Sub-orbital-scale variations of the East Asian winter monsoon (EAWM) and its mechanisms during the Holocene are controversial, partly due to the lack of high-quality records from Chinese loess. Here, we present high-resolution reconstruction of Holocene EAWM intensity based on optically stimulated luminescence dating and grain-size analysis from three loess sections taken from the Chinese Loess Plateau. The EAWM showed a persistent weakening trend during the early Holocene (ca. 11.7–6.5 kyr B.P.) and a strengthening trend during the mid- to late Holocene (since ca. 6.5 kyr B.P.). We propose that this was caused by changes in high-latitude Northern Hemisphere ice volume and middle- to high-latitude Northern Hemisphere atmospheric temperatures, respectively. We also observed an anti-correlation between EAWM and East Asian summer monsoon. Our findings provide a robust solution to the debate regarding Holocene EAWM changes and contribute to the understanding of potential future variations in EAWM intensity.


The East Asian winter monsoon (EAWM) determines the strength of cold and dry northwesterly and northeasterly surface flows over East and South Asia during boreal winters (Fig. 1). Intensified EAWM directly causes ambient temperature declines and frequent dust storms, resulting in significant economical, biological, and social impacts in this region (Chang et al., 2006).

Long-term (e.g., tectonic- to orbital-scale) and millennial-scale changes in EAWM intensity before the Holocene (ca. 11.7 kyr B.P.) are well understood, mostly based on loess records from the Chinese Loess Plateau (CLP) (Ding et al., 1995; Guo et al., 2002; Sun et al., 2012). Variations in Holocene EAWM intensity on multiple time scales have also been investigated using records including marine sediments (Huang et al., 2011; Zheng et al., 2014; Zhang et al., 2019), lake sediments (An et al., 2011; Wang et al., 2012), loess deposits (Stevens et al., 2008; Yang and Ding, 2014; Kang et al., 2018), and climate modeling (Wen et al., 2016). However, these reconstructed EAWM variations are inconsistent at a sub-orbital scale, such as weakening throughout the Holocene (Huang et al., 2011; Wen et al., 2016) or weakening in the early Holocene and strengthening in the mid- to late Holocene (Yang and Ding, 2014; Zheng et al., 2014). Moreover, the relationship between the EAWM and East Asian summer monsoon (EASM) is still under debate, with correlation (e.g., Wen et al., 2016) and anti-correlation (e.g., Zheng et al., 2014) both being proposed.

Loess on the CLP is the ideal and most commonly used material for reconstructing past EAWM intensity (Liu and Ding, 1998). However, high-resolution loess records of Holocene EAWM intensity are still lacking, particularly with reliable chronology. To reconcile the discrepancies between different records and determine accurate Holocene EAWM intensities, we conducted closely spaced optically stimulated luminescence (OSL) dating and grain-size analysis on three high-resolution loess sections from the CLP to produce a stacked reconstruction of Holocene EAWM intensity. We determined sub-orbital-scale variations in EAWM intensity and its possible mechanisms and compared these with other records. Moreover, the relationship between EAWM and EASM intensities was revealed.


Setting, Sections, and Sampling

Climate on the CLP is seasonally controlled by the warm-humid southeasterly EASM during the warm season and the cold-dry northwesterly EAWM during the cold season. The loess-paleosol sequences record the alternating dominance between the EAWM and EASM at multiple time scales over the past ∼2.6 m.y. (Liu and Ding, 1998). Aeolian loess is mainly transported by EAWM winds from arid deserts, dry riverbeds of the Yellow River (Nie et al., 2015), and alluvial fans of the Qilian Mountains (Derbyshire et al., 1998) in northwestern China and southern Mongolia (Fig. 1).

Loess sections WN2 (34°24′54.85″N, 109°33′44.18″E, 646 m above sea level [a.s.l.]; Kang et al., 2018) and GB (34°35′2.14″N, 110°36′24.14″E, 470 m a.s.l.) were collected from the southeastern CLP, while section LGG (35°45′29.4″N, 107°48′7.8″E, 1070 m a.s.l.) was collected from the central CLP (Fig. 1). Compared with the majority of Holocene sections previously investigated from the CLP (Stevens et al., 2008; Yang and Ding, 2014), these three sections showed relatively high dust accumulation rates (Fig. 2D) because of their low topographical setting or contributions from adjacent riverbeds (see the Supplemental Material1). This implies the potential for three high-resolution loess records. The loess deposits of all three sections are aeolian in nature according to both field observations and grain-size distributions (Figs. S1 and S2 in the Supplemental Material). OSL samples and bulk samples were densely collected at 10–25 and 1–2 cm intervals respectively (see the Supplemental Material).

Chronology and Proxy

We applied the conventional single-aliquot regenerative-dose OSL dating protocol (Table S1) to fine (4–11 μm) quartz grains from all loess sections (see the Supplemental Material, Figs. S3–S8). A total of 10, 24, and 18 OSL ages were obtained from sections WN2, GB, and LGG, respectively (Figs. 2A–2C; Table S2). The accepted OSL ages were incorporated into the Bayesian age-depth model in Bacon software (Blaauw and Christen, 2011) to produce the chronology (see the Supplemental Material, Fig. S9; Figs. 2A–2C). Chronology reliability was confirmed by various routine tests within the single-aliquot regenerative-dose protocol (Figs. S3–S8) and by inter-site proxy comparisons (Fig. 2D; Fig. S11).

