Last modified 2nd December 1997



+ Eurasia 18,000 14C years ago

+ Eurasia 8,000 14C years ago

+ Eurasia 5,000 14C years ago

+ Eurasia, present-potential.

List of References (separate document)

Key to the vegetation classification system used in the atlas

Return to QEN Home Page


The text is in two parts: North Asia, and Southern and Eastern Asia. The maps are of palaeovegetation from both regions combined.

  • Click here for Part Two, on Northern Eurasia
  • PART ONE. SOUTHERN AND EASTERN ASIA from the central Asian desert southwards and eastwards.

    18,000 radiocarbon years ago (Last Glacial Maximum, or LGM).

    Distribution of sites used towards reconstructing the LGM palaeovegetation distribution for southern and eastern Eurasia

    The region in general. Much drier, colder conditions. The general picture emerging for this region is of a considerable reduction and southwards retreat of the forest at the LGM, relative to the interglacial state. Deserts expanded, and even where arboreal vegetation did predominate, this tended to be of a more open character than during the Holocene. Lake level evidence agrees with the general picture from other sources, with many lakes drying up completely around 18,000 radiocarbon years ago (Fang 1991, Winkler & Wang 1993). However, some lakes (e.g. in NW Tibet) in fact show the opposite trend, towards moister conditions at the LGM (Winkler & Wang 1993).

    The overall picture of aridity in southern Asia fits in with the African evidence to suggest a general failure of the summer monsoon rains to penetrate as far north as at present, and a tendency to be less abundant where they did fall. However, the LGM vegetation of many parts of the south Asian region remains almost unknown. The greatest certainty on vegetation character at the LGM is for what are presently the cool moist temperate latitudes, although information is now becoming available for other areas too. For those of us from the West, there is the general problem that we cannot read the considerable literature on the Quaternary that has been published in Chinese, and must rely on our Chinese colleagues to translate and interpret this information on our behalf. My discussion and summary is thus based mainly upon reviews and tentative LGM summary maps published in English, and upon discussions with Chinese colleagues.

    An LGM vegetation map (18,000 - 15,000 14C yr. BP) for China has been published (in Chinese) in 1990 by An et al. (Sun Xiangjun pers. comm., 1992). Wang & Sun (1994) have published an LGM vegetation map showing a major reduction in forest vegetation and southwards retreat of climatic zones relative to the present-natural, although we are inclined to say that considering all the evidence together, conditions were more arid even than Wang & Sun suggest.

    An aridity maximum at 17,000-15,000 14C y.a.? There are now some signs that conditions around 17,000-15,000 14C years ago were the aridity peak in at least some parts of Eurasia. In Yunnan Province of the extreme SW of China (Liu 1991), the local warm temperate forest disappeared from the site at around this time. In eastern and central Turkey, lake levels were lowest shortly after the LGM - 17,000-15,000 4C y.a. (Landmann et al. 1997). In general, however, conditions in terms of vegetation cover most parts of Eurasia seem to have been similar throughout the interval 18,000-15,000 14C y.a.

    Sea level. For cartographic convenience, coastlines are mostly drawn at the -150 m contour, although this is certainly too low. In fact, a figure of -125m is suggested for the Indonesian region on the basis of the latest compilation of evidence (Hantoro et al., manuscript in preparation).

    Considerable stretches of low lying land were uncovered around the shores of China, and the Malaysian Peninsula became linked to the islands of Borneo, Java and Sumatra, and to the Philippines. The islands of Japan were linked together into a peninsula due to the lower sea level, but probably remained separated from the Asian mainland by the Korean channel. The Sea of Japan was almost entirely enclosed as a lake, its only outlet being the Korean channel.

    Ice extent.

    The Himalayan region is the only area in the region likely to have had significant ice cover. It is generally thought that there was not a large ice sheet over the Himalayan Plateau, but rather a scattering of glaciers and small ice caps. The extensive evidence of aridity over south Asia is thought to indicate that there would not have been enough moisture to support the build-up of such an ice sheet. Studying glacial tills and other remnants, Han (1991) considers that he and his colleagues have now found the necessary evidence of continuous ice cover during the early-to-mid Quaternary glaciations, but for the last glacial maximum he regards the glaciers as having been small in scale and mainly confined to small valley glaciers, due to aridity. Permafrost desert conditions appear to have existed in the unglaciated Himalayan regions (Thompson pers. comm.). This general scenario of limited glaciation is acknowledged in the CLIMAP (1981) map of the LGM, which shows only 10-20% ice cover in this region.

    However, Kuhle (1991) seems confident that he has assembled the necessary evidence for a very extensive ice sheet covering all of the Tibetan Plateau (from the Himalayas north to Kun-Lun Shan) during the last glacial. The lower equilibrium line of the ice sheet would have been around 4000m a.s.l. (1100-1600m lower than today), so all of the land above this altitude would have been ice-covered. This putative ice cap may have reached 2.7 km thickness at its centre (Kuhle 1991). Such an ice cap would require considerable precipitation over the Plateau during the LGM period, and seem to stand in surprising contradiction to the LGM aridity seen elsewhere in southern Asia.

    In the map reconstruction here, the more widely accepted view of Han (1991) is assumed; in any case there would have been effectively no vegetation in the higher plateau region for the LGM. For the Himalayan region in general, we suggest a topographic mosaic of 70% polar desert, and 30% dry montane tundra.

    Northern China (north of about 30N).

    Much colder, much drier. The general indications are of much colder, more arid conditions than at present. Lake level evidence almost everywhere indicates dry conditions, with complete drying-up of many lakes (Winkler & Wang 1993). On the basis of diverse palaeoevidence (summarised below), Petit-Maire et al. (1994) present a rough summary sketch diagram suggesting that the summer monsoon limit was shifted some 700 km to the south-east in Northern China; a shift of this order would bring the semi-desert steppe climates of eastern Mongolia down into central China.

    Eastward expansion of the central Asian desert belt The widespread distribution of loess in Northern China (and extending eastwards to Korea) indicates more extensive central Asian desert conditions at around the LGM (Pye & Zhou 1989). Although most loess sections have not been radiocarbon dated - due to the paucity of organic matter - correlation with marine isotope records indicates that the loess was deposited under glacial conditions, alternating with stable soil-forming conditions during warmer, moister interglacials. Magnetic susceptibility of loess sequences, tentatively correlated with ice-core climate evidence, suggests a much drier climate than now in the Linxia area on the western Yellow River loess plateau uplands (about 103 E, 36 N). Li et al. (1995) suggest that around 18-22,000 ('real' rather than 14C?) years ago there was a much drier climate than at present, about 100mm as opposed the present annual rainfall of 350mm. Mean temperatures seem to have been about 6 or 7°C lower than at present. Palynomorphs recovered from loess indicate that during each glacial interval the vegetation was much more open than today's, with the dominance of pollen of herbs and grasses in many loess horizons, in contrast to the arboreal types found during the present interglacial conditions. In a recent analysis of the isotope ratios of organic matter extracted from a loess section (109E, 37N) in central north China, Frakes & Jianzhong (1994) found that C4 plants were considerably more abundant than at present at the LGM, although C3 plants remained dominant both at the LGM (55-60%) and throughout the Holocene (>80%). In fact, over almost all of northern China at the LGM, there appears to have been 'severe desert with almost no pollen influx' (Sun X., pers. comm., Jan 1992). Cold, arid conditions are also indicated by molluscan palaeoecology (Frakes & Jianzhong 1994).

    Although the degree of soil development and organic matter content of the loesses in north-central and north-eastern China tends to be somewhat higher than for the European loesses (data in; Li et al. 1995), the lack of evidence for soil development (e.g. high CaCO3 content) in the loess which was deposited at around the LGM further supports the view that the whole area was cold and arid with low biological activity and a sparse herbaceous vegetation cover. Presumably the areas to the west, from where the loess was derived and blown by the winds, were even more arid. To a great extent, the source of the loess dust seems to have been the aridified uplands of the Tibetan Plateau (Li et al. 1995). Dust influx to the north Pacific from China reaches a strong peak at the LGM, possibly reflecting greater aridity over Asia at that time, although wind strength may have been greater and thus better able to carry the dust (Hovan et al. 1989). However, some other records indicate a dust peak in the Pacific during the Holocene and not the LGM (Street-Perrott pers. comm. 1995).

