Introduction

Contemporary research increasingly highlights the critical interplay between climate change and human societal evolution, with paleoclimatic reconstructions offering unique insights into how environmental shifts influenced cultural trajectories. The East Asian Winter Monsoon (EAWM), a key component of regional climate systems1,2, exerts significant influence on temperature and ecological stability—factors central to agricultural sustainability and cultural resilience. While summer monsoon hydroclimate extremes have been linked to historical societal collapses3,4, the role of EAWM variability in shaping Neolithic cultural trajectories remains poorly constrained, particularly due to limited high-resolution paleoclimatic records. This study focuses on the CEC region encompassing the middle and low reaches of the Yellow River Basin (MLYR)—the cradle of Chinese Neolithic civilizations, including the Haidai and Central Plain cultures5,6. Dominated by monsoonal climate, this ecologically sensitive zone witnessed some major cultural transitions during the Holocene, coinciding with archeological evidence of settlement abandonments and population migrations5,6,7,8. Previous hypotheses associate culture transitions or dynasty changes with abrupt cooling episodes and EAWM intensification3,9,10, yet the absence of quantitative, temporally precise EAWM reconstructions has hindered understanding of the mechanism of climate-society interactions.

To address this gap, we present a decadal-resolution sea surface temperature (SST) record in Core SD-01 (Methods) from the Shandong Mud Wedge (SMW) in the northwestern South Yellow Sea, reconstructing EAWM intensity using the TEX₈₆ paleothermometer. Combined with archeological radiocarbon dates, fecal stanol biomarkers and compiled δ13C values of charred plant grains, human and animal bones from archeological sites, this dataset elucidates feedbacks between EAWM variability, the intensity of Neolithic human activity, and cultural adaptation. Our multi-proxy approach reveals that EAWM variability can result in the cultural flourishing or the decline of human activity as well as societal upheaval, with a lag of ~120 ± 40 years; while China’s Neolithic cultures can sustain societal continuity through incremental innovation rather than migration or collapse in other ancient civilizations. These findings underscore how Neolithic societies navigated climatic pressures, balancing innovation and adaptive strategies in response to monsoon variability. By integrating marine and terrestrial proxies, this study advances understanding of climate-society interactions, offering lessons for modern resilience in an era of anthropogenic climate change.

Results

Core SD-01 exhibits an undisturbed, high-resolution record of Holocene EAWM variability, reconstructed through marginal sea SST (Fig. 1A). The SST fluctuated between 6.9 ± 0.1 °C and 14.1 ± 0.2 °C over the past 11.5 kyr, revealing three distinct phases of EAWM evolution. From 11.5 to 6.0 ka BP, SST gradually increased and then stabilized with climate warmth, peaking at ~6.0 ka BP, indicative of a long-term weakening EAWM. This trend reversed during the mid to late Holocene (6.0–1.5 ka BP), with SST declining steadily from 13.9 ± 0.2 °C to 6.9 ± 0.1 °C, reflecting progressive EAWM intensification that culminated in peak EAWM strength at ~1.5 ka BP. Subsequently, SST showed a slightly warming trend with a large-amplitude fluctuation and two important peaks of colder conditions. Pronounced SST minima at ~11.2, 8.1, 6.2, and 4.1 ka BP align with global climate events, including the Younger Dryas, 8.2 ka, and 4.2 ka events11. Despite minor discrepancies in proxy sensitivity and resolution, our SST reconstruction agrees with other independent EAWM indicators, such as Chinese loess12, riverine Zr/Rb ratios13, grain-size data14 in the Okinawa Trough, and South China Sea alkenone-derived SST15 (Fig. 1), highlighting regional Holocene monsoon consistency. Our high-resolution SST record, constrained by robust chronology and high resolution, can offer superior precision in capturing millennial and centennial-scale EAWM dynamics, which are often unresolved in terrestrial archives.

Fig. 1: Holocene SST in Core SD-01, and other paleoclimatic forcing proxy records.
Fig. 1: Holocene SST in Core SD-01, and other paleoclimatic forcing proxy records.The alternative text for this image may have been generated using AI.
Full size image

A SST in Core SD-01 (this study), the error bar was calculated based on 4 replicates of treatment. B Stacked normalized mean grain size (MGS) from Chinese Loess12. C Zr/Rb from river-terrace sediment core DDW on the western Chinese Loess Plateau13. D Grain-size index in core Oki02 from the Okinawa Trough14. E Alkenone-derived SST in the South China Sea15. F 231Pa/230Th values in Core ODP 106347. G Middle- to high-latitude Northern Hemisphere (90°–30°N) temperature anomaly46. H Northern Hemisphere annual insolation at 65°N46. I δ18O in ice core NGRIP148. J Holocene ENSO index49.

Spectral decomposition of the SD-01 SST record via Ensemble Empirical Mode Decomposition (EEMD)16 isolates variability across timescales (Supplementary Fig. 1). Decadal to multi-decadal oscillations (IMFs 2-1) coexist with centennial-scale variability (IMFs 3–4; 44% variance) and sub-millennial to millennial cycles (IMFs 5–6; 49% variance). Spectral and wavelet analyses identify dominant periodicities at ~1500, ~1000, 700, 500, 400, 300, 250, 226, 111, and 106 years (Figs. S1 and S2). Millennial-scale oscillations (~700–1500 years) dominated the early-to-mid Holocene, transitioning to centennial-scale variability (~200–700 years) after ~6 ka BP (Supplementary Fig. 2B). Strong correlations (r = 0.94, p < 0.01) between 250 and 700-year bandpass-filtered SST and combined IMFs 3–5 confirm the robustness of these decomposed timescales (Supplementary Fig. 1).

