Introduction

Climate change is a potential major threat to human civilization and our living environment1,2,3,4. Understanding the impact of past climate change on societal development and human activities may provide valuable insights for addressing current and future climate challenges5.

The Western Zhou (1045–771 BC)—one of the most important dynasties in ancient China—was the largest geopolitical unity ever achieved by a single power until the rise of the Qin Empire in 221 BC6. At its apex, the Western Zhou controlled regions from the Liupan Mountains in the west to the Shandong Peninsula in the east, and from the Yanshan Mountains in the north to the Huai and middle Yangze Rivers in the south6,7 (Fig. 1). Moreover, the fundamental concept and system of Chinese civilization were developed during the Western Zhou6,8,9. For example, much of the Confucian tradition centers around core texts that originated during this period6.

Fig. 1: Locations of hydrological proxy sites mentioned in the text.
figure 1

1—Miaodong cave (this study), 2—Daihai lake45, 3—Gonghai lake25, 4—Xianglong cave46, 5—lacustrine sediments of the Pu Tian core47, 6—Liang’ercun archeological site11, 7—Heshang cave49, 8—JH001 sediment core from the Jianghan Plain50, 9—Xianyun cave (this study). The yellow circle denotes the capital of the Western Zhou. The red dashed line represents the maximum area controlled by the Western Zhou6,7. Background is the summer (June–September) precipitation trend during 1950‒2010 AD, based on a precipitation dataset with a spatial resolution of 0.025° (named LZU002574).

Full size image

The collapse of the Western Zhou dynasty remains an enigma in the history of early human civilizations. After 274 years of reign, the Western Zhou collapsed in 771 BC, accompanied by prolonged political disorder, social unrest, and population migration6,10,11. An important mark of the end of the Western Zhou was the eastward relocation of King Ping—known as the “Ping Wang Dong Qian”, which not only shifted the political center but also a large-scale migration of royal family members, officials, and associated groups6. Archeological12,13,14 and genomic evidence15 further support the occurrence of substantial population movements during this transitional period. Explanations for the collapse and population migration typically focus on military defeats, invasions, internal politico-economic conflicts and rebellions6,16,17,18. However, the potential role of climate change in the collapse of the Western Zhou and the subsequent population migration has been largely ignored, despite the fact that this period may coincide with the documented weakened East Asian summer monsoon (EASM) event at 2.8 ka10,11,19. Until now, the lack of paleoclimate records from East Asia, with sufficient age control, temporal resolution, and spatial coverage comparable to the historical and archeological records for the Western Zhou, has precluded an assessment of whether regional climate change contributed to the dynasty’s collapse and associated population migration.

Here, we present two precisely-dated multi-proxy (stable isotope and trace elements) stalagmite records from Northeast and Southeast China, spanning the interval from 2.95 to 2.50 ka BP. These new records, along with existing climate data, provide information on the changes in atmospheric circulation and hydrology in the Chinese monsoonal region around the 2.8 ka event, and they enable us to examine the impact of climate change on the collapse of the Western Zhou and the associated population migration via comparison with archeological data.

Results

Paleoclimatic records

Stalagmites MD12 and CXY1701 are from Miaodong cave (41.05°N, 125.52°E, 193 m above sea level (a.s.l.)), in northeastern China; and from Xianyun cave (25.55°N, 116.98°E, 968 m a.s.l.), in southeastern China, respectively (Supplementary Fig. 1). The stable oxygen (δ18O) and carbon (δ13C) isotope records of MD12 and CXY1701 are based on 240 and 214 measurements, respectively. The trace element ratios Mg/Ca, Sr/Ca, and Ba/Ca were obtained using the line scanning mode and are presented at a spatial resolution of 50 μm for both samples (“Methods”). The chronological frameworks of MD12 and CXY1701 are constrained by 5 and 11 230Th dates, respectively. Three different age-modeling schemes (linear interpolation, COPRA20, and ISCAM21 (“Methods”)) are generally consistent within the analytical uncertainties, and our conclusions are not sensitive to the choice of age model (Supplementary Fig. 2). We thus adopted ISCAM age models, because this scheme: (1) increases the signal-to-noise ratio, (2) reduced the age uncertainty of the records during the overlapping time intervals, and (3) enables the construction of a composite δ18O record based on multiple profiles21.

