Introduction

Lakes provide diverse ecosystem services that benefit millions of people globally, including provisioning services, regulation and maintenance services, and cultural services1,2. However, many lake ecosystems are currently experiencing abrupt shifts in ecological structure under multiple stressors3. This issue not only raises environmental concerns but also has substantial economic and social implications4. Consequently, understanding the shifts in lake ecological systems has become a priority for ecosystem service management4,5.

Anthropogenic activities have been recognized as the primary driver of lake ecological shifts after the 1950s6. Most of these shifts are attributed to local or regional human activities, such as wastewater discharge7, agriculture8, and land-use changes9, especially in densely populated regions like Europe, North America, and East Asia (Fig. 1a)6,10. However, it remains uncertain whether pristine remote lakes, defined as those in high latitude or altitude regions free from direct human interference, have experienced an ecological shift in response to large-scale (e.g., hemispheric or global) anthropogenic forces. While climate change has been identified as the primary driver of ecological shifts in Northern Hemisphere remote lakes due to their isolation from local human impacts11, systematic sediment analyses of 25 such lakes revealed a coherent signal of anthropogenic nitrogen deposition on a hemispheric scale since 1895 ± 10 years12. This finding points to an important alternate and synergistic mechanism for global ecological change, as evidenced by studies of ecological shifts in alpine lakes on the Tibetan Plateau and in central China13,14. Increasing evidence underscores the escalating threats posed by long-range pollutant transport to alpine lake ecosystems15, necessitating the development of high-quality ecological records to evaluate potential regime shifts.

Fig. 1: Study area and sampling locations.
figure 1

a Distribution of Shuanghu Lake (red dot, remote site) and anthropogenically impacted lakes (yellow crosses) proximal to human settlements6. Background nighttime lights imagery from NASA Earth Observatory (https://earthobservatory.nasa.gov/features/NightLights/page3.php)10. b Geographical setting of Shuanghu Lake on the southern flank of the Altai Mountains (reference records marked by yellow triangles). c Coring sites SH12B and SHW21E.

Full size image

The Altai Mountains provide an exceptional natural laboratory for investigating large-scale anthropogenic impacts on remote aquatic ecosystems. Despite their location distant from major population centers (Fig. 1a), alpine lakes in this region receive substantial atmospheric deposition of pollutants transported from industrialized areas of Europe16,17,18. This unique combination of geographic isolation and pollutant exposure makes Altai lakes particularly sensitive indicators of ecological responses to hemispheric-scale anthropogenic forcings. However, while paleolimnological research in this region has made substantial advances in understanding hydrological or climatic changes19,20,21,22, critical gaps remain in documenting ecological changes. These limitations stem from both insufficient lengths of observational data and, more fundamentally, the absence of long-term biological records required to robustly separate anthropogenic impacts from climate-driven ecosystem variations3,23,24.

In this study, we present a well-dated diatom record spanning the past 2400 years from Shuanghu Lake, located within the core conservation zone of Hanas National Nature Reserve in the Altai Mountains (Fig. 1b, c). This sedimentary record encompasses both pre-industrial baseline conditions representing natural variability and the industrial period characterized by intensifying European anthropogenic impacts. By integrating diatom assemblages with geochemical data, vegetation reconstruction, and historical emission record from upwind Europe, this study is designed to: (1) establish the natural trajectory of lake ecological dynamics under climate variability; (2) determine the timing of any discernible ecosystem shifts; (3) evaluate potential linkages between ecological shifts and large-scale anthropogenic forcings.

