Reconstruction and prediction of climate and vegetation change in the Holocene in the Altai–Sayan mountains, Central Asia

Two quantitative methods were used to reconstruct paleoenvironments and vegetation in the Altai–Sayan mountains, Central Asia, during the Holocene. The ‘biomization’ method of Prentice et al (1996 Clim. Dyn. 12 185–96), applied to the surface pollen record, worked fairly well in the reconstructions of current vegetation. Applying this method to fossil pollen data, we reconstructed site paleovegetation. Our montane bioclimatic model, MontBioCliM, was used inversely to convert site paleovegetation into site paleoclimates. The differences between site paleo and current climates served as past climate change scenarios. The climatic anomalies for 2020, 2050, and 2080 derived from HadCM3 A1FI and B1 of the Hadley Centre, UK, served as climate change scenarios in the 21st century. MontBioCliM was applied directly to all climate scenarios through the Holocene to map past and future mountain vegetation over the Altai–Sayan mountains. Our results suggest that the early Holocene ca 10 000 BP was cold and dry; the period between 8000 and 5300 BP was warm and moist; and the time slice ca 3200 BP was cooler and drier than the present. Using kappa statistics, we showed that the vegetation at 8000 BP and 5300 BP was similar, as was the vegetation at 10 000 BP and 3200 BP, while future vegetation was predicted to be dissimilar to any of the paleovegetation reconstructions. The mid-Holocene is frequently hypothesized to be an analog of future climate warming; however, being known as warm and moist in Siberia, the mid-Holocene climate would likely impact terrestrial ecosystems differently from the projected warm and dry mid-century climate.


Introduction
The vulnerability of mountain ecosystems to climate change makes them the best areas to identify climate change impacts on the biota. Along altitudinal gradients, a wide range of different vegetation structure and functions can be found, 3 Author to whom any correspondence should be addressed. and altitudinal gradients are in many ways complementary to latitudinal gradients (Koch et al 1995, Korner et al 1991. Altitudinal gradients facilitate the examination of relations between environmental variability and biotic patterns that can be used to design conceptual and analytical models to predict the distribution of the biota (IGBP Report 43).
Over mountains, due to the complex topography (combinations of elevations and slopes) and to the labor-intensive character of collecting and analyzing pollen data, they are usually not sufficient for detailed vegetation and climate reconstructions. Yet, pollen 'deposits' in mountain areas represent many different ecosystems. It is difficult, if not impossible, to interpret vegetation change from fossil pollen which originated from locations distant from sample sites. Instead, a reliable paleoecological interpretation should be based on pollen of local origin (Solomon and Silkworth 1986).
Due to the complex topography, many flora and fauna species are able to survive unfavorable times in some habitats called refugia. For instance, some species of tertiary flora (Tilia cordata, Azarum europium, Brunnera sibirica, etc) are still found over the Altai-Sayan Mountains, in lowland warm and moist Pinus sibirica-Abies sibirica-Populus tremula forests, called 'chern' (black) in Russian, which are rich in flora biodiversity and ferns in particular (Pologii and Krapivkina 1985). Thus, mountain areas may serve as a good model for reconstructions of both paleovegetation and climates. These reconstructions can be translated into detailed maps only by using simulation approaches.
This study's objectives were (1) to reconstruct paleoclimate and vegetation during the Holocene (from 10 000 before present, BP, to the present) across the Altai-Sayan mountains, embracing about one million square kilometers over both the Russian and foreign parts; (2) to predict a future vegetation pattern by the end of the 21st century; and (3) to compare all vegetation mapped for the past, present, and future in order to search for analogs between them.

Material and methods
The study area is the Altai-Sayan mountains (figure 1) located in Central Asia, mainly in Russia (the northern half) and Mongolia (the southern part), with a small area in Kazakhstan (in the west) and China (in the southwest). The elevation is generally about 1000-2200 m, with the highest point being Mount Belukha (4506 m) in the central Altai. The current climate is of a continental type with cold winters and warm summers. Westerlies are the dominant factor of the atmosphere circulation, resulting in high annual precipitation: up to 1500-2000 mm on the northwestern windward slopes and as little as 200-300 mm of precipitation on the foothills of leeward slopes and the inner intermountain depressions of Tuva and Mongolia. Most of the precipitation (up to 90% in Mongolia) falls in the summer.
