Draining of an ancient lake in the Hexi Corridor 4500 years ago triggered migration of the Hei Shan civilization

Given that early human communities would have been greatly influenced by their natural environment^1,2,3,4,5, the exploration of how landscape evolution has impacted the development of civilizations is of considerable value. The Silk Road has been a crucial link between East and West Eurasia, enabling trans-Eurasian cultural exchange throughout prehistory and history^6,7. Along the Silk Road, there is much evidence (in some cases preserved as government-designated monuments) of early human activity, dating back to before the route was established^{8,9,10,11,12,13,14,15,16,17,18,19} (Fig. 1a). What prevented these ancient humans from facilitating the transmission and exchange between western and eastern civilizations is an important question, pertaining not only to the historical development of the Silk Road but also to the interaction between people and the environment. Archaeological evidence indicates that the Silk Road emerged in prehistoric times, before the establishment of the Qin and Han dynasties, around 2200 BP (note that ‘BP’ is used in this paper to denote years before 1950 AD)²⁰. The regions along the Silk Road were affected by geopolitical change, frequent wars between settlements, and changes to the natural environment, which together limited the potential for sustained interaction²¹. There has, however, been limited research specifically exploring the relationship between the evolution of civilization and environmental changes within the Silk Road region.

a The spatial distribution of prehistoric sites along the Silk Road within the Hei River catchment; the Hei drains northward across the Hexi Corridor. The white dotted line represents the ancient Silk Road²¹. GLS Gulang Section (climatic record). b Distribution of dated civilization sites dominated by agriculture and religion (purple, orange, and white dots) within the Hei River catchment. Sites from which climatic records have been obtained are also marked (white squares): BHZS Beihaizi section, THS Tiaohu section, YCS Yanchi section. c The spatial distribution of the Hei Shan civilization marked by the Hei Shan rock paintings within the Hei Shan Gorge, excavated by the Shuangjingzi River through the Hei Shan (Supplementary Note 4). The imagery in b is sourced from Google Maps ©2025 Google, with data from Landsat/Copernicus. The imagery in c is sourced from Google Maps (Image ©2025 Google), with data from Maxar Technologies, Airbus, and CNES/Airbus.

Full size image

The Hexi Corridor in North West China, an area with an arid or semi-arid climate, is a key segment of the Silk Road and preserves, along the Hei River catchment, considerable evidence of ancient human settlements from both the prehistoric and historical periods²², representing a great opportunity for exploring human–environment relationships. These settlements were generally short-lived²², however, raising questions about why that was the case and whether it resulted from the influence of environmental changes in the region.

The Hei Shan rock paintings are preserved on exposed rocks on both sides of the valleys of the Duanshankou catchment, belonging to the Hei River system^{23,24,25,26,27} (Fig. 1b, c). These rock paintings portray scenes of nomadic life (for the importance of the Hei Shan civilization within the Hei River Basin, see Supplementary Note 2), indicating that the Hei Shan region once had a rich natural environment, with forests, grasslands, rivers and abundant wildlife, including bison, large-horned deer, rhinoceros, tigers and sambar deer. These depictions suggest that the emergence of the Hei Shan civilization (as indicated by the rock paintings) dates back to ~10,000–6000 years ago, spanning the late Upper Paleolithic to Early Neolithic^24,25 (for a detailed introduction to the Hei Shan civilization, see Supplementary Note 3). The environment of that time was entirely different from that of today. The Hexi Corridor is currently extremely arid and completely unsuitable for hunting. Additionally, the Hei Shan petroglyphs are found only within the Duanshankou catchment, which raises the question of how the evolution of this river might have supported the development of this civilization and why the evidence is not more widespread. The term “Hei Shan Civilization” is a name we have coined in this manuscript to refer to the people responsible for the Hei Shan rock paintings. It describes the unique cultural and archeological characteristics identified in the Hei Shan region, which can be differentiated from surrounding cultures.

