1 Introduction

Wars and internal fighting for control of China discouraged explorers from visiting the Qinghai-Tibet Plateau (QTP) in the first half of the 20th century. Oldham et al. (1925) wrote about studies in the Mount Qomolangma (Everest) region. De Terra and Patterson (1939) examined the Quaternary glaciation of the western part of the Trans-Himalaya in Kashmir.

The opening up of travel on the QTP took place only after the establishment of the People's Republic of China in 1949, especially since the "Reform and Open-up" in the early 1980's. Geologists wanted to learn how the highest plateau in the world was built, its relationship to the northwards movement of the Indian Plate, and the building of the Himalayas, about which, they already had considerable information. Glaciologists wondered about the nature of glaciations there. Earlier, Gerasimov and Zimina (1968) had found evidence that glaciations were more widespread during the Early Pleistocene than in later times. Geocryologists were curious about the permafrost features and landforms occurring there, and quickly found that they were very different from those in the more humid permafrost areas of more northern latitudes (Harris et al., 1998a , 1998b).

Chinese scientists had great difficulty in carrying out research until the early 1960s, but had been steadily accumulating information. The Chinese Scientific Expedition to Mount Qomolangma investigated the evidence for glaciations between Mount Qomolangma and the Ronxua Zangbo River (Liu et al., 1962 ). Other expeditions followed, and Shi et al. (1979) wrote a paper on the effect of uplift of the plateau on China in the first volume of a new Chinese Journal on Glaciology and Cryopedology (Geocryology). This was followed by numerous papers dealing with glacier studies in China, e.g., Zheng (1989). From this work, they concluded that Gerasimov and Zimina were right, and that there were no widespread ice caps developed during cold events on the plateau during the latter part of the Pleistocene (Shi et al., 1990 , 2006, 2011).

Clashing with the evidence amassed by the Chinese is the idea of some German scientists that a major ice sheet developed over the QTP during the cold events of the latter part of the Pleistocene (Kuhle, 1988, 1998, 1999, 2001, 2002, 2004). Kuhle had taken part in the Sino-German joint Geomorphological and Bioecological Expedition to the Mount Qomolangma (Everest) and Mount Xixiabangma regions in 1984. The idea of an ice cap over the plateau seems to have been developed as a hypothesis in 1988, postulating that it altered the climatic regime of the world. Subsequent papers discuss the concept further but with limited field evidence. This contradicted the evidence amassed by Chinese glaciologists and the abundant geocryological evidence described in the literature. In spite of this, it appears in the international literature as late as 2004, and has influenced later climatologists modeling past climates in the area.

Subsequently, remains of numerous primary tessellons have been described from the northeast part of the QTP, e.g., Guo (1979). At first, they were confused with other structures (Harris et al., 2017a ), but this has been overcome. Cui (1980) published a very important inventory of the periglacial features found on the plateau together with details of their characteristics (Cui, 1982), which demonstrated extensive areas of permafrost of late Pleistocene age.

The purpose of this paper is to re-examine the arguments regarding glaciations during the last 60,000 years, discussing the available evidence for the dated extents of the last few glaciations. This is followed by a critical examination of the evidence from tessellons (generally referred to as "sand wedges" in China) together with their age and distribution. It indicates that large areas of the north and central plateau were very cold and dry during the Last Glacial Maximum event (LGM) identified in other parts of the world (Vandenberghe et al., 2014 ). An examination of the effects of lower sea levels in the area of the present China Sea explains these results and indicates a major change in the distribution of air masses and winds in the region.

2 History

The collision of the Indian Plate moving northwards with the Asian plate about 56 Ma B.P. resulted in the buckling of sediments of the former Tethys Sea and the uplift of the rocks to produce a wide, high plateau with the Himalayan Mountains on its southern margin (Robert, 2012). The movement causing the uplift is continuing today resulting in a continually changing environment on the Quinghai-Tibet Plateau (QTP). The following is a brief summary of some features.

2.1 Topography

The QTP is the highest plateau in the world, consisting of an area of about 2.6 million km². From Golmud to Lhasa is nearly 1,200 km, and represents only part of the north-south extent of the plateau. It is bounded by the Himalayas to the south, the Qilian Range, Altun Tagh and Kunlun Mountains to the north and northwest, the Karakoram Range and the 7,000 m Pamir Knot to the west, and the 6,000 m peaks of the Bayan Har, Anêmaqen (Anê Maqen), Hengduan and West Qinlin mountains as well as the Zoȋgé Plateau to the east. The average elevation of the plateau is about 4,500 m, and the Tanggula Range of mountains lies east-west across the middle of it, with peaks rising to well over 6,000 m. The height of the fifteen highest mountains in the Himalayas exceeds 8,000 m, while the mountain peaks of the ranges on the northern plateau rise above 6,500 m. Thus, the plateau is like a castle with topographic walls all around, as well as within the fortress.

The actual surface of the QTP is not as flat as the name suggests, having a series of east-west-trending mountain ranges rising above the general surface (Figures 1 and 2). The plateau also becomes higher and more arid to the west and northwest. It is essentially a desert or semi-desert consisting of alternating basins and ranges that have been uplifted to a great elevation, so that it has developed continuous and discontinuous permafrost on the higher land and north-facing slopes. Taliks occur in river basins and/or sunny slopes at lower elevations Both the Kunlun and Tanggula mountains exhibit relict permafrost and periglacial features at their higher elevations.

2.2 Onset of aridity

By the Late Oligocene (ca. 25 Ma), it is now thought that southern Tibet had become elevated to 4,500 m, and was beginning to be affected by the development of both the Indian and East Asian monsoons. The Indian Plate had now slowed to 57 mm/a, indicating increased resistance from a thickening crust and a deeper root below the elevated plateau. The ¹⁸O analyses suggest that southern Tibet had reached its current elevation of ca. 4,500 m by Miocene times, in contrast to the conclusions of Li et al. (2012) . Another period of uplift occurred between 26 and 15 Ma B.P., with a warm, humid climate until 13.5 Ma.

The western part of Tibet rose rapidly, becoming higher than the east side during the Oligocene. The uplift mainly involved compressive deformation within the terraces produced by erosion of the current average elevation of ca. 4,500 m a.s.l. (Robert, 2012). Associated with this uplift was the commencement of deposition of the aeolian red clay on the Loess Plateau at about 25.6 Ma B.P. (Qiang et al., 2011 ). This red clay is regarded as being the result of desertification of areas to the west, and ended about 4.8 Ma B.P..

About 11 Ma, the northward movement of the Indian Plate had slowed to 44 mm/a, signalling the start of the rise of the Himalayan Range, as well as the resultant rain shadow effect on the plateau to the north. The uplift rates were only 0.1–0.4 mm/a during the Eocene to Miocene times, 0.4–0.5 mm/a during the Pliocene and into the early Pleistocene, since when the uplift has averaged 4–15 mm/a. This agrees with the conclusions of Gerasimov and Zimina (1968).