Discussions

Determination of SST as an EAWM indicator

Thaumarchaeota, the primary producers of glycerol dibiphytanyl glycerol tetraethers (GDGTs), are found in specific environmental conditions that help refine the interpretation of TEX86-derived temperature values. These archaea thrive in coastal areas during the colder months with low light and a less stratified water column17,18,19,20,21. Core SD-01 (22 m depth) is located in a shallow coastal area where strong resuspension and enhanced coastal currents, driven by EAWM activity, lead to vertical water mixture (Supplementary Fig. 3) and sediment resuspension22. As a result, chlorophyll-a (Chl-a) concentrations are highest nearshore, particularly during the winter22 (Supplementary Fig. 3). A decreasing gradient of Chl-a is observed further from the coast (Supplementary Fig. 3), suggesting that primary production peaks in winter22. Consistent with this seasonal pattern, in situ measurements show that isoGDGTs concentrations are substantially higher in winter (8.3–251.4 ng/L) than in summer (0.7–5.3 ng/L) (Supplementary Fig. 4), aligning with observations from other marginal seas17,18,19,23.

To evaluate potential non-thermal influences on the TEX86 signal, we examined several diagnostic indices. Low Methane Index (MI < 0.3) and GDGT-2/Cren ratios (<0.2) in most water samples at in situ surveys (Supplementary Figs. 5 and 6) indicate minimal input from methanotrophic archaea24,25,26,27,28. In addition, branched isoprenoid tetraether index (BIT) values in Core SD-01 range from 0.03 to 0.28 (Supplementary Fig. 7), below the 0.3 threshold, suggesting negligible terrestrial GDGT contributions26.

During the winter of surveying period (June 2022–June 2024), ammonium (\({{{\rm{NH}}}}_{4}^{+}\)) and Nitrite (\({{{\rm{NO}}}}_{2}^{-}\)) concentrations in surface water (depth 5 m) at site SD-01 were highest (9–13 μmol/L and 1.4–2.6 μmol/L, respectively) (Figs. S8A and S8B), coinciding with an increasing trend of nitrate (\({{{\rm{NO}}}}_{3}^{-}\)) (Figs. S8A–S8C), which suggested net nitrification18,29. Furthermore, bottom sediment-trap data at site SD-01 revealed that isoGDGT fluxes peaked in January–February (Supplementary Fig. 8J), coinciding with winter maxima in ammonium and nitrite concentrations (Supplementary Figs. 8A–8C). And the terrestrial and marine biomarker ratio (TMBR) in sediment traps ranged from 0.17 to 0.32 (Supplementary Fig. 8G), indicating that the site is dominated by marine organic carbon30,31,32,33, which aligns with similar findings in the other coastal areas6,28,32,33,34. In addition, low MI values and GDGT ratios at most samples of sediment traps further confirm minimal terrestrial impact (Supplementary Fig. 8D), While low GDGT-2/Cren and GDGT-0/Cren ratios suggest limited terrigenous inputs (Supplementary Figs. 8E and 8F), aligning with the Chinese marginal seas where Euryarchaeota contribute little to isoGDGTs23,28,35,36,37,38. Low GDGT-2/ GDGT-3 ratios of <4 and Cren’ of <4 % in most survey samples (Supplementary Figs. 8H and 8I) are characteristic of shallow-water Thaumarchaeota communities, further supporting a dominant winter-season source23,28,38,39,40.

These conditions may favor the presence of planktonic archaea during winter, as strong winter winds enhance vertical mixing, enriching surface waters with nutrients that prime productivity. Additionally, phytoplankton and microphytobenthos contribute organic nitrogen17,18,41, which is mineralized ammonia, fueling nitrification by Thaumarchaeota18,42,43. Under strong EAWM conditions, the water column at Site SD-01 becomes thoroughly mixed with its low light levels, homogenizing temperature from surface to bottom, and further enhancing the activity of ammonia-oxidizing Thaumarchaeota28,44. This process, together with the winter-biased production of isoGDGTs by shallow-dwelling Thaumarchaeota, ensures that the TEX86 signal predominantly reflects winter SST rather than bottom-water temperature45. Therefore, the TEX86-derived SST record from Core SD-01 provides a robust and seasonally specific proxy for reconstructing Holocene EAWM variability.