The average temporal resolutions of stalagmites MD12 and CXY1701 δ18O are 1.3 and 2.7 years, respectively. Stalagmite MD12 δ18O values vary between −9.84 ‰ and −7.33 ‰, with the average of −8.44 ‰, spanning the interval from ~2.591 to 2.893 ka BP (Fig. 2a); and stalagmite CXY1701 δ18O values range from −8.15 ‰ to −6.93 ‰, with the average of −7.55 ‰, spanning the interval from ~2.384 to 2.963 ka BP (Fig. 2b). The trace element records from stalagmites MD12 and CXY1701 provide an average resolution of better than 0.3 years. The Mg/Ca, Sr/Ca, Ba/Ca ratios of stalagmite MD12 fluctuate within the ranges of 0.010‒0.055, 0.010‒0.042, and 0.0005‒0.0030, respectively (Fig. 2a). The Mg/Ca, Sr/Ca, Ba/Ca ratios of stalagmite CXY17 fluctuate within the ranges of 0.020‒0.085, 0.046‒0.260 and 0.0002‒0.003, respectively (Fig. 2b).

Fig. 2: Multi-proxy time series and their correlation for stalagmites MD12 and CXY1701.
figure 2

a Stable isotope (δ18O, δ13C), trace element ratios (Mg/Ca, Sr/Ca, Ba/Ca), and PC1 of the trace element ratios (X/Ca) of stalagmite MD12 from northeastern China. The vertical gray bar highlights the 2.8 ka event. b Same as (a) but for stalagmite CXY1701 from southeastern China. c, d Correlation heat map of the stable isotope and trace element data for MD12 and CXY1701, respectively. The correlation is calculated using 3-year bins when the records had different resolutions. “*” denotes p ≤ 0.05, “**” denotes p ≤ 0.01.

Full size image

The two stalagmite δ18O profiles exhibit consistent multidecadal- to centennial-scale oscillations (Supplementary Figs. 3, 4a–c), with the most notable positive excursion being during ~2.65–2.77 ka BP (Fig. 2a, b and Supplementary Fig. 4). There are significant positive correlations among the Mg/Ca, Sr/Ca, and Ba/Ca time series (hereafter X/Ca) for each stalagmite (Fig. 2c, d). The X/Ca series for MD12 show very similar temporal patterns to its δ18O record, with pronounced positive excursions during ~2.65–2.77 ka BP (Fig. 2a). In contrast, the correlation between the X/Ca ratios and stable isotopes is much weaker in CXY1701 than in MD12 (Fig. 2d). During 2.50–2.95 ka BP, the X/Ca ratios of CXY1701 exhibit an overall W-shaped structure, with a negative excursion during 2.65–2.74 ka BP (Fig. 2b).

Discussion

Interpretation of stalagmite δ18O

The climatic significance of stalagmite δ18O from the EASM region remains controversial (e.g., refs. 22,23,24,25,26). Previously, we suggested that stalagmite δ18O in the northern margin of the EASM area can be used as a proxy for EASM intensity and associated monsoon precipitation. More positive stalagmite δ18O values imply a weaker EASM, a southward shift of the monsoon rainbelt, and hence, less summer precipitation in northern China, and vice versa27,28,29. This interpretation is supported by the significant positive correlation (r = 0.44–0.51, Fig. 2c) between δ18O and the X/Ca ratios of stalagmite MD12 from northeastern China (Fig. 2a, c), where the X/Ca ratios primarily reflect local hydrological changes.