Harrison et al. (1992) and Molnar et al. (1993) argued that there were further significant increases in elevation between 10–8 Ma, as well as the onset of the Indian and East Asian monsoons based on marine sediments from the Indian and North Pacific oceans (An et al., 2001 ). These monsoons intensified with time, while there was increasing aridity in the Asian interior due to the rising Himalayan Ranges. This started the long, continuous loess deposition on the Loess Plateau, east of Lanzhou (Wang et al., 2013 ). Ren et al. (2009) suggested that the Australasian impact event jolted the Indian Plate enough to cause an increase in tectonic movements when a small (ca. 5 km) meteor crashed into the Indian Ocean at ca. 0.8 Ma B.P..

Enhanced aridity in the Asian interior and the onset of the monsoons occurred about 9–8 Ma B.P.. An et al. (2001) divide the subsequent time into three parts, viz., 6–3.6, 3.6–2.6, and 2.6 Ma B.P. until the present. The 6–3.6 Ma B.P. period corresponds to the time before the onset of the first North Pacific Glaciation that affected Alaska and Eastern Siberia (the Patagonian Glaciation at 3.5 Ma B.P. of Harris, 1994).

The time period between 3.6 and 2.6 Ma B.P. saw increased monsoonal intensification, together with increased dust transport to the North Pacific Ocean (Rea et al., 1998 ). During this period, three glaciations took place on lands on both sides of the North Pacific Ocean (Harris, 1994; Aubekerov and Gorbunov, 1999; Fotiev, 2009). Since then up to the present, fluctuations in the Indian and East Asian monsoons have occurred, probably correlating with major glacial cold events in the North American glacial sequence of Harris (1994). This contrasts with the European glacial sequence which was shorter (only 1.8 Ma—Ehlers and Gibbard, 2008) caused by El Niño entering what is now the Atlantic Ocean through a seaway between North and South American plates until 2.8–2.4 ka B.P., when it closed due to the collision of the American Continental plates.

3 Glaciations

Current thinking is that there were five glacial advances that can be discerned on the plateau (Zheng et al., 2014 ), with at least three occurring during the last 0.5 Ma. Figure 3 shows their official names and elevations of the mountain peaks at the time they were active. Unfortunately, evidence for any earlier events appears to have been destroyed by erosion, while evidence for the last glaciation and Holocene changes has not been described in detail. However, data from oxygen and hydrogen isotopes in a core from the Guliya Ice Cap on the western part of the Kunlun Mountains provides valuable information (Thompson et al., 1997 ). The sequence of events is different to that recorded at higher latitudes (Thiede et al., 2001 ; Hubberten et al., 2004 ), and it is becoming clear that there are considerable regional variations in the duration and timing of glacial events depending on climatic controls in each environment.

Zheng et al. (2014) provide the latest update on the most recent uplift of the northern part of the plateau, after reviewing recent results obtained by various dating methods (Figure 3). There was a Kunlun-Huanghe (Yellow River) tectonic movement between 1.2 and 0.6 Ma B.P. which raised the plateau between the peaks of the northern mountains to about 3,500 m a.s.l.. The Eastern Monsoon provided enough moisture to produce the maximum (Kunlun) glaciation between 0.75 and 0.6 ka B.P., but otherwise, the lack of evidence for glacial advances argues for a cold, arid climate in basins on the plateau. Glacial outwash gravels and associated ice blocks indicated at the base of the Menyuan section (Harris et al., 2017a ) most likely resulted from this glacial event. The next uplift was the Bayiquan tectonic movement between 0.3 and 0.2 Ma before present, at which time these areas were elevated to about 4,200 m a.s.l. (Zheng et al., 2012 ), causing climate on the plateau to become extremely cold and dry as indicated by rock tessellons (Harris and Jin, 2012). This cooling shows in oxygen and hydrogen isotope records from the Guliya Ice Cap at 6,200 m a.s.l. in the western part of the plateau (Thompson et al., 1997 ), and in the ages of thin sandy silt deposits dated at 29 to 19 ka B.P. that are resting on top of the glacial outwash in the Menyuan section on the northeastern margin of the Qilian Mountains. The final uplift was the Gonghe Basin uplift dated at 0.2 to 0.1 Ma B.P.. This raised the basins to at least 4,500 m a.s.l., and the area is still rising at a rate of 6–9 mm/a (Chen et al., 1994 ). For comparison, the Lhasa-Gangdise area is rising at a rate of 9–10 mm/a, the Himalaya at 10–15 mm/a, and Qomolangma at 37 mm/a. These results are very similar to those obtained by Zhang (1987, 1991) and Xu et al. (2000) .

The nearest good Late Pleistocene glacial sequence is that of Owen et al. (2009) who recognized the following moraines on the northern slope of Mount Qomolangma (Everest):

Last Neoglacial moraines, ca. 1.6±0.1 ka;

Neoglacial moraines, ca. 2.4±0.2 ka;

Neoglacial moraines, (6.8–7.7)±0.1 ka;

Rongbu Temple end moraines, 16.6±4.1 ka and 14.2±0.9 to 16.3±0.8 ka;

Jilong Temple end moraines, 24.3±3.8 ka, and 26.5±1.6 ka;

Dzakar moraines, 34.6±6.6 ka or 39.4±4.1 ka;

Tingri moraine, beside the high platform, 330±29 ka.

In this case, erosion has destroyed the evidence for the earlier glacial history. This sequence is important because of the limited detailed information about the late Holocene climatic changes on the QTP. The plateau was too arid to develop a full suite of glacial moraines, the nearest to it being on the north slope of the Kunlun and Tanggula passes. However, it would seem that evidence for the Jilong and Dzakar cold events are to be found in the dated geocryological features that were formed during the last 40 ka in the northeast QTP in the form of extensive permafrost (Harris and Jin, 2012; Vandenberghe et al., 2016 ; Harris et al., 2017a ). Fluvio-glacial gravels are absent from the valleys of the lower parts of the Qilian Mountains, being replaced by the development of ice wedges when the climate became less arid after 20 ka.

The existing evidence available for the sequence of cold events during the last 150 ka based on Jin et al. (2016) and Harris et al. (2017a) on the N.E. QTP is as follows:

ca. 150–130 ka B.P.: Intensive expansion of permafrost during the late stages of the Penultimate glaciation.

ca. 80–50 ka B.P.: Intensive intensification and expansion of permafrost in the basins during the early stages of the LGM, with a glacial maximum dated at 0.75–0.6 ka B.P. on the highest mountains.

ca. 30 ka B.P.: Development of extreme aridity resulting in greatly reduced glaciation and the development of rock- and sand-filled tessellons at elevations from 3,000 to 4,600 m a.s.l..

ca. 20–10.8 ka B.P.: Amelioration of the aridity and cold, resulting first in small glacial advances on the high mountains and the development of ice-wedges below about 4,800 m. Continued warming with fluctuating temperatures resulted in glacial retreat in the mountains and the formation of ice-wedge casts at lower elevations.

ca. 10.8–4 ka B.P.: A weak Hypsithermal/Megathermal event with weak permafrost expansion to 8.5 ka B.P., followed by loss of permafrost from the lower basin areas.

ca. 6–1 ka B.P.: Late Holocene cold period;

ca. 1–0.5 ka B.P.: Late Holocene warm period;

ca. 0.5–0.1 ka B.P.: Little Ice Age;

ca. 0.1 ka to the present time: Recent warming.

