieval Climate Anomaly (MCA), a warm climate episode around 1200–800 cal. yr BP that is clearly defined from the Kola Peninsula in north-western Russia (Kremenetski et al., 2004). In common with the present reconstruction, tree-ring-based temperature reconstructions from the western Urals suggests similarly a weak MCA signal (Briffa et al., 2013) indicating that the climate in our study area resembled more Siberian conditions than those of northern Europe.

After the warm phase, the chironomid-inferred temperatures decreased close to the values that were reconstructed prior to the ~700 cal. yr BP warming (Figure 6). This period corresponds with the general timing of the ‘Little Ice Age’ (LIA) in eastern parts of Northern Europe (Luoto and Helama, 2010) and has also been previously described from tree-ring records from Siberia (Briffa, 2000). The present record also showed that the temperatures began to rapidly increase about 300 years ago, after the LIA, and continued to increase towards the present, although part of this increase may be related to chironomid response to the limnological factors unrelated to climate. In fact, there is an overestimation in the inferred temperature for the surface sample compared with the modern observed (interpolated) value (reference period: 1971–2000 = 10.5°C), but these meteorological data display very high inter-annual variability and large estimation errors. In another study from the western side of the northern Ural region, a 200-year chironomid-based temperature reconstruction from Mitrofanovskoe Lake, situated in the permafrost belt within the Bol’shezemel’skaya Tundra, indicated a rapid increase since the early 1900s (Solovieva et al., 2005), thus suggesting that the present results can be realistic. Similar to our findings, also the tree ring-based evidence from the eastern side of the northern Urals indicated a progressive climate warming from ~350 cal. yr BP (Briffa et al., 1995). Again in common with our results, Briffa et al. (1995) showed that the recent warmth was unusual, with the summer temperatures of the 20th century (1901–1990) warmer than during any similar period over the last millennium. Therefore, it appears that the general trends in summer air temperatures were similar during the examined time period on opposite sides of the Ural Mountains.