Diatom community responses to long ‐ term multiple stressors at Lake Gusinoye, Siberia

Global freshwater systems are threatened by multiple anthropogenic stressors via impacts on ecological structure and function necessary to maintain their health. In order to properly manage freshwater ecosystems, we must have a better understanding of the ecological response to human ‐ induced stressors, especially in multiple stressor environments. When long ‐ term observational records are scarce or non ‐ existent, paleolimnology provides a means to understanding ecological response to long ‐ term stress. Lake Gusinoye is a large, deep lake in continental southeast Siberia, and has been subject to multiple human ‐ induced stressors since the 19th century. Diatom assemblages since the late 17th century were


| INTRODUCTION
Freshwater lakes are under threat globally from interacting stressors (Mills et al., 2017), including contamination from human activity (Malaj et al., 2014), enrichment from catchment agriculture (Mekonnen & Hoekstra, 2018) and aquaculture (Legaspi et al., 2015), increased warming of surface waters from climate change (O'Reilly et al., 2015), increased deposition of reactive nitrogen (Nr) (Bergstöm & Jansson, 2006), increased hypoxia due to human activities (Jenny et al., 2016), alterations to local biodiversity from biological introductions and invasions (Gallardo et al., 2016), hydrological modifications (Kominoski et al., 2018), and falling water levels from abstraction and shifting precipitation patterns (Wurtsbaugh et al., 2017). Due to long-range transport of pollutants, even remote lakes are no longer truly pristine (Catalan et al., 2013;Wolfe et al., 2013). Lakes also provide freshwater and protein for human populations, and therefore the history of a lake is often intertwined with the regional history of human activity (Dodds et al., 2013;Jackson et al., 2016a). However, even though the provision of freshwater ecosystem services is fundamental to human well-being, ecosystem functions are often compromised by human activity, threatening the health of the lake itself. Understanding which threats are most important for individual lakes is crucial to managing freshwater ecosystems, but because different threats have different causes, pinpointing when they first started to have a major impact on an ecosystem needs a long-term perspective (e.g., see Dubois et al., 2018 for a review). While long-term monitoring can provide some of these answers, such records are rather rare and not available for most lacustrine ecosystems. For example, although there are several long-term monitoring networks (e.g., The International Long-term Ecological Research [ILTR] network (Vanderbilt & Gaiser, 2017) and the Global Lake Ecological Observatory Network [GLEON] (Rose et al., 2016)), none of these are found in Russia. However, through careful exploitation of lake sedimentary archives, including preserved microfossils and chemical constituents, it is possible to reconstruct environmental impacts on lake ecosystems through time, whether they be caused by atmospheric pollutants, disturbance from agriculture and aquaculture, or impacts of global warming, leading to alterations in local ecology and biodiversity. Paleolimnological records can also provide useful insights into natural variability and baseline conditions before human-environment interactions (Dubois et al., 2018).
One of the most important river basins in Russia is the Lake Baikal Basin in southeast Siberia, which contains the world's oldest, deepest, and most voluminous lake, which is primarily fed by the Selenga River. The Selenga River originates in northern Mongolia and is the largest tributary feeding into Lake Baikal, accounting for over 60% of hydrological flow into the lake (Shimaraev et al., 1994). As over 75% of the plants and animals found in Lake Baikal are endemic, understanding and managing pollution inputs into the lake via rivers such as the Selenga are fundamentally important to retaining its World Heritage Site status. Geographically, much of the Lake Baikal Basin is positioned along the mountain taiga foreststeppe ecotone, a semi-arid transition zone sensitive to climate change (Wu et al., 2012) and fire (Tchebakova et al., 2009). The ecotone is underlain by extensive sporadic to semi-permanent permafrost, which together with the taiga forest represent carbon stores vulnerable to fire and oxidation. The basin sits in one of the most continental regions, but also one of the fastest warming regions on the planet (Jones et al., 2012), leading to the potential for significant ecosystem change.
The Lake Baikal Basin also contains the second and third largest cities in Siberia, Irkutsk, and Ulan Ude. The Selenga River flows through Ulan Ude which is the hub of economic activity in southeast Siberia and one of Russia's most polluted cities. Smaller industrial centres in southeast Siberia, such as Selenginsk and Gusinoozersk (Figure 1), are also highly polluted, due to open cast mining of brown coal, mining for aluminium and molybdenum, coal-fired power generation in Gusinoozersk, and pulp and paper manufacturing in Selenginsk. Long-range atmospheric transport of sulphur and nitrogen pollutants from these industrial centres are a major source of pollution for Lake Baikal (Obolkin et al., 2017). Lake Gusinoye is the second largest lake in the Lake Baikal Basin, and is the only source of drinking and industrial water in the Gusinoozersk region. However, poor treatment facilities result in the lake being highly polluted from industrial and domestic waste. One of the largest polluters is the coal-fired Gusinoozersk State Regional Power Plant (GSRPP), which discharges about 2 million m 3 of heated water into the northern part of the lake every day (UNOPS, 2015, p. 95). Affected waters may be up to 14°C higher than the rest of the lake, which not only prevents part of the lake freezing over every year (Batueva, 2016), but also has a significant impact on lake structure and function. For example, in this region of warmer water, cold adapted fish have disappeared (Pisarsky et al., 2005). In addition to point source pollution, the ecology of Gusinoye has been impacted by invasions of introduced non-local fish species since the 1950s, followed by the construction of fish farms in the 1980s, which benefit from the warmer waste waters from the GSRPP. Therefore, despite a remote Siberian location, the Gusinoozersk region and Lake Gusinoye are subject to many local anthropogenic stressors. Local and regional human activities have been well documented in the Lake Gusinoye region in the latter half of the 20th and early 21st centuries (e.g., Pisarsky et al., 2005), although studies into the impact of such activities on lake environments and biological responses to anthropogenic activities are much less common, with the response of primary producer communities particularly lacking. Paleolimnological analyses allow us to evaluate ecological sensitivities and thresholds, the nature of ecological responses, as well as evidence for transitions in aquatic ecosystems on account of human activities in the critically important Lake Baikal Basin. Here, we use Lake Gusinoye as a model to study the potential impacts of long-term multiple local and regional stressors on freshwater ecosystems, and utilise multiple components of the Lake Gusinoye sediment record to disentangle recent impacts. The aim is to test our overarching hypothesis that activities associated with economic development in the former USSR since 1945 have had a major impact on freshwater diversity, F I G U R E 1 (a) Map of Lake Gusinoye and Lake Baikal in southern Siberia; (b) Lake Gusinoye with the town of Gusinoozersk and major tributaries indicated. The location of sediment core GSNO is indicated by a star. ADAMS ET AL.