    Low lake levels over northern China generally support the view that conditions were more arid at the LGM than at present. Yet although the level of Dalai Lake (112E, 40N) was much lower at the LGM than during most of the Holocene, it was apparently higher than it is today (An et al. 1993). Pollen recovered from surfaces correlated as being approximately contemporaneous with the LGM, in the same area in North-western China, close to Mongolia (112E, 40N), shows that xerophytic plants (especially Artemisia and Chenopodiaceae) were virtually the only ones present (An et al. 1993), in a region that presently has a much denser grassy steppe cover. However, the proximity of this area to the present Ordos Desert perhaps suggests that in fact the spatial shift in climate might not have needed to be very large (editors' observation).

    Arid steppe in the north-east. For the north China plain and NE China ('Manchuria'), a cold and very sparse steppe-tundra dominated by Artemisia with grasses and chenopods was the predominant vegetation (Liu 1986, Wang & Sun 1994). Trees appear to have been absent during the period of the LGM itself. The aridity itself seems likely to have prevented pollen deposition in many sites at the LGM, but radiocarbon-dated pollen diagrams from Beizuanaguan (109 E, 34N) and also near Beijing indicate this sparse Artemisia vegetation.

    Mammoths (Mammuthus) also seem to have been absent from the Yellow River basin and the continental shelf area at around the LGM, possibly due to aridity leading to very sparse vegetation, although they had been were abundantly present in the area both before and after the LGM (An et al. 1991). Arid as well as windy conditions in north-east China are further indicated by sand dunes, detected by seismic profiling on the bed of the southern Yellow Sea (around 35N). This area is thought to have been land at the LGM, and a last glacial age is indicated by radiocarbon dates ranging between 10,000 and 20,000 years ago in the muddy deposits that overlie these dunes (Wang & Sun 1994). Further evidence of aridity in north-eastern China is provided by the occurrence of radiocarbon-dated LGM loess deposits (already referred to in passing, above) across the Bohai Bay (around 120E and 40N) and even southwards of the Yangtze River (25-30N) (Li & Zhou 1993).Dust (though not loess) deposits also occur on islands in the east China Sea (Wang & Sun 1994). Whilst loess deposition itself is thought to require the presence of some vegetation, it emphasises the arid nature of most of northern China. The only partially weathered nature of the LGM loess-soils also suggests that even where the loess was deposited and stabilised, it was in a semi-arid environment (Li & Zhou 1993).

    Treelines. Treelines in mountainous areas in both northern and southern China were about 1,700m lower than at present (An et al. 1991). For example, An et al. (1991) report treeline depressions of about 1,200m in Guizhou.

    East-central China (around 30-35 Deg.N). Deserts on the continental shelf, but parklands also? An et al. (1993) report that active dunes and desert or semi-desert conditions are widespread in the middle-lower Yangtze River basin at around 18,000 14C years ago. As mentioned above, loess deposition was also occurring in the Yangtze area at the LGM. However, there also seems to have been at least some parkland and/or boreal woodland vegetation in the lower Yangtze area, with Chen & Olson (1990) reporting that in east China there are LGM pollen assemblages with high percentages of Pinus, Abies, Picea, indicating a mean annual temperature depression of 9 - 13.5°C.

    Taiwan; steppe in some areas, forest in others. In Taiwan, which would have been connected to the Chinese mainland during the LGM, pollen cores on the western side of the island - presently covered by moist temperate evergreen forest - indicate dry steppe vegetation (87% herbaceous, with 36% Artemisia), although trees (especially Pinus and other Pinaceae) were present to some extent, perhaps in a wooded steppe (Lin & Liew 1986). Similarly, a new pollen record from a peat bog in the uplands (650m a.s.l.) of central Taiwan shows that at the LGM the present moist Castaneopsis forest was absent and its place taken by a herb-dominated assemblage (about 66% herbaceous pollen, consisting of grasses and sedges with Umbelliferae). The main tree present was a drought-tolerant species of Alnus (Liew et al. 1995, Liew et al. 1998). On what are presently the highest rainfall areas in the uplands of the north-east of the island, further pollen evidence indicates that at least some forest vegetation persisted through the LGM, but under rather cooler conditions than today, with Quercus instead of the more warmth-demanding Castaneopsis.

    An overview of Southern China; woodland absent from areas with present precipitation below 1600mm?. Liew et al. note that even in places in eastern China and Taiwan where tree cover was predominant at the LGM, from the presence of abundant grass and open-ground pollen and spore types, it seems to have been fairly open (perhaps more a 'woodland') in those areas. Liew et al. show a present-day annual precipitation map of southern China (p.93). It is interesting to note that two pollen sites which were apparently steppe at the LGM (Pearl River at about 23 deg.N, and Ningbo Plain at 26 Deg.N) have present precipitation above 1400 mm, but evidence of long-distant transport of tree pollen from uplands receiving about 1600mm or more. If we extrapolate this relationship across southern China-Taiwan for the LGM, with all areas with less than 1600mm present rainfall being steppe, and those above 1600mm being forest or open woodland, most of southern China would be steppe, with areas of wooded vegetation in the scattered uplands covering about a third of the region. Even this may be a conservative estimate in favour of forest cover; mid and low altitude sites on Taiwan with annual rainfall above 2000mm also seem to have lost their forest cover at the LGM.

    Open pine forests in the west. In west-central China, on the upland plateau region of the upper Yangtze (around 25N, 110 deg.E), a pollen core indicates pine (Pinus) forest at around the LGM. This was probably a more-or-less open forest or woodland, judging from the low pollen influx values (Sun Xiangjun 1990). This seems to fit with the pattern hypothesized above; that a rather open wooded vegetation survived in areas with greater than 1600mm annual precipitation.

    South China.

    Drier, with savanna development in the present rainforest zone in the south-east. Chen and Olson (1990) suggest that in south and south-east China generally, there were relatively insignificant differences in climate from the present at the LGM. Liu (1991) mentions that although there is little or no sign of temperature depression during glacials in southernmost China.

    However, Wang & Sun (1994) refer to the subtropical rainforest having disappeared from southern China and replaced by mixed conifer and evergreen broad-leaved forest, although they do not cite any specific sources. In places, conditions may have been more arid than this. Reviewing the pollen evidence at several pollen sites scattered through SE China (e.g. Min River Delta 26 deg.N, Pearl River Delta 23 deg.N), Liew et al. (1998) conclude that grasslands predominated in lowland areas, with cool temperate forest and open woodlands in upland areas. On the Leizhou Peninsula in southernmost China (20 deg.N), which presently has a very moist, warm climate and paratropical rainforest, a relatively northerly warm temperate pine/oak forest was present and with signs of more arid conditions (pollen of grasses and open-ground ferns among the trees, and sedimentary evidence of drier conditions). Winkler & Wang (1993) already suggest a belt of 'forest and grassland' across southern China at the LGM, but from the evidence of Liew et al. it seems necessary to conclude that the LGM vegetation map of Winkler & Wang (1993) shows too much forest in south-eastern China/Taiwan. Wang et al. (1977; published in Chinese and cited by Winkler & Wang 1993) obtained a pollen-bearing core from Beibu Bay (22N, 109E) showing undated wet-dry cycles in vegetation during the Quaternary that are suggested as paralleling the cold-warm cycles found elsewhere. During the dry intervals, pollen from savanna plants such as grasses and Artemisia was abundant, in contrast the predominance of tropical/subtropical trees during moist intervals. It may be appropriate to correlate these phases of savanna expansion with the steppic expansion seen on the adjacent shelf area west of Taiwan, indicating a general opening up of the forest vegetation, but such a conclusion is tentative. Cores from the South China Sea (around 20N) showing increased sediment deposition rates have been taken to indicate drier conditions at the LGM, but Wang & Sun suggest that this conclusion might not be justified and simply due to further extension of rivers out across the continental shelf. In the maps here, an opening up of forest vegetation in south China is tentatively shown, considering the pattern of aridity seen in Taiwan and other areas further to the north, and also to the south in Indo-China.

    Survival of warm temperate forests in the south-west. In the mountainous areas of Northern Yunnan Province, there are indications of snowline lowering, indicating a 4-5°C depression in temperature. Slightly moister-than-present LGM conditions are indicated by the levels of the two montane lakes Dianai Lake and West Lake (An et al. 1993) in Yunnan Province in the extreme south-west.