Though grain size may be affected by various processes (Újvári et al., 2016), such as EASM-related dust source distances (Ding et al., 2005), it has been commonly used as an index for EAWM intensity, with higher values (stronger winds) in glacial loess deposits and lower values (weaker winds) in interglacial paleosols (An et al., 1991a). Magnetic susceptibility (MS) is another commonly used proxy for EASM intensity, with the paleosols or loess layers showing higher or lower MS values, respectively (An et al., 1991b). Therefore, we used mean grain size (MGS) and low-frequency MS in this study (see the Supplemental Material).


Sub-Orbital-Scale Changes in EAWM Intensity

Our dating results showed disturbances in section LGG (above 175 cm depth) and depositional hiatuses in section GB (depths of 90–100 cm) (Figs. 2B and 2C), as previously found in Chinese loess (Stevens et al., 2008, 2018). After removing unreliable data, we reconstructed sub-orbital-scale changes in EAWM intensity for each section (Fig. 2D). We identified an early Holocene (ca. 11.7–6.5 kyr B.P.) persistent weakening of the EAWM (reaching a minimum at ca. 6.5 kyr B.P.) and a mid-Holocene (ca. 6.5–3.0 kyr B.P.) continuous strengthening, as demonstrated by records from sections GB and LGG. The EAWM further intensified during the late Holocene and reached a maximum during the Little Ice Age period, as shown by the section WN2 and GB records. After normalization and stacking of MGS and MS data (see the Supplemental Material), the early Holocene persistent weakening and the mid- to late Holocene strengthening of the EAWM could be clearly observed. The stacked records can be regarded as representative of sections WN2, GB, and LGG. In addition, the relationship between changes in dust accumulation rate and MGS (Fig. S12) implies that dust accumulation is not always correlated with EAWM intensity.

Comparisons with Other EAWM Records

Grain-size records from the western CLP suggest that the EAWM switched from a weakening trend to a strengthening trend at ca. 9 kyr B.P. (Maher and Hu, 2006). This switch was observed at ca. 7.5 kyr B.P. in the Chinese Loess Millennial-scale Oscillation Stack (CHILOMOS) record (an eight-section grain-size stack from the CLP; Yang and Ding, 2014), which is partly consistent with our records (Fig. 3B). Despite uncertainties in chronology and limited resolution, previous loess records partly support our EAWM reconstruction.

Changes in the Holocene EAWM that we identified (Fig. 3A) are also broadly consistent with various non-loess EAWM records, including river-terrace sediment (aeolian-dominant) Zr/Rb records from the western CLP (Fig. 3D; Liu et al., 2020), grain-size index variations from the northwestern Pacific Ocean (Fig. 3C; Zheng et al., 2014), changes in the proportion of mobile desert dunes in northern and northeastern China (Fig. 3E; Xu et al., 2020), the Uk′37 (sea-surface temperature proxy) record from the northern coast of the South China Sea (Zhang et al., 2019), and lake MGS records from arid northwestern China (An et al., 2011). The discrepancies in EAWM variation details between these studies are primarily due to chronological uncertainty, proxy sensitivity, and sediment resolution. We believe that our reconstructions are more accurate and detailed due to loess resolution, robust chronologies, and the sensitivity of loess at recording EAWM changes.

Influence of Ice Volume and Temperature on EAWM Intensity

Increased ice volume in the high-latitude Northern Hemisphere intensifies the Siberian High and/or displaces it southward via downstream atmospheric cooling in the middle latitudes. This further increases the wind and temperature gradient over East and South Asia, strengthening the EAWM (Ding et al., 1995; Stevens et al., 2018). Loess records have clearly demonstrated this linkage between ice volume and EAWM intensity on an orbital scale (Ding et al., 1995; Stevens et al., 2018). Changes in marine benthic δ18O (Fig. 4B; Lisiecki and Raymo, 2005) and ice-volume-equivalent sea level (Fig. 4C; Lambeck et al., 2014) show a delayed response of ice-sheet decay to the early Holocene Northern Hemisphere summer insolation maximum (Fig. 4D; Berger and Loutre, 1991). This gradually lowered ice volume probably contributed to the weakening EAWM trend during the early Holocene (Fig. 4A). However, high-latitude Northern Hemisphere ice volume did not influence EAWM intensity during the mid- to late Holocene because the ice volume showed little changes (Figs. 4B and 4C; Lisiecki and Raymo, 2005; Lambeck et al., 2014).