However, stalagmite δ18O in monsoonal China, especially in southern China, may not always be a reliable proxy for local rainfall amount, given that the principal influencing factors (e.g., rainfall amount, precipitation seasonality, and moisture source) are timescale dependent30. Indeed, the stalagmite CXY1701 δ18O record from southeastern China shows a weaker correlation with its corresponding X/Ca ratios on decadal-to-centennial timescales (r = 0.24–0.33, Fig. 2d).

Although the dominant factors affecting stalagmite δ18O may vary across different areas and timescales, the similarity between the δ18O records in the EASM regions, on multidecadal-to-centennial timescales, suggests that these factors are fundamentally linked to changes in large-scale monsoon circulation (Supplementary Fig. 5). Therefore, consistent with previous studies (e.g., refs. 22,26), we propose that stalagmite δ18O reflects the overall EASM intensity associated with large-scale atmospheric circulation and the concomitant north-south movement of the monsoon rainbelt.

Interpretation of trace element ratios

Stalagmite X/Ca ratios have been widely used as proxies of local hydrological changes31,32,33,34, as they are primarily influenced by karst hydrological processes31,32,35. During arid periods, decreased water storage and infiltration in the karst aquifer can intensify the ventilation within the karst fracture network, triggering more intensive CO2 degassing from water. As a result, the water becomes supersaturated, causing calcite to precipitate before the drip waters reach the growing stalagmite. This phenomenon, known as prior calcite precipitation (PCP)32,35, leads to the enrichment of Mg, Sr, and Ba relative to Ca in the seepage water and thus in the stalagmites. Furthermore, during arid periods, the increased residence time of groundwater allows for more extensive water-rock interaction. This interaction can enhance the leaching and dissolution of Mg and Sr from the surrounding rock, leading to the enrichment of Mg and Sr in stalagmites35,36,37. In contrast, during wet periods, the opposite process occurs, resulting in lower X/Ca ratios.

The covariation among the X/Ca ratios in stalagmites MD12 and CXY1701 (Fig. 2c, d) supports the view that PCP and water-rock interaction are likely the primary factors influencing these ratios. Therefore, we interpret the X/Ca ratios of stalagmites MD12 and CXY1701 as indicators of local hydrological conditions. To effectively integrate the climate records, we performed principal component analysis on the X/Ca ratios of each stalagmite (Fig. 2a, b). The first principal component (PC1) of the X/Ca ratios explained more than 72% of the variance for both stalagmites, with higher values indicating drier conditions and vice versa.

Hydrological changes in monsoonal China during the 2.8 ka event

The 2.8 ka event was first reported as a cold event that occurred in the North Atlantic area at ~2.8 ka BP, known as Bond 2 event38,39. During this event, the temperature in the Northern Hemisphere dropped by ~0.8 °C40 (Fig. 3a). In the Asian monsoon region, numerous studies have reported a weak monsoon event corresponding to the 2.8 ka cold event, which occurred roughly between 2.90 and 2.65 ka BP27,41,42. However, due to the limitations imposed by dating uncertainties, sample resolution, and disparities in defining the event, there are discrepancies in the timing and structure of this event among different studies27,41,42. Our new composite δ18O record, with a time resolution of 1–3 years and an average age error of 12 years (Supplementary Fig. 4d), provides a rare opportunity to analyze this weak monsoon event. To identify this event, we followed our prior approach of defining Holocene climatic events28. Anomalies were defined as the positive deviation of δ18O values from the long-term average, with the central extreme portions exceeding one standard deviation (1σ). As a result, our δ18O record indicates that the 2.8 ka event in monsoonal China occurred mainly between 2.77 and 2.65 ka BP, centered around 2.70 ± 0.01 ka BP, and lasted for ~120 years (Supplementary Fig. 4d).