    From a pollen sequence dated to just before the LGM (finishing at 20,000 14C yr. B.P.) at Xishuang-banna in the present subtropical rainforest zone in the uplands of Yunnan Province of the extreme SW of China, Liu (1991) suggests that the LGM climate was nearly warm as at present, but with much higher precipitation in winter (as indicated by the presence of the subalpine tree Dacrydium). The driest period at this particular site, in which pine-oak scrubland with abundant herbs replaced the forest apparently due to a reduction in winter rainfall, is suggested as occurring after 18,000 years ago rather than within the interval 20-18,000 years ago. Hence the map reconstruction here conservatively suggests a belt of moist evergreen warm temperate forest persisting, somewhat further south than at present due to temperature depression.

    A forthcoming overview of vegetation distribution in China from the BIOME 6000 group.

    A summary paper on LGM environments across China has been submitted to J. Biogeog. by 28 regional authors (Ge Yu et al.) for the BIOME 6000 vegetation reconstruction project. Its findings seem closely in agreement with picture outlined above.

    Japan. Open woodland. For Japan, the abundant plant fossil evidence for the LGM (more than 52 pollen sites and 34 macrofossil sites; Ooi et al. 1990) from both onland and offshore cores shows a general southwards shift of the vegetation zones under LGM conditions (Reynolds & Kanser 1990). Permanent ice seems to have covered the uplands of what is now the northernmost island (Hokkaido), with a belt of tundra and open boreal woodland (with extensive grassy openings) dominated by Larch (Larix) and Spruce (Picea) (Igarishi 1996).

    The open-ness of the LGM vegetation throughout Japan seems to have been added to by the fall of an extensive volcanic ash deposit just before the LGM (Ooi et al. 1990). Radiocarbon and ash-dated pollen-bearing sites near Lake Biwa in east-central Japan (Ooi & Sei-ichiro 1989) and at several other localities across the centre of Japan (work published in Japanese, cited by Ooi & Sei-ichiro 1989) indicates that the lowland vegetation at the LGM was Cyperaceae-rich grassland with sparse scattered stands of Alnus, Fraxinus and Salix. Forests of a rather open character with Quercus and Pinus seem to have been widespread in the mid-altitude uplands. Open boreal-type woodland (consisting mainly of Pinus, with Abies and Betula) covered much of Japan's uplands (Heusser 1990), from about the middle of the main island to the south of the linked chain of islands. Artemisia was often present, indicating dryness (Ooi et al. 1990).

    Warmer temperate elements of the flora (e.g. Cryptomeria) persisted only locally as minor components in lowlands towards the southern part of Japan, which seems to have had a dry cool-temperate climate with open vegetation and scattered woodland on the uplands (Ooi 1992). Mean annual temperature in SW Japan seems to have been about 7-9 Deg.C lower and precipitation was probably less than 1/3 of present values (Yasuda 1990).

    Indo-China and Malesia. There is little information from Indo-China and Malaysia in general. The records which do exist are often ambiguously dated.

    Drier conditions in Thailand. Morley & Flenley (1983) refer to undated pollen evidence for pine forest occurring in the present rainforest areas of Thailand and Malaysia, which they suggest as possibly being of LGM age. In the lower Mun river basin, north eastern Thailand, there is loosely dated geomorphological evidence of widespread desiccation associated with aeolian activity and an increase in ground water salinity after 20,000 years ago (Loeffler et al. 1984). The aeolian activity does not seem to have been sufficient to form true dune systems, but riverine sands and silts were blown as sand sheets and loess layers onto the slopes and uplands surrounding the Mun River. Loeffler et al. note that savanna species occur in pollen-bearing cores further south, but they do not state how much further south or what sources they are referring to.

    It is interesting also to note that a deep sea core (Core 35-5) taken at 7N, 112E shows a much higher terrigenous sedimentation rate during the LGM (correlated by oxygen isotopes) than the Holocene. This has been ascribed to aridity and sparser vegetation cover on the tropical land area to the west (southern Indo-China and Malaysia) allowing greater riverine erosion (Broecker et al. 1988). Wang & Sun (1994) suggest that in fact this increase in sedimentation might have been due to the extension of land area with the sea level fall (bringing the river mouths out closer to the core site), rather than an increase in aridity.

    Cooler but moist in uplands of Sumatra and Java. In Sumatra and west Java, two lake sediment records in the mid-altitudes (about 1300-1500 m a.s.l.) indicate drier climates, but there are no indications in the pollen records from these sites of any drying of climate, with everwet indicators persisting through the LGM (Stuijts et al. 1988). If the lake sediment evidence is to be believed, it may indicate a regional decrease in rainfall, that could have affected what are presently the more seasonally dry areas rather more severely. However, it is a matter of guesswork as to how much loss of rainforest there would have been. From these palynological records, there appears to have been a cooler-than-present climate, by at least 1.8°C and perhaps as much as 7°C (which would give a vegetation zone lowering of about 1200m).

    Loss of forest from West Java. A recently obtained (U/Th and 14C dated) pollen core in the intramontane Bandung Basin in West Java (665m a.s.l., 1700mm precip.) indicates that the present freshwater swamp forests were absent during the LGM and were instead replaced by a vegetation similar to the scrubland-grassland of southernmost New Guinea (S. van der Kaars & R. Dam. MS submitted for publication 1994). By their reckoning, this vegetation change would have required about a 30% reduction in annual rainfall. A lowering of the mean annual temperature at the site (presently 23.3°C) by about 4-7°C is also indicated.

    This site is of interest as the first well-dated lowland rainforest core in the region. It suggests that there was indeed a net loss of lowland rainforests in south-east Asia, and that the adjacent montane pollen records do not acknowledge this drying; possibly because their rainfall totals are too large to register effects from any similar diminution in rainfall that did occur in the mountains. In fact, if such significant drying could occur in an area relatively close to the moisture source of the sea, it suggests that there would have been an even more arid climate and sparser vegetation over most of the exposed central Sundaland mass.

    Loss of forest from western Borneo? In what is presently an ever-wet rainforest climate (3200-5000mm annual rainfall) area in lowland western Borneo (Kalimantan), there is evidence of savanna development, with high frequencies of grass pollen at some stage during the late Quaternary (Caratini & Tissot 1988). Thomas & Thorp (1992) note from the available radiocarbon dates that during the earlier Lower Pleniglacial (approx. 60,000 years ago) glacial period large amounts of angular sediment (e.g. the white sands) were deposited in coastal Kalimantan, apparently transported by braided streams with occasional flash-floods, probably in a savanna environment. Whether this earlier glacial phase resembled the LGM in this area is a moot point, although in many parts of the world where there are long records it is closely similar to the last glacial. A second phase of alluvial sedimentation seems to correspond to late glacial age; in these deposits there are two near-basal dates of 10,500 14C yr. B.P. (see Thomas 1994 p.233). This continuing slope wash into the lowermost Holocene may reflect the effects of continuing lower sea level in preventing moist air moving inland, or perhaps a lag in recolonisation of surfaces by closed forest vegetation.

    Rainforest surviving in northern Borneo. In the presently ever-wet rainforest region of Sarawak and Sabah (northern Borneo), there is apparently palaeozoological evidence for the persistence of rainforest through the LGM. At the Niah, Tingkayu and Baturong cave sites, T. Harrison & B. Harrison are reported to have found that humans during the LGM interval were hunting wild dog, tapir and Javan rhino, all of which are characteristic rainforest species (cited without references in Edwards 1994). However, more recent paleozoological work on a cave suggests that the LGM in northern Borneo was a period of open non-forested or woodland conditions (unpublished work in progress). If so, my map may show too much forest in Borneo for the LGM.

    The vegetation of central Sundaland; mangrove swamps? Bellwood (1990) refers to the work of Biswas in 1973, describing the results of core analyses from the bed of the South China Sea east of the Malayan Peninsula. Biswas suggested that during glacial periods there were brackish lagoons and bays at the centre of the Sundaland basin, bordered by widespread mangrove swamps and forests. However, the only published radiocarbon dates that have been obtained for peats from Sundaland fall within the age ranges >30,000 years ago and <12,000 years ago, so the mangrove swamps could well have been from these age bands. Biswas had only one radiocarbon date to base his case upon, and this was a peat sample 11,100 years in age and thus more nearly Interglacial than Glacial in age (though possibly indicating Younger Dryas-age vegetation; a potential analogy for deducing LGM conditions). As Morley (pers. comm., July 1994) suggests, this peat could just as well have come from a narrow fringe of strandline vegetation as an extensive mangrove swamp of the sort that Biswas proposes.