In interior Central Asia, with a continental climate, ecosystems vary from steppes in lowlands, with warm and dry climates, to forests at middle elevations to tundra and nival communities in cold/wet highlands. The current vegetation in the Altai-Sayan mountains varies on the leeward and windward macroslopes. The sequence of vegetation types along a south to north direction are as follows. From desert and steppe in the dry climate of Mongolia, north to light-needled (pine and Siberian larch) taiga upslope; then to tundra in the highlands, changing to dark-needled (Siberian cedar and fir with some spruce) taiga down slope; then to lush dark-needled 'chern' (which means 'black' in Russian) forests (Siberian cedar and fir with aspen), productive and rich in flora and ferns with some nemoral (temperate) tall herbs in lowlands; followed by light-needled (pine) and birch subtaiga; and again steppe in the northern foothills (Smagin et al 1980).
In our paleo studies, to construct retrospective pollen diagrams, nine long sediment cores were extracted with a square-rod piston sampler (Wright 1991) from the deepest parts of small mountain lakes or with a peat corer from peat deposits down to the underlying rocks in the intermountain hollows across the study area. The upper-most sediments from lakes were taken with a transparent plastic tube fitted with a piston and subsampled in the field while vertical. The length of collected sediment cores ranged from 230 to 450 cm. Samples for pollen analyses were collected with intervals of 6-9 cm and prepared according to the method of Grichuk and Zaklinskaya (1948) using HCL, KOH, HF, and acetolysis. The pollen analysis was carried out under 400× magnification. Pollen types were identified with the help of reference books of Kupriyanova (1965), Kupriyanova andAleshina (1972, 1978), Bobrov et al (1983), Moore et al (1997). In each sample, 300-500 pollen grains of tree taxa plus other types of pollen and spores were counted. The percentage of all taxa was calculated as a sum of all pollen and spore taxa without aquatic and wetland plants and redeposited pollen. Pollen diagrams were constructed with the program TILIA (Grimm 1991). Radiocarbon ages were determined by accelerator mass spectrometry (AMS) at the University of Colorado (USA) and by a bulk method at the Geological Institute of Novosibirsk (Russia). Terrestrial macrofossils were used for radiocarbon dating (5-15 dates for a core) when applicable. If macrofossils were absent, then the humic acid fractions of gyttja were dated. The series of pollen diagrams used in this study were published by Blyakharchuk et al (2004Blyakharchuk et al ( , 2008. Our pollen spectra were examined for five time slices of the Holocene: 10 000, 8000, 5300 3200 years BP and Table 1. Surface pollen-based reconstruction of vegetation by the expert method (from plant indicator) and by the 'biomization' method (Prentice et al 1996) the present. These time slices characterize relatively steady periods in the paleovegetation development when it was in equilibrium with the climates of the time. For these time slices we took pollen spectra dated with the use of depth-age models on the base of neighboring radiocarbon dates from each of nine pollen diagrams. We used an uncalibrated radiocarbon age, so that our results could be easily compared with other pollen data for Siberia. Montane paleovegetation across the Altai-Sayans during the Holocene was reconstructed based on nine pollen diagrams using the qualitative method of 'the principle of actualism' and the quantitative method of Prentice et al (1996). Our paleo orobiomes were qualitatively reconstructed from fossil pollen spectra using relationships between indicator plants of modern vegetation types in the mountains of southern Siberia (Smagin et al 1980) and their surface pollen spectra. The qualitative (to large extent subjective, expert-constructed pollen-climate relationships) method, called in Russia 'the principle of actualism' (Markov and Velichko 1967), is similar to the 'analog' method developed and used by Davis (1963) and McAndrews (1966), quantified by numerous paleoegeologists (Prentice 1980, Overpack et al 1985, Giuot 1990) a couple of decades later.
In order to simulate paleovegetation from pollen data at the biome level, this study used the method of Prentice et al (1996), the 'biomization' of pollen data successfully applied in many studies (e.g. Tarasov et al (1998), Herzschuh et al (2004). Pollen taxa are assigned to plant functional types from which biomes are constructed. We omitted this step and assigned the dominant taxa directly to orobiomes. Russian vegetation classifications are based on vegetation types much like the broad classes Prentice et al (1992) defined: tree versus shrub, broadleaved versus needled leaved, evergreen versus deciduous, steppe forb, desert forb, grass, sedge, etc. Thus, Boreal zonal/altitudinal vegetation classes (biomes/subbiomes): tundra, forest tundra, light conifer and dark conifer taiga, broadleaved, mixed, forest steppe, steppe, and desert constructed from plant species indicators are similar to those of Prentice et al (1992): tundra, cold deciduous, taiga, cold mixed, cool conifer, steppe, and desert constructed from plant functional types.