EAWM variation mechanism on millennial and centennial timescales

The long-term evolution of the EAWM during the Holocene was fundamentally shaped by orbital-scale forcing, upon which pronounced millennial- and centennial-scale variability was superimposed. Early Holocene warmth, driven by peak NH summer insolation46 (Fig. 1H) and intensified Atlantic Meridional Overturning Circulation47 (AMOC; Fig. 1F), progressively weakened EAWM intensity (Supplementary Figs. 9B-9G). This phase supported stable temperatures across mid- to high-latitude NH regions48 (Fig. 1G), as evidenced by NH temperature stacks and other temperature reconstructions (Supplementary Fig. 9B–O). A regime shift at ~6.0 ka BP as declining insolation reduced the meridional temperature gradients, initiating an orbital-scale trend toward EAWM intensification through the mid- to late Holocene46,47,48 (Fig. 1F, H, I). Enhanced Siberian High pressure, coupled with increased sea–ice extent and continental cooling, strengthened the meridional temperature gradients, thereby intensifying EAWM through the mid to late Holocene. Late Holocene EAWM variability peaked at ~1.5 ka BP, marked by large-amplitude SST fluctuations (Fig. 1A). Moreover, the simulated EAWM suggests the transition of EAWM intensity at ~6.0 ka BP, supported by our SST record (Supplementary Fig. 10; other referenced cores are listed in Supplementary Data 1).

The SD-01 SST record aligns with Chinese loess12 and other temperature reconstructions46,47,48,49, confirming EAWM sensitivity to NH high-latitude dynamics46 (Fig. 1 and Supplementary Fig. 9). A dominant ~1500-year cyclicity in the SD-01 SST record aligns with Bond events-recurrent NH cooling phases associated with ice-rafted debris in the North Atlantic, superimposed on the orbital-scale trend15,46,47,48 (Fig. 1E–I, Supplementary Fig. 2). While the pacing of these cycles may be linked to solar variability or internal climate dynamics, their expression in EAWM intensity was modulated by the background orbital state. The delayed EAWM response to insolation maxima46 (Fig. 1H) reflects the integrated effects of ice melt and oceanic thermal inertia, evidenced by lagged SST and loess records. This orbital-scale trend, characterized by progressive EAWM intensification, was amplified after ~6.0 ka BP as the accelerated decline in NH insolation further diminished mid-to-high-latitude temperature gradients14,15 (Fig. 1D, E). Wavelet coherence analysis demonstrates that SST lagged solar irradiance (ΔTSI) on millennial scales (Supplementary Fig. 11), a pattern also observed in Chinese loess12 and sea-level records50.

Millennial oscillations form the background framework of Holocene climatic changes, whereas superimposed centennial-scale abrupt cold events (e.g., 11.1 ka, 8.1 ka, 4.1 ka BP) are clearly revealed in the SST records, synchronous with North Atlantic cooling episodes marked by hematite-stained grains51 (HSG) and drift ice tracers51 (Fig. 2A–C). The most pronounced cooling (8.4–7.8 ka BP) coincides with the 8.2 ka event, the Holocene’s most abrupt cold phase52,53. Meltwater-induced weakening of the AMOC during this event expanded the Siberian High54, intensifying EAWM winds and driving SST cooling in the marginal sea (Fig. 2A). It was interesting that the AMOC was weakest at 8.2 ka; the SST of Core SD-01dropped sharply. And after the weakest AMOC by about one century, the SST of Core SD-01 fell to its lowest point. It is implied that the coupled relationships between the high-latitude forcing and the mid-latitude climatic and oceanographic synergistic effects may need some time to reorganize. This pattern is also shown in the Okinawa Trough, where the sedimentary grain size showed a decrease14 (Fig. 1D), with the meaning of the strong EAWM. Additionally, the AMOC overall has been in a weakened state in the period around 8.2 ka, suggesting that EAWM has always been strong, even as it exhibits an occasional peak of the AMOC. Thus, this long-sustaining strong EAWM can result in the minimum SST in the Core SD-01 around 8.2 ka. Post-8.2 ka BP, seven strong and three weak cold events recurred at ~500–600-year intervals (Fig. 2A), broadly synchronous with centennial-scale increases in North Atlantic ice-rafted debris51 (Fig. 2B). These variabilities correlate with solar irradiance minima (ΔTSI)55 and atmospheric Δ¹⁴C anomalies56 (Fig. 2D, E), implicating solar forcing as a primary forcing. Cross-spectral analyses reveal that EAWM shares some periodicities (~700, ~500, ~250 years) with solar activity (Supplementary Fig. 12), suggesting that the solar activity is a potential forcing to influence the centennial-scale EAWM variability. This is evidenced by other climate records57. Declining solar output can weaken the AMOC and amplify meridional temperature gradients, thereby expanding the Siberian High, strengthening mid-latitude westerlies, and ultimately intensifying EAWM intensity. Our SST record exhibits strong covariance with North Atlantic cooling events (Fig. 2B), supporting the hypothesis that high-latitude ocean-atmosphere feedbacks modulated by both solar forcing and internal variability constituted the primary transmission mechanism for centennial-scale EAWM changes during the Holocene. These dynamics also align with stalagmite δ¹⁸O1,5,57, pollen records58, reinforcing the role of insolation variability in modulating EAWM intensity.

Fig. 2: Holocene centennial fluctuation of the Core SD-01 SST record and other paleoclimatic proxies.
Fig. 2: Holocene centennial fluctuation of the Core SD-01 SST record and other paleoclimatic proxies.The alternative text for this image may have been generated using AI.
Full size image

A Detrended SST. B Detrended HSG51. C Detrended Drift ice tracers51. D Detrended ΔTSI26. E Detrended Δ14C27. F Detrended Stacked normalized MGS12. The thick lines mean 200-a running average.