Fig. 3: Comparisons of Northern Hemisphere temperature, stacked stalagmite δ18O, hydrological proxy records and archeological data.
figure 3

a Northern Hemisphere temperature anomaly record40. b Stacked stalagmite δ18O record across the monsoon region of China (see Supplementary Fig. 5 for details). c PC1 of the MD12 trace element records (X/Ca PC1) (this study). d Pollen-based precipitation reconstruction from Gonghai lake25 (lighter curve) and Daihai lake45 (darker curve). e Sr/Ca record for stalagmite XL16 from Xianglong cave in Northwestern China46. f Pollen-based precipitation reconstruction from lacustrine sediments of the Pu Tian core in the middle and lower Yellow River areas (Henan)47. g Number of water wells in the middle and lower Yellow River areas11. h Environmental magnetic parameter IRMsoft-flux (representing detrital inputs) in stalagmite HS4 from Heshang cave49. i Grain size end-member fraction (EM4) from the JH001 sediment core from the Jianghan Plain, a proxy of paleofloods50. j PC1 of stalagmite CXY1701 trace element records (X/Ca PC1) (this study). k Number of archeological sites in the monsoon region of China64,75. Specifically, N = 208 archeological sites and n = 449 radiocarbon dates are included in the dataset. l δ13C data of human bones of upper and lower socio-economic status from different archeological sites in Northern China61. ‘ב represents dry events recorded in historical documents (Supplementary Table 1). The vertical gray bar highlights the 2.8 ka event.

Full size image

During the 2.8 ka weak monsoon event, the regional hydrological changes in eastern China showed a “north dry‒south wet” dipolar pattern (Figs. 1, 3), like the spatial changes in modern summer precipitation in the region under monsoon weakening over the past few decades43,44. In northeastern China, our new stalagmite MD12 X/Ca PC1 reached its maximum during the 2.8 ka event, indicating severe drought conditions (Fig. 3c). The integrated analysis on modern observations and other paleo-hydrological records suggest the MD12 X/Ca record captures hydroclimatic variability representative of a broader region across main northern China (Fig. 3d‒g and Supplementary Fig. 6). Pollen-based precipitation reconstructions from Gonghai Lake25 and Daihai Lake45, in northern China, both show decreased precipitation during this event, within the age uncertainty (Fig. 3d). The Sr/Ca record from Xianglong cave in northwestern China indicates a dry climate during this event46 (Fig. 3e). A recent pollen-based precipitation reconstruction from lacustrine sediments of the Pu Tian core in the lower Yellow River region indicate a decline in precipitation during this event47 (Fig. 3f). Archeological evidence indicates that during the 2.8 ka event, the groundwater level in the lower Yellow River area was ~3 m lower than today, and the number of water wells increased significantly in the middle and lower Yellow River region11 (Fig. 3g). Furthermore, historical records19,48 document fourteen droughts during 2.6–3.0 ka, seven of which occurred during this period, suggesting a high frequency of droughts linked to this event (Fig. 3 and Supplementary Table 1). Due to limited historical records, these drought accounts provide qualitative support for climate anomalies, not quantitative evidence of drought frequency. The above results indicate that northern China experienced drought conditions during the 2.8 ka weak monsoon event. Conversely, in southern China, hydrological records from the Yangtze River valley indicate wet conditions during this event, as seen in the environmental magnetic record from Heshang cave49 (Fig. 3h) and a recent paleo-flood record from the Jianghan Plain50 (Fig. 3i). In southeastern China, our stalagmite CXY1701 X/Ca PC1 record also indicates an overall wet climate during the 2.8 ka event (Fig. 3j). This study confirms that the dipolar pattern of precipitation observed over the past few decades also occurred in the past, on the centennial timescale.

