Introduction

The East Asian Monsoon (EAM) is one of the world’s strongest climate systems and controls the water supply to over one-third of the global population1. It is composed of the East Asian Summer Monsoon (EASM) that dominates during interglacial periods and transports moisture and heat from the low-latitude oceans towards northern China, and the East Asian Winter Monsoon (EAWM) that dominates during glacial periods and is characterized by cold, dry Siberian air moving southward2. The EASM variability is recorded in the oxygen isotopes (δ18O) of Chinese cave speleothems, where low values reflect periods of a more intense EASM2,3,4,5. Speleothems from the Hulu and Sanbao caves have provided a high-resolution, well-dated record of EASM variability over the last 640 thousand years (kyr). This record is dominated by a strong precession signal (23 kyr cycle), which has been explained by the EASM strength being mainly driven by Northern Hemisphere (NH) summer insolation2,3,5. However, it has also been argued that the changes in the speleothem δ18O record contain an annual signal or contributions from the Indian Summer Monsoon6,7,8, or are rather forced by insolation variations at low-latitudes (30°N-30°S)9,10, where temperature gradients between land and sea, and between hemispheres drive the position of the Inter-Tropical Convergence Zone (ITCZ) and monsoon rainfall9,11. The latter scenario would also explain the absence of a clear signal of NH ice sheet changes in the speleothem records. Regardless, the presence of abrupt variations in the speleothem record coinciding with Heinrich (H) events does indicate an imprint of high-latitude climate dynamics on the EASM on sub-orbital timescales2,3,5. Past variation in EASM intensity has also been inferred from many proxy records from Chinese loess sequences, where alternating layers of loess and paleosols represent periods with a dominating winter (glacials) and summer (interglacials) monsoon, respectively12,13. The loess layers are characterized by larger grain size (GS) and lower magnetic susceptibility (MagSus), whereas the paleosol layers have smaller GS and higher MagSus. MagSus is traditionally used as a proxy for the intensity of the EASM which delivers ~80% of the precipitation to the Chinese Loess Plateau (CLP), assuming that MagSus increases during pedogenesis under warmer and wetter conditions prevailing during the summer monsoon season14,15. Similarly, GS is generally used as a proxy for the strength of the EAWM linked through wind intensity12.

Interestingly, MagSus records from the CLP show a dominant 100 kyr cycle related to the glacial-interglacial variability, which is not recorded by the speleothem δ18O records, whereas both proxies are generally interpreted as reflecting EASM variation1,2,11. A possible explanation for this discrepancy, known as “the Chinese 100-kyr problem”16, may be that the archives record slightly different aspects of the EAM climate. Indeed, the exact interpretation of the speleothem δ18O records is still under debate with different explanations posited for the effects of precipitation amount, moisture composition, moisture source, and pathway17,18,19,20,21,22. In addition, the temperature dependency of calcite/water fractionation of oxygen isotopes can possibly introduce a temperature overprint on the hydroclimate signal2,23. In this study, we will follow the interpretation of model outputs suggesting that orbital scale variations in the speleothem δ18O records primarily reflect changes in moisture transport and precipitation amount20, summarized as EASM intensity. Similarly, the MagSus can lose sensitivity to changes in precipitation due to changes in the loess deposition rate, where high deposition rates may cause a dilution of the MagSus signal, especially during glacial periods with dominating winter monsoon24, and low deposition rates, in combination with bioturbation, could smoothen out the shorter orbital cycles25,26,27,28. In addition, MagSus only responds to precipitation when the mean annual precipitation is higher than 200 mm/yr and lower than 2200 mm/yr29. Magsus is, next to precipitation, also determined by temperature, as well as soil organic content, carbonate content, type of organic matter and pH29,30.

