Earth and Planetary Science Letters 237 (2005) 45–55 www.elsevier.com/locate/epsl

Stepwise expansion of environment across northern China in the past 3.5 Ma and implications for monsoon evolution

Z.L. Ding a,*, E. Derbyshire b, S.L. Yang a, J.M. Sun a, T.S. Liu a

aInstitute of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, China bDepartment of Geography, Royal Holloway (University of London), Egham, Surrey TW20 0EX, UK Received 20 January 2005; received in revised form 26 May 2005; accepted 8 June 2005 Available online 19 July 2005 Editor: E. Boyle

Abstract

A systematic study of the last glacial cycle along three transects across the Chinese shows that sand-sized particle content within loess decreases rapidly from north to south, and that markedly high sand particle contents in loess horizons occur only in the northern part of the Plateau. This suggests that variation in the sand-sized particle fraction within loess near the desert margin is closely linked to migration of the southern desert border in northern China where sand grains move mainly in saltation or modified saltation mode near the ground surface. As desert margin shift is essentially controlled by the amount of monsoon precipitation, the sand-sized particle content within loess near desert margin is regarded as a new and readily applied proxy for variations in the strength of the East-Asian summer monsoon. A continuous record of sand content in loess along the loess–desert transitional zone shows that the Mu Us Desert migrated southward at 2.6, 1.2, 0.7 and 0.2 Ma, suggesting a stepwise weakening of the East-Asian summer monsoon during the past 3.5 Ma. This evolutionary pattern is significantly different from that previously inferred from loess magnetic susceptibility records, a widely used monsoon proxy. Our results further suggest that changes in global ice volume may have been an essential factor in controlling Plio–Pleistocene monsoon evolution, and that the anticipated future melting of polar ice cover may lead to a northward migration of the monsoon rainfall belt in northern China. D 2005 Elsevier B.V. All rights reserved.

Keywords: East-Asian summer monsoon; loess–red clay sequence; Mu Us desert; sand content of loess; Plio–Pleistocene

1. Introduction

Formation and evolution of can be regarded * Corresponding author. Tel.: +86 10 62008111; fax: +86 10 as a result of interactions between the atmosphere, 62010846. hydrosphere, biosphere and lithosphere [1–4]. Changes E-mail address: [email protected] (Z.L. Ding). in desert environment can, in turn, exert a significant

0012-821X/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.epsl.2005.06.036 46 Z.L. Ding et al. / Earth and Planetary Science Letters 237 (2005) 45–55 influence on the climate system by altering surface The long-term evolution of the deserts of northern albedo feedback and by supplying to the atmosphere China remains poorly understood because of the and oceans variable quantities of mineral aerosols that sparseness of directly extractable geological evidence have the potential to affect the radiative balance of the of suitable type and quality to be found within them. atmosphere and the global carbon cycle [5–7]. The Given that the loess deposits making up the Loess deserts of northern China, covering an area of about Plateau of China lie immediately south and southeast 1.5Â106 km2, together make up the worldTs largest of these deserts, and so are largely a product of winds mid-latitude, temperate, continental interior desert. from these dryland sources [10–12], it is considered The location of this desert zone is thought to be closely that well-dated, semi-continuous loess records may associated with the uplift of the Tibetan Plateau during provide valuable insights into the recent history of the late Cenozoic, a process that progressively hin- the deserts and their margins. In this study, three dered northward penetration of moisture-laden air loess transects covering the last glacial cycle, from from the Indian Ocean [8]. Another forcing mecha- Hongde (HD) to Yangling (YAL), Zichang (ZC) to nism leading to dryland environmental change is the Lantian (LAT), and Yulin (YL) to Weinan (WN), variation in the strength of the East-Asian summer were sampled and analyzed, together with a thick monsoon. A compilation of paleo-data from within eolian loess–red clay sequence at Jingbian near the the North China deserts has demonstrated that the Mu Us desert (Fig. 1). The aim was to use spatial southern desert margin migrated several hundred kilo- changes in loess particle size to establish a semi-quan- meters north of its last glacial maximum (LGM, ~20 ka titative relation between particle size and desert margin BP) limit in response to increased monsoon rainfall location, and to reconstruct shifts in desert margins during the Optimum (~8–4 ka BP) [9]. since the late Pliocene. The linkage between desert

