Sedimentary characteristics
The use of the freeze-corer made it possible to see that after 2013 the structure of the deposits changed from laminated to homogeneous. We assume that the decrease in monimolimnetic stability enhanced the movement of bottom water and, hence, led to resuspension of the uppermost sediments. The origin of the distinct dark layer deposited after 2013 is still unknown (Fig. 9). However, this layer can be a good marker for dating the recent sediments in further monitoring of the lake ecosystem dynamics.
The winter minimum of fluxes has been observed in many lakes and is obviously associated with both a decrease in the production of organic matter and the termination of the input of terrigenous components due to the freezing of the lake (Tylmann et al., 2012; Apolinarska et al., 2020).
The color transitions in the traps generally correspond to the typical alternation of light and dark layers described for other lakes. So, the sediment flux in Lake Shira was represented by a sequence of pale, whitish lamina deposited during spring and early summer, followed by dark, grayish or black lamina deposited in the autumn and winter. That sequence was caused by the seasonality of hydrological and production-biological processes, as well as the seasonality of the conditions of chemogenic mineral formation (Ojala et al., 2013; Kalugin et al., 2013; Apolinarska et al., 2020). However, the uppermost semi-liquid sediments at the water-sediment interface collected using the freeze-corer in March were light brown (Fig. 9).
Previously, it was reported that cores of modern bottom silts of Lake Shira have a well-defined submillimeter lamination of light gray, dark, and black layers (Kalugin et al., 2013). In Lake Shira, the light layer enriched with chemogenic carbonates consists of large (sandy-silty) particles deposited in the period of active water runoff during spring and summer. This layer is conventionally called the "summer" layer. The dark layer enriched with organic matter is composed of clay material deposited in the autumn-winter period and is conventionally called the "winter" layer. These two layers constitute a typical annual cycle of sedimentation (Kalugin et al., 2013). Using sediment traps, we revealed a violation of the typical alternation of colors: in 2012, the color of the sediment in the autumn trap was light, while in the autumn periods of other years it was dark. The light color in autumn 2012 coincides with a higher content of inorganic carbon: in autumn 2012, it was almost twice as high as in other autumns (Fig. 3). In addition, the contents of all components of Group B in the autumn trap of 2012 were much higher, while the contents of components of Group A in this trap were similar to other years (Fig. 6). We assume that in autumn 2012, a large amount of terrigenous matter was carried into the lake after abnormally high precipitation in August 2012 (Fig. 6). Similarly, in summer 2017, the high precipitation probably resulted in an increase in fluxes of terrigenous components (Fig. 3). Thus, we demonstrate that weather anomalies can cause irregularities in the formation of annual varves. For example, it was shown that heavy rainfalls might result in pulses of allochthonous input from the catchment causing thin minerogenic layers in varved structure (Petterson et al., 1993; Tiljander et al., 2003). These occasional episodes are most likely to occur during autumn (Ojala et al., 2013).
Note that biological production can also disturb the seasonal rhythm of sedimentation. For example, in Lake Suminko (Poland), two or even three peaks of deposition of autogenous calcite associated with outbreaks of phytoplankton blooms were observed in one year (Tylmann et al., 2012). However, in Lake Shira the increased salinity led to saturation for carbonates in all seasons as shown earlier (Tretyakov et al., 2012). Thus, in this study, we were not able to differentiate between biogenic and chemogenic carbonate precipitation in Lake Shira. In winter, due to the ice cover, the supply of terrigenous material decreases, which possibly causes a decrease in the sedimentary flow of carbonates.
Elemental composition
Assuming that during the “October–March” period, the flow is approximately constant, a component in the “March–May” trap constitutes about 35% of the same component in the “October–May” trap, in proportion to the time of incubation. A higher percentage may indicate both the increased flux in spring and accumulation of this component on the ice followed by its transfer to the traps after the ice melts. In all years, for all chemical elements except molybdenum, these proportions varied between 40% and 60% of the amounts in the “October–May” traps. Thus, these components could be partly transported to the traps from the melting ice, and, then, their winter fluxes estimated from the difference between traps (Fig. 6) could be somewhat underrated. The only exception was molybdenum, whose amounts in the spring traps were abnormally low – only 10, 10, and 13% of its amounts in the “October–May” traps in 2013, 2014, and 2017, respectively. Hence, the greatest amounts of molybdenum were entrapped during the “October–May” period, and, thus, this element was not transferred from the ice. The flows of the other elements calculated without separation into spring and winter ones were still considerably smaller than their summer and autumn flows. Thus, possible underestimates do not affect our conclusion about the winter maximum of molybdenum sedimentation (Fig. 7b) and the winter minimum of sedimentation of the other components.
The similarities and differences in the dynamics of sediment components are clearly visible from PCA (Fig. 4). Groups A and B are similar in that their flux is maximum in summer and autumn and minimum in winter and spring, as reflected in the coordinates along the PC1 axis (Fig. 4). Along the PC1 axis, all traps are divided by season, and the similarity of sedimentation flows in summer and autumn is clearly visible (Fig. 4). The only exception is autumn 2012, which stands out mainly due to the large number of components of Group B (Fig. 4, 6). Winter and spring traps belong to separate groups, and the content of the components of both groups is minimal in these seasons, which is reflected in the PC1 scores (Fig. 4).