Holocene EAWM exhibited orbital- to centennial-scale variability driven by a combination of external radiative forcing and internal ocean–atmosphere dynamics. Against the backdrop of an orbitally-paced long-term trend, millennial-scale oscillations were primarily modulated by high-latitude ice-ocean feedbacks and atmospheric circulation changes. Weakening AMOC during the mid-Holocene further intensified these gradients by reducing northward oceanic heat transport, fostering stronger westerlies and EAWM. Centennial-scale fluctuations were synchronized with solar activity minima and North Atlantic cooling events, where reduced irradiance weakened AMOC, expanded the Siberian High, and enhanced mid-latitude westerlies. These processes propagated cooling signals from high-latitude ice-rafting episodes to East Asia via atmospheric teleconnections, linking North Atlantic perturbations to EAWM strengthening. In addition to primary forcings as high-latitude processes, EAWM was modulated by low-latitude climate forcings through dynamic interactions involving solar irradiance, El Niño/Southern Oscillation (ENSO), and oceanic-atmospheric teleconnections, shaping the EAWM’s intensity and spatial patterns. The Core SD-01 SST record was an anti-phase relationship with ENSO49, except for the early Holocene (Fig. 1J). The early-Holocene ENSO weakening was primarily driven by precessional forcing, which enhanced austral winter insolation over the South Pacific, leading to easterly wind anomalies and weakened ocean–atmosphere feedbacks49,59. At the same time, the enhanced NH isolation resulted in weak EAWM due to interglacial small ice sheets12,46 (Fig. 1B, G). During El Niño-like phases, weakened trade winds reduced equatorial Pacific zonal SST gradients, suppressing convective heating in the western Pacific Warm Pool. This suppressed convection shifted the Walker Circulation eastward, leading to Arctic sea–ice increase and the strengthening of the Siberian high, thereby strengthening the EAWM60,61, as evidenced by clay mineralogical data62, grain size63 in marine sediment cores. Our SST record highlights the EAWM’s role as a critical mediator between high-latitude and low-latitude climatic systems, with its variability reflecting both hemispheric insolation gradients and oceanic circulation changes.

Climate-driven agricultural stress and cultural transitions

Centennial-scale fluctuations in the relative intensity of Neolithic human activity (inferred from RGR) across the CEC exhibit pronounced synchronization with EAWM variability reconstructed from coastal SST records (Supplementary Fig. 13). We therefore hypothesize that these synchronous patterns reflect a climate-driven modulation of socio-ecological pressure, with societal responses lagging behind climatic triggers. Relative growth rate (RGR), reflecting demographic growth dynamics or the rate of change in human population size over time, reveals dominant ~500-year cyclicity in both the potential proxy for human activity and EAWM intensity (Supplementary Fig. 14), with the trends of human activity lagging climate shifts by 120 ± 40 years (Fig. 3 and Supplementary Fig. 15). This lag reflects the cumulative impacts of climatic stress on agricultural systems, and represents the time required for cultural, economic, and technological transformations to manifest in the archeological record, rather than an immediate societal collapse or migration. This response lagging in Neolithic societies to environmental changes also existed in other ancient cultures, e.g., ancient Arabia, Eastern Mediterranean, Akkadian empire and Egyptian Old Kingdom64,65,66,67,68. The duration of the response lag is not uniform and inconstant, and depends on climate forcing, but also on the level of socio-technical complexity. And it is demonstrated that dynamic response lags to environmental change were a variable feature of ancient cultures. The duration of these lags, spanning from years to centuries, was primarily a function of societal flexibility and environmental buffering. However, high mobile hunter-gatherer and pastoralist societies, e.g., Northern Arabia cultures67, exhibited an immediate, rather than lagged, adaptive response to arid conditions. And consecutive unexpected extreme climatic events may push an intensity of human activity beyond adaptation and resilience practices, resulting in a semisynchronizing response of ancient societies to climate changes69. Compared to the abovementioned ancient civilizations with low social complexity and technical level, the highly developed Neolithic ancient civilizations of CEC might experience longer lags due to a high level of socio-technical complexity, continuation of the civilizations of CEC.

Fig. 3: Comparsion of the EAWM variability with the prehistorical cultural development in the CEC.
Fig. 3: Comparsion of the EAWM variability with the prehistorical cultural development in the CEC.The alternative text for this image may have been generated using AI.
Full size image

Detrended SST in Core SD-01 represented by 250–700 a band-pass filter of the detrended SST (A), density cultural 14C probability (B), and RGR (C) in central plain and eastern China represented by 200–1000 a band-pass filter of the detrended RGR.

The coherence between RGR and EAWM at ~500-year intervals (Supplementary Fig. 15) aligns with solar-forced AMOC perturbations, suggesting high-latitude climate teleconnections regulated East Asian agro-ecological stability. As the isolation decreased, weakened AMOC amplified meridional temperature gradients, strengthening EAWM, which resulted in a catastrophic cold climate. This mechanism explains the tight RGR-climate coupling in Holocene records, as expanding Neolithic human activity heightened sensitivity to monsoon variability. In addition, the human activity intensity can also modulate cultural trajectories68. Even during favorable climatic periods, human activity intensity as proxied by RGR peaked within ~300 years before declining, underscoring inherent limits to unchecked expansion. Notably, Neolithic societies exhibited resilience to abrupt climatic stress.