Fig. 1. The sampling localities and annual precipitation isopleths (mm) in the Loess Plateau. The southern border of the Mu Us Desert during the LGM lay in a location similar to that of the present [9]. The three loess transects studied here stretch from YL to WN, ZC to LAT and HD to YAL. Within the loess–desert transitional zone, few loess sections with complete last glacial loess deposits have been found. Z.L. Ding et al. / Earth and Planetary Science Letters 237 (2005) 45–55 47 changes in northern China and East-Asian monsoon for the three transects. The HD–YAL (upper panel), evolution is also briefly addressed. YL–WN (middle panel) and ZC–LAT (lower panel) transects consist of 12, 8, and 7 sections, respectively. All sections include the S1–S0 stratigraphic units. The 2. Spatial variation in grain size of S1 soil formed in the last interglacial period, and has a loess brownish or reddish color. In the northernmost part of the transects, this soil consists of three discrete pedo- Fig. 2 shows the down-section changes in median genic units and two intercalated loess horizons. Loess grain size and content of sand particles (N63 Am%) unit L1, which accumulated in the last glacial, can be

Fig. 2. Changes in median grain size and sand-grade particle (N63 Am%) content for the HD–YAL (upper panel), YL–WN (middle panel) [15] and ZC–LAT (lower panel) transects. The left curve in each pair is median grain size and the right curve is sand-grade particle content. The shaded zones indicate interglacials. The loess sections were sampled at 5–10 cm intervals. 48 Z.L. Ding et al. / Earth and Planetary Science Letters 237 (2005) 45–55 subdivided into 5 loess layers, namely L1-1, L1-2, L1- 3, L1-4 and L1-5. Chronological studies have shown that the L1-2, L1-3 and L1-4 units together accumu- lated in marine isotope stage 3 (MIS 3), and that L1-1 and L1-5 formed in MIS 2 and MIS 4, respectively [13–15]. In the YL section, the L1-1, L1-3 and L1-5 horizons are composed of medium sand particles, indicating southward advance of paleo-desert during these times [14]. The S0 soil, developed in the Holo- cene, has a darkish color because of relatively high organic matter. In some sections of the northern Loess Plateau, e.g. at ZC, the S0 soil has been partly or totally eroded (Fig. 2). Across the Loess Plateau, there is a consistent northward decrease both in present-day mean annual temperatures (from ~13 to ~8 8C) and mean annual precipitation (from ~600 to below 400 mm) (Fig. 1). The northward increase in aridity is essentially controlled by the systematic decrease in summer rainfall derived from the East-Asian summer monsoon over the Loess Plateau. The stratigraphy of the sections is correlative in the field, suggesting quasi continuous dust accumulation during the last glacial cycle. Particle size was deter- mined for all samples taken at 5–10 cm intervals down- section, using the method described in Ding et al. [16]. The following features are clearly evident in Fig. 2. (1) Thickness, median size and sand particle percentage all decrease consistently from north to south, indicating strong spatial differentiation of dust during subaerial transport. (2) Loess units L1-1 and L1-5 are character- ized by much coarser median grain sizes and higher sand contents than soils S0 and S1 and the middle part of L1 in each section. (3) Markedly high sand particle contents in the glacial loess (L1-1 and L1-5) occur only Fig. 3. Changes in the content of sand-sized particles southward in the northern part of the Plateau, a time when the from the southern border of the Mu Us Desert (Fig. 1) for L1-1, L1- desert margin was close by [9]. 4, L1-5 and S1. The values for each horizon in each section were averaged from between 4 and 20 samples (depending on estimated In order to demonstrate further the spatial differ- sedimentation rates) from the portions with coarsest (L1-1 and L1- entiation of eolian , changes in sand particle 5) or finest (L1-4 and S1) median grain size. Changes in these content (N63 Am%) with increasing distance south- values are assumed to represent an approximately synchronous ward of the present Mu Us Desert margin (Fig. 1) are differentiation of sand-sized particles on a millennial-scale average. shown for L1-1, L1-4, L1-5 and S1 (Fig. 3). All the cold-dry LGM (L1-1) was broadly similar to that horizons exhibit a consistent southward decrease in of today, whereas it retreated several hundred kilo- the sand particle content, the rate of decrease for L1-1 meters to the north during the warm-humid Holocene and L1-5 being more rapid than that for L1-4 and S1. Optimum. This implies that this desert margin expe- Both the L1-1 and L1-5 records show an abrupt rienced wide-ranging advance–retreat cycles in re- decrease near the desert margin and a gradual decrease sponse to climatic oscillations at orbital time scales. beyond. According to the reconstruction of Sun et al. It is clear that significant increase in sand percentages [9], the southern border of the Mu Us Desert during in