Of potential relevance are lacustrine sediments, preserved in the upper reach of the Duanshankou, extending over a length of 12 km within the Hei Shan Gorge (Figs. 1c and 2a). These lacustrine deposits terminate at the NNW-trending Jiayuguan Fault, which is an actively developing thrust with a dextral strike-slip. Both vertical uplift and crustal shortening rates have been estimated at ∼0.1 mm/a since ∼420 ka²⁸. Geomorphic markers and dated landforms indicate southeastward fault propagation and hanging-wall uplift from the Hei Shan toward the modern Beida River channel²⁸ (Fig. 1b). Thermoluminescence dating from exploratory trenches along the Jiayuguan Fault suggests that five earthquake events occurred during the latest Pleistocene to Holocene²⁹. In addition, exploratory trenches have also allowed documentation of Holocene activity of several NNW-trending subsidiary faults on the southwestern side of the main Jiayuguan Fault scarp^30,31. Tectonic activity and climatic fluctuation are probable causes of environmental changes that notably influenced the rise and fall of the civilization represented by the Hei Shan rock paintings in this area.

a Lake deposits overlaid onto Google imagery, showing the spatial distribution of the outcrops. b Elevation of the ancient channel and field photos of the outcrops. We can observe from the reach between outcrops 2 and 3 that the underlying bedrock has been uplifted above the general level of the ephemeral channel of the Duanshankou. To determine the longitudinal profile of the palaeochannel, we surveyed the elevation of the modern river valley, using differential GPS (Huace T12 PRO), with an average point spacing of 0.2 m (Fig. 2b). In addition, we performed single-point measurements at locations where lacustrine deposits interface with the underlying strata (gravel layers or bedrock) exposed within the valley. These data were used to correct the modern riverbed elevation and accurately restore the palaeochannel morphology. All data were processed and visualized using ArcGIS software. The longitudinal profile of the palaeochannel reveals channel deformation affecting the river. c Stratigraphic panel showing lithostratigraphy, sedimentology, and OSL ages at outcrops along the modern Duanshankou River. The imagery in a is sourced from Google Maps (Image ©2025 Google), with data from Maxar Technologies, Airbus, and CNES/Airbus.

Full size image

As a means of exploring the evolution of the Duanshankou catchment and its links to the Hei Shan civilization, particularly in relation to climate change and tectonic activity, we focus here on the lacustrine sediments preserved on both sides of the valley. We systematically examined the sediments from seven well-exposed outcrops, looking at sedimentology, geomorphology, and chronology in the late Holocene (Meghalayan) record.

Profile and chronological sequence based on optically stimulated luminescence (OSL) dating of sediments from cross-sections along the Duanshankou valley

The longitudinal profiles of rivers can provide valuable understanding of neotectonic perturbations in an area³². To ascertain the longitudinal profile of the Duanshankou, we surveyed the elevation of the ancient river channel employing differential GPS (Huace T12 PRO) (Fig. 2b). The valley currently occupied by the Duanshankou was originally occupied by the Shuangjingzi, a larger river that drained northwards from the Qilian Shan. Due to the coupled effects of tectonic activity and climate change, the Shuangjingzi altered its course, abandoning this section of valley, which continued to be drained by the ephemeral Duanshankou. The longitudinal profile reveals valley-floor deformation in this reach.

Furthermore, sand and silt deposits are preserved within the valley. To evaluate the sedimentary processes involved in their deposition, we conducted detailed sediment surveys at seven locations (Fig. 2), including outcrops 1 to 7 (Fig. 2b, c). A total of twelve OSL samples were taken from the seven outcrops (Fig. 2c and Table 1).