    There is apparently no other palaeo-evidence as to the nature of the vegetation in the large central area of shelf exposed by low sea levels during the LGM. Thomas & Thorp (1992) note Walker's speculation that the rapid oscillations in sea level (1.9-1.5m per century) at around the LGM would have meant that there would have given insufficient time for rainforest colonisation between these sea level oscillations and mangrove phases, especially due to the delays in colonisation caused by brackish soils left behind as the sea retreated.

    Geomorphological evidence for aridity in the south of the Malay Peninsula. Fluvial braided-stream deposits which occur in the southern Malay Peninsula and Singapore are generally regarded as having been deposited during typical 'glacial' phases, and geologists currently describe these deposits as 'semi-arid' (Morley pers. comm. 1990). Similar deposits around Palembang are also apparently glacial in age (Morley pers. comm. 1992, Thomas & Thorp 1992). Landforms in granitic terrain are described by Banka and Billington (cited by Morley pers. comm.) as being formed under seasonal climates that they hypothesise may have been glacial in age. Thomas & Thorp emphasise that one might expect that if there was a large area of shelf exposed in SE Asia, this central area would tend to receive less rainfall from the sea, and an initial loss of forest would be amplified due to a reduction in transpirational recycling of water across the land surface.

    Biogeographical evidence for dry forest expansion. Extension of the relatively dry monsoon forest right across the region during glacials is perhaps also supported by the present-day disjunct distribution of plant species (e.g. the neem, a typical monsoon forest tree) which occur in central Malesian areas such as Java (Whitmore 1983). This biogeographical evidence is relevant in the sense that it might be regarded as indicating the coincidence of drier conditions and lower sea levels in the past.

    Highlands were cooler but only slightly drier at LGM. However, Barmawidjaja et al. 1993 note (as mentioned briefly above) that the data in previously published studies from highlands of Sumatra, Java and New Guinea all suggest cooler temperatures (2-3°C) but not drier conditions in this region at the LGM. However, it is worth noting that coarser sediments from about 18,000 14C years ago onwards during the glacial may indicate somewhat drier conditions at these sites (Thomas & Thorp 1992). Van der Kaars (pers. comm.) suggests that in the Malay Peninsula's upland areas, precipitation did not go far below 2000mm.

    Marine sediments from the Sulu Sea, south-west of the Phillipines, have been taken as suggesting no decrease in rainfall at that time (Lindsey & Thunell 1990). However, on the basis of palynological evidence from cores taken in the northern Molucca Sea (Indonesia), Barmawindjaja et al. (1993) suggest that in fact conditions were somewhat more arid than today at the LGM. The actual degree of aridity that they envisage is not clear from their paper, but they do find a major decrease in rainforest elements and fern pollen at the LGM, perhaps partly due to decreased stream runoff as well as actual vegetation change. But instead of increased aridity, a lowering of the altitudinal vegetation zones on the nearby islands from which the pollen is derived may have been partly or largely responsible for this change (W.A. Van der Kaars pers. comm 1993); there is a large increase in pollen from submontane Fagaceae.

    Map interpretation. R. Morley (pers. comm., March 1992) has kindly drawn up for us a preliminary sketch map for Malaysia/Indonesia at around 18,000 years ago, showing rainforest largely confined to peripheral areas of Sundaland, in north and eastern Borneo, and western Sumatra. He suggests 'monsoon and deciduous forests' and savannas over large areas, and these we have interpreted as roughly equivalent to my woodland/monsoon forest category in terms of structure. He tells us that map he gives has been based on a mixture of pollen evidence, loosely dated sedimentological evidence and undated dry-climate landforms.

    Considering the evidence of drier conditions from around the exposed landmass of Sundaland, and the fact that the central region would have been relatively cut off from rainfall (as GCM predictions also consistently tend to indicate), we suggest a dry forest and savanna belt along the central region.

    India, Pakistan and Bangladesh. India seems to have been generally much drier and more sparsely vegetated at the LGM than it would naturally be at present. Evidence from pollen off the southern coasts of India suggests that savanna and open vegetation were more widespread than at any time during the Holocene (Erdelen & Preu 1990).

    North-western India; desert conditions in Rajasthan. In the north-west, Sarnthein (1982) maps sand-dunes as indicating fairly widespread desert conditions at around the LGM in north-west India. Goudie (1983) likewise maps extensive desert conditions indicated by active sand dunes in north-west India for the LGM. These reconstructions are based, for example, on the dated evidence of Bryson & Swain (1981), who find that what is now a lake in Rajasthan (NW India) was a dune field before the start of the Holocene. Other lake level evidence in northern India (lakes Didvara and Lunkaransar) shows a similar picture of LGM aridity (Singh et al. 1974, Agrawal 1988).

    Probable aridity in Kashmir?. In nearby Kashmir Province (in extreme north-western India), palaeosol development and a C3-dominated grass flora was thought to have been present at about 18,000 years ago, following an aridity maximum just before this time (Agrawal 1988). In the lowlands of Kashmir Valley, palaeosoils have been dated to around 18,000 years ago. Analysis of their carbon shows it was derived from C3 vegetation, indicating cool, moist conditions (in contrast with earlier palaeosol phases which were C4 dominated). Likewise, a bog at 3000m in Kashmir was thought to show fairly moist LGM conditions, with the localised presence of some temperate tree taxa (Agrawal et al. 1990). However, that there is currently some dispute about the dating of the moist phase, with it possibly in fact being older than 24,000 years ago and with LGM aridity. Street-Perrott (pers. comm. 1995) notes that two other more rigorously dated and recently obtained recent pollen diagrams suggest LGM aridity in this area.

    Evidence of aridity off the west coast of India. Salinity in the northern Arabian Sea appears to have been higher than today, indicating decreased input from rivers. On the basis of two offshore cores at 10 and 15N, Van Campo (1986) suggests that reduced rainfall and reduced runoff of rivers from the western Ghats (the hills running along the west side of India) was responsible for a great reduction in mangrove vegetation and an increase in herbaceous pollen (e.g. Gramineae, Chenopodiaceae). The LGM aridity along western India generally seems to have been severe; van Campo (1986) remarks that the LGM conditions on the eastern side of the Arabian Sea would have been about as arid as those on the western side.

    North-central India and Bangladesh; more arid than present. In north-central India, there is radiocarbon-dated sedimentological evidence that two rivers had much more sparsely vegetated catchment areas than at present, around the time of the LGM (Williams & Clarke 1984). These two rivers, the Son and Belan, originate at about 23N in the low Kaimar range of hills, and so presumably reflect the general shift in the environment in Northern India.

    So far, there seems to be no direct evidence on the vegetation of Bangladesh at around the LGM. In the northern Bay of Bengal, planktonic evidence indicates that sea surface salinity was much higher than today, resulting from decreased freshwater input from the Ganges-Brahmaputra-Irrawady river system. This provides a further indication of a reduction in summer monsoon rainfall (Cullen 1991).

    Although sediments of last glacial age have not been analysed from a palaeoenvironmental viewpoint, there are poorly dated mid-to-upper Quaternary palaeosol sequences from NW Bangladesh (24 deg.N, 88 deg.E) with carbon isotopic values and intermittant carbonate concretion phases that show strong wet-dry cycles (Alam et al. 1997). The wetter parts of the climate cycles suggest high annual rainfall (similar to the present 1244mm) and forest vegetation. The drier phases suggest less than 750mm annual rainfall, and dry C4-grassland and scrub vegetation. The drier parts of these cycles might correspond to the general glacial-interglacial pattern of aridity oscillations seen across southern Asia, but interglacial arid phases cannot be ruled out.

    On the basis of the marine cores and sedimentological evidence showing a decrease in river runoff, and the general pattern of south Asian aridity, we suggest that the present monsoonal forest belt would have been largely replaced by scrub or grasslands at about the LGM-late glacial.

    Southern India. In south-western India, a high altitude site in the Nilgri Hills (>2000m above sea level) at the LGM is marked by a strong dominance by C4 plants, and an absence of woody vegetation, as indicated by the isotope composition of peats (Sukumar et al. 1993). This is interpreted as indicating more arid than the present conditions, which have C3 grassland interspersed with stunted montane evergreen forest (a direct CO2 effect might also be involved, since the plants themselves would presumably have been rooted in year-round moist substrate conditions in order for the peat to have been preserved). By extrapolation, one can surmise that at lower altitudes of the Western Ghats there would likewise have been drier-than-present conditions and a retreat of forest.