Changes in middle- to high-latitude Northern Hemisphere and low-latitude air temperatures also influence the EAWM intensity by alternating the gradient of both meridional temperature and pressure cells. Marcott et al. (2013) suggested the occurrence of small changes in low-latitude temperatures throughout the Holocene (Fig. 4G). Therefore, the meridional temperature gradient between the middle- to high-latitude Northern Hemisphere and lower latitudes is primarily determined by the middle- to high-latitude Northern Hemisphere temperature (Figs. 4E–4H; Marcott et al., 2013; Kaufman et al., 2020). Decreases in middle- to high-latitude Northern Hemisphere temperatures can strengthen the Siberian High and subsequently increase the meridional temperature gradient between the Siberian High and Equatorial Low. Moreover, the gradient between the Siberian High and Aleutian Low also increases because of the difference between continental and oceanic thermal capacities. The EAWM intensity is therefore increased. In contrast, increases in middle- to high-latitude Northern Hemisphere temperatures weaken the EAWM.

Although decreasing ice volume was the dominant control on early Holocene EAWM weakening, increasing middle- to high-latitude Northern Hemisphere temperature from ca. 11.5 to 9.7 kyr B.P. (Figs. 4E and 4F; Marcott et al., 2013; Kaufman et al., 2020) likely also contributed to EAWM decline (Fig. 4A). The relatively stable and high temperatures from ca. 9.7 to 6.5 kyr B.P. (Figs. 4E and 4F) provided beneficial conditions for a weakened EAWM. With the decrease of Northern Hemisphere summer insolation (Fig. 4D; Berger and Loutre, 1991), the continuous decline in middle- to high-latitude Northern Hemisphere temperature during the mid- to late Holocene (Figs. 4E and 4F) was the possible cause for the strengthening trend of the EAWM (Fig. 4A).

Relationship between the EAWM and EASM

In contrast to the EAWM, our MS results demonstrated a strengthening EASM during the early Holocene (ca. 11.7–6.5 kyr B.P.) and a weakening EASM during the mid- to late Holocene (after ca. 6.5 kyr B.P.) (Fig. 2D). Our reconstructions are in agreement with the precipitation record from the northeastern CLP (Fig. 3G; Chen et al., 2015), stabilized dune proportion analysis in northern and northeastern China (Fig. 3H; Xu et al., 2020), and loess MS records from the CLP (e.g., Stevens et al., 2008, 2018; Lu et al., 2013). Moreover, recent stacked Holocene stalagmite δ18O records from China support a maximum EASM intensity at ca. 7 kyr B.P. (Fig. 3F; Yang et al., 2019). Therefore, a general trend shift in the EASM from increasing to decreasing at ca. 7–6 kyr B.P. can be clearly seen from various archives, though with different amplitude and relative strengths, possibly due to different sensitivities of the proxies used by various studies (Figs. 3F–3I; Liu et al., 2015). A delay (∼4 k.y.) in the maximum EASM to Northern Hemisphere summer insolation maximum (Berger and Loutre, 1991) was possibly modulated by ice cover–sea level–atmospheric CO2 forcing (Lu et al., 2013; Chen et al., 2015; Liu et al., 2015; Stevens et al., 2018). Our results support the conclusion that Holocene EAWM and EASM variations were anti-correlated at a sub-orbital scale (Fig. 3), possibly controlled by changes in ice cover, temperature, and summer solar insolation.


We propose an early Holocene (ca. 11.7–6.5 kyr B.P.) persistent weakening and a mid- to late Holocene (since ca. 6.5 kyr B.P.) strengthening of the EAWM, possibly caused by changes in high-latitude Northern Hemisphere ice volume and middle- to high-latitude Northern Hemisphere atmospheric temperatures, respectively. Moreover, our results indicate an anti-correlation between the Holocene EAWM and EASM intensity at a sub-orbital scale. Our study also provides a clue to changes in EAWM intensity and its dynamics before the Holocene.

Instrumental observations have implied higher temperature increases in the high-latitude Northern Hemisphere relative to the low latitudes over the past a few decades (IPCC, 2014). Model simulations also predict the same pattern in coming decades, implying severe melting of high-latitude Northern Hemisphere ice sheets under future global-warming scenarios (IPCC, 2014). We therefore predict a weakening long-term trend in EAWM intensity in response to reduced ice volume and increased high-latitude Northern Hemisphere temperature.


We thank Thomas Stevens and two anonymous reviewers for constructive comments and suggestions that greatly improved our manuscript. This study was supported by the National Natural Science Foundation of China (grant 41772177), the National Key Research and Development Program of China (grant 2016YFA0601902), the Second Tibetan Plateau Scientific Expedition and Research Program (grant 2019QZKK0101), the Strategic Priority Research Program of the Chinese Academy of Sciences (CAS) (grant XDB 40010100), the Youth Innovation Promotion Association CAS (grant 2018447), and the State Key Laboratory of Loess and Quaternary Geology. This work is a part of the “Belt & Road” project of the Institute of Earth Environment, CAS.

1Supplemental Material. Details of sampling, quartz OSL dating, Bayesian age-depth model, proxy measurements and implications, normalization and stacking, and original data for Figure 2. Please visit https://doi.org/10.1130/GEOL.S.12501833 to access the supplemental material, and contact editing@geosociety.org with any questions.
Gold Open Access: This paper is published under the terms of the CC-BY license.