Outcrop 1, located furthest upstream, showed five main strata and a total thickness of 3.4 m. The lower unit (0.5 m). It is composed of pebble–cobble gravels. From 2.6 to 2.9 m depth, there is a 0.15-m-thick medium sand, showing distinct horizontal stratification. Mixed deposits of fine to medium gravel and cobbles from 1.8 to 2.6 m are rounded and poorly sorted. From 0.6 to 1.8 m depth, a 0.8-m-thick layer of gray–black fine and very fine sand, and gray–brown coarse to medium–coarse silt displays rhythmic alternations. This is overlain by a 0.4-m-thick gray–brown, coarse to medium–coarse silt with rhythmic alternations. The upper unit is a 0.6-m-thick layer of gravel-bearing fine sand, exhibiting visible parallel stratification. Sample O1B, from the top of the basal unit, yielded an OSL age of 9.61 ± 1.22 ka, whereas Sample OIT, from the top of the upper unit, gave 4.58 ± 0.93 ka.

A ~3.1-m-thick sedimentary sequence is exposed at outcrop 2, comprising three lithological units: the lower unit comprises 0.7 m of mixed fine and medium gravels, generally poorly to modestly well rounded, with poor sorting. The middle unit is coarse silt interbedded with fine sand and showing worm burrows (Fig. 3). The upper unit consists mainly of coarse silt, with gravel. Three samples (O2B, O2M, and O2T) constrain the ages of lower, middle, and upper units, yielding ages of 9.31 ± 1.85 ka, 8.60 ± 1.10 ka, and 4.51 ± 0.69 ka, respectively.

Field photos of the outcrops in the Duanshankou valley.

Full size image

Outcrop 3, a 5.6-m-high cliff, exposes two lithological units, a lower unit composed of coarse gravelly deposits derived from the local bedrock and an upper unit of silt. The latter is dominated by thick layers of greyish yellow–brown medium–coarse to fine silt, interbedded with multiple layers of thin gray and yellowish gray fine to very fine silt. The silt layers exhibit well-developed horizontal bedding and rhythmic layering. At a depth of 1 m below the top, there is a 0.5-m-thick layer of fine to very fine horizontally bedded sand. An OSL age of 7.62 ± 0.76 ka (sample O3B) has been determined for the base of the sequence.

Outcrop 4, with a thickness of ~6.0 m, features a basal layer of well-rounded gravel, above which are layers of greyish yellow–brown fine silt, occasionally interbedded with thin layers of yellowish gray very fine silt. The silt layers exhibit a blocky bedding structure. Sample O4B, 5.9 m from the top, yielded an age of 7.10 ± 1.10 ka.

Outcrop 5 reveals five divisions. A 2.5-m-thick layer of silts sits above well-rounded gravel and is succeeded by a 1-m-thick coarse gravelly deposit derived from the phyllite bedrock. Above this, a greyish yellow-brown fine silt layer is interbedded with thin strata of greyish yellow medium silt, showing blocky bedding features. Coarse bedrock-derived deposits are preserved at depths of 0.3–0.4 m. Chronological analysis of samples collected from the well-rounded gravel unit and the uppermost silt deposits yielded ages of 45.89 ± 6.41 ka and 6.20 ± 0.79 ka, respectively (samples O5B and O5T).

Outcrop 6 (~7.2 m thickness) is dominated by thick layers of greyish yellow–brown fine to very fine silt, intercalated with thin layers of greyish yellow medium–fine silt. The silt layers exhibit blocky bedding. From 5.1 to 5.8 m depth, there are coarse bedrock-derived deposits. Additionally, the silt layers display horizontal and rhythmic bedding. We obtained ages of 6.71 ± 0.60 ka and 5.88 ± 0.47 ka (samples O6B and O6T) from the base and top, respectively, of the sequence.

Outcrop 7, located 3.8 m above the left bank of the Duanshankou River, revealed three divisions. The bottom and top consist of 0.8 m of coarse bedrock-derived deposits. The middle division predominantly comprises thick layers of greyish yellow–brown fine to very fine silt, interbedded with thin layers of yellowish gray medium–fine silt. The silt layers exhibit massive bedding. The sediments at the top of outcrop 7 are dated to 6.32 ± 0.55 ka (sample O7T).