L1-1 and L1-5 relative to S1 and L1-4 were con- Z.L. Ding et al. / Earth and Planetary Science Letters 237 (2005) 45–55 49 trolled at the first order by the desert advance during Pliocene red clay deposit, making a total thickness of their accumulation. Using the sand content–distance ~282 m. A previous magnetic polarity study showed relation of L1-1 (Fig. 3), it is therefore inferred that that this sequence has a basal age of 3.5 Ma [21]. sand particle contents of ~30% and ~15% within the Field observations have demonstrated that the devel- loess indicate a distance from the desert margin of opment of the paleosols within both the Pleistocene ~100 and ~200 km, respectively. and Pliocene eolian deposits is much weaker at Jing- In general, loess particles are transported by winds bian than in the main body of the Loess Plateau series, from the source area to the depositional area in two and that the Pleistocene loess-soil stratigraphy corre- modes, namely saltation and suspension. Theoretical lates well with the classic loess sections. To the best of and experimental studies by Pye [17] and Tsoar and our knowledge, the Jingbian section is the only desert Pye [18] concluded that, during low-level atmospheric margin eolian sequence known to cover the whole dust storms, sand-sized particles are usually trans- Pleistocene and the late Pliocene. Its proximity to ported by saltation or modified saltation near the the dust source region makes it ideal for the study desert surface, and that any sand particles transported of long-term desert changes. in suspension quickly settle back on the ground sur- In the field, samples were taken at 5–10 cm inter- face. Our observations show that significant volumes vals, making up a total of 3440 samples. The magnetic of sand particles are present within the wind-blown susceptibility and median grain-size records at Jingbian loess only near the desert margin during glacial per- are shown in Fig. 4. As seen in other loess sections, the iods, which is consistent with the Tsoar and Pye magnetic susceptibility values are 2–4 times higher in model. Moreover, our data suggest a statistical rela- paleosols than in loess horizons, and grain sizes are tionship from which may be derived a semi-quantita- much coarser in loess than in soils. Both the suscepti- tive estimation of the distance between dust sources bility and grain-size records further support our field and depositional areas. Of course, as previously sug- observation that this eolian sequence is a well pre- gested [11], other factors such as wind velocity play a served and almost complete record, making it possible part in sand particle transport. It is widely accepted to develop a time scale for this sequence by correlating that the Chinese loess was transported mainly by its loess-soil unit succession to the Chinese Loess winter monsoon winds driven by the Siberian–Mon- Particle Time Scale (Chiloparts) [22]. The Chiloparts golian high pressure system [11,19]. The mean and record was developed by stacking five individual loess maximum velocities of winter monsoon winds during grain-size records that were tuned to the obliquity and the LGM may well have been higher than during other precession records of the EarthTs orbits. intervals of the last glacial cycle, because of the likely The Jingbian magnetic susceptibility and grain-size intensification of the Siberian High associated with records, plotted on the Chiloparts time scale, are expanded polar ice sheets and a more extensive sea ice shown in Fig. 5, together with a composite marine cover [20]. It follows from this that source-to-sink oxygen isotope record [23–25]. The N63 and N125 estimates for loess horizons other than L1-1, based Am% records (Fig. 5C and D) at Jingbian show four on the N63 Am% to distance relationship provided by stepped increases in sand-sized particle content. The L1-1 (Fig. 3), should be regarded as upper limits. late Pliocene red clay (below L33) contains few sand particles, indicating that the dust was transported in suspension, mainly from a remote source. From ~2.6 3. Desert migrations during the past 3.5 Ma to ~1.2 Ma, sand content in interglacial soils remains low, whereas it varies generally between 18% and The Jingbian section (37840V54W N, 108831V15W E), 25% in glacial loess except for the case of L15 and at 1370 m above sea level, lies on the summit of the L16. This suggests that, during glacial periods, the Baiyu Mountains. It is located only ~12 km south of desert environment advanced to a location no more the present margin of the Mu Us Desert. No local than 200 km from the present northern margin of the sources of sand, such as river channels, are present in Loess Plateau. In the part of the section deposited this area. The section is composed of a ~252 m thick between ~1.2 and ~0.7 Ma, sand content increases Pleistocene loess-soil sequence resting on a ~30 m to ~12% in soils and to ~43% in loess with a sub- 50 Z.L. Ding et al. / Earth and Planetary Science Letters 237 (2005) 45–55