While this study focuses on quantifying the Holocene variability of the East Asian Winter Monsoon (EAWM) and its societal impacts, it is crucial to frame its influence within the broader monsoonal system. The EAWM exerts direct pressure on Neolithic agro-societies through temperature stress on crops and growing seasons. However, its more profound impact likely operated in tandem with the EASM. Periods of intensified EAWM, as recorded in our SST reconstruction, were often synchronous with or led to weakened EASM precipitation, creating synergistic “cold-dry” climatic regimes that severely challenged agricultural resilience. Thus, the EAWM serves not only as a direct climatic stressor but also as a key indicator and potential driver of integrated monsoonal instability that ultimately shaped human adaptations. Early to mid-Holocene warmth, marked by weakened EAWM activity, fostered agricultural stability in the CEC, supported by increased precipitation as indicated by speleothem δ18O in Magou at Henan Province (Supplementary Fig. 9G). Millet cultivation thrived in semi-arid loess soils, evidenced by carbon isotope data (δ¹³C) from charred grains and bones (Supplementary Figs. 16 and 17). During the Longshan period, rice irrigation expanded under humid conditions, supporting human activity intensity and settlement complexity (Supplementary Figs. 18–20). Concurrently, social stratification emerged, evidenced by Haidai jade artifacts and elite burials70, reflecting advanced cultural development. However, from the Middle Holocene (MH) onwards, the SST record indicates a generally intensified EAWM (Fig. 4A-4D). This intensification, occurring alongside a remarkable decline in the East Asian Summer Monsoon (EASM) and its precipitation (Fig. 4E, F), disrupted the stability of the monsoonal system since MH2,5,12,71,72. This shift toward an overall drier regime increased aridity and drought risk across the CEC. It is worth noting that the abrupt 4.2 ka event was characterized by a ~4 °C SST drop (intensified EAWM) coinciding with a sharp decline in precipitation (weakened EASM) across the Yellow River catchment5,12,71 (Fig. 4A–D). Meanwhile, the precipitation in the whole Yellow River catchment showed a sharp decrease5,72,73 (Figs. 4E, J–L), and even the whole Northern China has experienced drought2 (Fig. 4F). In particular, in the MLYR, this dry condition persisted for about a century5 (Fig. 4K), accompanied by a pronounced and sustained temperature decrease (Fig. 4G). This compound cold-dry shift inflicted severe losses on the agricultural foundation of the Longshan Culture in the lower catchment of the Yellow River (LYR). A subsequent period of increased precipitation and warmer temperatures from 4.15 to 4.02 ka at the MLYR5 (Fig. 4G, K) provided a temporary but favorable warm-wet interval for Neolithic societies. This brief climatic amelioration may have contributed to the social processes leading to the foundation of the Xia Dynasty. However, a return to a series of severe dry and cold conditions ultimately led to the decline of the Longshan Culture5,73 (Fig. 4G, K, L). The surviving populations underwent a century-scale societal reorganization, culminating in the emergence of the Yueshi Culture at the LYR. Throughout these abrupt climate changes, the socio-demographic trajectory of the Neolithic culture in the CEC exhibited corresponding fluctuations, as supported by the coporstanol and 5β-stigmastanol fluxes (Fig. 4H, I) in Core SD-01.

Fig. 4: Middle Holocene SST record in comparison with EAWM index, precipitation and temperature as well as coporstanol index.
Fig. 4: Middle Holocene SST record in comparison with EAWM index, precipitation and temperature as well as coporstanol index.The alternative text for this image may have been generated using AI.
Full size image

Middle Holocene SST in Core SD-01 (A, G), Speleothem δ13C in Magou of Henan Province5 (‰) (B, K), Stacked normalized mean grain size (MGS) from Chinese Loess12 (C), mean atmospheric temperature anomaly (MATA) in Shandong Province71 (MATA in Shandong) (D), Mean annual precipitation (MAP) in Yazihai72 (E, J) and Gonghai2 (F), Coporstanol (H) and 5β-stigmastanol (I) flux in Core SD-01 and Speleothem δ13C in Wuya at Loess Plateau (‰)73 (L).