Multiple studies have attempted to explain the discrepancy in dominant orbital cycles present in speleothem and loess proxy records. For example, a recent study produced a spliced loess record of MagSus for the interglacials and inversed sand content for the glacials to circumvent the dilution of the MagSus signal during glacials, based on the assumption that variations in EASM strength are then recorded by sand content, determined by the inverse relation of the EASM with the strength of the winter monsoon that delivers this sand to the CLP27. Spectral analysis of this record indeed indicated the presence of a dominant 23 kyr cycle, demonstrating that the strong response of EASM to summer insolation is also represented in loess-paleosol records from the CLP. Notably, another study aimed at solving this same discrepancy found a strong 100 kyr cyclicity in their EASM intensity record that they based on the carbon isotopic composition of snail shells (δ13Cshell) stored in a loess-paleosol sequence from the central CLP31. They assumed a link between δ13Cshell and that of the δ13C of the vegetation consumed by the snails, which is controlled by local precipitation and thus mainly the EASM32. Since snail shells are composed of secondary carbonate, they are, therefore, unaffected by changes in sourcing, pedogenesis, or deposition rate that could alter or smoothen the cyclicities present in the loess record31. The dominance of the 100 kyr cycle in that record, as also found in those of loess proxies, thus indicates a primary forcing by NH ice volume rather than local summer insolation. However, earlier studies that used the δ13C of inorganic carbonate (δ13CIC) to reconstruct EASM intensity for two high-resolution loess sequences in the arid northwestern CLP found that these records contain significant 23-, 41-, and 100 kyr cycles, although the latter is most dominant33,34. They linked δ13CIC with vegetation density, which, on the western CLP, is mainly controlled by precipitation, and thus EASM intensity35, although changes in the mixing ratio of detrital and pedogenic carbonates may also influence δ13CIC33. The same three cycles are also present in the Ca/Ti record from the western CLP, where this ratio was linked to precipitation-induced leaching intensity associated with summer rainfall36. The distinct cyclicities in each of these proxy records suggests that they likely reflect different aspects of the EASM, and also record hydrology differently than the δ18O of speleothems, to which they are compared. Thus, the dominant driver of EASM intensity, as it is recorded in loess-paleosol sequences, remains unclear.

We here aim to solve the orbital discrepancy by using hydrogen isotope compositions of plant waxes (δ2Hwax) stored in the Yuanbao section on the western CLP (Fig. 1a) to generate a direct hydrologic record of EASM intensity over the past ~130 kyr. The Yuanbao section is characterized by exceptionally high deposition rates (5 to 143 cm kyr−1)37 that allows the generation of high resolution records that capture millennial scale events. Since previous studies from the modern CLP have shown that temperature and evapo-transpiration have a small impact on soil water δ2H in non-desert regions where MAP>400mm38, as is the case in Yuanbao (Fig. 1b) and also shown in the global monsoonal system39, the δ2Hwax primarily reflects the isotopic composition of meteoric water supplied during plant growth40,41. This allows us to make the comparison of EASM proxy behaviour between archives, i.e., δ2Hwax vs. speleothem δ18O, as well as between proxies in the same archive, i.e., δ2Hwax vs. MagSus from the same loess section. Subsequent assessment of the orbital forcings of these proxies should lead to identification of the mechanism(s) resulting in the distinct cyclicities between loess and speleothem-based EASM records.

Fig. 1: Map indicating the position of the Chinese Loess Plateau and meteorological data from the area.
figure 1

(a) Map (modified from Lu et al., 201979) indicating the position of the Chinese Loess Plateau (CLP), the location of the study site Yuanbao and loess sections cited in the main text, the Hulu and Sanbao caves, and the modern vegetation coverage of the CLP. Red arrows indicate dominant wind directions of the East Asian monsoon system. (b) Monthly averaged meteorological data from 2016 to 2017 at Yuanbao: air temperature (red line) and monthly precipitation amount (purple bars) derived from China Meteorological Data Service Center, http://data.cma.cn/en.