Jin et al. (2016) provide descriptions of the evidence for recognising each of these events, but the sequence does not really fit into the marine isotope stages of Lisiecki and Raymo (2005), except where noted below. Voris (2000) summarizes evidence that indicates that much lower levels of the Pacific Ocean caused much of the South China Sea as far north as Japan to become dry land during this time, which caused the East Asian Monsoon to either fail completely or to turn north over the newly exposed sea bed before reaching the present coast of China.

4 Tessselons and their interpretation

Cryological tessellons are structures arranged in a polygonal pattern on the landscape, unlike infilled ice-block casts. They are caused by seasonal contraction of the ground in extremely cold weather (Harris et al., 2017b ). Among the geocryological landforms inventoried by Cui (1980), tessellons are the most abundant, widespread group that have the potential of establishing both the occurrence of cold climatic conditions and the nature of the environment at specific moments in time. A key concept for differentiating wedges of different origins is the use of height-to-width ratios of wedges where the full vertical extent is preserved (Cheng et al., 2006 ). Washburn (1979) provided a clear diagram of the diagnostic features for differentiating primary tessellons from ice-wedge casts. Primary tessellons exhibit vertical foliation in the infilling as well as upturning of the host sediment along their margins. They also have a polygonal arrangement, unlike infillings of former blocks of ice.

In spite of this, tessellons have been generally referred to as sand wedges or ice-wedge casts in the literature, but unfortunately, other features have been grouped with these. Table 1 lists all the features that have been included in "tessellons" or primary sand wedges and shows how they may be differentiated. An example of "sand wedges" that are probably ice-wedge casts would be those described from the Hexi Corridor (Figure 4) by Wang et al. (2003) . In that case, the height to width ratio is that of ice-wedge casts, and the infilling lacks the vertical laminations or bedding of primary tessellons.

Primary sand-filled, loess or rock tessellons imply a dry, cold climate with aeolian sand blowing over the surface in winter. Dates on the infilling will indicate when active cracking in winter was actually occurring. Dates from the host sediment indicate a date prior to the commencement of cold conditions resulting in the cracking. Primary rock tessellons indicate no sand or loess moving past that site, but frost shattered rock was entering and infilling the cracks before they closed up. This requires cold and very dry conditions.

Secondary ice-wedge casts indicate thermal contraction cracking under cold but more humid conditions, where the sediment providing the infilling was moving over the surface of the ground during the thawing of the ice. Thus, dates of the infilling correspond to the time of thawing of the ice in the tessellon, i.e., the presence of a warmer climate. The age of the host sediment constrains the earliest age of development of the cracking resulting in the formation of the ice-wedges. Since no ice remains, it is impossible to date the actual time of formation or the degree of cold by using the δ¹⁸O methodology.

Massive ground ice casts and load casting features indicate the former presence of permafrost, but lack the polygonal form and are not related to cracking of the ground. The infilling of the former ground ice masses indicates the time the ice mass was thawing, i.e., the presence of climatic warming. The load casting materials potentially date the onset of deep, water logged parts of the active layer under warmer conditions. Thus, all these features are evidence of past permafrost in the ground, but they provide quite different information about past environments. The features are important because the available evidence indicates that during the last 30–40 ka B.P., uplift of the plateau had caused waning of the glaciers due to the more arid conditions that result in the formation of geocryological landforms.

5 Occurrences of geocryological features indicating past cold and arid events 5.1 200,000–100,000 years B.P.

By 300 ka B.P., Zheng et al. (2014) regard the plateau as having reached an elevation of 4,000 m a.s.l. Prior to 100 ka B.P., precipitation was higher as indicated by a slightly higher shoreline and terrace around the playas in the Qaidam Basin, and continuous deposition of alternating salts (halite and gypsum) since 1,201 ka B.P.. The first appearance of mirabilite in the Qaidam Basin was at about 189 ka B.P., indicating the commencement of desertification.

The earliest evidence for the presence of permafrost on the QTP reported to date comes from the ice-wedge pseudomorphs in the gravel layer of fluvio-alluvial sediments on the Da'Heba sandbar in Xinghai, Qinghai Province (135,700±10,500 B.P.; Pan and Chen, 1997), and ice-wedge pseudomorphs in the loess strata at North Yangsigeguzui village, Junger Banner, Inner Mongolia (132±13 ka B.P.; Zhou et al., 2000 ). These dates come from the material infilling of former ice-wedges, indicating the time of thawing of the ice within the wedges. Subsequent erosion has apparently destroyed additional evidence elsewhere. Thus, some permafrost was present around the glaciers on the QTP during the Penultimate Glaciation.

5.2 The Interglacial dated at ca. 100,000–80,000 years B.P.

The climate appears to have been warm and dry in the Qaidam Basin at a present-day elevation of 2,700 m a.s.l. (Han et al., 2014 ). Although it has been suggested that the basin supplied large quantities of loess to the Loess Plateau (Kapp et al., 2011 ), and salt crusts on the remains of at least two of the playa lakes in the Qaidam Basin formed about 1.0 ka B.P. so that subsequent erosion of about 800 m of strata from the basin could not have taken place. Strata in the basin are over 3,000 m thick, and yardangs are present on ridges near the western end (Goudie, 2007). It was these yardangs that prompted the suggestion that this area was the source of the loess on the Loess Plateau.

Further north at Qinghai Lake located between the Kunlun and Qilian mountains, Madsen et al. (2008 , Figure 8) suggested that the lake level rose to almost 3,260 m a.s.l., i.e., about 66 m higher than today during this period. This implies that there was enhanced precipitation on the Qilian Mountains to the north, in contrast to the Qaidam Basin to the west. There is additional strong evidence for substantial differences in water balance in the nearby basins, producing asynchronous glaciation of the higher mountains of the Tibetan Plateau.

5.3 The cold marine isotope stage (MIS) 4 (80,000–57,000 years B.P.)

Liu and Lai (2012) dated an ice-wedge/sand wedge from west of Qinghai Lake at Tianjun (ca. 3,300 m) at 62,400±5,700 years B.P.. The central part of the infilling lacked vertical bedding, but that bedding was present near edges of the wedge. Another sand tessellon was found nearby. They therefore concluded that it was formed during the cold marine isotope stage (MIS) 4. Pan and Chen (1997) regarded this stage as being when permafrost was expanding onto the northeastern QTP. Madsen et al. (2008) thought that lake levels in Qinghai Lake were as much as 16 m lower than today around 70 ka B.P., which suggests lower precipitation around the lake at that time.