Collapsing rice systems and sharp shifts of floods and drought reduced human activity intensity suggested by the RGR, by ~120 years (Fig. 3 and Supplementary Fig. 21). In the Haidai region, societies adapted through crop diversification, shifting to drought-tolerant millets and marine resource exploitation on the coast—evidenced by elevated δ¹³C/δ¹⁵N values in human remains74. Defensive settlements75 and reduced ceremonial jade production70 signal societal downsizing (Supplementary Fig. 18). The archeological visibility and intensity as suggested by RGR of the Yueshi Culture sharply decreased (Supplementary Fig. 20) in support of the lower coporstanol and 5β-stigmastanol fluxes of the Yueshi Culture than ones of the Longshan Culture (Fig. 3H, L). This demonstrates that the social development level of Yueshi Culture in the Haidai region shows a decrease due to the extreme cold and drought events. Conversely, the Central Culture prioritized technological innovation that rudimentary irrigation stabilized millet agriculture76, enabling urban centers like Erlitou and increasing population (Supplementary Fig. 20). Elite-controlled granaries77 and dietary disparities (20–30% higher millet consumption among elites78,79) underscore centralized governance emerging from climate-induced scarcity. In addition, these divergent strategies highlight regional adaptability due to spatially heterogeneous climate impacts. Haidai societies leveraged ecological flexibility, while the Central Plain invested in organizational complexity (Table 1). Both trajectories were shaped by solar-forced EAWM variability, primarily forced through high-latitude ice-ocean feedbacks and mediated by low-latitude climate forces. The ~4.0 ka EAWM and EASM regime shifts catalyzed innovations of adaptation strategies—crop diversification, storage infrastructure, and political centralization—that buffered climatic stress. Such adaptations highlight that Neolithic societies in China can navigate viability thresholds through institutional and technological co-evolution, in which monsoon systems may be as non-linear amplifiers of climate–culture interactions.

Table 1 Adaptation strategies of Haidai Culture and Central Plains Culture under abrupt climate changes
Full size table

This observed synchronicity and phase relationship between EAWM variability (proxied by SST) and fluctuations in human activity intensity (proxied by RGR) lead us to hypothesize that centennial-scale climatic deterioration exerted significant pressure on Neolithic socio-ecological systems. The apparent ~120 ± 40 year response lag likely represents the time required for cumulative agricultural stress to manifest and for societies to implement adaptive strategies, which are visible in the archeological record. While methodological constraints prevent precise population quantification from radiocarbon databases alone, the convergence of multiple proxy indicators demonstrates a statistically significant correlation between EAWM intensification phases and transformations in socio-ecological systems across the CEC. This pattern supports the hypothesis that climate forcing acted as a critical external driver of adaptive responses during key transition periods in Neolithic cultural development. In addition, the ~120 ± 40 year lag may represent an aggregate, regional-scale signature derived from integrating archeological data across the CEC, and can be regarded as a regionally integrated lagging signal. Despite underlying local variations, identifying a statistically significant, coherent lag at the regional scale remains a meaningful finding. It suggests that centennial-scale EAWM variability acted as a pervasive, synchronizing pressure across the CEC, within which diverse local adaptations unfolded. The aggregate lag likely reflects the characteristic timescale required for climate stress to permeate agricultural systems and for broadly visible socio-cultural reorganizations (e.g., settlement pattern shifts, material culture changes) to manifest in the regional archeological record.

Our high-resolution EAWM record reveals that its centennial-to millennial-scale intensification acted as a potent climatic forcing on Neolithic societies. Importantly, this forcing was most effective when amplified by concomitant EASM weakening, leading to the dual stressors of cold and aridity. Therefore, the climatic forcings driving major population and cultural transitions in monsoonal East Asia were not exclusively the winter monsoon dynamics, but rather the integrated state of the seasonal monsoonal cycle, in which EAWM variability served as a critical and measurable component. This underscores the necessity of multi-proxy approaches that capture both temperature and hydrological changes to fully understand past human–climate interactions.

Holocene climate adaptations in China: a paradigm of resilience in global context

The Neolithic agricultural and cultural transitions in China’s Haidai and Central Plain regions exemplify a distinct pathway of societal resilience under Holocene climate stress. The EAWM reorganization at ~4.0 ka BP, driven by solar forcing and high-latitude ice-ocean interactions, triggered divergent adaptation strategies among sedentary Chinese Neolithic societies that differ from global patterns of climate-induced cultural collapse. Unlike migration pastoralist groups (e.g., Yamnaya steppe pastoralists) that relied on large-scale migrations to buffer ecological stress80, or monsoon-dependent civilizations (e.g., Mayan city-states81,82,83 and the Indus Valley84) that collapsed or deurbanized, under climatic upheaval, Chinese societies pioneered in situ adaptations that avoided systemic collapse. This adaptive dichotomy highlights a critical divergence in Holocene human–environment interactions: while many civilizations prioritized specialization or mobility, East Asian groups achieved sustainability through sedentary agroecological innovation and institutional flexibility.

The Haidai Culture and Central Plain Culture in China exemplify East Asia’s ecological diversification strategy. Facing unstable coastal climates, they developed a crop-diversification subsistence system combining millet cultivation, rice paddies in wetlands, and marine resource exploitation, as evidenced by carbon and nitrogen isotopic data consuming millet, rice and marine food74. This polyculture system buffered against monsoon-driven aridity and erratic rainfall, ensuring food security even during climatic extremes. This contrasts with the Yamnaya pastoralists’ reliance on horse-mounted mobility across the Eurasian steppe—a strategy effective for avoiding localized droughts but requiring constant territorial expansion85. China’s multispecies farming system provided a “portfolio effect,” spreading risk across crops with differing environmental tolerances—a strategy absent in these cases. By contrast, Yamnaya pastoralists abandoned fixed settlements during climatic deterioration86,87. The Maya, reliant on a monsoon-driven seasonal cycle, developed complex urban centers supported by intensive maize agriculture. During episodes of protracted drought, this specialized system, coupled with environmental degradation, contributed to political fragmentation and urban81. Similarly, in the Indus Valley, a society dependent on both monsoonal rainfall and riverine floods, a sustained weakening of the summer monsoon around ~4.1 ka is coeval with a widespread settlement shift and a decline in urban integration. These cases illustrate vulnerabilities associated with highly centralized and specialized subsistence systems in monsoon environments84,88.