Full size image

Results and discussion

The Yuanbao δ2Hwax record

The δ2Hwax (compensated for ice-volume induced changes in the global hydrological cycle) ranges from −119‰ to −201‰ (Fig. 2d), where the lowest values occur during interglacial periods and glacial periods are characterized by less negative δ2Hwax. Since plant waxes are mostly synthesized during the growing season40,41, the δ2Hwax likely represents the isotopic composition of their main water source during this season, i.e., soil moisture41. On a global scale, the δ2H of plant source water follows that of (amount weighted average) meteoric water41. Given that ~80% of the precipitation at the CLP falls during summer42, this implies that δ2Hwax at Yuanbao reflects the isotopic composition of EASM associated precipitation, and should thus vary according to the same processes which control speleothem δ18O. Hence, variations in the plant wax record presumably reflect a combined effect of upstream depletion, precipitation amount, and atmospheric circulation changes3,20, where lower δ2Hwax values indicate an increased EASM intensity, with more precipitation sourced from the Pacific Ocean, whereas higher δ2Hwax values indicate less precipitation, derived from more local sources3,20. However, δ2Hwax may also be influenced by vegetation change, as the fractionation of hydrogen isotopes during leaf wax synthesis differs between vegetation types43. For example, larger fractionation occurs in grasses than in shrubs41, and may thus introduce variations in δ2Hwax independent of climate. The stable carbon isotopic composition of bulk organic carbon and plant waxes indicates that on part of the CLP, vegetation changed between C3 and C4 plant types over glacial-interglacial cycles44,45,46,47,48. At Yuanbao, however, the δ13Cwax record only reveals minor scale variation ranging from −30.5‰ to −32.9‰ (Fig. 2e). These values all fall within the range of C3 plants, likely due to the high elevation of the site (2177 m asl) causing insufficient temperatures for C4 plants to be favoured44,45,49. This indicates that at Yuanbao, changes in δ2Hwax are primarily caused by changes in precipitation δ2H.

Fig. 2: Comparison of proxy records for the Yuanbao loess section with global climate records.
figure 2

(a) Sea level stack69, (b) Magnetic susceptibility, (c) Mean grain size, (d) Ice-corrected δ2Hwax (dark green; δ2HC31 and δ2HC29 in light and lighter green, respectively). Error bars indicate standard deviation based on at least duplicate analysis. VSMOW – Vienna standard mean ocean water, (e) δ13Cwax. Error bars indicate standard deviation based on at least duplicate analysis. VPDB – Vienna Peedee belemnite, (f) Comparison between insolation and half precession signal between North and South hemisphere (35° N–35° S). The grey lines indicate monthly mean insolation in March and monthly mean insolation in September, the maximum values of these two are in red dashed lines, (g) Composite speleothem δ18O record from the Hulu2 and Sanbao5 caves (red), and June 21st insolation at 35°N80. Marine Isotope Stages (MIS) and Heinrich events (H) are indicated with grey bars. Alternating loess (L) and paleosol (S) layers are indicated at the bottom of the lower panel.

Full size image

The exceptionally high resolution of our record facilitated by the high deposition rate at Yuanbao reveals millennial scale variations in EASM intensity, in particular during the interval corresponding to Marine Isotope Stage (MIS) 3 (Fig. 2d). The Yuanbao δ2Hwax record clearly shows excursions towards more positive values at times that match with H events (H2-H6) similarly recognized in the speleothem record (Fig. 2d, f). The imprint of these North Atlantic climate events in the Yuanbao record suggests that strengthening of the Westerlies as a result of the slowdown of the AMOC reduced the inland penetration of the EASM during these intervals12,50,51, and thus blocked the delivery of moisture from remote sources and a more negative isotope signal to the CLP. In addition, the glacial conditions reduced sea surface temperatures and the size of the West Pacific warm pool52, leading to a weakening of the EASM intensity and making way for polar air from high latitudes to reach the CLP. Notably, these millennial scale events are not resolved in other available δ2Hwax records from the CLP (e.g., Weinan47, Xifeng53,54, Lingtai49; Supplementary Fig. 1), likely due to the lower sedimentation rates on the central and southern CLP, where these sites are located, introducing a smoothing effect on the record55, or simply by lower sampling resolution. In addition, the trends in the δ2Hwax records from Weinan and Lingtai on the southern edge of the central CLP are different from those in the Yuanbao record. This can possibly be attributed to contributions of additional moisture sources and/or local recycling that influenced δ2Hwax at these southern sites47. Indeed, the trends in δ2Hwax recorded at Xifeng that is located further north are more similar to those from Yuanbao, although the resolution of the latter record remains unprecedented.