As Prentice et al (1996) did, we designed a 'biome × taxon' matrix indicating which pollen taxa may occur in each biome: '1' was assigned to an orobiome if the taxon could occur and '0' was assigned if it could not. Our 'biome × taxon' matrix included 15 orobiomes and 39 dominant taxa identified in surface spectra. To evaluate the work of the matrix, it was applied to 22 pollen surface samples each of which was averaged from 2 to 5 samples located closely to each other, counting in total 132 samples. Modern orobiomes at these sites were identified in situ from plant indicators and were presumed not to be much disturbed over the mountains. An affinity score was calculated for all pollen samples according to the formula of Prentice et al (1996), and each pollen sample was 'biomized' according to its maximal affinity, determined by an affinity score. If several biomes had the same score then the '1' was assigned to a sample which had a smaller number of taxa (table 1).
Two methods, qualitative and quantitative, of the reconstruction of orobiomes were compared. From 22 pollenreconstructed orobiomes, 73% agreed with actual vegetation, 22% fell into a close orobiome, e.g. dark-needled taiga might be 'biomized' as subalpine dark taiga, light-needled taiga might be 'biomized' as subtaiga, chern taiga might be 'biomized' as subtaiga/forest steppe, because geographically these pairs of orobiomes are located next to each other and pollen can easily intermix (table 1).
Using the constructed 'biome × taxon' matrix, each paleo orobiome was then simulated from a pollen spectrum for 3200, 5300, 8000, and 10 000 BP in nine sites across the Russian part of the Altai-Sayans. Simulated reconstructions corresponded well to our expert reconstructions. Only those vegetation reconstructions that coincided were then used in our paleoclimate reconstructions.
A mountain vegetation bioclimatic model, MontBioCliM, was developed to quantify climate-vegetation relationships. MontBioCliM is a set of climate envelopes limiting each altitudinal vegetation belt (an orobiome, Walter 1985). Three climatic indices were employed to limit an orobiome: growing degree days above 5 • C, GDD 5 characterizing plant requirements for warmth; negative degree days below 0 • C, DD 0 , characterizing plant resistance to cold; and annual moisture index, a ratio between growing degree days and annual precipitation, AMI, characterizing plant tolerance to water stress. Our choice to use a ratio between heat and water instead of a direct measure of available water (precipitation) was because borders between vegetation zones were shown to correspond well to specific values of ratios between heat and available water (Dokuchaev 1948). Since then, various different expressions of those ratios have been developed and used: simple GDD 5 to annual or summer precipitation (Selianinov 1930, Rehfeldt et al 1999; actual to potential evaporation (Prentice et al 1992); potential evaporation to annual precipitation (Budyko 1974; deficit to potential evaporation (Thornthwaite 1948); actual evaporation and deficit (Stephenson 1998); and the difference between annual precipitation and potential evapotranspiration (Hogg 1997). However, both potential and actual evaporation are not usually recorded in most weather stations. Instead, they are calculated from temperature, precipitation, air humidity, sunshine or cloudiness, albedo, etc.
These empirical relationships are regional and work rather well over plains, although they are not without problems over mountains (Budyko 1974). We avoided this problem by only using variables that are direct functions of observed temperature and precipitation, thus arriving at a straightforward annual moisture index.
Climate limits of each orobiome (table 2) were derived from the ordinations of approximately one hundred Altai-Sayan weather stations (with assigned orobiomes) in two pairs of climatic indices: GDD 5 -AMI and GDD 5 -DD 0 .
A validation of MontBioCliM was done by comparing predicted vegetation with mapped actual vegetation (Samoylova 2001) using kappa statistics (Monserud 1990, Monserud and. The overall kappa was 0.4 and varied between 0.2 (e.g. light conifer, subalpine, steppe) and 0.6 (e.g. tundra, 'chern', semidesert) for individual orobiomes, which showed a fair match of the model with actual vegetation (figure 2). The worst match was found for the Mongolian part of the mountains, poorly represented by climatic data and thus poorly interpolated climatic surfaces. Additional disagreement was caused by the fact that the model was designed based on the Russian part of the mountains, and was also applied to Mongolia, representing the other half of the mountains.