The techno-social innovation was characterized by China’s cultural groups. The Central Plain’s response to East Asian Monsoon-driven climate changes included state-coordinated hydraulic engineering and granary systems89, implying early centralized governance. At Xinzhai, Erlitou and Shanggangyang sites, elites centralized grain storage in large ceramic silos, redistributing surplus during shortages77,90, contrasting with fragmented responses in other cultures. The deurbanization of the Indus Valley after ~3.9 ka BP was ascribed to lacking centralized storage or adaptive governance to manage water scarcity84,88. The Yamnaya culture (Eurasian steppe) relied on high mobility and relocated herds and populations to evade drought, rather than investing in permanent infrastructure85. In contrast, China’s emphasis on territorial stability and resource redistribution enabled societies to endure climatic stress without abandoning settlements. Ceramic evidence shows standardized storage vessels proliferated in Longshan Culture before the 4.2 ka BP event91, indicating proactive rather than reactive adaptation.

In addition, China’s cultural groups achieved cultural continuity through institutional memory. Unlike the Indus Valley’s abrupt deurbanization, which erased hydraulic engineering knowledge for millennia, China’s adaptation strategies became embedded in long-term cultural memory. The Erlitou culture inherited and scaled Longshan innovations, developing the earliest bronze ritual vessels for grain taxation—a system that institutionalized climate resilience through centralized resource pooling.

The resilience of Haidai and Central Plain cultures underscores a broader pattern that sedentary societies can thrive under climate stress through incremental innovation rather than migration or collapse (Fig. 5). Suffering coupled with EAWM intensification and EASM weakening, these societies did not collapse or widely abandon their regions. Instead, as our multi-proxy records indicate, they implemented a suite of in-situ strategies: ecological diversification (millet-rice-marine resources), techno-social innovation (centralized storage, rudimentary irrigation), and social reorganization. This integrated approach buffered against the specific “cold-dry” stress imposed by the synergistic weakening of the seasonal monsoon cycle. These adaptations contrast with the Mayan reliance on rigid infrastructures, the Yamnaya’s mobility-dependent pastoralism, and the Indus Valley’s urban inflexibility. This study documents that the Neolithic societies in Central and Eastern China navigated mid-to-late Holocene monsoon variability through a strategy predicated on diversification, storage, and social coordination, which facilitated long-term continuity. While the specific manifestations of resilience are context-dependent, this East Asian case expands the known repertoire of successful human adaptations to monsoon climate variability. It underscores that sustainability in such environments can be achieved through in situ innovation—rooted in ecological flexibility and social coordination, offering a deep-time perspective on adaptation pathways for monsoon-dependent regions facing anthropogenic climate change.

Fig. 5: Schematic drawing of different cultural adaptations responding to climate change.
Fig. 5: Schematic drawing of different cultural adaptations responding to climate change.The alternative text for this image may have been generated using AI.
Full size image

Schematic drawing of different cultural adaptations responding to climate change.

Methods

Core location, sediment sampling, and chronologies

Core SD-01 was drilled in July 2020 off the Shandong Peninsula in the southern Yellow Sea (122°32.22’E, 36°32.93’N, water depth of 22 m; Fig. 6) by the marine drilling vessel Kan 407. The information on the lithology, sedimentary grain size, color, and radiocarbon chronology is reported elsewhere92, shown as in Supplementary Table 1 and Supplementary Fig. 22. According to the age model, the 3427 cm of the core covers the last 11470 years. The average sedimentation rate of the core is as high as 4.66 mm a−1, with a high resolution of ∼11 a for the SST record.

Fig. 6: Locations of Core SD-01 and other referred sites.
Fig. 6: Locations of Core SD-01 and other referred sites.The alternative text for this image may have been generated using AI.
Full size image

YSCC is the Yellow Sea Coastal Current. YSWC is the Yellow Sea Warm Current. CDW is Changjiang Diluting Water. ZMCC is the Zhe-Min Coastal Current. The location data of the referred sites are shown in Supplementary Data 1.

210Pb and 137Cs chronologies also were determined in the shorter core (SD-01-1), shown in Fig. S22. 210Pb and 137Cs were determined by gamma spectrometry. About 5 g of dried ground sample was placed into polyethylene calibrated geometries (length = 40 mm, diameter = 10 mm) sealed with rubber caps and Teflon tape. Samples were stored for three weeks to allow secular equilibrium between 226Ra and 214Pb. Samples were measured in a gamma spectrometry system with a high-resolution, low-background HPGe well-type detector (Ortec-Ametek) for at least 48 h. The excess 210Pb (210Pbex) activities were calculated by subtracting 226Ra (via 214Pb) from the total 210Pb activities. Analytical quality was assured by replicate analysis of the reference material IAEA-257. The 210Pb age model based on the constant rate of supply (CRS) model93 was constrained by the 137Cs peak at 1963 AD (Supplementary Fig. 22).