Results and discussion

Diatom assemblages in Shuanghu Lake

A total of 90 major taxa from 21 genera were identified from the 138 lake sediment samples. The major taxa fell into three ecological groups: planktonic (e.g., Aulacoseira ambigua (Grunow) Simonsen and Aulacoseira subarctica (Otto Müller) E.Y.Haworth), facultative planktonic (e.g., Staurosira construens Ehrenberg and Staurosirella pinnata (Ehrenberg) D.M.Williams & Round), and benthic (e.g., Achnanthidium minutissimum (Kützing) Czarnecki and Mayamaea atomus (Kützing) Lange-Bertalot). Given the shallow water environment of the study site, facultative planktonic taxa were analyzed together with the benthic group (Fig. 2). Throughout the record, the diatom assemblages were dominated by small benthic fragilarioid species, particularly S. construens and S. pinnata, which together account for ~45‒92% of the composition. Four diatom assemblage zones were recognized based on cluster analysis (Fig. 2), and they are described below.

$${{\rm{Zone}}} \, 1\left( \sim 460 \, {{\rm{BCE}}}-870 \, {{\rm{CE}}}\right)$$
Fig. 2: Diatom assemblages from core SH12B in Shuanghu Lake.
figure 2

All taxa can be classified into planktonic, small benthic fragilarioid and other benthic categories. Four zones were identified based on cluster analysis. The dark green dotted line marks the ecological shift of Shuanghu Lake at ~1870 CE.

Full size image

The percentages of S. construens and S. pinnata are relatively stable in this zone. Another dominant benthic taxon, A. minutissimum, is much less abundant than S. construens and S. pinnata in this zone, with relatively high percentages during ~330–230 BCE, ~50–190 CE, and ~430‒560 CE. Other taxa are present at low abundances, with the peak percentages of planktonic taxa occurring during ~330–230 BCE, ~50‒190 CE, and ~430‒560 CE.

$${{\rm{Zone}}} \, 2\left( \sim 870-930 \, {{\rm{CE}}}\right)$$

This zone is characterized by pronounced decreases in S. construens and S. pinnata; and by increases in A. minutissimum, Rossithidium pusillum (Grunow) Round & Bukhtiyarova, Fragilaria capucina (Grunow) Lange-Bertalot, and planktonic taxa.

$${{\rm{Zone}}} \, 3\left( \sim 930-1870 \, {{\rm{CE}}}\right)$$

Notable features of this zone are the frequent and inverse pattern of fluctuations in the percentages of S. construens and S. pinnata. Another dominant taxon, A. minutissimum, increased abruptly and reached its maximum abundance (23.2%) during ~1640‒1800 CE.

$${{\rm{Zone}}} \, 4\left( \sim 1870 \, {{\rm{CE}}}-{{\rm{present}}}\right)$$

Staurosira construens, S. pinnata, and A. minutissimum decreased rapidly to a low abundance in the late 19th century. Subsequently, S. construens increased continuously while S. pinnata and A. minutissimum remained relatively constant. Notably, both M. atomus and Achnanthes acares M.H.Hohn & Hellerman increase markedly and reach their maximum abundances in this zone.

Response of diatom assemblages to changes in ice-cover duration before the 1870s

The diatom communities in Shuanghu Lake over the past 2400 years were dominated by S. construens and S. pinnata, with a low presence of planktonic species (Fig. 2). This can be attributed to a short growing season resulting from prolonged ice cover. Small benthic fragilarioid taxa, like S. construens, S. pinnata, and Pseudostaurosira brevistriata (Grunow) D.M.Williams & Round in Shuanghu Lake, are well-known for their rapid reproductive ability and competitiveness in shallow, ice-covered environments with short growing seasons in Arctic and alpine regions11,25,26,27,28,29,30,31,32. Prolonged ice cover leads to the development of a long-lasting ice platform in the central part of lakes28, which provides suitable habitats for small benthic fragilarioids in the shallow lake margins while restricting the development of planktonic diatoms25,32. S. construens, in particular, is well-adapted to low light conditions beneath the ice cover27,32,33,34. Thus, the high abundance of small benthic fragilarioids and the low number of planktonic diatoms are recognized as indicators of prolonged ice-cover in high-latitude or high-altitude lakes11,27,35,36. With a cool, short summer and a cold, long winter in Shuanghu Lake, the extended ice cover lasting nearly six months creates favorable conditions for the growth of small benthic fragilarioid species. Consequently, the frequent fluctuations in small benthic fragilarioids, with notable declines during ~50‒170 CE, ~430‒560 CE, ~830–970 CE, and ~1640–1800 CE (Fig. 3a), are mainly influenced by changes in ice cover duration in Shuanghu Lake. Additionally, fluctuations in A. minutissimum abundances and taxonomic diversity in Shuanghu Lake exhibit clear responses to changes in ice cover duration. As a common epiphytic diatom that firmly attaches to aquatic vegetation24,33,37, A. minutissimum serves as an indicator of improved habitat conditions linked to reduced ice cover and increased light availability31. This relationship is particularly pronounced during ~50‒170 CE, ~430‒560 CE, ~830–970 CE, and ~1640–1800 CE (Fig. 3b), characterized by declines in small benthic fragilarioids and enhanced diatom α-diversity (Fig. 3a, c). These synchronous changes highlight the pronounced impact of ice cover duration on diatom assemblages in Shuanghu Lake.