MontBioCliM was inversely used to predict average climatic indices (growing degree days above 5 • C and an Vegetation key. BOREAL:1-tundra, 2-subalpine dark conifer and meadow, 3-subalpine light conifer, 4-montane dark conifer, 5-montane light conifer, 6-subtaiga and forest steppe, 7-dark-conifer 'chern', 8-steppe, 9-dry steppe, 10-semidesert/desert, 11-cryosteppe; TEMPERATE: 12-temperate mixed conifer broadleaved;, 13-forest steppe, 14-steppe. annual moisture index) for each simulated paleo orobiome in each time slice of the Holocene as Monserud et al (1998) did. We converted these climatic indices correspondingly to average July temperatures and precipitation. The July temperature was derived from growing degree days, base 5 • C, which are strongly correlated (R 2 = 0.9). Annual precipitation was derived from the ratio between the growing degree days and the annual moisture index. Climate change (anomalies) in each Holocene time slice was evaluated as the differences between the July temperature and annual precipitation between contemporary and paleoclimates (3200, 5300, 8000, and 10 000 BP).
On average over the mountains, July temperature anomalies with respect to the contemporary climate were negative −2 to −3 • C 3200 BP and especially greatly negative −2 to −5 • C 10 000 BP. July anomalies were positive (2-4 • C) both 8000 and 5300 BP. For comparison, in the Minusink depression located in the north of our study area, (Koshkarova 2004) reconstructed the July temperature as being 3 • C higher in 7000-5800 BP than today and 1 • C lower in 3000-2400 BP than today based on the macrofossil analysis. Zubareva (1987) used palynology to reconstruct July temperature anomalies at approximately −2 • C about 3000 BP.
Next, our differences were scaled down to 1 km grid cells using the Surfer software (www.ssg-surfer.com) and were added to the current July temperature and annual precipitation surfaces which were mapped using Hutchinson's (2000) thin plate smoothing splines on our base map at a resolution of 1 km. Thus, detailed paleo July temperature and annual precipitation maps were obtained based on which paleoclimatic indices. Coupling paleoclimatic indices were calculated with MontBioCliM, we reconstructed and mapped vegetation over the Altai-Sayan mountains for the major Holocene time slices: the present, SubBoreal, Late Atlantic, the end of Boreal, and PreBoreal.
MontBioCliM was also used to predict future mountain vegetation in the 21st century. The future bioclimatic indices were calculated using climatic anomalies for 2020, 2050, and 2080, derived from two climate change scenarios, the HadCM3 A1FI and B1, of the Hadley Centre in the UK based on the SRES Emission Scenarios (IPCC 2000). These scenarios reflect opposite ends of the SRES range, the largest temperature increase from the A1FI scenario and the smallest temperature increase from the B1 scenario.
Similarities/dissimilarities in mountain vegetation during the Holocene starting from 10 000 BP through the present to 2080 were found using kappa statistics (Monserud 1990, Monserud and in order to identify possible analogs in vegetation which are commonly used in paleoecological reconstructions (Budyko 1986, Overpack et al 1985, Giuot 1990, Jackson and Williams 2004. The qualitative descriptors of Monserud and Leemans (1992) were used to evaluate the agreement between paired maps: 'excellent' and 'very good' for a kappa > 0.7, 'good' for a kappa 0.55-0.7, 'fair' for a kappa 0.4-0.55, and poor for a kappa < 0.4.

Results
Our pollen-based expert reconstructions of vegetation patterns during the Holocene across the Russian part (where pollen samples were collected) in the Altai-Sayan mountains included the following: before 10 000 BP-most of the Altai-Sayan mountains (see figure 1 for geography) were covered by treeless steppe and tundra vegetation. Only in the Kuznetzky Alatau mountains were there islands of spruce and larch forests spread along river valleys and around lakes.
Intermountain depressions were covered by cold steppe with islands of spruce forests. In central Altai and south-western Tuva, vast areas were occupied by grass Artemisia steppe with subalpine shrubs upslope. Small islands of forests started to spread. Steppes in the southern part in the mountains were drier compared to those in the northern part.
8000 BP-The Kuznetzky Alatau was covered by thick dark conifer forests with Abies sibirica and with tall herbs and ferns in the ground layer. The lower Batenev Ridge of the Kuznetzky Alatau was covered by forests with Pinus sibirica and Picea obovata which stretched to lower elevations along river valleys. In the drier central Altai, larch forests with an admixture of Pinus sibirica and Abies sibirica covered a large area. On the western ridges of Altai, the portion of Pinus sibirica increased in the forests. At the upper elevations, mountain tundra and meadows developed. Lower than 1800 m, real steppe occurred. Forests of Larix sibirica mixed with Picea obovata were spread further to the dry south-eastern Altai and south-western Tuva. In those forests Pinus sibirica was also found. 5300 BP-Vegetation of the Kuznetzky Alatau did not change much compared to 8000 BP, but in the forests of the Batenev Ridge, Picea obovata disappeared. Also from the forests of the central Altai, Abies sibirica disappeared. The steppe vegetation of intermountain depressions became more xerophytic with a greater role played by Artemisia.