| 3 of 19
and that the impact of multiple stressors has led to contemporary novel primary producer communities. To test this hypothesis, the following objectives were undertaken: 1. Reconstruct geochemical evidence of increased nutrients, catchment erosion, and pollution, and combine these with documentary evidence of local aquaculture. 2. Reconstruct diatom assemblages and compositional change (β diversity) for the past 200+ years, to place into context any impact from anthropogenic activities with baseline reference conditions, and undertake appropriate statistical analyses to determine if compositional changes observed are important.
2 | METHODS 2.1 | Study area and regional setting Lake Gusinoye is located in southeast Siberia, in the region of Buryatia, and is situated at the foot of the Khamar-Daban ridge at an altitude of 551 m (Figure 1a). The lake was formed at the end of Pleistocene as a result of tectonic activity (Bazarov, 1969). It is the second largest lake in southeast Siberia (after Lake Baikal), with a surface area of 164 km 2 , a coastline that extends over 65 km, and a catchment area of 924 km 2 . Lake Gusinoye is morphologically divided into a southern and a northern basin, and has a maximum depth of 25 m (southern basin). Until the early 18th century, the northern and southern basins were two separate, smaller lakes. Increased inflow to the lakes c. 1730 as a result of termination of the LIA led to increased lake levels, resulting in a merging of the two small lakes into one (Pisarsky et al., 2005). The lake-level regime of Lake Gusinoye is dependent on its hydrological inflows (Khardina, 2002), which includes 11 currently inflowing streams. Tsagan-Gol River is the main tributary and enters the southern basin ( Figure 1b). The second largest tributary is the Zagustai River, which enters in the northern basin, and on the mouth of which sits the GSRPP. Gusinoye's only outflow is located in the southern basin and forms the headwaters of the Bayan-Gol River, a tributary of the Selenga River.
The town of Gusinoozersk is located on the northeast shore of Lake Gusinoye (Figure 1b), and was founded in 1939 as a settlement for the local coal industry. The settlement grew rapidly, owing to industrialisation, and in 1953 was given the status of town. In 1976 the largest coal-fired power plant in southern Siberia was built on the shores of Lake Gusinoye, near the town of Gusinoozersk.

| Field sampling and sediment core collection
A sediment core (GSNO) was collected from the northern basin of Lake Gusinoye in October 2013 ( Figure 1b). The northern basin was chosen due to its proximity to anthropogenic development in Gusinoozersk. The core was collected from a depth of 22 m, the deepest point in the basin, using a Uwitec gravity corer fitted with a 6.3 cm internal diameter Perspex tube. The GSNO core was stored intact and upright at the Limnological Institute of the Siberian Branch of the Russian Academy of Sciences (LIN SB-RAS, Irkutsk, Russia) until extrusion. The GSNO core was sectioned at 0.2 cm intervals using a vertical extruder in September 2014 at LIN SB-RAS. Total core length for the GSNO sediment core was 66.0 cm, however for the purposes of this study we focus on the top 10 cm only. Extruded sediment samples were stored in Whirlpak bags, shipped to University College London (UCL, London, UK) and stored at 4°C until processing.

| Radioisotope dating
Radiometric dating techniques were used to date the upper sediments from the GSNO core. Approximately 0.5 g of freezedried sediment samples were analysed for 210 Pb, 226 Ra, and 137 Cs by direct gamma assay in the Environmental Radiometric Facility at UCL, using ORTEC HPGe GWL series well-type coaxial low-background intrinsic germanium detectors. 210 Pb chronologies were constructed using the constant rate of supply (CRS) dating model (Appleby, 2001;Appleby & Oldfield, 1978), and independently verified using 137 Cs.