Fig. 3: Response of diatom assemblages in Shuanghu Lake to changes in ice cover duration.
figure 3

a Variations in the abundance of small benthic fragilarioids and b A. minutissima from Shuanghu Lake before the 1870s, along with c the Shannon index (alpha diversity) of diatom spices, were mainly influenced by changes in ice cover duration. This relationship is supported by winter temperature reconstructions based on d tree-ring data in East Asia over the past 700 years41, where the grey line represents annual temperature variations and the pink line indicates the 130-point smoothing, as well as e the cellulose δ18O record from the Big Black peatland in the Altai Mountains19,42. f The DCCA results (green line) and the RSI (blue bars) of the diatom assemblages in Shuanghu Lake indicate an ecological shift around 1870 CE. The grey/yellow shaded bars indicate the periods of reduced ice cover duration and the new regime of diatom assemblages, respectively. The grey dots and square at the top correspond to ten AMS 14C dates with a 1σ uncertainty interval, and a 210Pb/137Cs date from Shuanghu Lake, respectively22. The red triangles in e represent the five 14C dates from the Big Black peatland42.

Full size image

The dominant role of ice cover in shaping diatom taxa in Shuanghu Lake is further corroborated by variations in winter temperature (summer temperatures excluded due to their minimal impact on ice phenology). The duration of ice cover is primarily affected by heat exchange with the atmosphere and the heat stored in the water38. Exchange with the atmosphere is primarily determined by air temperature: lower temperatures lead to earlier ice formation, thicker ice, and delayed ice break-up; while higher temperatures have the opposite effect11,39,40. Winter temperature reconstructions based on tree-ring data in East Asia41 and cellulose δ18O record from the Big Black Peatland in the Altai Mountains19,42 (Fig. 1b) show elevated winter temperatures during the shaded intervals in Fig. 3d, e. This implies a reduced duration of ice cover in Shuanghu Lake, corresponding to a decline in small benthic fragilarioid taxa and an increase in A. minutissimum abundance and diversity during ~50‒170 CE, ~430‒560 CE, ~830–970 CE, and ~1640–1800 CE (Fig. 3a–c). These findings reinforce the response of diatom assemblages to changes in ice cover duration in Shuanghu Lake.

However, the observed patterns of percentage variation in these diatom taxa do not apply to the period from the 1870s to the 2010s CE (Fig. 2). During this interval, both small benthic fragilarioids and A. minutissimum exhibited low abundances, while taxa such as A. acares and M. atomus became more prevalent. Results from detrended canonical correspondence analysis (DCCA) and regime shift index (RSI) indicate a statistically significant regime shift in diatom composition at ~1870 CE (Fig. 3f), probably due to abrupt environmental changes linked to intensified human activities since the 1870s. This will be further discussed below.