    With lower sea level, Sri Lanka would have been connected to mainland India via the Palk Strait. Erdelen & Preu (1990) mention work by Moore on mammalian biogeography, suggesting that Sri Lankan rainforest squirrels dispersed across continuous rainforest on the Palk Strait during low sea level glacial phases. However, Erdelen & Preu take a contrasting view that in fact the last glacial maximum was more arid than present, citing presently unobtainable work by Deraniyagala published in 1981, on animal fossils and archaeological sites showing that lowland rainforests on Sri Lanka were contracted and restricted to the periphery of the Central Hills (they suggest however that conditions around 30,000 years ago might have been moist enough for rainforest on the exposed shelf). Erdelen & Preu thus instead suggest that the exposed Palk Strait would have been covered in dry forest or savanna-like vegetation, and although they refer to some 'poorly dated' pollen data it is not clear from their wording whether it is actually from the Palk Strait itself.

    The general picture of greater aridity is backed-up by various oceanographic studies showing evidence of a weaker summer monsoon wind flow over the Indian Ocean (e.g. Rostek et al. 1993). This reduction in the northwards air flow would be expected to have given less rainfall over the region as a whole.

    Conclusion; A fairly arid India at the LGM. From all this scattered evidence, it does seem overwhelmingly likely that all or most of India was considerably drier than present at the LGM. However, the general nature of vegetation at the LGM remains highly speculative, as there is not enough direct evidence to show precisely what degree of aridity occurred over most of the land surface.

    Pakistan; lack of evidence. There seems to be no evidence from Pakistan and Afghanistan, although it is reasonable to suppose that the regional pattern of greater aridity would have extended across the southern part of this area. However, moist (winter rainfall) conditions may have prevailed in the north, as in Iran to the west.

    Other parts of the Middle East. There is relatively little evidence from the area between western Iran and western India. A pollen site, and the lake Zeribar site, from the Zagros mountains of western Iran shows drier and cooler than present conditions at around the LGM, with a semi-desert Artemisia and chenopod steppe predominating (Bottema & van Zeist 1988). N. Roberts (Loughborough University, pers. comm.) suggests that such arid steppe would have predominated over the whole area, in the place of the woodland that is the present-potential vegetation. The central Iranian Plateau seems to have been more arid than present with, extensive dune fields were active on the Plateau at around the same time; due to a combination of greater aridity and stronger wind speeds (Krinsley 1966). Thomas et al. (1997) find further evidence of cooler and drier conditions on the central Iranian Plateau around the LGM (with the driest phase ending around 17,000 y.a., according to optical dating) in the form of frost-shattered and windblown sediments.

    8,000 14C years ago (early Holocene). .

    Sea Level Given that relative sea level was still lower than present in some parts of the world, partly due to the remaining parts of the Laurentide ice sheet, it is likely that coastline in many areas of southern Asia extended slightly further seawards than at present. However no maps seem to be available for this time slice at 8,000 14C y.a., so present-day coastlines are illustrated here. In the Yellow River delta of China, in contrast, relative sea level seems to have been higher than today (possibly due to a legacy of reduced sediment build-up during the preceding glacial phase). Here, the sea apparently came in 80-100 km further inland than now, between 8,000 and 5,000 years ago (Yang & Wang 1990).

    Moister and warmer than present, across the monsoon belt. At 8,000 years ago, the south Asian region in general seems to have been strikingly moister and slightly warmer than at present. The greater moistness fits in with a general pattern extending across into northern Africa, reflecting greater summer monsoon rainfall at that time. Lake level evidence over all of China and Mongolia shows conditions moister than present between around 9,500 and 5,000 years ago (e.g. reviewed by Winkler & Wang 1993, Petit-Maire et al. 1994). However, Gasse & Van Campo (1994) find evidence from lakes in Tibet and Rajasthan of a major dry phase somewhere between around 8,000 and 7,000 years ago. This probably extended across much of the monsoon belt, as they note that it also shows up in lake levels across west Africa. However, my reconstructions here assume the 'background' moister state occurred at the 8,000 years ago time slice itself.

    China. Warmer and moister. For China, vegetation maps published by An et al. (1990) and Winkler & Wang (1993) are based on fossil and sedimentological evidence that at 8,000 years ago conditions were significantly warmer and moister than present. A northwards shift of the forest belts is shown; this was the result of a sudden rise in high forest tree species at the expense of Betula just before 8,000 years ago (Winkler & Wang 1995) (perhaps reflecting either climate warming or ongoing ecological succession) and also a westward expansion of forest and grassland hundreds of kilometres towards the now-arid central Asian interior.

    In the Loess-Plateau area of north-central China, the vegetation seems to have been existing under moister conditions than at present. The proportion of drought-tolerant C4 plants in the vegetation seems to have reached its lowest point during the early-to-mid Holocene, before increasing slightly towards the present (Frakes & Jianzhong 1994). At the Linxia site in the western part of the Loess Plateau, magnetic susceptibility suggests a rainfall of around 460mm compared to the present 350mm, between about 8,000 and 3,000 years ago, with temperatures about 2°C warmer than at present (Li et al. 1995). Other recent magnetic susceptibility work suggests that in the drier western part of the Loess Plateau, annual rainfall was 60-100% higher than present at around 9,000-6,000 14C y.a., but towards the east of the Loess Plateau it has remained about the same since 8,000 14C y.a.

    Many forest tree species extended their ranges further north and west than at present between 8,000 and 5,000 years ago, indicating precipitation about 100mm higher than today in many areas of China, and temperatures were perhaps 2-4°C warmer (An et al. 1990, Winkler & Wang 1993). For example, conditions according to palynological evidence seem to have been around 3-4°C higher than now in Beijing and 2-4°C higher in the lower Yangtze river area (Sun & Chen 1991), reflected in terms of the northward movement of tree taxa. The deciduous forest of north-eastern China seems to have been expanded a couple of hundred kilometres northwards into the Russian far east (Velichko 1991). In this north-eastern area, Betula and Pinus pollen were at lower percentages than at present, their place being taken by other, warmer-climate broad-leaved species.

    Along the east China coast down to around Shanghai, it seems that the sea came in about 100km inland relative to the present coastline (Tong & Shao 1991).

    The maps shown here are based on the various sources of evidence discussed by An et al., including pollen cores from lakes. Winkler & Wang (1993, p.245) also present a map for the early-to-mid Holocene moist phase with an extension of forest vegetation hundreds of kilometres westwards into the Inner Mongolian and Tibetan steppe zone.

    Central Asian desert belt.

    Significantly moister with steppe vegetation.

    Petit-Maire et al. (1994) suggest on the basis of the combined palaeoevidence that the summer monsoon extended a further 300 km or so north-westwards into inner Mongolia, relative to its present limits. This generally moister-than-present phase began around 9,000 years ago and ended around 5,000 years ago.

    Winkler & Wang (1993) discuss the ambiguous (poorly dated) evidence of pollen and buried soils in Inner Mongolia that may suggest moister conditions at various stages during the Holocene. Jaekel (1995) reviews evidence from various sites in Inner Mongolia (China) (at around 100 E and 41-43 N) which suggest that conditions were moist enough for permafrost to be present up until 7,000 years ago. They suggest on this basis that steppe vegetation covered the region, instead of the present-day desert and semi-desert vegetation. J. Olson (University of Arizona, pers. comm. May 1995) has recently found pollen evidence of moist steppic phases (so far undated) from the most arid parts of the Mongolian Desert. He is of the opinion that these were probably lower Holocene in age.

    In a vegetation map presented for 6,000 years ago, but possibly applicable to the whole period 8,000-5,000 years ago, Winkler & Wang 1993 p.245), most of the present-day grassland zone on the Tibetan Plateau is depicted as being covered by a forest-grassland mosaic, reflecting warmer and moister conditions. Likewise a forest-grassland zone is shown as impinging from the north over much of the Mongolian arid steppe zone. The result is much reduced desert/semi-desert zone in central Asia.

    More moist deciduous vegetation in SW China. In Sichuan Province, SW China (around 30 N), Jarvis (1993) also finds pollen evidence for a considerably stronger summer monsoon than at present between 9,100 and 7,800 14C years ago, with deciduous oaks (Quercus) being more abundant than sclerophyllous evergreen ones in the mid-altitude forests close to the edge of the Tibetan plateau.