| Trace elements and metals
Sediment was analysed for trace and major element concentrations. Following air drying, dried sediment was sub-sampled for analysis at 23 intervals within the top 10 cm of the core. Each sample was ground to a fine powder using an agate mortar and pestle. Approximately 1.0 g of dried, finely powdered sediment was weighed into a sample cuvette lined with polypropylene film. Samples were then analysed for trace and major elements using a Spectro X-Lab 2000 energy dispersive X-ray fluorescence spectrometer (ED-XRF) with a Si(Li) semiconductor detector in the Department of Geography at UCL. Certified standard reference materials (SRM), Buffalo River sediment (SRM 2704;Epstein et al., 1989) were analysed every 10 samples during analysis. Accuracy of standards was within 10%. Trace element and metal enrichment factors (EFs) were calculated for each sample by first normalising the concentration of an element or metal (M) to the conservative lithogenic element Ti within the sample, and then normalising to the background ratio within the core. EFs were calculated using the following equation from Weiss et al. (1999): Background was taken as the average of the oldest five samples from the sediment core, as this was assumed to be the most minimally impacted. An enrichment factor of 1 throughout the record is equivalent to background concentrations, values above 1 indicate enrichment, while EFs greater than 3 indicate definite anthropogenic contamination of the metal (Boës et al., 2011).

| Spheroidal carbonaceous particles
Sediment was analysed for spheroidal carbonaceous particle (SCP) concentrations following Rose (1994). Samples underwent a sequential attack of mineral acids to break down organic matter, carbonates, and silica components. A sub-sample of the resulting SCP residue was then transferred to a coverslip and allowed to evaporate, mounted onto microscope slides and the SCPs counted under a light microscope at 400× magnification. SCP identification criteria followed Rose (2008). Mean recovery rate for this method is 95.2%, with detection limits of 80-100 SCPs/g dry mass (Rose, 1994). Reference sediment (Rose, 2008) was analysed concurrently to the samples. SCP concentrations are reported in units of number of particles per gram dry mass of sediment (gDM −1 ).

| Loss-on-ignition and sediment densities
Combustion of sediments (loss-on-ignition -LOI) at high temperatures provides a first-order estimation of organic matter and carbonates in lake sediments. LOI at 550 and 950°C was conducted on GSNO sediment following Heiri et al. (2001). Mass loss between 105 and 550°C, and between 550 and 950°C was calculated and converted to % organic matter (LOI 550 ) and % carbonate (LOI 950 ) of the dried sediment, respectively (Heiri et al., 2001). It was assumed that organic carbon comprised 58% of the total organic matter determined through LOI 550 (e.g., Nelson & Sommers, 1996;Schumacher, 2002). Total organic carbon density was calculated using sediment bulk density and organic carbon content. Carbon accumulation rates were then calculated using total organic carbon density and sedimentation rates (Loisel et al., 2014).

| Diatoms
Approximately 0.1 g of wet sediment (weighed to four decimal places) was sub-sampled at approximately 0.5 cm intervals for diatom analysis, and processed following standard procedures according to Battarbee et al. (2001). For diatom slide preparation, the final cleaned diatom sample was topped up to 10 ml with distilled water and mixed well. One millilitre of the final, cleaned diatom solution was pipetted into a new tube and mixed with 1 ml of 8.0 × 10 4 microsphere/ml solution. The new diatom sample was then topped up to 10 ml with distilled water and mixed well. Approximately 1 ml of the diatom and microsphere solution was then pipetted onto a round 19 mm diameter cover slip, and allowed to settle and evaporate overnight. The coverslip was then permanently mounted onto a slide using the resin, Naphrax. Diatoms were observed at 1,000× magnification under oil immersion using a Leica DMLB. A minimum of 300 valves were counted for all samples, with the exception of samples 6.5, 7.1, and 7.5 cm depth, in which 200 valves were counted due to low overall concentrations on the slides. Diatom species identification was conducted following Bacillariophyceae Vol. 1-4 (Krammer & Lange-Bertalot, 1986, 1988, 1991a, 1991b, Diatoms of Europe Vol. 2 (Lange-Bertalot, 2001), Diatoms of Europe Vol. 3 (Krammer, 2002), and Diatoms of Europe Vol. 5 (Levkov, 2009). Diatom raw counts were converted to percent relative abundance per sample. ADAMS ET AL.
| 5 of 19 2.8 | Statistical analyses As the total number of diatom valves counted varied between samples, species richness was determined through rarefaction using 200 valves. Diatom species rarefaction richness was conducted on the diatom raw counts using the vegan package in R (Oksanen et al., 2018;R Core Team, 2018). Hill's N2, an index of the diversity of very abundant species in a single sample, was calculated on the full diatom dataset using C2 (Juggins, 2014). Diatom valve concentrations were calculated using the microsphere method, and normalised to the weight of sediment of the original sample (no. valves/g dry weight) (Battarbee & Kneen, 1982). Diatom valve accumulation rates were calculated using sedimentation rates (no. valves cm −2 year −1 ).
Unconstrained ordinations were employed to detect major patterns of variation in the diatom assemblage data. While the gradient length of axis 1 was shorter than 2.0 standard deviation units (SD), unimodal methods (detrended correspondence analysis -DCA) were used, due to the obvious horseshoe pattern resulting from the use of linear methods (principal components analysis -PCA). DCA was performed in Canoco5 (Ter Braak & Šmilauer, 2012), on the untransformed diatom dataset containing only those species present at over 2% in a single interval, to eliminate the influence of very rare species and examine only major trends. Compositional change (β diversity) in the diatom flora was estimated by detrended canonical correspondence analysis (DCCA) (Birks, 2007;Smol et al., 2005), using estimated and extrapolated ages obtained from the 210 Pb CRS age model as the constraining variable. Relative abundance data were square-root transformed to stabilise variances, but rare species were not down-weighted. Detrending was done by segments and non-linear rescaling. Monte Carlo permutation tests for temporally ordered data were used to determine significance levels (n = 499).
Constrained cluster analysis (based on chord square distance) and breakpoint analysis were performed on the diatom dataset to determine diatom assemblage zones. Broken stick analysis was performed to determine the number of significant zones based on the constrained cluster analysis. Cluster analysis and broken stick analysis were performed in R using the rioja package (Juggins, 2017), and breakpoint analysis was performed in R using the segmented package (Muggeo, 2008). All stratigraphical plots were constructed using C2 (Juggins, 2014).