Impact of European air pollutants on the shift in lake ecology since the 1870s

The drivers of the observed diatom assemblage shift in Shuanghu Lake since the 1870s require systematic investigation to elucidate potential anthropogenic and climatic forcing mechanisms. The increase in A. acares (Fig. 4b), a small diatom often associated with slightly acidic waters in Canadian subarctic lakes43, from the 1870s to the 2010s suggests an acidification process in Shuanghu Lake. Additionally, the substantial occurrence of the acidophilic Tabellaria flocculosa (Roth) Kützing44,45,46, identified in core SHW21F obtained from the lake’s center in 202147, supports this acidification process (Fig. 4b). Although the tychoplanktonic T. flocculosa could theoretically benefit from shorter ice-cover periods, its relative abundance (Fig. 4b) showed no correlation with fluctuations in ice cover duration (Fig. 4a) over the past two millennia, indicating insensitivity to ice phenology in Shuanghu Lake. The absence of T. flocculosa in core SH12B, in contrast to its presence in core SHW21F, can be explained by habitat differentiation between the two sites, with water depth (8.2 m vs. 15.3 m) being a potential contributing factor. Lake acidification is commonly associated with several drivers, including long-term natural acidification48,49,50, land-use changes (e.g., afforestation and the decline of burning and grazing)45,50, and atmospheric acid deposition45,50,51. However, natural acidification driven by catchment processes such as base-cation leaching and catchment paludification typically occurs over post-glacial timescales48,49,50. Although current warming-induced glacial melt and subsequent soil leaching can increase terrigenous inputs52 and drive lake acidification49, geomorphological evidence reveals the absence of modern glaciers within the catchment53, ruling out meltwater influence. Similarly, while increased precipitation could theoretically promote lake acidification through enhance leaching, hydroclimatic reconstruction from arid central Asia indicates a pronounced regional drying during the Current Warm Period (Supplementary Fig. 1)54, thereby excluding it as a potential driver. Furthermore, the lake’s pristine surroundings, largely undisturbed by human activity, precludes land-use change as a contributing factor. We therefore conclude that the acidification of Shuanghu Lake was primarily caused by atmospheric acid deposition.

Fig. 4: Ecological regime shift in Shuanghu Lake around 1870 CE and associated drivers.
figure 4

Variations in the abundance of a small benthic fragilarioids (this study), b A. acares (black line) (this study), T. flocculosa (grey diamonds)47, c M. atomus (this study), and d Cladopelma22 in Shuanghu Lake reflect the processes of acidification and eutrophication during ~1870-2012 CE (yellow shaded bar). e Rb/Sr ratios in sediments (this study, gray line: raw data; red line: 80-point smoothing) and f reconstructed vegetation cover (this study) show no evidence of increased erosion intensity during ~1870-2012 CE, suggesting minimal erosional contributions to lake acidification and eutrophication. g Anthropogenic CO2 (black line) and SO2 (grey diamonds) emissions in Europe during 1750–2022 CE62,63 show consistent patterns with the acidification and eutrophication processes in Shuanghu Lake, highlighting the pronounced impact of European air pollutants. Changes in h nitrate concentration (black line) and δ15N (grey diamonds) from a Greenland ice core reveal a coherent signature of anthropogenic pollutant deposition in remote Northern Hemisphere regions since the 1870s67.

Full size image

The pronounced rise in M. atomus, a eutrophic species55 linked to elevated nitrate levels56,57, since the 1870s (Fig. 4c), indicates the eutrophication of Shuanghu Lake. This is supported by the increase in Cladopelma, a chironomid taxon associated with mesotrophic aquatic environments58, observed in the same core22 (Fig. 4d). Lake eutrophication events are primarily driven by elevated nutrient inputs from increased catchment erosion and/or atmospheric nitrogen deposition24. According to the modern-process analysis of Rb/Sr in Shuanghu Lake59, the relatively low Rb/Sr ratios from the 1870s to the 2010s (Fig. 4e) reflect a decline in catchment weathering. Consistent with this interpretation, the pollen-based vegetation cover reconstruction in the Shuanghu Lake area (validated against MOD44B data, Supplementary Fig. 2) shows increasing vegetation cover over the same period (Fig. 4f), implying a concurrent reduction in physical erosion. Thus, we conclude that atmospheric nitrogen deposition, rather than catchment-driven processes, was the primary driver of eutrophication in Shuanghu Lake from the 1870s to the 2010s.