Lipid biomarker analysis

A 2–5 g freeze-dried sediment from SD-01 core (n = 329) was extracted for GDGT analysis by the accelerated solvent extraction method (ASE 200 system, Dionex, USA) at 120 °C and 100 bars with dichloromethane/methanol (9:1, v/v). The total lipid extract was subsequently separated into polar and apolar fractions using a column packed with Al2O3 using hexane/dichloromethane (9:1, v/v) and dichloromethane/methanol (1:1, v/v) as eluents. The polar fraction was then filtered through a 0.45 μm, 4-mm diameter PTFE filter prior to injection. The GDGTs were measured on the Agilent 1200 series high-performance liquid chromatography via an atmospheric pressure chemical ionization coupled with mass spectrometry (HPLC-APCI-MS)26. Both isoprenoid and branched GDGTs were analyzed within a single acquisition run for each sample. Quantification was performed using the Analyst software, and the peaks were integrated manually for each sample multiple times. The TEX86 and branched and isoprenoid tetraether (BIT) indices were calculated using the published equations26,94,95.

RGR in central and eastern China

It is important to acknowledge that summed probability distributions (SPDs) of radiocarbon dates and derived metrics like the RGR are widely used but indirect proxies for past human activity, inferring past demographic trends from archeological radiocarbon dates. Their trajectories represent a composite signal influenced by factors including taphonomic processes, regional variations in research focus, archeological visibility, and the wiggles in the radiocarbon calibration curve. These factors can introduce peaks or troughs that may not correspond directly to demographic changes. Therefore, in this study, we interpret temporal shifts in SPDs and RGR primarily as indicators of relative changes in the intensity of human activity preserved in the archeological record. Inferences about underlying demography are treated as testable hypotheses, evaluated against complementary lines of evidence such as fecal stanol fluxes (coprostanol, 5β-stigmastanol). According to previous archeological 14C dates, the compiled dataset in this study has 553 dates that were collected from 125 sites in the CEC referred from previous studies7,96,97,98. This 14C database has 553 dates, with errors of 35.2 a, as shown in Supplementary Data 2. The uncalibrated 14C dates from the archeological sites were screened by the previously published methods98,99. Prior to calibration, the dates underwent a rigorous screening process to exclude unsuitable materials, such as shells, soils, and samples from temples, pagodas, and canoes10. The redundant dates of archeological sites were combined by the OxCal v4.2.3 with the command of R_Combine. The averaged dates were calibrated with a 95.1% confidence, the SPDs of these archeological sites in the central and eastern China were calculated with the CALIB 7.04 program100 and the IntCal13 calibration curve101 using the Sum function. Summed probability distributions were calculated at decadal intervals and plotted along the timeline. The resulting probability distributions reveal distinct peaks and troughs, which are interpreted to reflect fluctuations in human activity, when integrated with settlement patterns and subsistence evidence, provides insights into demographic trends102. This approach allows for a nuanced understanding of the linkage between climatic changes and human responses over time. Causal analysis is facilitated by transforming SPDs to relative growth rates with RGR = {SPD(t + Δt)−SPD(t)}/{SPD(t)Δt}68.

Sterol determination

Sterols were measured in sediment (n = 45) from the core SD-01. According to previous studies103, samples were freeze-dried and homogenized before solvent extraction using a Dionex accelerated solvent extraction (ASE 350) system at 120 °C and 1200 psi with dichloromethane:methanol (DCM:MeOH; 9:1 v/v). The total lipid extract was separated using silica gel flash chromatography to obtain the alkane (hexane), ketone (DCM), and polar (MeOH) fractions. The resulting ketone fraction was analyzed for sterols and stanols.

Sterol and stanol samples were identified and quantified using select ion monitoring with an Agilent 7890B gas chromatography (GC) system coupled to an Agilent 5977B quadrupole mass spectrometer (MS). The samples were injected onto a ZB-1MS (30 m x 320 μm x 0.25μm) column in pulsed splitless mode (320 °C, 1.3 psi). The oven temperature was increased from 40 °C for 1 min to 255 °C at 20 °C min-1 and further increased to 315 °C at 4 °C min−1 where it was held isothermally for 10 min. The ionization energy for the MS was set to 70 eV and a scan range of 50–650 mz−1 was used. 5-cholestan-3-one was used as an internal standard.

Climate model simulations

The Community Earth System Model version 1.2 (CESM1.2) integrates five core components: the Community Atmosphere Model (CAM5.3), Community Land Model (CLM4.0), Community Ice Code (CICE4.0), Parallel Ocean Program (POP2), and River Transport Model (RTM), linked via a central coupler104. The atmospheric component (30 vertical levels) and land component (15 soil layers) were configured at a horizontal resolution of 1.9° × 2.5°105, while the ocean component operated at a nominal 1° grid. Simulations incorporated orbital parameters, greenhouse gas concentrations (CO2, CH4, N2O), and ice sheet configurations from ICE-6G106, each calibrated to its respective time periods. Each run continued until achieving a top-of-atmosphere energy imbalance below 0.1 W m−2. To derive annual mean temperatures, we applied paleo-calendar corrections by weighting monthly values to account for orbital-induced seasonal fraction changes. Final climatologies were computed using the last 50 years of each simulation.