Fig. 4. The magnetic susceptibility and median grain-size records of the Jingbian section, located in the northernmost part of the Loess Plateau (Fig. 1). The Pleistocene loess (above L33) at Jingbian is ~252 m thick, and the Pliocene red clay ~30 m thick. The red clay accumulated during the time interval ~3.5–2.6 Ma, as suggested by magnetic polarity studies [21]. The Jingbian section was sampled at 5–10 cm intervals. The major loess-paleosol units are labeled for the section. stantial increase in N125 Am particles, implying a Chinese loess, namely strength of the transporting large-scale advance of the desert margin during both winds, specifically those associated with the winter glacial and interglacial times. Throughout material monsoon [11,19], and the location of the desert mar- deposited in the interval ~0.7–0.2 Ma, N63 Am parti- gin [15,22]. If the source-to-sink distance remains cles range from ~30% in soils and ~55% in loess unchanged, an increase in the winter monsoon inten- units, with the N125 Am particles exceeding 8%. sity will result in the deposition of coarser dust parti- This suggests that the distance between the Loess cles. On the other hand, if the desert margin advances Plateau and the present desert margin was less than southwards, the loess deposited at a specific site on 100 km. During the last two glacial periods, eolian the Plateau will also be coarser. In this study, we show sand was directly deposited at Jingbian, indicating a that loess particle size is much coarser during glacial further southward desert shift. periods than during interglacials. This may have resulted from either or both desert margin advance and wind intensity increase due to cold-dry climatic 4. Discussion and conclusions conditions. However, a significant sand content within glacial loess horizons is present only in the northern In previous studies, two factors have been pro- part of the Loess Plateau, i.e. near desert margins posed as controls on grain-size variations within the where saltation or modified saltation may be the Z.L. Ding et al. / Earth and Planetary Science Letters 237 (2005) 45–55 51