Comparison of δ2Hwax with speleothem δ18O

The δ2Hwax and δ18O speleothem records presumably record the same aspect of hydroclimate, i.e., the isotopic composition of EASM precipitation driven by changes in moisture transport and precipitation amount5,40,41,56, and are thus expected to show similar behaviour over time. The two records indeed vary in concert on both precessional and millennial timescales (Figs. 2d, g, 3). The strong similarity between the δ2Hwax and speleothem records thus suggests that the main driver of their variability is the same. Assuming that δ2Hwax is a growing-season signal, this implies that speleothem δ18O indeed also represents a summer monsoon signal, as previously suggested based on summer insolation as main driver of the EASM and the fact that the majority of annual precipitation (~80%) falls during summer2,57,58,59. The phase analysis between δ2Hwax and δ18O shows that both records are in phase during MIS 5, but that δ2Hwax is leading speleothem δ18O during MIS 3 (Supplementary Fig. 2). This could be the result of the low insolation forcing during that time, giving way to other factors that may influence precipitation isotopes at Yuanbao, such as high-latitude climate variability (shown by the millennial scale H events), or the distance between the location of the caves and Yuanbao, and possibly the high elevation and the vicinity of Yuanbao to the Tibetan Plateau that can influence the atmospheric teleconnections bringing EASM precipitation inland. Regardless, in line with the similarities between the records, the main periodicity present in the δ2Hwax record is precession (23 kyr; Fig. 3). The period from 132-70 ka, corresponding to the last interglacial (MIS 5), also shows half-precessional cycles (Fig. 4), which is something that has so far only been observed in loess proxy records from the Caotan section, also located on the northwestern part of the CLP60. The presence of half-precession cycles indicates an influence of low-latitude insolation forcing on the EASM, likely through ocean-atmospheric connections transporting moisture from the West Pacific warm pool to inland China60, where the sea surface temperature of the West Pacific is highly dependent on both NH and southern hemisphere (SH) summer insolation61. The fact that insolation from spring and autumn equinoxes are out of phase by half a precession cycle creates this pattern, as there are two insolation maxima in the tropics during one precession cycle62. Half-precessional cycles are also weakly present during 50–20 ka, although they are below the 95% confidence interval. The expression of these cycles during interglacials can be explained by the amplitude of NH and SH insolation (35° N and 35° S, Fig. 2f), which is higher during interglacial periods than during glacial periods63. The EASM thus relays low latitude signals from the Pacific to the CLP60,64,65. Furthermore, Sun and Huang (2006) noted that the speleothem δ18O could record changes in moisture from the Pacific to the Indian Ocean, but that the moisture coming from the Indian Ocean does not reach the CLP due to its position relative to the Himalayas, which blocks this moisture transport. Instead, changes in δ2Hwax should then show variations in moisture from the South China Sea (related to the Pacific) vs. the West Pacific warm pool (tropical Pacific), indicating the source of the moisture associated with the half-precessional cycles60.

Fig. 3: Spectra of time series of the proxy records for the Yuanbao loess section compared to global climate records.
figure 3

Spectral analysis of (a) δ18O benthic stack69, (b) Magnetic susceptibility from Yuanbao (this study), (c) Mean grain size from Yuanbao, (d) Ice-corrected δ2Hwax from Yuanbao, and (e) Hulu/Sanbao speleothem δ18O records2,5. Orange lines represent the 99% significance level bending power law (B.P.L.). All spectral analyses for the Yuanbao section, the speleothem δ18O record and the benthic stack span to 132 ka. Vertical bars highlight primary orbital periods of the 23 and 100 kyr cycles. Frequencies <0.025 are not significant due to the length of our record (132 kyr). Half-precessional cycles are recognized by a frequency of ~0.1.

Full size image
Fig. 4: Wavelet power spectrum of ice-corrected δ2Hwax record from Yuanbao.
figure 4

Warm and cold shadings indicate the power of the correlation. Significant sections are enclosed by contours. Black dashes represent half-precession (10 ka), precession (23 ka) and obliquity (41 ka) cycles, respectively.