In the various outcrops, we have identified fluvial, lacustrine, and colluvial (piedmont) facies. Fluvial facies appear both below and above the lake sediments. The basal gravel is well sorted and has moderate roundness, and its compositional characteristics and sedimentary structures indicate a high-energy fluvial depositional environment, suggestive of a large river draining from the northern Qilian Mountains. We attribute these deposits to the Shuangjingzi, which once occupied the Duanshankou valley (Supplementary Note 4). The laminated sands and silts, exhibiting clear lacustrine affinity, provide the record of the palaeolake, including features such as burrows.

Above the lacustrine facies are further sediments attributable to fluvial deposition, albeit differing from the basal gravels in grain size, which we interpret as representing Duanshankou drainage, following the demise of the lake. The final facies consists of layers and “stringers” of bedrock-derived detrital material that has accumulated at the edge of the valley, potentially coeval with the lacustrine silts in the valley centre.

In general, the variation in grain size directly reflects the lacustrine laminae in the seven outcrops and is closely consistent with the field observations (Fig. 2b, c). The sediments in these outcrops can be correlated based on the chronological sequence and grain size (Fig. 2c). Samples (O1B and O2B), from the bottom of lacustrine sediments in outcrops 1 and 2, yielded ages of 9.61 ± 1.22 and 9.31 ± 1.85 ka, whereas samples O1T and O2T, from the top of lacustrine sediments in the outcrops 1 and 2, were dated to 4.58 ± 0.93 and 4.51 ± 0.69 ka. Thus dates from the bottom and top of the lacustrine sediments in both outcrops are of mutually similar ages, suggesting sedimentation during the same period. This is further supported by the consistent sequence of variations in sediment grain size observed in both outcrops. The ages of samples O3B and O4B, from the bases of outcrops 3 and 4, respectively, gave ages of 7.62 ± 0.76 and 7.1 ± 1.1 ka, also indicating synchronous deposition. Sediments spanning outcrops 1–4 demonstrate coherence in grain-size distribution (Fig. 2c). Furthermore, OSL ages from samples at the bottom and top of outcrops 5–7 are predominantly 5–6 ka. Notably, within outcrop 4, a layer of very fine sands is observed at a depth of 4.5–5.0 m, potentially contemporaneous with the coarse bedrock-derived deposits in the lower parts of outcrops 5–7.

The lithostratigraphy thus consists of basal gravels, with modest rounding, attributed to fluvial transport from the Qilian Shan (Fig. 1b), followed by parallel or massive fine sand and silt, marking the formation of the lake. The uppermost fluvial gravel records the draining of the lake and reestablishment of the fluvial regime.

According to the chronological sequence and lithological correlation obtained from the outcrops, combined with the spatial distribution of lacustrine sediments, it is possible to reconstruct landscape evolution in the Duanshankou valley during the Holocene (Fig. 4g). The longitudinal profile of the palaeochannel reveals deformation, indicating the formation of a natural dam that resulted in lake formation and lacustrine sedimentation.

a Palaeoseismic events of the Jiayuguan Fault and the surrounding faults^{29,30,31,33,34}. The bars indicate chronological intervals in which palaeoseismic events occurred. b Climatic record from magnetic susceptibility (MS) and lightness of sediments in the Yanchi Section⁴². In China, higher/lower magnetic susceptibility (MS) is generally accepted as an indicator of ameliorated/deteriorated combinations of temperature and moisture availability in response to strengthened/weakened East Asian summer monsoon (light green). Higher/lower lightness represents lower/higher organic matter content, a reflection of drier/wetter conditions (dark green). c Fe/K ratio and mean grain size of GL loess⁶⁰. A low Fe/K ratio and High mean grain size indicate a relatively dry climate. d Climate proxies (δ¹³Corg and Clay) from the Beihaizi section⁴³. High clay and negative δ¹³Corg (organic carbon isotopes) values indicate a relatively humid climate in Beihaizi paleolake. e Climatic record from Fe/Mn and Rb/Sr ratios of sediments in the Tiaohu Section⁴⁴. A high Rb/Sr ratio indicates weak chemical weathering intensity within the watershed, pointing to arid climatic conditions. Similarly, a high Fe/Mn ratio reflects a lowering of lake levels, suggesting a decrease in humidity in the study area. f Profile evolution of the valley-floor. g Distribution of the lacustrine sediments along the Duanshankou valley. The imagery in g is sourced from Google Maps (Image ©2025 Google), with data from Maxar Technologies, Airbus, and CNES/Airbus.