Fig. 5. Changes in magnetic susceptibility (A) and grain-size data (B, C, D) at Jingbian, and correlation with a stacked marine d18O record (E) [23–25]. The time scale of the Jingbian section was developed by correlating the loess-soil units with the Chiloparts record [22]. dominant mode of transport. It is conceivable that, Humidity is the most important factor affecting regardless of the wind intensity, loess deposited close desert margin shift, since desert margin advance can to desert margins will contain a high sand-sized par- occur only when the vegetation cover there is ticle fraction, and that this fraction will rapidly de- destroyed with a critical reduction of humidity. In crease with distance from the margins, as shown in the loess–desert transitional region of northern Fig. 3. It is thus concluded that changes in the sand- China, about 80% of rainfall is attributable to the sized particle fraction within loess can be used as an East-Asian summer monsoon. When this summer indicator of advance–retreat cycles along desert–loess frontal rainfall belt penetrates farther to the north, transition zones. the desert margins show a corresponding northward 52 Z.L. Ding et al. / Earth and Planetary Science Letters 237 (2005) 45–55 shift, as in the case of the Holocene Optimum [9].In The Jingbian sand-sized particle record clearly this context, the sand-sized fraction in the loess near demonstrates that, superimposed on the glacial–inter- desert margin can be regarded as a proxy of summer glacial oscillations, the deserts in northern China monsoon intensity. experienced significant expansion at ~2.6, ~1.2, In most previous studies, loess magnetic suscepti- ~0.7 and ~0.2 Ma, directly implying a stepwise bility [26,27] and various chemical element ratios [28] southward retreat of the monsoon rainfall belt, asso- of the Chinese loess-paleosol series have been used as ciated with a complementary reduction in summer summer monsoon proxies, a practice based, respec- monsoon strength, in the past 3.5 Ma. This evolu- tively, on the recognition that the susceptibility values tionary pattern conflicts with current understanding of interglacial paleosols are 2–4 times greater than based on loess magnetic susceptibility records, which those found in the loess units laid down in glacials, show a significant increase in susceptibility values in and that paleosols show much more advanced chem- sediments from early to late Pleistocene in age ical weathering than is seen in adjacent loess. How- [26,34], thus implying a progressively enhanced ever, the processes and factors affecting variations in monsoon precipitation. This suggests that factors loess-paleosol susceptibility remain controversial [29– affecting variations in loess magnetic susceptibility 32]. Recent studies have shown that increased mag- require further investigation. netic susceptibility in paleosols may not result solely Two forcing factors may have driven the long-term from intensified pedogenic processes arising from evolution of the East-Asian monsoon, namely changes increased monsoonal precipitation [31,32]. Here, we in the elevation of the Tibet Plateau [3,35] and chang- propose that changes in elemental concentrations ing global ice volumes [20,34]. Both meteorological within loess-paleosol sequences may be determined observations [33,36] and numerical model experi- largely by three factors: chemical weathering in ments [35,37,38] indicate that uplift of the Tibet source regions, grain-size distribution, and post-depo- Plateau plays a critical role in initiation and mainte- sitional weathering. It is only the signature resulting nance of the Asian monsoon system. Geological stud- from post-depositional weathering that can be linked ies show that the main body of the Plateau had to summer monsoon intensity. Unfortunately, the de- reached its highest elevation by ~14 Ma [39] or ~8 tailed studies required to distinguish the major con- Ma [40], although a late phase (~3.6 Ma) of uplift of trolling factors in loess element ratios are still its northern parts may have occurred [41]. This pending. implies that Tibetan tectonic changes were not a sig- The East-Asian summer monsoon affects vast areas nificant influence on Plio–Pleistocene monsoon evo- of the continent. Essentially, monsoon precipitation is lution. In addition, any substantial uplift of the a product of the interaction between warm-moist Plateau, if such occurred during the past 3.5 Ma, southerly airmasses and cold northerly airflows from would be expected to have caused an increase in middle and high latitudes. Generally, the more north- monsoon strength. The Jingbian record shows that erly the penetration of the frontal rainfall belt the this was not the case. greater is the intensity of the summer monsoon [33]. The main causal factor in the stepwise weakening It follows that desert margin retreat in northern China of the summer monsoon, therefore, was probably the is consistent with a stronger summer monsoon. Thus, increase in global ice volume. The marine oxygen the linkage described above between the sand-sized isotope record [23–25] shows a gradual, oscillatory particle fraction within loess, desert margin location increase in global ice volume from the late Pliocene and summer monsoon intensity forms the basis of a until replaced by a dstationaryT fluctuation at around novel, and readily applied monsoon proxy. As