| Radioisotope dating
Total 210 Pb activity reached equilibrium with supported 210 Pb activity at a depth of around 6.5 cm in the GSNO core (Figure 2a). Unsupported 210 Pb activities declined irregularly with depth ( Figure 2b). The CRS dating model placed 1963 at around 4.1 cm, which is in agreement with the depth suggested by the 137 Cs record (Figure 2c). Sedimentation rates in the core show a very gradual increase in the 20th century, also indicated through a constant increase in unsupported 210 Pb from 6 to 3.5 cm. There was little net change in unsupported 210 Pb in the top 2.6 cm, likely to be derived from dilution caused by an increase in sedimentation rates in the 21st century ( Figure 2d). Ages from 5.9 to 10 cm were extrapolated assuming constant sedimentation rates.

| Trace metals and elements
Trace elements and metals have the potential to provide information on anthropogenic contamination, and catchment disturbance. Elemental concentrations remained steady and unchanged in Lake Gusinoye prior to the 1930s. Beginning in the late 1930s, Ti-normalised EFs of Cu, Zn, and Pb increased above background levels ( Figure 3). In the years closely following the end of World War II, EFs for Cu increased to 2.0, and EFs for Zn and Pb began a continuous increase above background (1.0) (Figure 3). EFs for Cu, Zn, and Pb peaked between 1970 and 1980, followed by slight declines but relatively steady values to the surface. Ti-normalised EFs for P showed first signs of an increase post 1940s, with accelerated increase in EFs beginning c. 1970, and peaking at 1.7 at the surface (Figure 3).

| Spheroidal carbonaceous particles
Spheroidal carbonaceous particles are unambiguous indicators of high-temperature fossil fuel combustion, especially from power stations. SCPs appear in the sediment record at Lake Gusinoye c. 1950 at~500 SCPs/g dry weight, and are fairly steady in concentration and flux until c. 1980 (Figure 3). Post 1980, SCP concentration and flux increased, and peaked between the mid-1990s and early 2000s at~1,200 SCPs/g dry weight and 12 SCPs cm −2 year −1 . SCP concentrations and fluxes have declined in recent years (Figure 3).

| Carbon accumulations
LOI 550 was steady at between 13.1% and 14.5% from the mid-17th century until the mid-1970s, at which point values began to increase, reaching 20.4% in the most recent sample (Figure 3). Carbon mass accumulation rates exhibit a four-fold increase since the late 19th century, from~7 g m −2 year −1 to a peak of over 30 g m −2 year −1 in recent years (Figure 3).