Full size image

OSL dating shows that the sediments in outcrops 1 and 2 date back to 9.5–4.5 ka, whereas those in outcrops 3–7 are concentrated between 7.6 and 6 ka. Furthermore, the longitudinal profile reveals uplifted terrain between outcrops 2 and 3 (Fig. 2b), suggesting tectonic disruption. These observations indicate that the lacustrine sediments can be divided into three deposition phases. During an older phase (9.5–7.6 ka), the lake formed in the area between outcrops 1 and 2. At 7.6 ka, this region became infilled, and sedimentation expanded to the area of outcrops 3–7, lasting until ~6 ka. In a younger phase (6–4.5 ka), sedimentation continued in the area between outcrops 1 and 2, with the lake eventually disappearing at 4.5 ka (Fig. 4f, g).

Landscape evolution of this valley was indeed controlled by fault movement and climate change. The gravels beneath the lake sediments date to ~45 ka, indicating that the valley remained unblocked at that time, despite probable earlier earthquakes (Figs. 2c and 4a, f). Palaeoseismological evidence suggests that uplift resulting from fault movement could have formed a blockage (dam), obstructing upstream drainage and runoff, and thereby leading to lake formation¹. The present research area is intersected by the Jiayuguan Fault, with five palaeo-seismic events recorded as occurring at 58.1 ka, 45.0–35.5 ka, 20.0–23.2 ka at ~12.9 ka and between 4.3 and 5.3 ka^29,30,31 (Fig. 4a). Furthermore, there are records of palaeoseismic events from the same periods affecting other faults in the vicinity of the Jiayuguan Fault^33,34 (Figs. 1b and 4a). Earthquakes occurring before 12.9 ka, and coupled with drying climate, likely contributed to deformation of the Duanshankou valley. At 9.5 ka, the magnitude of vertical displacement exceeded the river’s incision capacity, resulting in damming and lake formation. Climatic records from the Yanchi (YCS), Gulang (GLS), Beihaizi (BHZS), and Tiaohu (THS) sections (Figs. 1a, b and 4b–e) in the Hexi Corridor indicate that the climate was warm and humid at 9.5 and 6 ka. The abundant precipitation would have promoted lake formation as well as the input of copious sediment, eventually resulting in infill and potential overflow of the lake. Seismic activity between 5.3 and 4.3 ka caused further deformation, increasing the accommodation space and facilitating continued lacustrine sedimentation until ~4.5 ka. This coincided with a period of arid climate, marked by decreased precipitation and reduced sediment supply, which ultimately caused the cessation of lake sedimentation. As the climate became progressively wetter after 4.5 ka, the Duanshankou River began to incise, eventually leading to the destruction of the lake.

Therefore, the evidence for lacustrine conditions preserved in the Duanshankou catchment may have resulted from a combination of tectonic activity and climate change. There are similar examples of lake formation from other parts of the world. For instance, a lake located within the frontal thrust belt of the Betic Cordillera, SE Spain, is dammed by an antithetic fault¹. Similarly, this phenomenon is also observed in Mississippi drainage².