desert 0.8 Ma (Fig. 5E); this is broadly consistent with the margin shift is a geographical phenomenon, the sand- trend in East-Asian monsoon evolution (Fig. 5C). sized particle content in loess near desert margin Numerical modeling results also suggest that Pleisto- should be regarded as a more direct and sensitive cene glaciation should lead to a substantial weakening proxy for monsoon strength than either loess magnetic of the East-Asian summer monsoon [35]. Here, we susceptibility or elemental ratios derived from physi- propose that the global ice volume signatures were cal and chemical characteristics of loess, respectively. transferred into the monsoon system by means of Z.L. Ding et al. / Earth and Planetary Science Letters 237 (2005) 45–55 53 three well-known mechanisms. (1) Development and lies within the Asian continent, rather than within the expansion of polar ice sheets and sea ice cover would oceans, providing further support for a link between lead to increased temperature and atmospheric pres- ice volume and monsoon sure gradients between the polar regions and the strength, as argued here. middle-low latitudes, thereby impeding northward The close linkage between the East-Asian mon- movement of moisture-laden monsoon airflows. (2) soon and global ice volume, on both long and glacial Ice accumulation on continents would cause sea-level cycle time scales, may have important implications lowering and exposure of vast continental shelves in for assessing the consequences of human-induced the Pacific marginal seas [42], thus resulting in en- global warming for ecosystems in the semi-arid hanced continentality (increased distances to ocean regions in northern China, where devastating desert- moisture sources, affecting monsoon rainfall belt mi- ification now threatens. If the anticipated future gration towards the inland desert margins). Given the melting of polar ice sheets and sea ice cover [45] steep monsoon precipitation gradients, particularly leads to a large-scale reorganization of atmospheric over northern China (Fig. 1), such changes would circulation patterns, a northward shift of the summer have the effect of weakening the summer monsoon. monsoon rainfall belt is highly likely, thus raising the (3) The expansion of polar ice sheets and sea ice cover possibility of reversing the current trend towards would result in the intensification of the northerly ecological deterioration. winter monsoon winds, imposing a downstream cool- ing on the low-latitude oceans. This would thereby lead to weakened oceanic water evaporation and a Acknowledgments decreased summer monsoon season via delayed onset. The abruptness of desert margin shift (Fig. This research is financially supported by the CAS 5C and D) also suggests that threshold mechanisms (grant KZCX2-SW-133) and NNSFC (grants should be recognized as fundamental factors in as- 90202020 and 40021202). We thank Dr. Steven Clem- sessment of long-term monsoon evolution. ens for valuable comments on an early version of the By generating a multiple-proxy data set, Clemens manuscript. et al. [43] found that the growth of ice sheets in the Northern Hemisphere over the past 3.5 Ma has weak- ened the Indian summer monsoon and increased the References aridity of subtropical Asia and eastern Africa. With [1] S. Clemens, W.L. Prell, Late Pleistocene variability of Arabian respect to long-term evolution, the phase of summer sea summer monsoon winds and continental aridity: eolian monsoon maxima relative to global ice volume max- records from lithogenic component or deep-sea sediments, ima has changed significantly over both the preces- Paleoceanography 5 (1990) 109–145. sion and obliquity bands, with major changes [2] P.B. deMenocal, Plio–Pleistocene African climate, Science occurring at about 2.6, 1.2 and 0.6 Ma. This would 270 (1995) 53–59. [3] D.K. Rea, H. Snoeckx, L.H. Joseph, Late Cenozoic eolian lead us to expect a similar evolutionary history for deposition in the North Pacific: Asian drying, Tibetan uplift, closely coupled though discrete East-Asian and Indian and cooling of the Northern Hemisphere, Paleoceanography monsoons, although such abruptness of events shown 13 (1998) 215–224. here for China is not evident in the Indian monsoon [4] M.A. Cane, P. Molnar, Closing of the Indonesia seaway as a records. The Indian subsystem has a north–south precursor to east African aridification around 3–4 million years ago, Nature 411 (2001) 157–162. land–ocean configuration with a latent heat source [5] J. Martin, Glacial–interglacial CO2 change: the iron hypothe- in the southern hemisphere part of the Indian Ocean. sis, Paleoceanography 5 (1990) 1–13. In contrast, the East-Asian subsystem has a significant [6] X. Li, H. Maring, D. Savole, K. Voss, J.M. Prospero, Domi- east–west land–ocean configuration with a major Pa- nance of mineral dust in aerosol light-scattering in the North cific Ocean latent heat source (e.g. [44]). The close Atlantic trade winds, Nature 380 (1996) 416–419. [7] A.J. Watson, D.C.E. Bakker, A.J. Ridgwell, P.W. Boyd, C.S. coupling between the Indian and East-Asian monsoon Law, Effect of iron supply on southern ocean CO2 uptake and subsystems indicates that the common mechanism implications for glacial atmospheric CO2, Nature 407 (2000) driving the long-term evolution of these subsystems 730–733. 54 Z.L. Ding et al. / Earth and Planetary Science Letters 237 (2005) 45–55