Full size image

The relative amplitude of glacial-interglacial variability is minor compared to that in sea level stack and global ice volume records (Fig. 2a). And although the δ2Hwax shows a weak 100 kyr cycle (Fig. 3), this is not significant due to the restricted length of the record (~130 kyr). This cyclicity could come from an additional sensitivity to NH ice volume that does not directly impact the speleothem record. The changes in the δ2Hwax record are often abrupt and show a sawtooth pattern that is reversed to that in the global benthic stack, i.e., δ2Hwax can rapidly become more positive during glacial inceptions, whereas the benthic stack reflects a slow buildup of ice sheets66 (Fig. 2). Similar rapid changes into glacial periods are present in the Yuanbao GS record, suggesting that they could be linked to the strength of the EAWM, which is closely linked to NH ice volume on orbital timescales67,68. The location and high elevation of Yuanbao relative to that of the caves could explain the higher sensitivity of δ2Hwax to high-latitude climate forcings compared to that of speleothem δ18O.

Comparison of δ2Hwax record with MagSus at Yuanbao

The generation of a MagSus record for the exact same section as used for δ2Hwax analysis allows for a direct comparison of the two records, which are both assumed to reflect EASM intensity. However, the amplitude and trend of the MagSus record for the Yuanbao section are different from that of the δ2Hwax record, as well as from the δ18O speleothem record (Fig. 2b, d, g). The MagSus record clearly follows the pattern of a MIS 5e that is higher than 5c which is again higher than 5a, followed by a subsequent maximum during MIS 1 that is typically also seen in sea level (Fig. 2a, b) and thus the structure of the benthic δ18O stacked record69. This in turn suggests that next to precession, NH ice volume and global climate also exert an influence on the MagSus signal25.

Indeed, the MagSus record contains the expected 23 kyr cycle, however, the 100 kyr cycle is also prominent in this record (Fig. 3). Although this cycle is not significant due to the length of our record, longer MagSus records from the CLP all contain a prominent 100 kyr signal1,47,67,70. The precession signal may be introduced by the influence of NH summer insolation on EASM intensity that is recorded by MagSus specifically in the paleosol layers27,33. These layers are presumably well resolved and relatively little impacted by smoothing processes due to the high deposition rates at Yuanbao37, which facilitates their registration of precession-to-millennial scale EASM intensity signals. Regardless, the MagSus and δ2Hwax records show different trends also within the paleosol layers representing interglacials (Fig. 2). Given that δ2Hwax mostly records the growing season, any offsets with MagSus are likely introduced during winters, and thus imply that MagSus represents both summer precipitation and annual temperature signals rather than solely EASM intensity11,16,26. The contribution of an annual temperature signal, and thus the imprint of the NH ice volume, would then explain the 100 kyr periodicity in the MagSus record. Hence, we suggest that the difference in the MagSus and δ2Hwax records can be explained by the distinct sensitivity of these two proxies to summer precipitation and annual temperature.

Conclusions

Since δ2Hwax is linked to the growing season of the vegetation from which the leaf wax lipids are derived, the similarities between precession-to-millennial scale variabilities of precipitation isotopes recorded by plant waxes in the Yuanbao section of the CLP and those by speleothem δ18O indicate that both records reflect a summer signal, i.e., that of EASM intensity. The dominant 23 kyr cyclicity in both records points to NH summer insolation as their main driver. The additional presence of half-precession cycles in the Yuanbao δ2Hwax record during interglacials indicates that low-latitude insolation forcings reach further north than generally anticipated. In contrast, MagSus recorded in the same loess section and traditionally also assumed to reflect EASM intensity through precipitation-induced pedogenesis, follows the trends and amplitude represented in global sea level, implying a strong control of NH ice volume with a prominent 100 kyr cycle. The different amplitude and periodicities in the MagSus record are thus explained by the contribution of a winter temperature signal, implying that MagSus is recording the effects of both summer precipitation and annual temperature on pedogenesis. The discrepancy in the main cyclicities between the loess proxies and the water isotopes (δ2Hwax and speleothem δ18O) thus has high potential for distinguishing the orbital and glacial imprints on EAM variability for further studies.