Possible mechanism of the 2.8 ka event

The 2.8 ka event coincided with a grand solar minimum (Homeric minimum) during the late Holocene51 (Supplementary Fig. 7a). This synchroneity supports the view that this event was driven by a change in solar activity27,41,52. The climatic impact of solar variability may be amplified by the “top-down” stratospheric ozone mechanism, and/or the “bottom-up” coupled atmospheric–oceanic mechanism53,54. Previous research based on annually laminated sedimentary records from Meerfelder Maar Lake in Germany, and on model simulations, has revealed that the low solar activity during the 2.8 ka event led to the strengthening and southward shift of the westerly jet (Supplementary Fig. 7c), via the “top-down” mechanism52. This strengthening and southward shift of the westerly jet can limit the northward extent of the southernly monsoon airflow penetrating into East Asia34,55. As a result, the major rainbelt remains in the south, causing increased precipitation in southern China and decreased rainfall in northern China. However, the response of tropical ocean processes to solar radiation may also play an important role53. During the Homeric minimum, an increase in El Niño-Southern Oscillation (ENSO) activity was observed56 (Supplementary Fig. 7d), which may have contributed to the dipolar precipitation pattern in eastern China. During the decaying phase of El Niño, cooler sea surface temperatures of the West Pacific may lead to the southward movement of the subtropical high and the rainbelt, resulting in more rainfall in southern China (especially in the Yangtze River valley) and reduced rainfall in northern China57,58.

Collapse of the Western Zhou Dynasty and population migration

Climate and civilization are closely linked. Climate influences the development and sustainability of societies via its effects on key factors, including agriculture, water resources, and economic stability. A favorable climate is a prerequisite for the maintenance of food production and water supplies, which are essential for the prosperity of civilization; while adverse climatic conditions—like droughts, floods, or temperature shifts—often led to societal stress or collapse2,3,4,30,59. Previous studies have shown that changes in the East Asian monsoon had a profound impact on the succession of Chinese dynasties59,60. However, the detailed impacts of changes in monsoon circulation and the regional hydroclimate on the collapse of dynasties, as well as on population migration, on the centennial timescale, are poorly understood. The 2.8 ka event coincided with the historical transition from the Western Zhou (1045–771 BC) to the Spring and Autumn Period (771–476 BC) in ancient China, which was a period of political and social turmoil10,11. The synchroneity of the 2.8 ka event and social unrest points to a potential causal relationship between them, and this climatic event thus offers the opportunity to examine the impact of climate change on the collapse of the Western Zhou and subsequent population migration.

Agricultural production was the mainstay of the economy of the Western Zhou17,18. Verses in ancient texts such as “With your ten thousand men all in pairs” (from the Yixi ode) and “And the ground is laid open by their ploughs; In thousands of pairs they remove the roots” (from the Zaishan ode) reflect the intensity of agricultural activity during the Western Zhou period. However, agricultural output was highly susceptible to climate variability, particularly in northern China, where rainfall was limited. During the 2.8 ka event, drought and cold in northern China likely had a severe impact on agriculture. When drought restricted growing-season moisture availability, cooling may have further impaired crop development by reducing growing-season temperatures or duration. Carbon isotope values from human skeletal remains in northern China show a decreasing trend after ~800 BC, indicating a dietary shift from primarily C₄-based foods (e.g., millet) to increasing reliance on C₃-based foods (e.g., wheat)61 (Fig. 3l). This transition likely reflects changes in crop availability and adaptive survival strategies. Furthermore, historical records from the “Zhu Shu Ji Nian” describe unusual seasonal phenomena—peaches and apricots ripened two months late in October (772 BC), suggesting marked climatic anomalies that would have disrupted agricultural productivity. These lines of evidence demonstrate that drought and cold during the 2.8 ka event had a direct and adverse impact on agricultural systems, particularly in the semi-arid regions of northern China. Therefore, the 2.8 ka event likely presented the northern nomadic tribes with resource shortages, forcing them to migrate southward to obtain resources and food. This may have led to frequent conflicts and wars between the nomadic tribes and the Western Zhou; indeed, inscriptions on bronze artifacts indicate frequent wars between the two during the late Western Zhou6. Therefore, the combination of adverse climatic conditions and wars seemingly weakened the Western Zhou, eventually leading to its collapse in 771 BC (2.721 ka BP), when nomadic tribes attacked its capital. The survivors of the Zhou Dynasty abandoned their homeland and moved ~400 km eastward across the gorges of the Yellow River to the Luoyang Basin6,10.