The processes of fluvial response to climate fluctuation and tectonic activity have played key roles in the evolution of ancient civilizations³⁵. Based on the sediment chronology, the lake within the Duanshankou catchment was formed at ~9.6 ka and disappeared at ~4.5 ka. It would have provided a habitat for a wide range of wildlife and nurtured the civilization responsible for the Hei Shan rock paintings^24,25 (Fig. 4g). The rock art depicts scenes of ancient humans primarily engaged in hunting, with illustrations of them using bows and arrows (Supplementary Fig. 2). Yao et al.¹⁹ reported the presence of arrowheads in this region dating back to 7–5 ka, further supporting linkage between the Hei Shan civilization and the evolution of the Duanshankou River. Moreover, the depiction of ferns in these rock paintings has important environmental implications, as such plants were only present in the Hexi Corridor during the warm period from 11,000 to 7000 years ago²⁴. This timeframe provides an ecological basis for dating the rock paintings. Notably, the arrangement of dancers in sequences and the design of their headdresses, featuring dance motifs found in the Hei Shan rock paintings, bear similarities to those in various cultures across different periods and geographical regions³⁶. These patterns bear a striking resemblance to the painted pottery designs of the Halaf and Ubaid cultures in Mesopotamia from 6000 to 7000 years ago, as well as to rock paintings, painted pottery, and other decorative artefacts found in the Indus Valley, West Asia, Southern Europe, the Americas, and Russia³⁶. This transcontinental cultural similarity, combined with the unique geographical ___location of the Hexi Corridor and its network of rivers flowing through mountain passes, strongly suggests that the region may have served as a crucial hub for Eurasian cultural exchanges during prehistoric times.

The challenges associated with inferring the age of rock art solely from motif content should be carefully noted, however. Specifically, motif-based analysis is constrained by factors related to cultural diffusion, individual artistic expression, and regional variation, making chronological attribution based solely on style potentially unreliable³⁷. Furthermore, many motifs, such as human figures, animals, and hunting scenes, are widely used across different time periods and regions, leading to ambiguity in chronological attribution³⁸. Additionally, cultural continuity or revival may result in similar symbols appearing in different periods, further complicating precise dating. The preservation state of the rock art also poses challenges, as weathering, erosion, and other natural processes over time can obscure or damage critical details necessary for stylistic comparison³⁹. Most importantly, the absence of direct dating methods or associated archeological materials for the rock art itself means that chronological interpretation based on content remains speculative without independent chronological control⁴⁰.

Nonetheless, the three lines of evidence (bow and arrow technology, fern depictions, and dance motifs) mutually reinforce each other, collectively illustrating how prehistoric humans might have adapted to the natural environment of the Hexi Corridor while simultaneously developing complex social networks through technological diffusion and cultural exchange. With the lake disappearing at ~4.5 ka, abundant water resources appeared in the lower reaches of the Beida and Hei rivers (Fig. 1b), leading to migration of the civilization from the foothills of the He Shan and Qilian Shan to the downstream delta, where agriculture flourished (Fig. 1b). During this period, the upper reaches of the Beida River also stabilized⁴¹. The stability of the river system and abundant water resources facilitated the further development of human civilization, giving rise to the agricultural civilization of the Hexi Corridor. With their basic needs met, the humans of the delta began to explore religious culture, and their long-term settlement eventually facilitated the establishment of the Silk Road. This indicates that localized environmental changes had a profound impact on the succession of prehistoric settlements, with environmental and landscape change forcing migration of human habitation. Similar cases have been documented in Indus Civilization, human migration in the Taklimakan desert, and Loulan Civilization in the Lop Nur^3,4,5.

Lacustrine sediments in the area near the Duanshankou (Such as at Yanchi, Beihaizi, and Tiaohu: see Fig. 1b) have been investigated previously, with particular focus on chemical indicators and climatic reconstruction; there is a lack of data from microfossils^42,43,44. Moreover, studies of lacustrine sediments from the broader region, such as those in Lake Qinghai^45,46, Lake Bosten⁴⁷, Lake Daihai⁴⁸, and Lake Donggi Cona^49,50, have also provided valuable data. This suggests that there is substantial potential for further research on the deposits of the Duanshankou palaeolake.