5.4 The period from 57,000–ca. 30,000 years B.P.

Although this period (MSI 3) is represented by interglacial conditions in many other parts of the world, northeast of China appears to have suffered from quite different climatic conditions (Jin et al., 2016 ). Palaeo-vegetation records from there (Cui et al., 2011 ) indicate a cold and dry climate between 65 and 40 ka, followed by a moister period. Pollen was absent between 30 and 20 ka B.P.. On the northeastern part of the QTP, Rhode et al. (2010) reported a loess-filled wedge dated at 45 ka B.P., but Liu and Lai (2012) could not find it. If it is a valid observation and date, it suggests that the uplift of the plateau during the 57–40 ka B.P. period prevented the area from being entirely permafrost-free during this interglacial. A TL-date of 50,200 years B.P. has been reported from a palaeosol from the eastern QTP. Madsen et al. (2008) suggested that Qinghai Lake gradually rose to about 6 m below present levels before abruptly dropping to about 3,166 m a.s.l. at around 40 ka B.P.. This suggests that the Eastern Monsoon failed to reach that lake at that time.

There is evidence for contemporary permafrost in the form of glacial outwash gravels burying large ice-blocks at the base of the exposed section at Mengyuan on the northeast margin of the Qilian Mountains (Harris et al., 2017a ). These lie beneath sandy silts dated at between 29 ka and 19 ka B.P., and must therefore be older than 29 ka B.P.. Since the psuedomorphs are infilled with sandy silts that descended into the gravels through loadcasts during the thawing of the active layer in the section after about 19 ka B.P., the section must have remained frozen continuously from the deposition of gravels and ice blocks until the thawing of sandy silts after 19 ka B.P.. The rate of deposition of the sandy silts was 2 to 4 cm/century, indicating that they probably represent the accumulation of the nuclei of snowflakes. Taken together with the evidence for the existence of permafrost discussed above, these indicate that glaciers were active in the higher mountains until about 29 ka, after which, the region suffered from only minor precipitation and extreme cold for about a millennium. Thus, glaciation of the bulk of this area was followed by widespread permafrost. Even today, permafrost areas are more widespread than glaciers in this region, so there has been a transition from glacier-type cold conditions on these mountains to cold, dry permafrost at about 30 ka.

5.5 MIS 2 (ca. 30,000–19,000 years B.P.)

The period of 25–17 ka B.P. was taken as representing the Last Glacial Maximum by Vandenberghe et al. (2014) and Shi et al. (1997) based on interpretations of past permafrost distribution in the world, whereas Cui et al. (2011) using pollen data from Northeast China suggested that there was a cold, dry period from 65 to 40 ka B.P.. After reviewing the evidence, Jin et al. (2016) concluded that the "Last Glaciation, LG" spanned the time between 65 and 11–8.5 ka B.P. in Northeast China. By this time, the northeastern part of the QTP had risen to 4,500 m a.s.l. (Zheng et al., 2014 ). Hao et al. (1998) obtained a date of 32 ka B.P. for a palaeosol on the eastern side of the plateau. Cheng et al. (2005a) reported a TL-date of 39.83 ka B.P. for the cryoturbated and frost-shattered bedrock at 4,301 m a.s.l. along the Ngöring access road to Niutoushan, 35°00'17.7"N, 97°35'58.3"E, northeastern Qinghai-Tibet Plateau (Figure 4). It was cut by rock tessellons (Harris and Jin, 2012), one of which was truncated by a sand-filled ice wedge cast TL-dated at 13.49 ka B.P.. The fact that the rock tessellon cut through cryoturbated rock material clearly implies that the host material was no longer moving when the tessellon was formed except to produce polygonal contraction cracking in winter. The cryoturbated material must have formed prior to 30 ka, but was subsequently frozen so as to act as a solid sheet during this period of extreme cooling and aridity. Loose surface rock fragments slid down the open cracks to form the infilling. There is no sign of loess or of aeolian sand in the infilling, implying that all sources of sand or silt in the area were frozen solid. This must therefore correspond to the period during the LGM when the area was at its coldest, bracketing the age of the formation of the rock tessellons at between 39.83 ka and 13.49 ka B.P.. Since the presence of the thawing ice-wedges implies climate amelioration and the onset of wetter conditions prior to 13.49 ka B.P., the rock tessellon must date from the early part of this extreme cold period.

A second section with rock tessellons in shattered, contorted bedrock was found southeast of Maqü in a road cut at the crest of a low hill above the Yellow River at 3,447 m elevation (Figure 5). Once again, the bedrock had been shattered to a depth greater than 4 m, after which, rock tessellons developed in it. Subsequently, faulting occurred and the frost-shattered rock on the higher side of the fault scarp moved down slope over the tessellons as a thin bedded layer. Subsequently two layers of loess were conformably deposited on the top, with a well-developed soil developed in the surface of the lower one. These rock tessellons again indicate dry, cold conditions with the frost shattered bedrock having been frozen solid. No soil developed before the deposition of the lower loess, but there was a period with strong soil formation characterized by freeing iron oxides before the arrival of the final loess. Thus, the tessellons and the bedded shattered rock above it probably formed during the coldest period of the LGM.

An additional section with tessellons developed in the bedrock was seen at the western road cutting along the road access to the Yellow River bridge at the Yellow River village (34°36'10.8"N, 98°15'52.6"E) at an elevation of 4,214 m a.s.l.. The loess infilling exhibited vertical bedding, i.e., they were loess tessellons. Thus, the zone between 3,450 m and 4,300 m on the northern part of the eastern slope of the plateau was cold and dry during the LGM with the sediments and bedrock frozen solid, facilitating ground cracking, except for localized movement of loess close to the Yellow River above 4,200 m a.s.l..

Zheng et al. (2003) reported four TL-dates of 20,200±3,600 years B.P., 18,100±3,600 years B.P., 14,870±2,970 years B.P. and 11,360±2,270 years B.P. for aeolian sand on the eastern part of the QTP, implying that windblown sand was moving in that area during these times. This indicates that the climate on the eastern side of the plateau was starting to ameliorate by about 20 ka B.P..

Conditions along the northern side of the QTP along the Hexi Corridor were rather different. The "sand wedges" of Wang et al. (2003) have a structureless sandy infilling together with the height-to-width ratio of ice-wedge casts, instead of sand wedges. The two TL-dates measured from the top and bottom of the infilling of one of the wedges are 22,500±190 years B.P. and 19,100±125 years B.P.. These are in the corridor through which the loess was transported by strong westerly winds from the Taklimakan Desert in the west, along the narrow, low Hexi Corridor (1,100–1,500 m elevation) to the Loess Plateau, east of Xining and north of Lanzhou. This provides evidence that more moist conditions of Northeast China at this time had expanded southwards into the Hexi Corridor. The transport of sand along the corridor must have started by 22,500 ka B.P. and has been continuous ever since, though the rate of transport has varied. Massive sand dunes were left behind in the deserts to the west.