    Indo-China. Early Holocene humidity in Thailand. In the Chi River basin, northeast Thailand, Tamura (1992) regards sedimentation rates and sediment grades as indicating that a more humid than present climate from early-to-mid Holocene time, up until around 3,500 years ago. Likewise, the evidence from palaeo river channels in the northern part of the central plain of Thailand (Yom River) is that water discharges were three to four times greater than at present during the early to mid Holocene (Bishop & Godley 1994). The present-natural vegetation of these regions is rainforest, and given a still moister climate at 8,000 years ago, the vegetation must presumably have been rainforest at that time.

    S.E. Asia. Similar or moister vegetation conditions in rainforest zone. The south east Asian tropical forest zone seems to have had higher-than-present rainfall during the early Holocene (Flenley pers. comm.), in at least some places close to the equator. Barmawindjaja et al. (1993) note that in the northern Molucca Sea (Indonesia) the indications are that essentially present-day vegetation conditions have existed since around 14,000 years ago. Morley feels that by 8,000 14C years ago the vegetation would have been generally very similar to the present (Morley pers. comm. March '92). For example, in mid-altitudes of central Sumatra, he finds pollen evidence of essentially similar-to-present conditions by 8,600 years ago.

    India. Moister than present. There are various sources of evidence (lake level, sea surface salinity, alluvial sedimentation) showing peaks of humidity in northern and western India at around 11,000-10,000 14C years ago and 7,600-6,000 14C years ago (e.g. Van Campo 1986, Petit-Maire et al. 1994), but still much moister than present at around 8,000 years ago. However, there do not appear to be any clear indications of the vegetation conditions at that time. The most direct information relating to the ecology appears to be from the assemblages of early-to-mid Holocene animal bones from an archaeological site at Mohenjo-daro in the Sind region at the western edge of the Thar Desert (Hyams 1976, p.69), in the lower Indus valley. The Sind presently has a tamarisk and scrub vegetation and a rainfall of around 150mm. The diverse Holocene vertebrate fauna seems to indicate, in contrast, either humid park-land or rainforest, with annual rainfall of at least 1200 mm. It is unfortunate that Hyams only mentions this information in passing, without giving details of the dating of this fauna, nor the literature sources that he has used (except to say that the information is from Piggott op. cit.).

    From cores taken in the Indian Ocean, there is evidence of a stronger-than-present monsoon flow over the Indian region (Van Campo 1986), which would be expected to have given moister conditions at 8,000 years ago, although no quantitative estimates of the rainfall change are available.

    By 8,500 years ago, the present-natural temperate deciduous forest had returned to a site at 3000m, on the southern edge of the Himalayas in Nepal. This forest was still in place at 5,000 years ago, and up until the beginning of anthropogenic deforestation about 2,000 years ago (Yasuda & Tabata 1988).

    C3/C4 balance in upland (>2,000m) peats from the Nilgri Hills in southern India record suggest herbaceous communities under conditions with about the same moisture levels as today at 8,000 years ago, although during the next few thousand years there was a strong arid period (Sukumar et al. 1993).

    5,000 14C years ago.

    The available evidence points to significantly warmer and wetter climates across the monsoon belt, with the summer monsoon rainfall reaching further north than at present. Lake level evidence from widely scattered areas across Eurasia (western Siberia, Mongolia, Yakutia and China) also suggests moister than present conditions at this time (Harrison et al. 1996). In China, sea surface height may have reached its Holocene maximum at around 5,000-6,000 years ago, being 10m or more above its present level (Winkler & Wang 1993).

    Similar-to-present rainforest distribution. Morley suggests that by 5,000 years ago the vegetation across the tropical rainforest region would have been generally very similar to the present - natural (Morley pers. comm. March '92). In southern Borneo (Kalimantan), various studies on the coastal peat swamps have shown peat deposition starting around 5,000-6,000 years ago (Morley 1981). At the upper lowland/montane rainforest boundary in central Sumatra, Morley (1982) also suggests that the first evidence of forest clearance occurs at about 4,000 years ago.

    Warmer and moister conditions in China. Conditions across much of China at 5,000 years ago seem to have been warmer than present, but perhaps cooler than in the early Holocene (Winkler & Wang 1993). In apparent contradiction to this, Sun & Chen (1991) note that conditions throughout China seem to have remained similar at 5,000 years ago to what they had been at 8,000 years ago, with warm temperate forest extending hundreds of kilometres further north than at present. Sun & Chen (1991) note that over much of China, palynological records indicate temperatures 2-4°C warmer than at present (perhaps 5°C higher in the Tibetan Plateau), cooling after about 4,000-3,000 years ago. In the north-east of China ('Manchuria'), peat deposition seems to have begun mainly around the mid-to-late Holocene, coincident with a cooling of climate just after around 5,000 years ago. Lake levels indicate conditions moister than present over most of China up until 3,500 years ago (Fang 1991), and the same picture is indicated by magnetic susceptibility of loess profiles (Li et al. 1995). The presence of Neolithic agriculture in north-western regions of China currently too arid for crop-growing is further testimony of the moister climate which prevailed at around 5,000 years ago (Petit-Maire et al. 1994). Agriculture was already present and expanding throughout the south-east Asian region, but deforestation in southern China and in the monsoon zones of Indo-China does not appear to have been significant until after around 4,000 years ago (e.g. see Tallis 1990).

    Moister in NW India. Likewise, in North-western India (Rajasthan) there is lake level, anthropological and other evidence of higher-than-present rainfall at around 5,000 years ago (Bryson & Swain 1981). Alekseeva (1991) suggests on the basis of palaeochannels of rivers that precipitation in winter exceeded the present by about 200-300mm at around 5,000-4,000 years ago. A figure of 500mm greater is suggested by Singh et al. (1974), who on the basis of plant fossils and the molluscan fauna reconstruct a savanna-grass steppe environment for Rajasthan at that time, in contrast to the present semi-desert. Hyams (1976, p.69) mentions vertebrate fossil evidence from Sind, in the lower Indus Valley, indicating a rainforest or rainforest-savanna environment during the mid-Holocene (though the forest was presumably restricted to riverine borders). The original source material that Hyam cites has not yet been obtainable. To the north-west of this area, in southern Haryana, much moister-than-present conditions are also indicated by palaeolake levels (Bhatia & Singh 1988).

    Names of QEN participating experts (named in the text above) who have made direct contributions to this work on South Asia:

    J.R. Flenley, Department of Geography, Massey University, Palmerston North, New Zealand.

    C.J. Heusser, New York University, New York, USA.

    Liew P.-M., Department of Geology, National Taiwan University, Taipei, Taiwan.

    Liu K.-B, Department of Geography and Anthropology, Louisiana State University, Baton Rouge, LA 70803-4105, USA.

    W.A. van der Kaars, Department of Geography, Royal Holloway College, Egham, Surrey, UK.

    R.J. Morley, c/o Geography Department, University of Cambridge, Cambridge, UK.

    J. Olsen, University of Arizona, USA.

    N. Roberts, Department of Geography, Loughborough University, UK.

    Sun X. , Institute of Botany, Academia Sinica, Beijing 100093, China.

    E. Zelikson, Insitiute of Palaeogeography, Russian Academy of Sciences, Staromonetny per 29, Moscow 109017, Russia.

    Zhou L.-P., Godwin Lab., University of Cambridge, Free School Lane, Cambridge CB2 3RS, UK.

    PART TWO. NORTHERN EURASIA mostly Russia east of the Urals, south to the central Asian desert.

    18,000 14C years ago, The Last Glacial Maximum or LGM.

    Distribution of sites used towards reconstructing the LGM palaeovegetation distribution for northern Eurasia (N.B. Various additional sites have come to our attention since this map was compiled)

    Ice extent. Grosswald's view of large ice extent. There has been a fair amount of controversy over the 18,000 y.a. ice extent in northern Eurasia, and the differences in viewpoint are well summarised by Dawson (1992). One perspective on ice extent is presented by Grosswald (1980, 1995), and favoured by Denton & Hughes (1981), suggesting relatively extensive glaciation, with a large ice lobe extending eastwards across northern Siberia to about 110E. There would have been large proglacial lakes along the southern edge of this ice in western Siberia. See this unpublished manuscript for discussion of the view that there may have been a single very large lake covering most of the West Siberian Basin.