The acidification of Shuanghu Lake occurred several decades before the first recorded acid deposition event in China (Southwest China, 1920s)60. This temporal discrepancy, combined with the fact that lake eutrophication predates the onset of anthropogenic reactive nitrogen emissions in China (early 1980s)61, strongly suggests an external pollution source. Our analysis indicates that the acid and nitrogen inputs likely originated from long-range pollutant transport, potentially from upwind Europe. This interpretation is evidenced by: (i) the absence of local anthropogenic influence, (ii) the established fact that pollutant deposition preceded regional emissions, (iii) the temporal correspondence between recorded ecosystem changes and European industrial emissions (refer to anthropogenic CO2 and SO2 emissions in Europe during 1750–2022)62,63 (Fig. 4g), and (iv) the well-documented transcontinental transport of air pollutants from Europe to Asia via the Westerlies16,18,64,65. This transboundary pollution phenomenon since the 19th century is not an isolated event. For example, the NO3– concentration in an ice core from Dasuopu, Tibet (Fig. 1b), revealed the transport of nitrate from Nepal and India to the Tibetan Plateau since 1870 CE66. The record of nitrate concentration and δ15N of nitrate (Fig. 4h) in an ice core from Summit, Greenland, suggests the transport of Northern Hemisphere pollutants since the 1870s67. Although intensified dust deposition during 1870s–2010s68 may have enhanced nitrogen inputs, its ecological impact remains uncertain given the absence of eutrophication signals during historical dust-active periods.

Based on the above arguments, the decline of small benthic fragilarioids, indicative of an alkaline and nutrient-poor environment31,43, during ~1870‒1940 CE (Fig. 4a), can be attributed to acidification and eutrophication rather than a reduced ice-cover duration. The decline of A. minutissimum, a diatom associated with aquatic vegetation, alongside decreased diatom diversity in Shuanghu Lake, may also result from deteriorating water quality. Conversely, the late 20th-centruy increase in small benthic fragilarioids (Fig. 4a) and the decreases of A. acares, M. atomus, and Cladopelma (Fig. 4b–d) are attributable to decreased pollutant transport and reduced acid and nitrogen deposition in Shuanghu Lake. This likely represents a direct response to the decline in pollutant emissions from upwind Europe62 (Fig. 4g), analogous to the chemical recovery documented in European surface waters during the late 20th-century following similar emission reductions69,70,71,72,73,74. Additionally, the weakening of the Westerlies in the study area over recent decades75,76 may also have contributed to decreased acid deposition in Shuanghu Lake. While Shuanghu Lake has exhibited a discernible mitigation of acidification and eutrophication since the late 20th century, eastern China has concurrently experienced intensifying anthropogenic air pollution over the same period77,78. This pronounced spatial and temporal contrast reinforces the inference that transcontinental pollutant transport—rather than regional emissions—has been a primary driver of acidification and eutrophication in Shuanghu Lake.

In summary, changes in the diatom community structure in Shuanghu Lake since ~1870 CE were mainly influenced by acidification and eutrophication resulting from transcontinental transport and deposition of European air pollutants. This reflects a shift from naturally regulated processes to a more human-influenced ecosystem in the alpine lakes of the Altai Mountains (Fig. 5). Notably, recent rapid warming has been identified as a driver of ecological shifts in remote lakes, with effects mediated by extended growing seasons and associated limnological changes29. However, the absence of discernible acidification or eutrophication during past warm periods such as the Medieval Warm Period in Shuanghu Lake (Supplementary Fig. 1) demonstrates that climatic warming has not triggered such a regime shift in the lake’s history. When combined with the evidence that modern mean warming in arid northwest China did not surpass the intensity of the Medieval Warm Period (Supplementary Fig. 1)79, the hypothesis that climatic warming was the primary driver of the ecological shift around 1870 CE is insufficiently supported.