The differences between Groups A and B are clearly evident along PC2, explaining the next 13% of the variance. Group A includes all organic components (with the exception of okenone and bacteriochlorophyll a), nitrogen, as well as elements Br, Zn, Cu, As, S, and Cl. It is well-known that these elements are mainly associated with organic matter; therefore, they correlate with organic components (Fedotov et al., 2010; Vilarrúbia et al., 2018; Sorrel et al., 2021).
Group B contains only inorganic components including inorganic carbon, rock-forming elements K, Ca, Ti, Fe, and Mn, and trace elements Rb, Sr, Y, Zr, Cr, Ni, Pb, and Th associated mainly with chemogenic and terrigenous-clastic processes and indicative of the allochthonous clastic input (Kasper et al., 2012; Vilarrúbia et al., 2018; Sorrel et al., 2021). The peaks of these elements often correspond to severe floods. Inorganic carbon is associated with the deposition of chemogenic carbonates, as well as with their input from the surrounding territory. Sr and Ca indicate endogenic carbonate precipitation (Kalugin et al., 2013; Sorrel et al., 2021). The Mo scores along the PC1 correspond to its predominantly winter sedimentation and along the PC2 axis, its affiliation with Group A, i.e. organic matter (Fig. 4).
Mo is readily scavenged under sulfidic conditions, and therefore is a well-established proxy of the presence of sulfide in the water column (Dahl et al., 2010; Wirth et al., 2013; Zhen et al., 2020). For example, in the bottom sediments of meromictic Lake Cadagno, the variations in Mo content reflected changes in redox conditions in the water column. In particular, a weakening of euxinic conditions indicated by decreased Mo burial seems to be associated with episodic ventilation of the bottom waters (Wirth et al., 2013). In our case, the Mo minima occurred in early spring, when the hydrogen sulfide content in the lake water was minimal due to deep mixing of the mixolimnion, while the maxima corresponded to the ice periods, when the hydrogen sulfide content was maximal as a result of its accumulation under conditions of stable stratification (Fig. 7). Consequently, in the Lake Shira sediments, the Mo peaks can be used as winter signals when analyzing the seasonal structure of the varves. However, the data obtained from traps are insufficient to assert Mo inter-annual fluctuation, in particular, during the decline in hydrogen sulfide in 2015–2016, since data on traps for 2015, unfortunately, are not available, and in 2016 the Mo flux was comparable to Mo fluxes in the other years (Fig. 8).
While Mo enrichment within the sediments is a good indicator of sulfidic conditions, elevated sedimentary manganese (Mn) concentrations were shown to be indicative of the suboxic redox regime (Wirth et al., 2013). Mn enrichments were shown to indicate generally anoxic depositional environments that are episodically perturbed by oxygenation events (Wirth et al., 2013). In our case, the winter Mn flux slightly increased from 2013 to 2014 (data not shown), which may probably reflect the increase in bottom ventilation (Fig. 2). However, the annual Mn dynamics reflected seasonal terrigenous input similar to other elements of Group B (Fig. 3) rather than redox-sensitive behavior.
Photosynthetic pigments
The dynamics of Chl a and carotenoids (except okenone) in traps was typical for temperate water bodies. These pigments showed clear seasonal dynamics, obviously due to seasonal fluctuations in temperature and light (Fig. 8). The carotenoids lutein and zeaxanthin are isomers; that is why they were not separated chromatographically in our study. Lutein is a carotenoid of green algae and higher plants, as well as cyanobacteria (Leavitt, 1993). Zeaxanthin is characteristic of cyanobacteria and green algae but prevails over lutein in cyanobacteria (Leavitt, 1993; Overmann et al., 1993). Previous studies have shown that green algae and cyanobacteria dominate the summer phytoplankton in Lake Shira (Gaevsky et al., 2002; Kopylov et al., 2002; Degermendzhi et al., 2003); therefore, the presence of lutein and zeaxanthin in sediments of the lake is natural. Alloxanthin is a specific carotenoid of cryptophyte algae (Leavitt, 1993). Cryptophytes accumulate near the chemocline in many stratified water bodies including Lake Shira (Prokopkin et al., 2014; Barkhatov et al., 2021). Beta-carotene is a pigment of all plants and some bacteria (Leawitt, 1993), and therefore its presence in the lake is quite natural.
Although year-to-year changes of these pigments are not clearly visible in the traps (Fig. 8), they obviously tend to increase (except okenone) in the frozen upper bottom sediments (Fig. 10). This is probably caused by an increase in lake primary production resulting from the release of nutrients during mixing of the monimolimnion (Melack et al., 2017; Biemond et al., 2021). In particular, the sharp increase in the Cryptophyta abundance in 2015 and 2016, reported earlier (Rogozin et al., 2017), could lead to an increase in the alloxanthin content in the uppermost layer of sediments in 2017 compared to 2016. (Fig. 10). Our results, therefore, confirm that the analysis of recent bottom sediments using high-precision sampling (in our case, freeze-corer) is a sensitive tool for monitoring recent changes in the primary production of the lake (Renberg and Hansson, 1993; Håkanson and Jansson, 1983).