[8] Z.T. Guo, W.F. Ruddiman, Q.Z. Hao, H.B. Wu, Y.S. Qiao, on ODP site 677, Trans. R. Soc. Edinb. Earth Sci. 81 (1990) R.X. Zhu, S.Z. Peng, J.J. Wei, B.Y. Yuan, T.S. Liu, Onset of 251–261. Asian by 22 Myr ago inferred from loess [25] N.J. Shackleton, M.A. Hall, D. Pate, Pliocene stable isotope deposits in China, Nature 416 (2002) 159–163. stratigraphy of site 846, Proc. Ocean Drill. Program Sci. [9] J.M. Sun, Z.L. Ding, T.S. Liu, Desert distributions during the Results 138 (1995) 337–355. glacial maximum and climatic optimum: example of China, [26] G. Kukla, Z.S. An, Loess stratigraphy in central China, Episodes 21 (1998) 28–31. Palaeogeogr. Palaeoclimatol. Palaeoecol. 72 (1989) 203–225. [10] T.S. Liu, Loess and the Environment, China Ocean Press, [27] Z. An, G.J. Kukla, S.C. Porter, J. Xiao, Magnetic suscepti- Beijing, 1985, 251 pp. bility evidence of monsoon variation on the Loess Plateau of [11] Z.S. An, G. Kukla, S.C. Porter, J.L. Xiao, Late quaternary central China during the last 130,000 years, Quat. Res. 36 dust flow on the Chinese loess plateau, Catena 18 (1991) (1991) 29–36. 125–132. [28] J. Chen, Z.S. An, J. Head, Variation of Rb/Sr ratios in the [12] E. Derbyshire, X. Meng, R.A. Kemp, Provenance, transport loess-paleosol sequences of central China during the last and characteristics of modern aeolian dust in western 130000 years and their implications for monsoon paleoclima- Province, China, and interpretation of the Quaternary loess tology, Quat. Res. 51 (1999) 215–219. record, J. Arid Environ. 39 (1998) 497–516. [29] G. Kukla, F. Heller, X.M. Liu, T.C. Xu, T.S. Liu, Z.S. An, [13] Y. Lu, J.R. Prescott, G.B. Robertson, J.T. Hutton, Thermolu- Pleistocene climates in China dated by magnetic susceptibility, minescence dating of the Malan loess at Zhaitang, China, Geology 16 (1988) 811–814. Geology 15 (1987) 603–605. [30] L.P. Zhou, F. Oldfield, A.G. Wintle, S.G. Robinson, J.T. [14] J.M. Sun, G.M. Yin, Z.L. Ding, T.S. Liu, J. Chen, Thermolu- Wang, Partly pedogenic origin of magnetic variations in Chi- minescence chronology of sand profiles in the Mu Us Desert, nese loess, Nature 346 (1990) 737–739. China, Palaeogeogr. Palaeoclimatol. Palaeoecol. 144 (1998) [31] J. Sun, T. Liu, Multiple origins and interpretations of the 225–233. magnetic susceptibility signal in Chinese wind-blown sedi- [15] Z.L. Ding, J.M. Sun, N.W. Rutter, D. Rokosh, T.S. Liu, ments, Earth Planet. Sci. Lett. 180 (2000) 287–296. Changes in sand content of loess deposits along a north– [32] H.Y. Lu, T.S. Liu, Z.Y. Gu, B.Z. Liu, L.P. Zhou, J.M. Han, south transect of the Chinese loess plateau and the implication N.Q. Wu, Effect of burning C3 and C4 plants on the magnetic for desert variations, Quat. Res. 52 (1999) 56–62. susceptibility signal in soils, Geophys. Res. Lett. 27 (2000) [16] Z.L. Ding, J.Z. Ren, S.L. Yang, T.S. Liu, Climate instability 2013–2016. during the penultimate glaciation: evidence from two high- [33] S.Y. Tao, L.X. Chen, A review of recent research on the East resolution loess records, China, J. Geophys. Res. 104 (1999) Asian Summer Monsoon, in: C.P. Chang, T.N. Krishnamurti 20123–20132. (Eds.), Monsoon Meteorology, Oxford University Press, Ox- [17] K. Pye, Aeolian Dust and Dust Deposits, Academic Press, ford, 1987, pp. 60–92. London, 1987, 334 pp. [34] Z.S. An, T.S. Liu, Y.C. Lu, S.C. Porter, G. Kukla, X.H. Wu, [18] H. Tsoar, K. Pye, Dust transport and the question of desert Y.M. Hua, The long-term paleomonsoon variation recorded by loess formation, Sedimentology 34 (1987) 139–153. the loess-paleosol sequence in central China, Quat. Int. 7/ [19] T. Liu, Z. Ding, Chinese loess and the paleomonsoon, Annu. 8 (1991) 91–95. Rev. Earth Planet. Sci. 26 (1998) 111–145. [35] Z.S. An, J.E. Kutzbach, W.L. Prell, S.C. Porter, Evolution [20] Z.L. Ding, T.S. Liu, N.W. Rutter, Z.W. Yu, Z.T. Guo, R.X. of Asian monsoons and phased uplift of the Himalaya– Zhu, Ice-volume forcing of the East Asia winter monsoon Tibetan plateau since Late Miocene times, Nature 411 variations in the past 800,000 years, Quat. Res. 44 (1995) (2001) 62–66. 149–159. [36] D.Z. Yeh, Y.X. Gao, Meteorology of the Qinghai–Xizang [21] Z. Ding, J. Sun, T. Liu, Stepwise advance of the Mu Us desert Plateau (in Chinese), Science Press, Beijing, 1979, 278 pp. since late Pliocene: evidence from a red clay–loess record, [37] J.E. Kutzbach, O.J. Guetter, W.F. Ruddiman, W.L. Prell, Sen- Chin. Sci. Bull. 44 (1999) 1211–1214. sitivity of climate to late Cenozoic uplift in southern Asia and [22] Z.L. Ding, E. Derbyshire, S.L. Yang, Z.W. Yu, S.F. Xiong, the American West: numerical experiments, J. Geophys. Res. T.S. Liu, Stacked 2.6-Ma grain size record from the Chinese 94 (1989) 18393–18407. loess based on five sections and correlation with the deep-sea [38] X.D. Liu, Z.Y. Yin, Sensitivity of East Asian monsoon climate d18O record, Paleoceanography 17 (2002) 1033, doi:10.1029/ to the uplift of the Tibetan Plateau, Palaeogeogr. Palaeoclima- 2001PA000725. tol. Palaeoecol. 183 (2002) 223–245. [23] N.J. Shackleton, N.G. Pisias, Atmospheric carbon dioxide, [39] M. Coleman, K. Hodges, Evidence for Tibetan plateau uplift orbital forcing, and climate, in: E.T. Sundquist, W.S. Broecker before 14 Ma ago from a new minimum age for east–west