Fig. 5
figure 5

Schematic diagram depicting the regime shift from climate-driven to human-driven ecology in Shuanghu Lake since the 1870s.

Full size image

These findings provide valuable insights for ecological management and policy formulation in the context of global environmental changes, viewed from a long-term and large-scale perspective. For ecological management, this implies a necessary shift from a localized “watershed management” paradigm to an integrated framework that incorporates regional and global environmental drivers. Establishing an alpine lake monitoring network, for instance, could provide a cost-effective early-warning system against worldwide environmental threats. From a policy standpoint, our study supplies direct evidence from a 2400-year ecological record of a vulnerable ecosystem, supporting the implementation of stricter international emission standards.

Conclusions

This study investigates the large-scale anthropogenic impacts on the ecology of extremely remote lakes using a 2400-year diatom record from Shuanghu Lake in the Altai Mountains, supplemented by Rb/Sr ratios, vegetation reconstruction, and European emission data. We identify a distinct ecological shift around 1870 CE, marked by pronounced changes in diatom assemblages and corresponding RSI values. Prior to the 1870s, variations in diatom assemblages were predominantly influenced by changes in ice-cover duration, however, post-1870s, these variations became primarily driven by lake acidification and eutrophication processes. Considering the limited natural acid and nitrogen inputs and the absence of local pollution sources, we attribute these changes to long-range atmospheric transport and deposition of pollutants from upwind Europe. This inference is supported by the rapid recovery of the diatom community following late 20th-century European emission reductions. These findings highlight the vulnerability of even pristine remote alpine lakes to transcontinental pollutant transport, carrying important implications for ecological management and policy formulation from a long-term and large-scale perspective.

Materials and methods

study site

Shuanghu Lake (48°52'37″N, 87°01′52″E, 1515 m a.s.l.), located on the western side of Kanas Lake, consists of two small elongated lakes (Fig. 1c) formed by moraine damming processes53. The two lakes are in close proximity, ~150 m apart. Each lake is ~1180-m long and 330-m wide, with a surface area of ~0.3–0.4 km2 and a maximum depth of ~17 m (Fig. 1c). The catchment is underlain predominantly by biotite granite80. Climatically, the Shuanghu Lake area experiences short, cool summers and long, cold winters, with mean temperatures of 15.9 °C in July and –16 °C in January81. Annual precipitation varies between 400 and 700 mm82, with 40% falling as snow during winter and spring. In addition to direct precipitation, Shuanghu Lake is intermittently recharged by a seasonal stream along its western shoreline. The lake remains frozen from early December to late April. Surrounding vegetation consists primarily of forest species such as Pinus, Picea, and Betula, and alpine steppe22.

Coring and sediment age-depth model

In September 2012, we collected a sediment core (SH12B) from Shuanghu Lake using a Uwitec piston corer. The 1.55-m-long core was retrieved from a water depth of 8.2 m. The sediment composition consists predominantly of dark clay without distinct stratigraphic variations. In the laboratory, the core was cut lengthwise, and one half was sliced at 0.5-cm intervals, with the samples subsequently freeze-dried.

A previous study established the chronology for core SH12B using ten accelerator mass spectrometry (AMS) 14C dates provided by Beta Analytic (Florida, USA)22. These dates were obtained from samples of bulk organic matter, which had a reservoir effect of ~500 years. The age-depth model was constructed using CLAM v2.5.083 in R v4.4.184. In September 2021, a supplementary sediment core (SHW21E) was retrieved from an adjacent site (Fig. 1c), and the top 12.5 cm of sediment was dated using ²¹⁰Pb/¹³⁷Cs geochronology22. Due to the robust correlation between the ²¹⁰Pb/¹³⁷Cs age at 12.5 cm (SHW21E) and the ¹⁴C age at 11.5 cm (SH12B)22, we harmonized the age models by applying SHW21E’s ²¹⁰Pb/¹³⁷Cs chronology to SH12B’s top 11.5 cm.