At Shabanliang, Datong, Shanxi Province, the infilling of a sand wedge was dated at 29,500±2,200 years B.P. (Cui et al., 2002 ) and the filling of a sand wedge west of Dongsheng (now called Ordos City), Inner Mongolia, gave an age of 16,900±1,300 years B.P.. Thus, the onset of movement of aeolian sand in these regions was probably diachronus.

South of Shandan on the southern side of the Hexi Corridor (38°06'01.4"N, 110°20'27.8"E, 2,891 m a.s.l.), loess-filled tessellons were found on the lee slope of rolling hills while ice-wedge casts are located on the lower, west, windward side of the hill (Figure 6). The difference between them is obvious in Figure 7. The margins of both types of wedges are turned upwards, the ice-wedge casts are much wider, and only the primary wedges exhibit vertical bedding. Presumably there was only enough moisture on the lower parts on one side of the ridges to form ice-wedges.

The "sand wedges" or ice-wedge casts from the east end of Qinghai Lake at an elevation of 3,194 m a.s.l. today, were TL-dated at 26,890±2,100 and 25,570±1,970 years B.P. (Liu and Lai, 2012, Figure 2) and have the height-to-width ratio of ice-wedge casts, based on the photographs in their publication. Thus, there had to be some moisture being carried into the area at some period prior to this time or moisture was being recycled from local lakes. Accordingly, the dates obtained by these authors and by Porter et al. (2001) that fall in the time span of the LGM and just later, are evidence of thawing ice-wedges in that topographically lower area becoming filled with loess during the period 23,300±1,900 and 17,600±2,000 years B.P..

At lower elevations on the Ordos Plateau of Inner Mongolia (105°E–115°E, 37°N–40°N), Cui et al., (2002) and Vandenberghe et al. (2004) found a network of polygonal wedges south of Wushenqi with a diameter of 8–9 m. The infillings of these wedges provided dates as old as 33,440±2,540 years B.P.. A case is described where this is transected by a later wedge dated at 23,210±1,790 years B.P. from a different network. A faint vertical lamination is sometimes present in these older wedges, and the lower part of the wedge is narrow. They exhibited up-turning of the margins of the host sediment and were interpreted as being sand tessellons. Near Dongsheng (Ordos Plateau), a narrow wedge with vertical laminations occurred from 1.0–1.9 m, but was overlain by a structureless sandy infilling consisting of a pan-shaped wedge 3 m wide at the top. The bottom laminated sandy infilling gave a TL-date of 26,890±2,100 years B.P., while the structureless infilling immediately below the surface soil was dated at 11,100±840 years B.P.. The upper filling was regarded as being due to seasonal frost. Near Dunhuang, both cryoturbations and shallow sand-filled wedges were found in roadcuts (Figure 8). Desiccation cracks were seen in the vicinity consisting of narrow (5–8 cm) wedges, less than 1 m deep and lacking up-turning of the margins of the host sediment. The infilling consists of weathered surface soil.

A buried pingo scar was reported from Mount Maomoa in Tinzhu County, Gansu Province at 2,540 m elevation by Zhao et al. (2010) . Unfortunately, dating results had too large an error to help in reconstructing the palaeoclimatic history of the area.

Shen et al. (2005) report that the climate was very cold and dry before 16,900 years B.P., based on multiple non-cryogenic proxies. Figure 9 shows the locations of sand and loess tessellons, ice wedge casts and mirabilite reported to date from this time period across the Hexi Corridor. Westward is the Taklimakan Desert from which the loess in the Loess Plateau comes today. The evidence infers that, before 17 ka B.P., there were colder (at least 13 °C cooler) and more moist conditions (ca. 100 mm/a) along the Hexi Corridor and its peripheries.

5.6 Amelioration in temperatures (19,000–10,500 years B.P.)

The exact timing of the commencement in amelioration of the climate is uncertain. Rising sea levels probably commence flooding of the outer portion of the dry China Sea bed about 19 ka B.P., resulting in an increase in the summer precipitation. However, the mean annual air temperature may not have increased much until about 17 ka B.P.. Sometime after the formation of the rock tessellon shown in Figure 4, conditions on the QTP changed and there was enough moisture to allow ice-wedges to be formed (Figure 4). Using multiple proxies, Shen et al. (2005) found that the climate at Qinghai Lake became warmer and wetter after 14,100 calendar years B.P., although frequent fluctuations took place. By contrast, Madsen et al. (2008) concluded that there was a single pronounced short-lived peak in precipitation at about 14.5 ka B.P., but otherwise, the lake level remained about 3,170 m a.s.l.. Thermal contraction cracking continued to occur in winter, but there was a sufficient increase in precipitation that ice-wedges formed in a polygonal pattern, cutting the previously existing rock tessellons. Cheng et al. (2005b , 2006) dated the structureless infilling of these wedges at 13.49 ka B.P.. Sandy infillings of former ice-wedges developed on the Hulun Buir High Plain, Northeast China between 13.5 and 9.5 ka B.P. (Jin et al., 2011 ). Accordingly the ice-wedges must have formed prior to that date, but after the formation of the rock tessellon was finished much earlier. These wedges may be widespread across the northeastern part of the plateau.

In the middle and southwestern part of the Qaidam Basin, fluvial sands were being deposited at 12,600±800 years B.P. (Yu and Lai, 2012; Yu et al., 2013 ). The area lies at about 2,800 m and is surrounded by the Kunlun, Qilian and Altai mountains, rising above 5,000 m. This was followed by a dry, cold period without appreciable fluvial or aeolian action until the Neoglacial period. In the Gonghe Basin much further east, there was a cold, dry phase from 11.8 ka that continued until 10.1 ka, probably related to the Younger Dryas (Liu et al., 2013 ). A similar cold phase at around 11 ka B.P. was reported from Inner Mongolia by Vandenberghe et al. (2004) producing broad, shallow wedges with a mean diameter of 3–4 m. These wedges were often developed in the tops of a much older network whose infilling was dated between 20 and 26 ka B.P.. Vanderbughe et al. considered the shallow (0.6–1.0 m deep) wedges to be ground wedges formed by seasonal freezing, although they could also be the infillings of former ice blocks (see Table 1). Thus, there is evidence for a short period of expansion of cold, dry conditions as in Northeast China (Jin et al., 2007 ).