    The larger ice extent was used by CLIMAP (1976) and is widely presented in texts on the Quaternary. Grosswald (1995) suggests that recent data from the ODP Leg 145 cruise support his view of extensive glaciation across northern Siberia. The recent evidence of the position of terminal moraines (Grosswald & Hugues 1995) and also ice-dammed lake sediments summarized in the unpublished manuscript above, also seems to confirm a relatively large ice extent, although not perhaps as large as Grosswald initial advocated.

    Velichko et al.'s view of smaller ice extent. Another view - which was used as a working hypothesis when the LGM here was being drawn up - is held by Velichko and colleagues (e.g. Velichko et al. 1984). They suggest that LGM glaciation was much less extensive, and this view is reflected in their recent maps. The relatively conservative view seems to be more widely accepted, but it is apparently put in doubt by the evidence of a greater LGM ice extent and a very large ice-dammed lake (Goncharov 1989, and also see the link above) existed in the part of western Siberia that Grosswald believes was glaciated. Ice sheet extent in the map given here is taken from maps by Velichko & Kurenkova (1990). Although in their scenario most of Siberia was free of ice sheets at the LGM, the area would presumably have been permafrosted if it was not ice covered.

    It is also suggested by Grosswald (pers. comm., 1991) that various large ice-caps also existed on mountains in the north-east of the Siberian region (e.g. in the Putorana and Verkhoyansk Mountains). However, Velichko et al. (1984) view these uplands as having only scattered glaciation, in the form of mountain glaciers.

    A mega-lake in west Siberia? Goncharov (1989) has found that the whole west Siberian basin was covered by continuous lacustrine and fluvio-lacustrine clays and silts. These molluscan and diatom-bearing deposits are 14C dated as being close to the LGM. The lake, dammed by a north-west Siberian ice sheet, may have flowed out into the Caspian Sea through Kazakstan and the central Asian rivers. Goncharov also suggests that the northern part of the area that Grosswald suggests was glaciated was in fact covered by a marine transgression (due to down-warping of the crust under the weight of the ice). This scenario is shown - for example - as a variant in the recent set of maps produced by Velichko's group in Moscow (Landscapes of Russia during the Late Quaternary, A.A. Velichko et al. in press) and in the earlier map of Velichko and Kurenkova (1990), although the most recent dating evidence indicates that the true extent of the lake may have been underestimated in these reconstructions.

    Sea level. Shorelines at -150m bathymetric contour. Sea level was probably 120 - 140 m lower, and would roughly correspond to the -150m contour. The Black Sea was also somewhat lower; the Caspian Sea was higher and more extensive. Their shorelines here are from Velichko & Kurenkova (1990).

    The region in general. Cold, arid conditions with sparse vegetation.

    There are few pollen records from anywhere within this vast region that are radiocarbon dated from the LGM - a fact that is taken by some (e.g. van Campo et al. 1993) as licence to reconstruct moist tundra and boreal forest over the whole region. However, the case against such vegetation existing on any broad scale is overwhelming when one considers the circumstantial evidence surrounding the hiatus in pollen data for the LGM. The level of understanding of the late-glacial flora, fauna and geomorphology of this area is much greater than for other areas in which direct LGM data are lacking, such as the Amazon Basin.

    The extensive literature on the Siberian region is mainly published in Russian and thus inaccessible to most of us from the west. To some extent, it has been expressed in published summary compilations by Russian authors (e.g. in Frenzel et al. 1992) However, colleagues from the Russian Academy of Sciences (in particular, A.A. Velichko, E. Zelikson & C. Kreminetski) have carefully explained to me their understanding of the literature concerning northern Eurasia during the late glacial and LGM. This distillate of the evidence is based mainly on conversations with them and other Russian Quaternary experts (named within the text), against the background of summary maps such as those published by Velichko & Kurenkova (1990) and in Frenzel et al. (1992). Grichuk (1992) has published a map that seems at odds with those of the others published for this region; large areas of apparent 'forest' cover are shown across southern Siberia and in the Caucuses. However, as Zelikson (pers. comm, 1992) has explained to me, Grichuk's map is not in fact intended to deal with vegetation as we would define it here, but instead biogeographic 'source' areas which contained at least some populations (often very localised) of the species of each particular biome.

    The current consensus amongst Russian Quaternarists appears to be that conditions almost everywhere across northern Eurasia were much colder and drier than at present, with a virtual absence of woody vegetation and only a sparse herbaceous ground cover, if any. The region has been worked on thoroughly enough to show that conditions became progressively colder and more arid after about 26,000 14C years ago, after which sedimentary deposition and fossil preservation seem to have largely stopped. Radiocarbon dated pollen records do not start appearing until after 15,000 14C years ago in the north Eurasian region, and even they show a dry, sparse vegetation in which the open-ground coloniser Artemisia is an important component. The lack of any surviving peat, pollen or other fossil deposits may in itself be taken to indicate the extremely cold and arid nature of the LGM climate. If there was much moisture around at that time, one would expect at least some pockets of stream deposition of plant fossils to have survived - yet it seems that there are none. C. Kreminetski (pers. comm., May 1994) has compiled data on the records of fossil woody material or other plant macrofossils across northern Siberia, and finds that although there are large numbers of radiocarbon-dated records from the period before about 25,000 years ago and after about 15,000 years ago, records become progressively rarer and then are absent completely for several thousand years in the period that includes the LGM. Thus for example, woody fossils in the Ineisei Valley (central Siberia) are not found after 24,000 14C years ago (Drozdov et al. 1995), with a gap in occurrence until several thousand years after the LGM. Kreminetski also obtains a similar picture from the northern Siberian records of fossil mammoths and other mammals, of which none have been found for the LGM period.

    Geomorphologic indicators from across northern Siberia (north of about 57 deg.) include widespread ventifacts (wind-sculpted desert pebbles) and deflation pavements, indicating dry, very sparsely vegetated and windy conditions (Spasskaya 1992). These features are thought to have been formed during glacial periods, and the most recent layers of these are used as a stratigraphic marker of the LGM (Kolpakov 1995). Where loess deposition was occurring during the last glacial, the lack of soil development within this loess is interpreted as a further indication of the low biological activity of the northern Eurasian region during the LGM period (Spasskaya 1992). Pollen records that may represent the LGM period itself have been taken from within the loess, but there is too little organic matter for reliable dating. They show a similar type of dry, very sparse vegetation dominated by Artemisia.

    After the resumption of fossil preservation at around 15,000 14C years ago, an increasing abundance of fossil sites and of woody species indicates progressively warmer and moister conditions towards the Holocene (C. Kreminetski pers. comm. May 1994).

    As an additional source of opinion, A. Andreev (pers. comm. Inst. Geography, Academy of Sciences, Moscow. April 1993) is adamant that conditions all across Siberia at the LGM must have been very dry and cold, without moist tundra or boreal forest. As further evidence he cites the situation in western Siberia during the Younger Dryas when the first dated cores are obtained, which does not seem to have been as cold and dry as the preceding LGM, but which had a periglacial steppe-tundra with the only arboreal vegetation being an open parkland which was confined locally to the floodplains of river valleys. The colder LGM would thus presumably have had even less of a woody component than the Younger Dryas. In particular as a compilation of the evidence and vegetation distributions for the LGM and early Holocene, he cites another more recent monograph by Velichko (1993).

    West, Central and Northern Siberia. The general picture across this vast area - justified above - is of treeless, arid and extremely cold landscapes at the LGM. The only indications from pollen in loess are of a dry flora, of which Artemisia is a major component. According to the predominant opinion amongst Russian palaeoecologists, most of Siberia was covered with a virtually tree-less 'steppe-tundra' or 'periglacial steppe'; a dry, open vegetation incorporating elements of present-day steppe and tundra environments. Towards the north, and at higher altitudes, this would have given way to polar rock desert and ice.