| Diatoms
The diatom assemblages at Lake Gusinoye were dominated throughout the record by planktonic species. Constrained cluster analysis, DCA, breakpoint analysis, and broken stick analysis indicated the presence of three important zones within the diatom assemblages of the past 200 years, and significant shifts occurred c. AD 1920, and in the late 1970s. Zone 1 of the GSNO sediment record occurred from the base of the record to c. 1920. The diatom assemblages from the earliest part of the record were dominated (up to 70%) by Aulacoseria granulata, while subordinate species included small benthic fragilarioids of the Pseudostaurosira-Staurosira-Staurosirella complex, particularly Pseudostaurosira brevistriata (Figure 4). Declines in the relative abundance of A. granulata began in the 18th century (Figure 4). A steady decline in A. granulata continued into the 19th and 20th centuries, concurrent with increases in abundance of two other planktonic species, Fragilaria crotonensis and Lindavia ocellata (Figure 4). Fragilaria crotonensis first increased in the early 19th century, reaching F I G U R E 3 Trace metal and element enrichment factors for P, Cu, Zn, and Pb, Fe/Mn ratios, SCP concentrations and accumulation rates, carbon mass accumulation rates, and sedimentation rates for sediment core GSNO. Zones presented are those from diatom assemblage analyses.
F I G U R E 4 Diatom assemblages of Lake Gusinoye from 17th century to present day. Relative abundances (%) of all diatoms present at over 5% at any one time in the sediment record. (+) indicates species present in sample at less than 1% relative abundance. Diatom concentrations (no. valves/g dry weight) and diatom accumulation rates (no. valves cm −2 year −1 ) are plotted, along with species rarefaction, Hill's N2 diversity, detrended correspondence analysis (DCA) axis 1 scores, and detrended canonical correspondence analysis (DCCA) (β diversity) axis 1. Zones are based on DCA, breakpoint, broken stick, and constrained cluster analyses. abundances of~20% c. 1820 and again c. 1880. The second increase in F. crotonensis is concurrent with the increase in L. ocellata to~5% abundance. Alongside late 19th and early 20th century changes in species abundances, species richness (rarefaction) and Hill's N2 measure of diversity increased (Figure 4). Zone 2 of the diatom assemblages occurred from c. 1920 to the late 1970s, with the first significant breakpoint in the diatom DCA c. AD 1920. By the start of Zone 2, the previously dominant A. granulata had declined to~20% relative abundance, replaced as dominant species by both F. crotonensis and L. ocellata. Relative abundances of A. granulata remained between 20% and 30% through Zone 2. Between c. 1920 and the 1970s, both L. ocellata and F. crotonensis continued to increase in abundance, seemingly at the expense of small Fragilarioids (Figure 4). Lindavia ocellata abundances increased to~15% c. AD 1930, and remained steady throughout Zone 2, while F. crotonensis increased through Zone 2, to an abundance of 35% at the end of the zone. As species richness and Hill's N2 diversity began to decline c. AD 1950, diatom concentrations and fluxes began to increase.
Zone 3 of the diatom assemblages occurred from the late 1970s until AD 2013. Beginning in the 1970s and concurrent with the second breakpoint in the diatom DCA analyses c. AD 1977, increased abundances of planktonic Asterionella formosa were observed, while abundances of Lindavia praetermissa and Stephanodiscus hantzschii began to increase c. AD 1990. Abundances of F. crotonensis and L. ocellata increased at the onset of Zone 3 to~40% and 20%, respectively. Since the onset of Zone 3, abundances of F. crotonensis and L. ocellata have remained relatively steady, and the assemblages have since been dominated by these two planktonic species, peaking at almost 50% and 30%, respectively (Figure 4). At the beginning of Zone 3, abundances of A. granulata declined once again to fluctuate between 5% and 15% through the zone. Diatom accumulation rates peaked c. AD 2000, however total diatom concentrations and fluxes, and all species fluxes have declined in the past two decades. Significant diatom compositional turnover was observed from Lake Gusinoye (β diversity = 1.300 SD; p = 0.002). Between the mid-17th century and early 19th century, species composition was stable, with little change in β diversity during this time. Beginning c. AD 1930, at the onset of diatom Zone 2, greater species turnover was observed with declines in β diversity through diatom Zones 2 and 3.

| DISCUSSION
Lake Gusinoye has undergone several ecological shifts in response to multiple regional and local stressors in the past 100 years. Previous research has indicated increases in zoobenthic biomass from an annual average of 18.97 g/m 2 in the 1940s to 32.78 g/m 2 by the 1990s, and unprecedented changes in the structure of the zooplankton community at Lake Gusinoye since the 1980s from large cladocera and copepods to predominantly rotifers and small Daphnia (Borisenko et al., 1994;Pisarsky et al., 2005). These changes to the zooplankton community occurred concurrently to fish diversification, but also loss of native fish species (Borisenko et al., 1994). The new contaminant and diatom sedimentary records from Lake Gusinoye provide important insights into ecosystem change in response to multiple anthropogenic stressors on this economically and ecologically important freshwater ecosystem in southern Siberia.

| Anthropogenic contamination of Lake Gusinoye
Enrichment of trace metals started during the early 20th century, concurrent with known historical records of increasing local economies, including the development of coal-mining operations around the lake (Pisarsky et al., 2005). The increases in trace metal and SCP concentrations between 1950 and the 1980s reflect the intensification of both local and regional economic activity, including the construction of GSRPP (Pisarsky et al., 2005). Other regional lake sediment records of anthropogenic contamination indicate similar timing of events, including organic pollutants and SCP records from the south basin of Lake Baikal (BAIK6) (Rose et al., 1998), and from shallow lakes in the Selenga Delta (SLNG04), and Black Lake, a small lake adjacent to Lake Gusinoye (Adams, 2017;Adams et al., 2018) (Figure 5). Rapid increases in SCP concentrations in the south basin of Lake Baikal in the 1950s and 1960s reflected local increases in industrialisation across southeast Siberia (Rose et al., 1998), while small increases in Baikal's north basin reflected regional sources because they were of a similar magnitude to hemispherical background increases. Increased concentrations of polycyclic aromatic hydrocarbons (PAHs) in the Lake Baikal Basin in the 1940s-1960s also indicate a high likelihood of atmospherically deposited contaminants to Lake Gusinoye from both local and regional sources in the mid-20th century (Adams et al., 2018;Shirapova et al., 2013). Close temporal matches between the Lake Gusinoye SCP record and SCP and PAH concentrations from lake sediment cores from the south basin of Lake Baikal (BAIK6), the Selenga Delta (SLNG04), and neighbouring Black Lake indicates the primary period of anthropogenic contamination from industrialisation occurred between the 1940s F I G U R E 5 A comparison of anthropogenic contamination and diatom records from Lake Gusinoye with regional and local lake sediment records of anthropogenic contamination from Lake Baikal south basin (BAIK6), Selenga Delta lakes #4 (SLNG04) and neighbouring Black Lake, and Irkutsk, Russia temperature change since the late 19th century.