We suggest that palaeo-seismic activity coupled with climate change, by disrupting drainage in the Duanshankou valley to form a dammed lake, had a profound influence on the prehistoric civilization in this region. This lake would have offered long-lived (5 ka) favorable conditions that provisioned the Hei Shan civilization, well-known for the rock art they left. However, when the lake disappeared, the civilization was forced to migrate. Our results firmly suggest that the absence of early transmission between Eastern and Western civilizations was largely attributable to this disruptive and rapid landscape evolution, driven by tectonic activities and climate change at the northeastern margin of the Tibetan Plateau.

Sampling

As one of the main characteristics of deposits, grain size can reflect the dynamic type and transport mode of a transportation medium. In addition, granularity indices can be used as a basis for the analysis and explanation of genetic environments⁵¹. Analyzes of particle-size distribution have been widely used in the field of sedimentology as a key part of research on the evolution of geological environments⁵² and for the reconstruction of palaeoclimate and palaeoenvironment⁵³.

To reconstruct the evolution of the Duanshankou valley, seven representative sections of lacustrine sediments, namely the sampling sites outcrops 1–7, were chosen for grain size analysis (Fig. 2b). A total of 454 samples were taken at 0.05-m intervals from the upper lacustrine sediments in the seven outcrops. Moreover, 12 optically stimulated luminescence (OSL) dating samples were collected from the lower and upper parts of the outcrops. The OSL samples were collected from fine sand and silt layers in lacustrine sediments. During sampling, after cutting back 30 cm of surface material, a 30 cm-long stainless-steel tube was hammered horizontally into the newly excavated vertical section. Immediately after removal, the tube, containing the sample, was sealed at both ends with tinfoil and tape.

Grain-size analysis

All samples for grain-size measurement were analyzed at the Key Laboratory of Western China’s Environment Systems, Ministry of Education, China. They were boiled at 180 °C with 30% hydrogen peroxide (H₂O₂) to remove organic matter and then with 10% hydrochloric acid (HCl) to remove carbonates. Furthermore, ultrasonic pretreatment, coupled with the addition of 20% (NaPO₃)₆ solution, was also used to disperse the particles from each other for subsequent grain-size determination. Finally, these processed samples were measured with a Malvern Mastersizer 2000 to obtain grain-size distribution⁵⁴ (Fig. 2c).

OSL dating

Sample preparation for luminescence measurements took place at the Qinghai Institute of Salt Lakes, Chinese Academy of Sciences. From the innermost material in the sampling tubes, quartz grains (90–125 μm) were extracted using conventional sample preparation procedures (HCl, H₂O₂, sieving, density separation, and etching for 60 min with 40% HF). The purity of the quartz extracts was confirmed by the absence of a significant infrared stimulated luminescence (IRSL) response at 60 °C to a regenerative β-dose. The sensitivity to infrared stimulation was considered to be significant if the resulting IRSL signal amounted to more than 10% of the corresponding OSL signal⁵⁵ or if the OSL IR depletion ratio deviated by more than 10% from unity⁵⁶. The luminescence measurements were made with a Risø TL/OSL-DA-20 reader, equipped with high-power blue diodes emitting at 470 ± 20 nm and IR diodes emitting at 830 nm. All luminesce32nce emissions were detected through a 7.5 mm thick Hoya U-340 UV filter. Irradiations were carried out with a ⁹⁰Sr/⁹⁰Y β-source mounted on the reader. The equivalent dose (De) was determined using the single-aliquot regenerative-dose (SAR) and standard growth curve (SGC) protocol as described by Murray and Wintle⁵⁷. The environmental dose rate was calculated from the concentrations of U, Th, and K in the samples and from the contribution of cosmic rays⁵⁸. The water content was estimated according to the natural water content and saturated water content. The annual dose can be calculated according to the formula and parameters provided by Aitken⁵⁹. All OSL sample information and age results are presented in Table 1⁵⁴.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.