    Far eastern Siberia. Steppe tundra with a woody shrub component in the extreme east. Recently obtained pollen records from the Kolnya River region, west of Kamchatka, (Lozhkin et al. 1993) span the LGM. The pollen spectra show drier and colder conditions than at present, with Artemisia dominant but also Pinus, Betula and Alnus (probably in their shrub forms, as they normally occur at present in this region) collectively an important component in the pollen record, probably as a shrubby mosaic in moister places, with Artemisia most common in the drier microsites. These far eastern sites appear to differ from the steppe-tundra in other localities further west and north, in the relative abundance of woody species and in the fact that Gramineae and other herb species do not seem to have been abundant (Lozhkin et al. 1993). This area would appear to have been subject to influxes of moist air from the north-west Pacific. The localities discussed by Lozhkin et al. are in the upper valleys of northward-draining rivers, but quite possibly a similar vegetation might have existed across the broad lowland shelf area exposed to the south. At higher altitudes and latitudes, in the mountain ranges that cover most of the Russian Far East, the vegetation would presumably have been more sparse than this, consisting mainly of polar desert due to low temperatures and aridity.

    Southern Siberia. Dry, generally treeless conditions in southern Siberia. A belt of boreal forest extending across southern Siberia at about the latitude of Baikal has been suggested by Grichuk (e.g. Grichuk 1992). This view has also been taken up in other reconstructions, such as those of Crowley (1994) and van Campo et al. (1993). Whilst it appears likely that the hills and mountains of southern Siberia were in some sense a refugium for the tree species that constitute the present boreal forest (Frenzel 1992, Kreminetski pers. comm.), there is no direct evidence of their occurrence there at the LGM. The only pollen diagrams for this region are published in Russian, one in the lower-to-mid altitudes of the mountains to the east and another to the west of Baikal (Kreminetski, pers. comm. May 1994). These pollen diagrams indicate a dry, sparse vegetation dominated by Artemisia both before and after the LGM, but with a hiatus during the LGM period. Tree pollen is completely absent for several thousand years before and after the LGM. If trees were present, they must have been confined to very isolated pockets.

    Animal fossils of approximately LGM age occur scattered across this south Siberian belt (approximately 50-57N, 120E to 40W) (Baryshnikov & Markova 1992). Instead of indicating a boreal forest or woodland environment, the species which occur are typical of the "mammoth tundra-steppe" assemblage (including Saiga, Mammuthus and Equus) which occur in the treeless LGM landscapes further west, for which better pollen evidence is available. This would tend to indicate an open, dry and sparsely vegetated environment all across this belt. The fossil record of smaller mammals in the Baikal area (Alexeeva 1995) for the late Quaternary indicates phases in which central Asian desert and steppe mammal species were present in what is now a forested and wooded region. However, the chronology is not exact enough to show whether these in fact corresponded to glacial phases.

    At Lake Baikal, the occurrence of drier-than-present conditions is also suggested by an increased aeolian sediment input at the time corresponding to the Last Glacial Maximum, dated according to a correlation with oceanic oxygen isotope records (Peck et al. 1994). Although increased wind speeds may have been a factor, Peck et al. are of the opinion that there was also a drier climate. Further indications of a dry, very open vegetation conditions and a windy climate in the south Siberian belt is the widespread occurrence of ventifacts (wind-eroded pebbles) of last glacial age, in areas both west and east of Lake Baikal (Spasskaya 1992). Further west at the same latitude, also within the belt assigned to boreal forest by Grichuk and others, deflation hollows alternate with loess, again indicating a predominantly sparse vegetation in a dry, windy climate (Spasskaya 1992). Velichko & Spasskaya (1991) point out the scattered presence of ephemeral internally-draining lakes (alternating with deflation episodes) between about 52 and 54N.

    In a broad belt extending across Siberia between 52 and 60N, Velichko & Spasskaya (1991) map large areas of wind-blown sands, and the predominance of desert conditions at the LGM. Sand dunes appear to have been active in what are presently the south-west Siberian steppe regions, in the valleys of the Tobol, Irstysh and Ob Rivers (Volkov & Zykina 1984). These sand ridge dunes were deposited on top of alluvial terraces dating from about 20,000 years ago, and are older than about 15,000 years ago when the first dated soils formed on top of them.

    Central Asia. Expanded central Asian desert. Desert in central Asia seems to have extended further northward and eastward than at present, with the aridity within the core of that zone being more severe than at present. In a sense, it seems to have extended northwards to cover much of southern Siberia (see above), and also southwards across the Iranian Plateau (see above). Loess deposition appears to have occurred widely across east-central Asia during glacial époques (though this correlation poorly constrained by dating), implying some vegetation cover to stabilise the loess, but there was insufficient biological activity for soil development at such times (Dodonov 1988), in contrast with interglacial periods.

    Lake levels of the Aral Sea and Caspian Sea seem to have been higher at the LGM than in the modern-day world (Velichko & Spasskaya 1991) suggesting greater available moisture in the western Asian desert area. However, it is difficult to understand where the extra moisture came from, unless periodic storms with rapid runoff (leaving litle opportunity for vegetation to benefit from it) were the source. The Aral Sea is currently fed from the western edge of the Himalayan Plateau, so its higher level at around the LGM could reflect greater winter monsoon rainfall in these mountain areas, combined with lower evaporation rates in the cooler lowland desert. However, Kashmir - a present source area for the Aral Sea - is now generally thought to have been substantially drier at the LGM. The Iranian Plateau (which shows evidence of drier conditions during the Last Glacial) seems unlikely to have contributed significant moisture to the Caspain. Also, there appear to have been very large areas of wind-blown sand sheets active in the lowlands of the south-west Asian desert (Velichko & Spasskaya 1991), implying much more arid and sparsely vegetated conditions than at present. Many sources of evidence also suggest semi-arid conditions across the Ukrainian and west Siberian Plains to the north.

    The view that there was greater glacial-age aridity all across Central Asia is discussed by Velichko and Kurenkova (1990) and also in the chapter on Central Asia in the same volume (Soffer and Gamble 1990). Lake-level and pollen data from north-west China adds to the general picture of greater aridity across the central Asian desert belt at the LGM (Sun pers. comm. 1990); (see the section on southern Asia, above). There is a notable lack of any archaeological evidence of human habitation from the plains area between the Caspian and Aral Seas and Lake Balkhash, despite the fact that dated sites occur a couple of thousand years before and after the LGM period (Madeskya 1990). This too may reinforce the general impression of greater aridity in the central Asia area. However, in many other fairly arid areas (e.g. the Zagros mountains) human habitation seems to have continued right through the LGM.

    8,000 14C years ago (early Holocene)

    Siberian peatland areas. Extensive peat initiation. From large numbers of dated columns of peat across the Siberian region, it appears that peat initiation in most areas began around 9,000 - 8,000 years ago (C. Kreminetski, May 1994). Likewise, Tallis (1990), in a diagram on page 348 (collated from a range of primary sources), suggests peat build-up initiating around 9,000 - 10,000 years ago, though without any quantitative indications of how extensive this peat initiation would have been by 8,000 years ago.

    Siberian vegetation in general. Greater northern forest extent in Siberia. Forest had reached the arctic coasts in the north by 9000 14C years ago and is here shown extending somewhat further northwards than at present, onto the continental shelf in the north still exposed by the slightly lower sea level. There was a general northwards shift of high-latitude vegetation belts relative to the present, and an expansion of mesic vegetation out into the central Asian desert zone. This follows maps compiled by Velichko (pers. comm.), adding to earlier maps by Khotinsky (1984) which are based on large numbers of pollen and plant macrofossil sites. In eastern Siberia, various plant fossil sites suggest that boreal scrub was present over most of the eastern peninsula (due to warmer and drier conditions?) in an area now occupied by open boreal forest (Tallis 1990).

    5,000 14C years ago

    Although there were some differences in treeline position, conditions were probably similar to the present. A moister central Asian desert belt seems still to have existed, by extrapolation from moister conditions elsewhere.

    Names of QEN participating experts (named in the text above) who have made direct contributions to this work on Northern Asia:

    A. Andreev, Insitute of Palaeogeography, Russian Academy of Sciences, Staromonetny per 29, Moscow 109017, Russia.

    V. Asthakov, Institute of Remote Sensing Methods for Geology, Birzhevy Proyezd 6, 199034, St Petersburg, Russia.

    C. Kreminetski, Insitute of Palaeogeography, Russian Academy of Sciences, Staromonetny per 29, Moscow 109017, Russia.

    X. Sun, Institute of Botany, Academia Sinica, Beijing 100093, China.

    A.A. Velichko, Insitute of Palaeogeography, Russian Academy of Sciences, Staromonetny per 29, Moscow 109017, Russia.

    E. Zelikson, Insitute of Palaeogeography, Russian Academy of Sciences, Staromonetny per 29, Moscow 109017, Russia.