ADAMS ET AL.
| 11 of 19 and 1990s ( Figure 5). Post-1940, increases in sediment P may reflect the gradual, long-term increase in P inputs to Lake Gusinoye, initially as a result of early agricultural practices in the region (Bazhenova & Kobylkin, 2013), then of increases in wastewater effluent into the lake from the GSRPP and the town of Gusinoozersk, and the impacts of fish farming, which were established in the lake in the 1980s (Pisarsky et al., 2005). While P chemistry is complex, making it difficult to interpret sedimentary changes in P as directly related to changes in P in the water column, there is little evidence for changing sedimentary redox conditions at this time in Lake Gusinoye (Figure 3), which may indicate that changes in sedimentary P in the 20th century provides an indication of directional change in P inputs to the lake. The sedimentary enrichment of P since the 1940s coincides with the measured increases in phosphate observed by Borisenko et al. (1994) between the mid-20th century and early 1990s in Lake Gusinoye.

| Biological responses
Over the past 150 years, undisturbed, temperate lakes show little change in diatom β diversity (only~1 SD) (Smol et al., 2005). These sites act as good reference points for comparing diatom β diversity in lakes in other regions with contrasting impact histories. Diatom composition in Lake Gusinoye shows significant change over the past few hundred years (Figure 4). Overall, Lake Gusinoye β diversity (1.3 SD) exceeds mean reference conditions from unimpacted lakes in North America and northern Europe (Hobbs et al., 2010;Smol et al., 2005), western Greenland (Hobbs et al., 2010), the Tibetan Plateau (Wischnewski et al., 2011), and nearby East Sayan Mountains (Mackay et al., 2012). However, Lake Gusinoye β diversity is lower than found in impacted lakes in Svalbard (Holmgren et al., 2010) and permafrost thaw slump-affected lakes in the Canadian Arctic (Thienpont et al., 2013). Prior to the 1800s, diatom assemblages were very stable in Lake Gusinoye (Figure 4), which we suggest represents reference conditions and baseline assemblages for the lake prior to anthropogenic development in the region. Evidence for diatom compositional turnover starts in the early decades of the 19th century, and accelerates in the lake after AD 1950. Whereas in polar regions increased β diversity is attributed to the limnological impacts of climate change, we argue that in Lake Gusinoye, species turnover is driven by multiple factors, notably impacts of development around the lake and its catchment over the past 100 years, alongside global warming in the past few decades. It is interesting to note that high β diversity was also observed in alpine lakes in the American Cordillera linked to increased Nr availability (Hobbs et al., 2010).
The early part of our record is dominated by the heavily silicified diatom, A. granulata (Figure 4). It is a meroplanktonic species, which means it is adapted to being suspended in the water column during periods of turnover (Kilham, 1990). When lakes stratify, A. granulata sinks out of the photic zone, but because it forms resting spores from which it can regenerate (Schelske et al., 1995) once mixing recommences, for example due to strong autumnal winds, it can become entrained back into the water column. One of the most striking features of the diatom record is the gradual decline in abundance of A. granulata from approximately AD 1750 to the late 19th century (Figure 4). There are good historical records for 18th century development of Lake Gusinoye. When A. granulata dominated the sequence, the lake was divided into two separated, smaller lakes. However, after AD 1730, water levels in the region rose and the basin was flooded. So prior to AD 1750, conditions in the region must have been such that deep mixing of Lake Gusinoye occurred, which allowed regeneration of nutrients such as silica for A. granulata to take up. Elsewhere in the region, paleolimnological records indicate that climate was very cold at this time, linked to the latter stages of the LIA (Mackay et al., 2005(Mackay et al., , 2012. After ice break up in spring on Lake Gusinoye, cold waters would have been easy to mix, so supporting A. granulata. Additionally, growing season at the time would have been shorter, leading to shorter and weaker periods of stratification. Increasing river flow and flooding of the lake after AD 1750 was likely caused by warming, as the region came out of the LIA, causing widespread hydrological instability (Pisarsky et al., 2005). Regional warming will also have caused the growing season and period of surface water stratification in lakes to lengthen. In Lake Gusinoye, this may have limited upwelling of nutrients, resulting in the decline in A. granulata. F. crotonensis is a common indicator of increasing nutrient influx and increasing trophic level in freshwater lakes (Akcaalan et al., 2007;Saros et al., 2005), and the early increase in F. crotonensis, especially after AD 1800 is concordant with both flooding of Lake Gusinoye, and earliest agricultural expansion and widespread deforestation in the Selenga River basin, likely leading to increased mobility and delivery of nutrients (Bazhenova & Kobylkin, 2013). Increased diatom turnover after AD 1800 therefore is a response to regional warming after the cool conditions of the LIA. Hill's N2 diversity values also increase at this time, indicative of increases in available resources (Interlandi & Kilham, 2001), likely brought in with meltwater floods.