Methods

Study site

Yuanbao is situated on the western part of the CLP at 2177 m above sea level (asl; 35.63°N, 103.17°E, Fig. 1a). The mean annual air temperature (MAAT) in Linxia, the closest weather station (35.15°N, 103.63°E)37, is 7.3 °C71. This translates into a MAAT of 5.7 °C at the elevation of the sample site (based on a lapse rate of 0.6 °C 100 m−1) with temperatures of −6.1 °C in winter, and 12.2 °C during the summer growing season (April‒October). The mean annual precipitation (MAP) is 500 mm yr−1 of which 80% falls between May and September (Fig. 1b)72. The Yuanbao section is characterized by exceptionally high deposition rates (5 to 143 cm kyr−1)37 that enabled the generation of high resolution records. Samples from the upper 40 m were collected at 5 cm intervals in August 2019 and stored in geochemical sampling bags for transport to the laboratory in Utrecht, where they were freeze dried and homogenized prior to analysis.

Magnetic susceptibility and grain size analysis

Magnetic susceptibility and grain size were analysed at 5 cm resolution (n = 120) at the Institute of Earth Environment, Chinese Academy of Sciences37. Low-frequency mass magnetic susceptibility was measured with a Bartington Instruments MS2 meter. After removal of organic matter (10 ml, 10% H2O2) and carbonate (10 ml, 10% HCl), the grain size distributions were determined using a Malvern 2000 laser diffraction instrument. Measurement uncertainty was <10% for magnetic susceptibility and <2% for mean grain size.

Chronology

The upper 40 m of the Yuanbao section covers the past ~132 kyr based on the alignment of GS and MagSus to the quartz GS and MagSus records from the drill core collected at Yuanbao in 201737. The latter GS record was matched with the benthic δ18O record66, selecting tie points to link the loess (L)/paleosol (S) boundaries to the glacial/interglacial transitions. This method is commonly accepted to establish age-depth models for loess-paleosol sequences from the CLP12,33,37. This age-depth model was supported by optically stimulated luminescence (OSL) dates from a nearby loess-paleosol sequence73. In short, the S0/L1, L1/S1, S1/L2 boundaries were linked to the transitions of marine isotope stage (MIS) 1/2 (14 ka), 4/5 (70 ka), 5/6 (132 ka), respectively. Three age control points were additionally chosen to match L1SS1-2 to MIS 3 (27–59 ka). Finally, four additional tie points were added to link H events in the GS and speleothem δ18O records (Supplementary Fig. 3). This age model was assessed by a phase analysis between Yuanbao GS and speleothem δ18O, showing that both records are in phase (Supplementary Fig. 4).

Lipid biomarker analysis

Long chain n-alkanes were extracted (3x) from ~30 g loess (n = 251) with dichloromethane (DCM): MeOH (9:1, v-v) using a MEX microwave extractor. The total lipid extracts (TLEs) were filtered over a paper filter (Whatman grade 42 Ashless Filter Paper, 55 mm diameter), then dried under N2. Subsequently, TLEs were split into an apolar and polar fraction over an activated Al2O3 column eluting with hexane:DCM (9:1, v-v) and DCM:MeOH (1:1, v-v), respectively. An internal standard was added to the apolar fraction, consisting of 64 μL squalane.

The apolar fractions, containing the n-alkanes, were analysed on a Hewlett Packard gas chromatograph (GC) equipped with a CP-sil 5CB fused silica column (30 m, 0.32 mm, 0.10 um) coupled to a flame ionization detector (FID). Helium was used as a carrier gas (1 mL/min at constant flow). The GC oven temperature program was as follows: 70 °C to 130 °C (at 20 °C/min), to 320 °C (at 4 °C/min), at which it was held isothermal for 20 min. The n-alkanes were identified and quantified by using the internal standard, and manual integration of the peak areas using Chemstation software B.04.03.