In Lake Shira, the group of anoxic photosynthetic bacteria is mainly dominated by purple sulfur bacteria (PSB) of the genus Thiocapsa (Chromatiaceae) containing Bchl a and carotenoid okenone (Pimenov et al., 2003; Lunina et al., 2007; Baatar et al., 2016). Another group of phototrophic sulfur bacteria, green sulfur bacteria (GSB), was shown to be a minor group in this lake (Lunina et al., 2007); therefore, its specific carotenoids (Bchl c, chlorobactene, isorenieratene) were not found either in the bottom sediments or in the sediment traps.
As shown previously, in the relatively deep chemocline of Lake Shira (11–15 m) the seasonal variations of temperature and light intensity are small, that is why the seasonal variation in the abundance of purple sulfur bacteria is also small (Rogozin et al., 2009). Therefore, neither okenone nor Bchl a demonstrated any pronounced seasonal pattern in sediment traps (Fig. 7). Our data show that the flux of okenone into the traps and sediments sharply decreased after 2012 probably due to increased ventilation of the lake deep waters from 2013 to 2016. Likewise, the flux of PSB main pigment, Bchl a, was below the detection limit after 2014 (Fig. 7). During this period, the abundance of PSB in the water column was also below the detection limit and did not recover after the restoration of meromixis, which indicated the inertia of these microorganisms in the lake. However, we saw that okenone was present in both the traps and sediments during the period when PSB were not detected in water either by microscopy or by pigments (Fig. 7). Consequently, PSB were present in the lake in minor amounts despite the disturbance of meromixis. The low PSB abundance is obviously explained by the deterioration of their living conditions. Indeed, hydrogen sulfide disappeared for a short time in spring, and then, for the rest of the year, it was present in deep waters at a concentration lower than in 2013 (Fig. 7). In addition, the oxic-anoxic interface was located at a greater depth, which led to a decrease in illumination. However, Bchl a was not detected in traps after 2014, probably due to its higher degradation rate (Leavitt, 1993) and lower sensitivity of the spectrophotometric method compared to HPLC.
Variations in okenone content in sediments are usually interpreted as changes in redox conditions. For example, in Lake Dudinghausen, a small dimictic lake (Germany), the okenone peak was interpreted as an episode of enhanced stratification and development of the hydrogen sulfide zone due to eutrophication. The subsequent decrease in okenone was interpreted by the authors as an indicator of a decrease in water transparency due to eutrophication (Dressler et al., 2007). In Lake Hamana (Japan), the disappearance of okenone in bottom sediments since 1955 coincided with an increase in circulation in the lake and deepening of the redox zone (Itoh et al., 2003). In Lake Montcortes (Spain), Vegas-Vilarrúbia et al. (2018) assessed the evolution of the oxic/anoxic shifts over the last 500 years using several independent proxies and showed that periods of stable meromixis coincided with an increased content of isorenieratene and okenone – carotenoids of sulfur bacteria. In Lake Mahoney (Canada), the concentration of okenone in bottom sediments exceeded the concentration of other carotenoids. This fact is in good agreement with the dominance of PSB in this water body. The authors concluded that the lake had been meromictic over the past several thousand years (Overmann et al., 1993). In meromictic Lake Cadagno, the composition of the carotenoids of phototrophic sulfur bacteria was traced for 9.5 thousand years, and it was shown that the hydrogen sulfide zone had existed all this time, and both PSB and GSB were present in this lake. It has also been shown that Mo and okenone correlate and reflect anaerobic conditions deep in the lake (Wirth et al., 2013).
As shown previously, the main cause for deep mixing in 2015 was the weakening of the salinity gradient due to the strong wind impact and early ice retreat in the spring of 2014 (Rogozin et al., 2017). Hydrogen sulfide was constantly present in monimolimnion during the period of regular seasonal observations from 2001 to 2015 (Rogozin et al., 2017). During 2002–2007, the level of the lake rose annually due to the influx of fresh water with atmospheric precipitation, which led to the emergence of a salinity gradient and stable meromixis. Between 2007 and 2016, the average annual level of the lake remained constant, which gradually led to the "erosion" of the salinity gradient and weakening of stratification from year to year. The cold and windy spring of 2014 was especially conducive to deep mixing, which weakened the density gradient in the subsequent autumn and winter and led to the final destruction of the meromixis in the spring of 2015 (Rogozin et al., 2017). Since 2017, the new transgression of the lake surface level has been taking place, which probably contributed to the meromixis restoration. Therefore, the recent Lake Shira can be termed oligomictic rather than meromictic, and the holomictic states could be probably reconstructed from the non-laminated intervals in down-core studies. In addition, our data provide evidence that in a closed lake, the salinity gradient may weaken during periods of the level remaining constant or decreasing, which will cause deeper mixing and a decrease in the content of hydrogen sulfide. Consequently, redox conditions can reflect the climate-induced fluctuations in the lake level.