Prior to the end of the 19th century, shifts in diatom communities between A. granulata and F. crotonensis were gradual but persistent, as highlighted by DCA axis 1 scores (Figure 4). It was not until the start of the 20th century that we start to see critical ecological transitions occurring in the diatom record. The first transition marks the stabilisation of A. granulata, but rapid increases in two other planktonic taxa, F. crotonensis and L. ocellata c. AD 1920 (Figure 4). This switch in community is concurrent with increases in diatom fluxes, trace metal enrichment, and SAR (Figure 3). These early 20th century changes in the diatom communities are likely an ecological response to land-use change from increasing livestock populations (Bazhenova & Kobylkin, 2013) and mining (Pisarsky et al., 2005) resulting in increased soil erosion, against a backdrop of increasing regional temperatures ( Figure 5). Fragilaria crotonensis is known to respond to increases in nitrogen, and has been considered as an indicator of increasing Nr in previously oligotrophic lakes in the western United States, as a result of increased nitrogen deposition (Hobbs et al., 2010;Saros et al., 2005;Wolfe et al., 2006), as well as rising nitrate concentrations (Stoermer et al., 1978;Wolin et al., 1991), and therefore may be indicative of increasing nutrient flux to the lake. Lindavia ocellata may indicate a response to continued warming temperatures, weaker lake turnover, and strengthened stratification (Liu et al., 2017;Malik & Saros, 2016). Although it may also represent a switch in the lake from N to P limitation (Winder & Hunter, 2008). Therefore, the first significant change in the diatom community of Lake Gusinoye, post baseline conditions, was likely a response to intensifying anthropogenic activities and nutrient enrichment from N, against a background of warmer temperatures in the early 20th century ( Figure 5).
The second significant change in diatom community composition occurred in the late 1970s (Figure 4), when livestock populations and concomitant soil erosion in the region were at their highest (Bazhenova & Kobylkin, 2013). Aulacoseira granulata abundances decline, reaching their lowest levels for the whole record, L. ocellata and F. crotonensis reach their highest abundances, and taxa including A. formosa, L. praetermissa, and S. hantzschii become persistent in the record, all of which may be indicative of increasing nutrient levels, both N and P, in Lake Gusinoye. Total diatom fluxes reach their highest values for the whole record, before declining in the most recently deposited sediments. The significant changes occurring in the late 20th century occur concurrently with peak SCP concentrations and P enrichment above background levels, indicative of high economic and agricultural activity in the region ( Figure 5). Asterionella formosa in particular has been linked in large freshwater lakes to increasing anthropogenic development and associated nutrient enrichment, particularly nitrogen (Bergstöm & Jansson, 2006;Stoermer et al., 1991;Wolin et al., 1991), and competes well for P in high Si environments (Bradbury, 1988;Saros et al., 2005). The concurrent high abundances of F. crotonensis and A. formosa have been used previously to infer increased nutrient loadings and anthropogenic disturbances in lake catchments (Anderson et al., 1995;Forrest et al., 2002;Fritz et al., 1993;Garrison & Wakeman, 2000;Hobbs et al., 2010;Wolin & Stoermer, 2005). Further, both F. crotonensis and A. formosa have been shown to respond primarily to increases in silica and nitrogen in both temperate and alpine lakes (Bennion et al., 2011;Saros et al., 2005). Therefore, the combined high abundances of F. crotonensis and A. formosa, alongside low abundances of Stephanodiscus spp., may represent low but increasing P conditions and moderately abundant N and Si (Michel et al., 2006). Indeed, spot measurements from Lake Gusinoye indicate concentration increases for several nutrients between the mid-20th century and the early 1990s, including phosphate (from 0.003-0.03 up to 0.09 mg/L), sulphate (from 8 to 68 mg/L), nitrate (from 0.035 to 0.9 mg/L), and ammonium (from 0.001 to 2.4 mg/L) (Borisenko et al., 1994;Khakhinov et al., 2005). Furthermore, Epithemia adnata, which contains endosymbiotic cyanobacteria in its cells to assist in nitrogen fixation, completely disappears from the Lake Gusinoye sediment record at this time, lending support to our suggestion that nitrogen supply was plentiful (DeYoe et al., 1992).
Ecological changes in the late 20th century are also likely a response to the onset of aquaculture economies (primarily Cyprinus carpio and Acipenser baeri baicalensis) in the northern basin of Lake Gusinoye c. AD 1980. Aquaculture is known to result in increases in organic matter deposition in aquatic systems due to fish food deposition and increased faecal matter, and inorganic nutrient enrichment, particularly ammonium, nitrate, nitrite, and phosphate (San Diego-McGlone et al., 2008). Primary producers have been observed to respond strongly to increasing nutrient levels related to fish farming activities, manifested, for example, as transitions to new primary producer communities, and increased occurrences of harmful algal blooms (e.g., Jiang et al., 2013;San Diego-McGlone et al., 2008;Wang et al., 2009). Thermal effluent from the GSRPP in recent decades provided excellent conditions for aquaculture to thrive (Pisarsky et al., 2005). The presence of thermal discharge from the GSRPP and the onset of aquaculture in Lake Gusinoye coincided with further declines in A. granulata, and our recorded increase in A. formosa, L. praetermissa, and S. hantzschii, and peaks in diatom flux. However, declines in fluxes of all diatoms since c. 2000 may be evidence for either their decline in the water column linked to increasing stratification, increased grazing, or increased competition for resources from other algal groups. Therefore aquaculture and thermal pollution from the GSRPP are the latest in the multitude of 20th century anthropogenic stressors impacting the ecology of Lake Gusinoye, which resulted in significant shifts in diatom assemblages during the past several centuries.