δ2H analyses of n-alkanes were conducted on a MAT253 isotope ratio mass spectrometer (Thermo Fisher Scientific) coupled via a GC IsoLink operated at 1420 °C to a GC (TRACE, Thermo Fisher Scientific) equipped with a PTV injector and a HP-5ms column (30 m, 0.25 mm, 1 μm). Each sample was measured at least in duplicate. δ2H values were calibrated against H2 reference gas of known isotopic composition and are given in ‰ VSMOW (Vienna Standard Mean Ocean Water). Accuracy and precision were controlled by a lab internal n-alkane standard calibrated against the A4-Mix isotope standard (provided by A. Schimmelmann, University of Indiana) every six measurements and by the daily determination of the H3+ factor. Measurement precision was determined by calculating the difference between the analysed values of each standard measurement and the long-term mean of standard measurements, which yielded a 1σ error of 3‰. The analytical reproducibility (1-σ standard deviation) of δ2H analyses for the n-C29 and n-C31 alkane was 2 and 3‰ VSMOW on average, respectively, with maximum values of 6 (n-C29) and 5 (n-C31) ‰ VSMOW. H3+ factors varied between 4.9 and 5.2 (mean ± s.d., 5.1 ± 0.1). Precision (1-σ standard deviation) of the squalane internal standard was 6‰. δ13C analyses of n-alkanes were conducted on a MAT252 isotope ratio mass spectrometer (Thermo Fisher Scientific) coupled via a gas chromatograph-combustion (GC-C) interface with a nickel catalyzer operated at 1000 °C to a GC (Trace, Thermo Fisher Scientific) equipped with a PTV injector and a HP-5ms column (30 m, 0.25 mm, 0.25 μm). Each sample was measured at least in duplicate if sufficient material was available. δ13C values were calibrated against CO2 reference gas of known isotopic composition and are given in ‰ VPDB (Vienna Pee Dee Belemnite). Accuracy and precision were determined by measuring n-alkane standards calibrated against the A4-Mix isotope standard every six measurements. The difference between the long-term means and the measured standard values yielded a 1σ error of 0.3‰. The analytical reproducibility (1-σ standard deviation) of δ13C analyses for the n-C29 and n-C31 alkane was 0.1‰ VSMOW on average with maximum values of 0.5‰ VSMOW for both compounds. For samples with single analysis, the long-term precision was assumed as analytical error. Precision (1-σ standard deviation) of the squalane internal standard was 0.2‰.

From the detected n-alkane distributions and their stable isotopic composition, records of the hydrogen isotopic composition of meteoric water (δ2Hwax), as well as of vegetation change (δ13Cwax) were generated based on the weighted mean of the isotopic composition of the most abundant plant waxes (C29 and C31 n-alkanes). The δ2Hwax record was corrected for ice volume changes which affect isotopes in the global hydrological cycle assuming a Last Glacial Maximum change in global δ18O of seawater of 1‰ to scale the benthic δ18O record74,75 and using the following equation:

$${\delta }^{2}{{{{{{\rm{H}}}}}}}_{wax}\,_{ice-corrected}\,=\frac{1000+{\delta }^{2}{{{{{{\rm{H}}}}}}}_{wax}}{8\times 0.001\times {\delta }^{18}{{{{{{\rm{O}}}}}}}_{ice}+1}-1000$$
(1)

Frequency analysis

Frequency analysis was carried out using Acycle 2.3.1 software76, using the periodogram function (default settings; angular frequency of π/4 rad/sample with additive N (0,1) white noise). The wavelet analysis was also conducted under Acycle 2.3.177 with the following parameters: period range from 1 to 100, pad = 1, discrete scale spacing = 0.1, Mother = Morlet. All analysed records were standardized and detrended (rLOWESS; Locally Weighted Scatterplot Smoothing). Cross-spectral analyses were performed with the Blackman-Tukey approach using arand-master software after interpolation to a constant time step of 1 ka. Cross-spectral results are also obtained between precession, ice volume, GHG and East Asian monsoon proxies. The coherency spectra of East Asian monsoon proxies with ice volume, GHG and precession are compared with the 80% non-zero coherency level, implying different roles of ice volume, summer insolation and GHG forcing on four proxy records. The following parameters were used to optimize bias/variance properties of spectrum estimates: number of lags = 40 (~1/3 length of record) and samples per analysis = 120. The phase analysis (wavelet coherence; WTC) was performed in a Monte Carlo framework (n = 1000)78, using the MATLAB code available at http://grinsted.github.io/wavelet-coherence/. The WTC helps to detect the period in different frequency bands where the two time series co-vary (but does not necessarily have high power). The black arrows in the figures represent the phase relationship between the two time sequences with rightward, upward and downward arrows indicating in phase, leading phase and lagging phase, respectively. The color scale indicates the amplitude correlations between the two datasets36.

Data availibility

The data set underlying this research is available at https://doi.pangaea.de/10.1594/PANGAEA.961583.