5.7 Early Holocene climate changes, 10,800 to 8,500 years B.P.

At Qinghai Lake, Shen et al. (2005) found that the trend to warmer and wetter conditions continued until 6,500 calendar years B.P., whereas Madsen et al. (2008) using additional data regarded the level of the lake as having been stable at 3,170 m a.s.l.. In general, the main geocryological features that have been dated from this period occurring below 4,700 m were cryoturbations indicating widespread thawing of permafrost, e.g., 8,860±200 years B.P. in the Zaxitang gully, Zoȋgé Basin, West Sichuan (33°54'N, 102°35'E, at 3,300 m a.s.l.; Li et al., 2012 ); and 9,970±135 years B.P. in the Wumqü River Valley in Damxung, central Tibet Autonomous Region, China (30°30'N, 91°25'E, at 4,600 m a.s.l., Jin et al., 2006 ). Sand infillings of ice-wedge casts were reported by Jin et al. (2006) from Highway Maintenance Station (HMSS) 82 on the southern foot of Fenghou Mountain (34°40'N, 92°45'E, 4,800 m a.s.l.) along the Qinghai-Tibet Highway (National Highway 109), dated at 9,160±170 years B.P.. However, these appear to be unique, high altitude cases. This period was transitional to the Holocene Altithermal/Hypsithermal (Megathermal) period, and the vegetation on the middle to upper slopes of the sides of the plateau (ca. 3,000–4,000 m a.s.l.) changed from mixed needle-leaf and broad-leaf forest to grassy steppe at about the end of this period (Ma, 1998).

Yu and Lai (2012) reported 28 OSL-dates from sand-loess sections in the eastern QTP, suggesting that the aeolian sand mainly accumulated at 12,400–11,500 and 10,000–8,500 years B.P.. The latter was the main phase of loess deposition in the area.

In the Gonghe Basin (ca. 2,800–3,300 m a.s.l.), the climate was wetter and warmer (Liu et al., 2013 ). However, cold intervals were recorded at 9.8–9.5 ka B.P. and 9.2–8.7 ka B.P. North of Mongolia in the region around Lake Baikal, boreal forest developed starting at 8.1 ka B.P. (Tarasov et al., 2007 , 2009), reflecting a change from colder and drier conditions before 9 ka B.P. to milder weather.

5.8 Megathermal (Altithermal/Hypsithermal) period, 8,500 to 6,000 years B.P.

Shen et al. (2005) noted that at 8,200 year B.P., a short but cold event is recorded in sediments of Qinghai Lake. Madsen et al. (2008) found evidence for a sharp rise in the level of the lake to 3,198 m a.s.l., followed by a slight lowering of the water level. This was the warmest part of the Holocene, but peat and humic soils were formed in wet places along the plateau at altitudes from 4,404–4,900 m a.s.l., e.g., at 4.4 m depth in Borehole No. 8 in Xidatan (35°45'N, 94°15'E; Jin et al., 2006 ) and more than 5 m in the source areas of the Yellow River. Li et al. (2012) have also found cryoturbations in sandy silts at the second terrace of Heihe River in the Zoigé Basin (33°52'N, 102°35'E, at 3,300 m a.s.l.), and similar features are recorded from Shazhuya Valley in the Qaidam Basin (7,750±90 years B.P.; 36°20'N, 100°10'E) and from the Tara Mesa in Gonghe Basin, Qinghai (7,890±185 years B.P. at 36°12'N, 100°30'E, at 3,300 m; Xu et al., 1984 ). The climate was more humid on the mountains northeast of Lanzhou. In addition, it would suggest that the easterly summer monsoons were coming far inland at this time, and probably extending onto the lower parts of the eastern plateau above 4,400 m a.s.l.. In the Gonghe Basin, the climate became relatively dry, with cold periods between 8.1–7.6 ka and 7.0–6.1 ka (Liu et al., 2013 ).

Then, there is a gap in time until about 6,000 years B.P. without any recorded geocryological landforms. Thus, the Megathermal event of Jin et al. (2007 , 2016) was not very warm on the northeast part of the Qinghai-Tibet Plateau and was short-lived, being more than counterbalanced by other controls such as uplift of the plateau and the regional pattern of an earlier onset of the succeeding Neoglacial cold fluctuations. Modelling by Liu and Jiang (2016) suggests that these areas were cooling during the earlier part of this time interval by 0.5 °C, compared with a warming of 2–4 °C in other parts of the world.

5.9 The Neoglacial period (6,000 years B.P. – the present)

Although the Neoglacial period is regarded as having started at around 4,500 years B.P. in many parts of the world, this is not obviously consistent with the findings on the plateau that have been obtained to date. Shen et al. (2005) certainly concluded that the warm, wet conditions ended abruptly about 4,500 calendar years B.P. around the east end of Qinghai Lake on the lower slopes of the QTP, resulting in the climate becoming colder and drier. Madsen et al. (2008) suggested that the lake level fluctuated within 5 m of its present level. Thereafter, peat and humic soils became reasonably common. However, in the Gonghe Basin, the cold periods range from 5.3–4.7 ka B.P., 3.1–1.5 ka B.P. and since 0.7 ka B.P. (Liu et al., 2013 ). At Ximen County in the Nianbaoyuze Mountains (4,300 m, 33°25'N, 101°07'E; Li et al., 2012 ), a peat layer is recorded at 5,422±94 years B.P.. Cheng et al. (2006) reported a group of sand filled depressions about 2 m wide and 1 m deep, that gave dates of 5,690–5,430 years B.P.. Their dimensions are similar to those resulting from the infilling of melted-out blocks of ice (Table 1).

The 6 ka date for the commencement of the Neoglacial period also fits with evidence from the Lake Baikal region (Tarasov et al., 2007 , 2009). Similarly Hong et al. (2001) and Yang et al. (2015) reached the same conclusion for areas of Northeast China. Jin et al. (2011) dated soil wedges on the Hulun Buir High Plain as forming between 5.4 and 2.3 ka B.P. They suggest that the southern limit of latitudinal permafrost may have been between 41°N and 43°N in North China at this time.

Yang and Jin (2011) and Yang et al. (2015) measured the stable isotopes and pollen in an ice and sedimentary core from Yituli'he in northern Northeast China, and concluded that there was a colder than present climate there from 6,400–1,400 years B.P., with the lowest temperatures occurring between 6,300 and 4,000 years B.P.. This resulted in the growth of ice-wedges, some of which have survived to represent the southernmost occurrence of ice-wedges in the northern Da Xing'anling Mountains (50°32'N, 121°45'E; Jia et al., 1987 ). These are the most southerly ice-wedges existing today in the Northern Hemisphere. Other older ice-wedges persist at Wuma at 52°45'N, 120°45'E (Tong, 1993). The fact that the δD and low δd excess form a general pattern changing across Eastern Asia supports this correlation with the colder climate of the northeastern plateau (Yang et al., 2015 , Figure 7). This suggests that the idea of the QTP being the centre of a major cold region is correct, but it is correlated with cold dry conditions rather than major expansion of glaciers.