| The nature of ecological shifts in multi-stressor systems
Diatom assemblage changes at Lake Gusinoye during the 20th century now indicate that primary producer communities have undergone significant shifts relative to pre-industrial assemblages. Our findings of primary producer changes correspond with previous studies of increasing zooplankton biomass and change in zooplankton community structure and fish diversification since the mid to late 20th century (Borisenko et al., 1994). Therefore, rapid and extensive 20th century anthropogenic development in and around the Gusinoozersk region resulted in whole-ecosystem shift of Lake Gusinoye to historically novel ecological communities for the lake. Major shifts at all trophic levels in Lake Gusinoye have come as a result of the multiple anthropogenic stressors impacting the lake ecosystem. In multi-stressor environments, it is possible that the ecological effect of multiple disturbances may be individual, with each stressor inciting an individual response, with an overall additive effect of all stressors. However, the interaction between multiple stressors may result in exceptional ecological shifts and novel ecological communities and trajectories (Bogan & Lytle, 2011;Christensen et al., 2006;Jackson et al., 2016b;Sala et al., 2000), promoting more extensive and/or rapid change, as well as unexpected or unprecedented response to stressors. At Lake Gusinoye, this includes an assessment that the record of diatom assemblages since the mid-17th century provides evidence for a current novel trajectory for algal communities. This includes significant shifts in the planktonic diatom community and declines in the littoral community, with no evidence for recovery, and overall decline in diatom flux since c. 2000, likely due to shifts in trophic dynamics or stronger stratification regimes.
Overall, diatom community changes indicate that increasing nutrient loadings to Lake Gusinoye since the early 20th century have had a greater impact on primary producer communities than increasing regional temperatures. This is in contrast to recent shifts observed in diatom communities from many other lakes in cold regions, which primarily experience increases in small Discotella/Lindavia species as a result of reduced ice cover and/or enhanced thermal stratification due to increasing temperatures (e.g., Rühland et al., 2003Rühland et al., , 2008Smol et al., 2005;Sorvari et al., 2002). This dominant shift was not observed at Lake Gusinoye. Therefore, Lake Gusinoye appears to either be more sensitive to nutrient increases than other northern lakes, or indicates that, where the opportunity allows due to local or regional sources, diatom responses to nutrient increases will dominate over responses to warming temperatures. However, the rate of compositional turnover (βdiversity; 1.3 SD), and major shifts in diatom assemblages, that is concurrent increase in F. crotonensis and L. ocellata with declines in A. granulata, fragilarioids, and E. adnata, followed by concurrent increases in A. formosa and L. praetermissa, may indicate a simultaneous response to increasing nutrients and temperature, with a switch from deep mixing and nutrient upwelling with plentiful silica supply to a lake with high nitrogen levels, abundant silica, and longer stratification times. Moreover, temperature and nutrients may interact with each other at several ecological levels, from individual to whole ecosystem, resulting in feedbacks (Cross et al., 2015). Therefore, the post AD 1920 diatom assemblage shifts observed may be the result of synergistic effects of increasing nutrients and temperature at Lake Gusinoye.

| CONCLUSION
Lake Gusinoye is a multi-stressor environment, and numerous human-related changes in the Gusinoozersk region have been documented through the 20th century, all with the potential to impact ecological community structure and function. Trace metal and element records combined with SCP concentrations indicate early impacts of regional and local development c. AD 1920, concurrent with the onset of increasing livestock numbers and deforestation resulting in increased regional soil erosion. Previous studies have indicated shifts in zooplankton and fish communities since the 1980s as a response to human-related stressors. Now, we show that primary producers have also responded with significant shifts in community structure to the multitude of developments in the Lake Gusinoye region. Diatoms first underwent assemblage shifts as a result of hydrological changes, including increased river flow and flooding, and regional warming leading to lengthening growing season and strengthening of stratification, due to the termination of the LIA in southern Siberia. Since the early 20th century, diatom communities have changed more profoundly as a result of multiple anthropogenic stressors, including nutrient influx, aquaculture, and wastewater discharge from the Gusinoozersk State Regional Power Plant. Lake Gusinoye is an economically and ecologically important freshwater system in southern Siberia. These records of significant ecological change in the second largest lake in southern Siberia reveal that anthropogenic stressors have had a significant impact on primary producers within the lake. However, despite declines in animal husbandry over the past few decades, diatom assemblages remain indicative of a persistent enriched system, which is likely related to increasing dependency on local aquaculture in the lake, and potential interactive effects between increasing regional temperatures and nutrients.