To further explore the impact of the 2.8 ka event on this population migration, we analyzed the spatial changes in the density of radiocarbon dates of archeological sites across the monsoon region of China before (2.87‒2.77 ka BP), during (2.77‒2.65 ka BP), and after (2.65‒2.55 ka BP) the 2.8 ka event (Fig. 4). This follows the assumption that the density of radiocarbon dates at archeological sites can reflect past fluctuations in human population size (e.g., refs. 62,63,64,65). The results show that, before the 2.8 ka event, the radiocarbon dates of archeological sites were concentrated in the administrative center of the Western Zhou Dynasty (Feng-hao, Fig. 4a), located in present-day Xi’an, indicating that the population was concentrated in this region. However, a remarkable shift occurred during the 2.8 ka event, when the number of archeological sites and associated radiocarbon dates decreased in previously inhabited regions, and the spatial distribution of radiocarbon dates of archeological sites became more dispersed (Figs. 3k, 4b). Concurrently, there was an increase in the number of archeological sites along the middle and lower reaches of the Yangtze River, in the southeastern region (Fig. 4b), indicating population migration to this area. After the 2.8 ka event, the numbers of archeological sites and radiocarbon dates increased significantly (Figs. 3k, 4c), and the middle and lower reaches of the Yangtze River became a new population center (Fig. 4c). In addition, recent studies provide archeological and genomic support for population migration during the transition from the Western Zhou to the Spring and Autumn period. For example, Li pottery, a cooking tripod originally from the Yellow River basin, spread southeast into the Jianghuai region after ~2.8 ka BP14. At the same time, agriculture in the Jianghuai region shifted from mainly rice farming to the mixed cropping pattern, which suggests that people from northern dryland farming areas migrated into this region and brought their agricultural practices with them14. Furthermore, ancient genomic data from 13 individuals at the Shangshihe cemetery in Yima, Henan, reveal genetic admixture from the Yellow River, Yangtze River, and Eurasian Steppe regions15. This indicates increased population migration and interactions between different regions during this period.

Fig. 4: Probability distribution of radiocarbon dates (14C) of archeological sites before, during, and after the 2.8 ka event (dots and shading).
figure 4

a Before the 2.8 ka event (2.87‒2.77 ka BP). The yellow circle denotes the capital of the Western Zhou (Feng-hao). b During the 2.8 ka event (2.77‒2.65 ka BP). c After the 2.8 ka event (2.65‒2.55 ka BP). Archeological radiocarbon dates are summarized from Wang et al.64 and Dong et al.75.

Full size image

The dispersal and migration of populations in China during the 2.8 ka event were arguably a consequence of adverse climatic conditions and wars (Figs. 1, 3). During the 2.8 ka event, prolonged drought and cold in northern China may have reduced crop yields, which would have led to food shortages and famine. Concurrently, conflicts with nomadic tribes exacerbated this situation, forcing the majority of the inhabitants of the Feng-hao area to abandon their settlements in search of more hospitable environments. In contrast, southern China experienced relatively favorable conditions during this period (Fig. 3h–j), which prompted the southward migration of population and led to an increase in the number of archeological sites in the middle and lower reaches of the Yangtze River (Fig. 4b). After the 2.8 ka event, the recovery of the monsoon, improved climatic conditions, and adjustments in crop structure in response to environmental stress collectively contributed to population growth across the monsoon region, and the middle and lower reaches of the Yangtze River became a new population center (Fig. 4c).