In the central and southwestern portion of the Qaidam Basin, palaeo-sand dunes have been dated at between 4,000 and 500 years B.P. (Yu and Lai, 2012). Periods of stabilization include 1,900–1,700 years B.P. and 1,600–0 years B.P., both periods corresponding to glacial advances in the adjacent mountains. Varve records from Sugan Lake indicate a cold and humid climate (Shao et al., 2006 ). Zeng (2006) indirectly dated two aeolian sand deposits using ash at an archaeological site in Nuomuhong, northern QTP at 3,229±82 years B.P. and 3,477±63 years B.P..

This was a period of fluctuating but cold temperatures with gradual desertification occurring on the plateau. There is a gradation eastwards reported from the Gonghe Basin by Liu et al. (2012) showing increasing moisture from the East Asia monsoon, that only penetrated a limited distance on to the eastern slopes of the QTP. They attributed this to a weakening East Asia monsoon, the thermodynamic effect of the plateau in regulating the regional atmospheric circulation, and the high temperature of this period, causing more evaporation than precipitation as in northern China (An et al., 2006 ). Anthropogenic actions may also have played a role in this, as they have during the last century on the plateau (Harris, 2013).

Using oxygen isotope records of five stalagmites, Wang et al. (2001) produced a high resolution, absolute-dated Late Pleistocene monsoon record close to present sea level from Hulu Cave near Nanjing, China. For the last 30 ka, the record is very similar to that found in Greenland ice cores. Earlier results show a progressive increase in the difference between the two data sets. They concluded that climatic variations recorded in both areas in the last 30 ka were essentially the same with only minor differences (a few centuries). In other words, during this period, the meridional transfer of heat and moisture from the warmest part of the ocean, the original location of the East Asian Monsoon, varied in concert with that in the North Atlantic.

6 Eustatic sea level changes

When water accumulates on land during a glacial advance, sea level will drop because of the loss of considerable quantities of sea water for producing the ice. There can also be changes in the volume of the ocean basins that produce sea level changes, as well as local tectonic movements. However, the latter will only produce changes of sea level relative to the land in the localized area where tectonics are taking place. Elsewhere, the sea level will change much the same amount over vast areas as water is sequestered on land in the form of cryospheric ice. Usually, it is hard to find a good record of the decrease in sea level when the glaciers are advancing, but once they reach their maximum, it is possible to date the timing of the glacio-eustatic sea level rise resulting from the melting of glaciers and ground ice around the world.

Fairbanks (1989) and Voris (2000) examined the evidence for the regional change in sea level for Southeast Asia from the LGM to the present day. Sathiamurthy and Voris (2006) provide maps of the inferred sea level changes using the present-day bathmetry. At about 21.0 ka B.P., sea level was approximately 120 m lower than today, resulting in a very different shoreline at that time. The Sunda and Sahul shelves including most of the floor of the South China Sea were dry land during this period, and Malaysia, Indonesia and the Philippines became a broad peninsula called Sundaland (Molengraaff, 1921; Molengraaff and Weber, 1921).

The low sea level of the LGM at ca. 22 ka B.P. was followed by a gradual rise in sea level, so that around 17 ka B.P., the sea level was at ca. 105 m below the present. This would not have changed the shoreline of the Pacific Ocean of eastern China by very much. However, by 10.8 ka B.P., the sea level had risen to about 37 m below the present-day level, and a considerable area of the present-day South China Sea had been inundated. By 8.5 ka B.P., sea level had risen to within 12 m of its present level and the Chinese shoreline would be approaching that of today. By 6 ka B.P., the glacio-eustatic rise in sea level was essentially complete.

7 Changes in Air Mass distribution

Due to latitudinal variations in temperature and moisture, as well as the underlying surface characteristics (particularly moisture), different air masses form around the Earth. They are labelled using a system in which the first letter of each air mass identifies its moisture status (maritime or continental) and the second its temperature characteristics (Tropical T, Polar P, Arctic or Antarctic A, Equatorial E). They occur over large areas of the globe and tend to be dominant factors in determining the type of climate occurring in any given locality. In the Northern Hemisphere, there is a circulation of warmer air aloft in the troposphere north of about 30°N which sinks in the polar regions, cooling to form a dense, cold air mass (cA/cP) referred to as the Mongolian or Siberian High pressure air mass, which has a specific upper surface, and tends to be imprisoned north of about 50°N during winter (Yu et al., 2015 ). After a substantial build up in mass, it usually breaks out, moving southwards in Eastern and Central Siberia and/or across the Arctic Ocean to northern Canada. After remaining stationary for a few days, the dense air mass breaks out as a "polar vortex" moving south to latitudes as low as 30°N. These outbursts may occur in both continents at once, or independently.

Currently, the climate of China is the result of three major interacting air masses (Figure 10). The fortress-like effect of the QTP is obvious. The Siberian High (cA/cP air) affects the climate of northern China, maintaining permafrost from Mount Huangangliang, the summit of the Da Xing'anling Mountains, and northwards as part of the main Siberian/Mongolian zone of latitudinal permafrost. The southward outbursts result in it linking up with similar cold dry air in northern Tibet, reinforcing the cold air mass in that area in winter. Because of the altitude of the QTP, cold temperatures persist from November to April, unlike the coastal areas (Tu, 1939). In the north, the upper surface is about 4,000 m a.s.l., but this decreases southwards in the coastal area to 2,000 m a.s.l. near Nanjing, and even lower further south. In summer, the cP/cA air mass retreats northwards into Russia and western Mongolia, and is replaced by westerly winds across part of the area and the East Asian Monsoon incursions (mT). The cP air mass is characterised by a clear sky and intense cold (Chang, 1957). Lower snowfall on the northern part of the QTP probably explains part of the colder ground temperatures compared with lowlands in Northeast China during the latter part of the LGM.

The Westerlies (cT) have been crossing the dry deserts to the west, with limited opportunities to pick up moisture from the large remnants of the Tethys, e.g., the Caspian Sea. Subsequently, they rise over the mountains of the Pamir Knot or Tian Shan, where they provide sustenance to glaciers above 4,500 m a.s.l.. They then continue eastwards as dry winds that help maintain desert conditions on the western part of Tibet. East of Tien Shan, they descend to about 1,500–2,000 m a.s.l. on the Taklimakan Desert, arriving as strong, hot, dry air masses, similar to Foehn/Chinook winds. Since the bulk of the air is funneled through the relatively low Hexi Corridor between the Qilian Mountains and the uplands of Inner Mongolia to the north, they erode the finer sediments along their path and carry them eastwards to be deposited as the loess on the Loess Plateau. This leaves spectacular erosion features along the western part of the corridor, as well as yardangs (Figure 11) and some of the largest sand dunes in the world on the floor of the desert basin to the west. This is also a major long distance source of aerosol pollution in Beijing (Wehner et al., 2008 ).