(Eds.), The Carbon Cycle and Atmospheric CO2: Natural extension, Nature 374 (1995) 49–52. Variations, Archean to Present, Geophysical Monograph Se- [40] T.M. Harrison, P. Copeland, W.S.F. Kidd, A. Yin, Raising ries, vol. 32, American Geophysical Union, Washington, DC, Tibet, Science 255 (1992) 1663–1670. 1985, pp. 303–317. [41] P. Tapponnier, Z. Xu, F. Roger, B. Meyer, N. Arnaud, G. [24] N.J. Shackleton, A. Berger, W.R. Peltier, An alternative astro- Wittlinger, J. Yang, Oblique stepwise rise and growth of the nomical calibration of the lower Pleistocene timescale based Tibet Plateau, Science 294 (2001) 1671–1677. Z.L. Ding et al. / Earth and Planetary Science Letters 237 (2005) 45–55 55

[42] P. Wang, Response of West Pacific margin seas to glacial [44] B. Wang, S.C. Clemens, P. Liu, Contrasting the Indian and cycles: paleoceanographic and sedimentological features, East Asian monsoons: implications on geologic timescales, Mar. Geol. 156 (1998) 5–39. Mar. Geol. 201 (2003) 5–21. [43] S.C. Clemens, D.W. Murray, W.L. Prell, Nonstationary phase [45] IPCC, IPCC Third Assessment Report: 2001: of the Plio–Pleistocene Asian monsoon, Science 274 (1996) The Scientific Basis, Cambridge University Press, Cambridge, 943–948. 2001, 881 pp.