Conclusion

We provide the first robust evidence to show that climate deterioration was the underlying factor in the demise of the Western Zhou and subsequent population migration. Two precisely-dated multi-proxy (stable isotope and trace elements) stalagmite records from Northeast and Southeast China, covering the interval of 2.95–2.50 ka BP, were yielded. Our results show that the 2.8 ka weak monsoon event primarily occurred during 2.77‒2.65 ka BP, with a duration of ~120 years and centered on ~2.70 ± 0.01 ka BP. Combining this evidence with other palaeohydrological records, we observed a “north dry–south wet” dipolar pattern in the Chinese monsoon region during this weak monsoon event. During the 2.8 ka event, severe drought and cold in northern China profoundly affected the core regions of the Western Zhou, contributing to its decline. Concurrently, the drought and cold forced the southward migration of northern nomadic tribes in search of resources, which intensified conflicts with the Western Zhou, and these combined pressures ultimately led to the demise of the Western Zhou in 771 BC. Furthermore, by integrating archeological data, we found that adverse climatic conditions and conflicts in northern China may have triggered significant population migration toward southern China, where the climate was relatively humid.

Methods

Cave sites, modern climate, and samples

Miaodong cave (41.05°N, 125.52°E, 193 m above sea level (a.s.l.)) is in Huanren county, Benxi City, Liaoning Province, in northeastern China. This region has a temperate monsoon climate, with warm/wet summers and cold/dry winters. The annual mean rainfall and air temperature of the two meteorological stations near Miaodong cave are 752 ± 157 mm and 8.4 ± 0.7 °C, respectively (1951–2012) (Supplementary Fig. 1d). Over 80% of the total annual precipitation occurs from June to September (Supplementary Fig. 1a). Miaodong cave is developed in Ordovician and Cambrian carbonate rocks, at a depth of 30–70 m below the surface. The cave is 500 m long with a cave air temperature of ~11.5 °C and relative humidity greater than 95% all year27. Two-year (2012–2013) cave monitoring indicated that the drip water δ18O mainly inherits the signal of summer precipitation δ18O27.

Xianyun cave (25.55°N, 116.98°E, 968 m a.s.l.) is in Laiyuan county, 260 km southwest of Fuzhou City, Fujian Province, in southeastern China. Today, this region is dominated by a typical subtropical monsoon climate. Two meteorological stations near Xianyun cave show that the annual mean rainfall and temperature are 1641 ± 314 mm and 20.5 ± 0.4 °C, respectively (1951–2016) (Supplementary Fig. 1e). More than 50% of the total annual precipitation occurs from June to September (Supplementary Fig. 1a). Xianyun cave is developed in Permian limestone at a depth of 10–60 m. The cave is ~2500 m long and has a year-round air temperature of ~17.5 °C and relative humidity close to 100%66,67.

Stalagmite MD12 was collected from Miaodong cave, 400 m behind the entrance. MD12 is 169 mm long and composed of calcite27. Stalagmite CXY1701 was collected in the southern branch gallery of Xianyun cave, ~600 m beyond the entrance (Supplementary Fig. 1c). CXY1701 is 145 mm in length and composed of calcite. In this study, we focus on the lower 80 mm of MD12 and the 30–145 mm interval of CXY1701 (Supplementary Fig. 2).

230Th dating

The results of 230Th dating of stalagmite MD12 have been reported previously27. For the new stalagmite CXY1701, calcite samples with weights of ~15 mg were obtained along the growth axis using a 0.3-mm carbide dental burr. The U and Th isotope ratios were measured at the Institute of Geology and Geophysics, Chinese Academy of Sciences, using a multi-collector inductively coupled plasma-mass spectrometer (MC-ICP-MS) (Thermo Scientific Neptune Plus). The chemical procedure for separating U and Th68,69, and details of the instrumental procedures are given in previous studies70. 230Th ages relative to AD 1950 (with 2-sigma errors) were calculated using the half-lives in Jaffey et al.71 and Cheng et al.70, assuming an initial 230Th/232Th atomic ratio of (4.4 ± 2.2) × 10−6, and the values for a material at secular equilibrium with the bulk earth 232Th/238U value of 3.8.