The third major air mass affecting China is the East China Monsoon (mT). This is the chief source of summer rainfall on the extensive, fertile coastal plain on which the bulk of the population of China live. In the south, it has sufficient thickness to penetrate to the lower southern parts of the plateau south of the Tanggula Mountains, its intensity and frequency decreasing with altitude. However, northwards its upper surface decreases in altitude so that it only climbs up to about 4,500 m a.s.l. on the northeast slope of the plateau at latitude 35°N, e.g., to the south shores of Qinghai Lake (Madsen et al., 2008 ), but rarely extends south over the mountains to the Qaidam Basin at 3,000 m a.s.l.. The vegetation along its upper margin on the northeast side of the plateau is becoming eroded by wind, though it is unclear whether this is due to anthropogenic causes or to climate change (Harris, 2013). The East Asia Monsoon does show up in the Qinghai Lake area which is at lower altitude, and brings some moisture to the higher land along the Hexi Corridor. At the western part of Mongolia and in southern Russia, the Monsoon interacts with the cP/cA air mass producing higher precipitation. North of there, it brings summer moisture to Northeast China, Korea, Japan, and coastal Russia.

There are also air masses of minor importance. The Indian Monsoon (mE) is only about 4,000 m thick and has been blocked by the rising Himalayan Range, except in extreme southeast China. Modified cT air sometimes moves north from the southern slopes of the Himalayas and the south of Tibet, flowing northwards over the Tanggula Range to provide a limited amount of summer precipitation.

The glaciers of Tian Shan, south to the Karakorum Ranges are nourished primarily by the westerly winds (Li and Li, 2014; Meyer et al., 2014 ). Li and Li (2014) report that the quantity of precipitation is the primary control on the health of these glaciers, which are currently shrinking. Those at lower elevations are shrinking fastest, while larger glaciers show less change in area. In the south-central part of the QTP north of Lhasa, the glaciers are affected by the East Asian summer monsoon which brings in moisture and by the cold dry westerlies in winter (Zhang et al., 2013 ). A critical factor to their health seems to be the type of summer precipitation, i.e., the relative proportions of rainfall and snowfall. Near Lijiang, the glaciers on Mount Yulong depend on snowfall from the Indian winter monsoon as well as the East Asian Monsoon. The climate at Lijiang shows cyclic fluctuations (2–3 years and 11–12 years) which are correlated with minor advances and retreats of the glaciers (He et al., 2012 ). Otherwise the glaciers are currently stable.

8 Comparison of the present climate with that of the last Neoglacial event

There has been a deterioration of the vegetation on the QTP in the last 50 years that is probably partly due to anthropogenic causes (Harris, 2013). However, the widespread retreat of glaciers across northern China indicates that there has been significant warming and/or reduction in precipitation during the last 150 years. At Yituli'he, the change in mean annual air temperature was inferred to be 4 °C by Yang et al. (2015) . Coupled with changes in precipitation amounts, this has produced dramatic changes in areas of smaller glaciers scattered across the higher Chinese mountains. The Equilibrium Line Altitude (ELA) of the glaciers in the Central Tian Shan Mountains has risen by an average of about 100 m at present (Li and Li, 2014). The Baishui No.1 Glacier on Mount Yulong had retreated 1,250 m horizontally near Lijiang by the 1980s with a rise in altitude of the terminus of 450 m by 2002.

Likewise, there is a significant and ongoing decrease in the area of permafrost that has been measured along the Golmud-Lhasa and Xining-Yushi transportation corridor (Jin et al., 2006 , 2007).

9 Comparison with the LGM

Zhao et al. (2014) provides a map of the distribution of permafrost in China today, with their estimated distribution about 21,000 years B.P.. They regard the local LGM as lasting from 35,000 to 10,500 years B.P., though the evidence discussed above suggests that the cold period started much earlier, and by 17 ka, the climate was ameliorating as the sea level rose, allowing the East Asian Monsoon to bring in moisture from the less-extensive China Sea. The southern limit of latitudinal permafrost in northern China advanced southwards to at least 38°N–40°N in the west and the absence of pollen indicates that extremely cold conditions extended to latitudes between 37°N and 39°N in the east. The total area of permafrost in all of China during the LGM is estimated to be 4.3 million km² compared with 1.75 million km² today. This is based on the known distribution of primary and secondary wedges and other proxies, though it is unclear whether the area mapped consisted of continuous permafrost.

Permafrost could exceed 1,000 m in thickness on very high mountains such as Mount Everest. We have no data on permafrost thicknesses on the Gongga Mountains where Har Goolun Glacier is located, nor whether there is permafrost there. No permafrost is shown in the Takilimakan, Badain Jaran and Tengger deserts, though the subsequent erosion and deposition of sediment may have destroyed the evidence for or against it having been there. Curiously, Shan (1996) reported "sand wedges" from the Tengger Desert.

Wang et al. (2003 , Table 3) provide a summary of estimates of the lowering of mean annual air temperature at the time of the LGM based on various proxies by various authors. The exact dating of the proxies is often rather wide, e.g., 27–10 ka B.P. for "sand wedges" in northern Shanxi (Su and Ma, 1997). Certainly, it would have varied from place to place, but a reasonable range of climate cooling appears to be about 8–14 °C.

Undoubtedly the East Asian Monsoon became greatly reduced at this time and may have either moved or failed altogether (Figure 12). In Europe and Africa, there were enormous changes in the air masses, with the westerlies (mT air) moving south to the area around the Mediterranean Sea (Fairbridge, 1964). As a result, they would have proceeded across the deserts of Southern Asia, providing considerably enhanced precipitation to these regions. Unfortunately, China lies in the rain shadow of the Pamir Knot and the Tian Shan Mountains. Although the winds continued to blow, they did not add appreciable additional precipitation. As a result, the largest ice cap to form on the plateau occurred on the Tanggula Mountains, but even that was relatively small (Zhang et al., 2013 ). The Indian Monsoon affected the south side of India and Sundaland, so there was no possibility of developing a major ice cap on the QTP during the last glaciation.

10 Conclusions

The sequence of cold events on the QTP is greatly influenced by the tectonic history, which in turn, determines the elevation of the land and influences the movements of the air masses. A further complication is the shallowness of the South China Sea, which resulted in a great extension of the land eastwards for much of the last 0.1 Ma. These factors resulted in a switch from glaciers dominant on the mountain ranges that rise above the general surface of the Qinghai-Tibet Plateau to widespread permafrost across the northern plateau beginning about 60 ka B.P.. The LGM (30–20 ka B.P.) was extremely cold and dry. The warmer Altithermal/Megathermal event only lasted from about 7.5–6.0 ka B.P. and was followed by three cold events with temperatures about 4 °C colder than today, which explains why the most southerly ice wedge in the Northern Hemisphere is found in northern China. In general, fluctuations in temperature during recent climatic changes have been amplified on the plateau compared with elsewhere, presumably due to the effects of the high altitude and aridity, and the climatic events do not closely correlate with the sea temperature chronology due to the local environmental changes taking place.