Diatom analysis

A total of 155 sediment samples were collected at 1-cm intervals for diatom analysis, following established standard protocols85. Sample preparation included: (1) carbonate removal using 10% HCl at 80 °C for ~15 min; (2) oxidation of organic matter through gentle heating with 30% H2O2; and (3) separation of diatoms from mineral particles via sieving or heavy liquid flotation. The purified diatom suspension was then pipetted onto coverslips in a tray. After drying, each coverslip was inverted and permanently mounted onto a slide using Naphrax®. A minimum of 400 valves per sample was counted under oil immersion (OLYMPUS BX-51) at 1000× magnification to ensure robust statistical representation, with 138 samples meeting this criterion. All diatoms were identified with reference to Krammer and Lange-Bertalot86,87,88,89.

All analyses were performed on “major taxa” defined as those present in at least two samples with an abundance >2%, with square-root transformation applied to abundance data. Cluster analysis was conducted using Tilia v2.0.290 to identify statistically significant diatom assemblage zones. The alpha diversity of diatom spices in Shuanghu Lake (i.e., the Shannon index in this study) was analyzed using R v4.4.184. DCCA was employed to evaluate compositional changes using sample ages as the sole constraint29,91,92,93,94, which was performed using CANOCO 595 with detrending by segments and non-linear rescaling. To detect abrupt shifts in the diatom community, Sequential t-test Analysis of Regime-Shifts algorithm (STARS) was conducted on DCCA axis 1 (cut-off length = 5, p < 0.01)96,97,98,99, and the RSI was determined to quantify the magnitude of each breakpoint.

Elemental data

Rb/Sr ratios are often used as a proxy to assess changes in regional weathering intensity100,101,102 and their potential impact on lake ecology. In this study, X-ray fluorescence element scanning was utilized to measure the variations in Rb and Sr as counts. Scanning was conducted on the surface of the split sediment core using an Avaatech X-ray fluorescence Core Scanner, with the sediments scanned at 5 mm intervals. X-ray fluorescence scanning was conducted at the MOE Key Laboratory of Western China’s Environmental Systems, Lanzhou University, China.

Vegetation cover reconstruction

The Random Forest model, an ensemble machine learning algorithm based on decision trees, is effective for handling high-dimensional data and capturing nonlinear relationships103. In this study, we developed a Random Forest model to characterize the relationship between pollen assemblages and vegetation cover, using 905 modern pollen samples from vegetation types across the Altai Mountains104,105 and corresponding vegetation cover data from the MOD44B v6.1 Vegetation Continuous Fields product106. This trained model was then applied to reconstruct past vegetation cover based on a fossil pollen record from the SHW21F sediment core in Shuanghu Lake107. Model performance was evaluated by randomly withholding 20% of the modern pollen samples as an out-of-bag test set. To ensure robustness, the model was configured with 500 decision trees, and predictions were repeated 1,000 times with different random seeds. The model instance with the highest coefficient of determination (R2 = 0.65) and lowest root mean square error (RMSE = 11.57) was selected for application to the fossil pollen data. All Random Forest analyses were conducted in R (v4.4.1) using the randomForest package (v4.7-1.2)108. To assess the reliability of the pollen-inferred vegetation cover, it was compared with MOD44B-derived vegetation cover within a 40-km pollen source radius around the lake for the period 2000–2012 CE. However, due to the limited temporal resolution of the sediment record, the pollen-based reconstruction provides only two data points since 2000 CE, which is insufficient for a direct year-to-year comparison with the MODIS time series. Therefore, we compared the pollen-based reconstruction with the mean MODIS vegetation cover for the period 2000–2012 CE.