Age models

We used point-wise linear interpolation, COPRA (Constructing Proxy Records from Age models)20 and ISCAM (intra-site correlation age modeling)21 schemes to construct the age models for stalagmites CXY1701 and MD12. The COPRA algorithm was used to generate 1000 random age realizations within the dating uncertainties, and the median ages are then used. The ISCAM approach uses a piecewise cubic Hermite interpolating polynomial between dates to obtain the best correlation among the CXY1701 and MD12 δ18O time series, calculated from 10,000 Monte Carlo simulations (within the age uncertainties). Confidence levels of the age-depth models were calculated against a red-noise background using 1000 pairs of artificially simulated first-order autoregressive time series (AR1).

Stable isotope analysis

The oxygen isotope data of stalagmite MD12, at a 1 mm resolution, have been reported previously27. To further improve the resolution, we milled calcite powder continuously along the stalagmite’s growth axis using a stainless-steel surgical blade. Twelve sub-samples per millimeter were taken and each weighing 120–150 µg. Sub-sampling of stalagmite MD12 was measured every third interval, equating to three samples per millimeter. For stalagmite CXY1701, we used a 0.3 mm carbide dental burr to drill 120–150 µg sub-samples at 0.5 mm intervals along the growth axis. A total of 481 sub-samples, 214 from CXY1701 and 240 from MD12, were measured. Stable isotope measurements of both stalagmites were performed on a MAT-253 mass spectrometer linked to an online carbonate preparation system (Kiel-IV) at the College of Geological Science, Fujian Normal University. The international standard IAEA-603 (δ18O = −2.37‰, δ13C = 2.46‰) and in-house standard CAI-13 (δ18O = −9.46‰, δ13C = −10.73‰) were bracketed to evaluate accuracy and precision. The one-sigma uncertainty was better than ±0.06‰ for δ18O and ±0.03‰ for δ13C. All isotope values are reported in per mill (‰) relative to the Vienna Pee Dee Belemnite Standard (VPDB).

Trace element analysis

The trace element ratios Mg/Ca, Sr/Ca, and Ba/Ca were measured by Laser ablation-inductively-coupled plasma-mass spectrometry (LA-ICP-MS) at the National Research Center for Geoanalysis, Beijing, China, using a New Wave UP213 laser ablation system (wavelength = 213 nm; Fremont, USA), coupled to a Thermo Finnigan Element Ⅱsector field ICP-MS. Helium was used as the carrier gas due to its advantages in signal enhancement72. The laser ablation was conducted in line scanning mode, targeting the growth axes of the stalagmites. The laser repletion rate was 10 Hz, with a laser spot diameter of 50 µm and scan speed of 10 μm/s. The length of a continuous sample scan is 2 mm each time. NIST SRM 610 and NIST SRM 612 were used for the external calibration standard and data quality control. The analysis yielded a total of 44,372 trace element measurements: 20,912 from MD12 (ranging from 2.5 to 75 mm) and 23,460 from CXY1701 (ranging from 50 to 138 mm). The trace element data for the stalagmites were then binned to achieve a spatial resolution of 50 μm.

Stacking the stalagmite δ¹⁸O records

To construct a composite δ¹⁸O time series reflecting the overall EASM intensity associated with large-scale atmospheric circulation, we stacked multiple high-resolution stalagmite δ¹⁸O records from the East Asian monsoon domain during the 3.0–2.4 ka BP interval. All individual δ¹⁸O records were first standardized (z-score) to remove site-specific differences in mean and amplitude. Each standardized record was then binned into 10-year intervals, as all included records have temporal resolutions better than 10 years. For each time bin, the median of the available δ¹⁸O values was calculated to represent the composite δ¹⁸O record, thereby reducing the influence of outliers. Additionally, we computed the standard deviation across all records within each bin to estimate the variability and uncertainty of the stacked curve.

Statistical analysis

Pearson correlation coefficients between the time series were calculated (Fig. 2 and Supplementary Figs. 3, 4). The statistical significance of the correlations was determined using the “isopersistent” test. The “isopersistent” test assesses the significance via a non-parametric (Monte Carlo) simulation of correlations with a large number (e.g., 10,000) first-order autoregressive (AR1) processes, with identical persistence properties as the original time series73.

Reporting summary

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