3200 BP-Dark conifer 'chern' forests retreated from central areas of Kuznetzky Alatau to its western macroslope. In the central Altai, vegetation cover was similar to that which occurred in 5300 BP. In the lower mountains, adjacent to Altai and West Sayan, the role of Betula pendula increased.
The above picture of vegetation change across the Altai-Sayan mountains reflects a relatively rapid change of climate from dry and, possibly, cold in the early Holocene to wet and warm from 8000 BP to 5300 BP. Then, a gradual change to a more continental and less humid climate took place in the second half of the Holocene. A more pronounced climate and vegetation change occurred in the leeward south-eastern and eastern parts over the Altai-Sayan mountains and a less pronounced change on windward western macroslopes. Such changes can be explained by the weakened Atlantic cyclone activities in the SubBoreal (Blyakharchuk et al 2004).
At 10 000 BP and 3200 BP, the climate was both cooler and drier, especially in the early Holocene, than the present, favoring cold tundra and subalpine (open light conifer taiga) to expand in highlands and semidesert to expand in lowlands (table 3). Kappa statistics show (table 4) that the vegetation distribution 10 000 BP and 3200 BP was similar (κ = 0.62).
The 8000 BP and 5300 BP climates were predicted to be warmer and wetter, respectively to the present, that were favorable for dark conifer taiga including 'chern' taiga and subtaiga (table 4). At these times, cold orobiomes like tundra disappeared, and subalpine open forests substantially decreased. Boreal steppe shrank; wetter climates were more suitable for temperate broadleaved forests, forest steppe and steppe. In total, Boreal and temperate steppes dominated a quarter of the mountain area in both the mid-Holocene and present. Holocene vegetation at 8000 BP and 5300 BP was predicted to be very similar (κ = 0.80). Khotinsky (1977) interpreted the maximal occurrence of spruce in the south of West Siberia during the Boreal and Late Atlantic with a warm period in the Holocene. He associated 4500 BP with the start of cooling. Khotinsky (1977), in the south of West Siberia, and Tarasov et al (1998), in western Mongolia, both believed that the boundary between forest and steppe did not shift substantially in the mid-Holocene; thus they concluded that in the south, the past climate was as dry as it is today. However, a vast current area of steppe, rather than contemporary semidesert/desert, is mapped in the mid-Holocene in the south of Mongolia that suggests a wetter climate of the mid-Holocene (figure 3).
The climate in the 21st century, as predicted from GCMs, will be warm and dry, which will cause a decrease in forests, a disappearance of tundra, and a considerable increase of grasslands (forest steppe, steppe and semidesert, table 3). Habitats for temperate vegetation would expand from none, in the current climate, to 15% by 2080, which is comparable to its area in the mid-Holocene (table 3). Current refugia for broadleaved forests, like the 5000 ha Tilia 'island' in the foothills of Kuznetzky Alatau (Polikarpov et al 1986), may expand somewhat by the end of the century, but will likely not be as large as in the Late Atlantic because future climates are predicted to be drier than the past climates reconstructed for the mid-Holocene. Both B1 and A1FI climate change scenarios yield similar results regarding the

Discussion
Mountains are useful areas of study for monitoring and modeling vegetation changes in both past and future climates because diverse ecosystems are characteristic of a rather small area. In mountains, mapping environments and related vegetation patterns is a difficult exercise due to complex topography and a lack of proper data. The task becomes more difficult for the past because acquiring paleodata necessitates even more time and labor. Paleoenvironmental studies must therefore rely on comprehensive modeling approaches based on sufficient and available data.
The method of modern analogs and a similar method, 'the principle of actualism' (as described in Russia), which have been employed in studying quaternary vegetation and climates, follow the logic that modern and fossil pollen or macrofossil assemblages match and the relationships between modern vegetation and related pollen assemblages and paleovegetation and fossil pollen assemblages match as well. How well they match, paleoecologists judged subjectively in earlier studies (Davis 1963, McAndrews 1966, Savina and Khotinsky 1984 or, in studies following a couple decades later, assessed objectively using numerical tools (Andrews et al 1980, Bartlein et al 1984, Giuot 1990, Solomon and Bartlein 1992. It is known in paleoecology that fossil pollen assemblages may lack modern analogs, e.g. for the late glacial period in eastern North America and other regions (Jackson and Williams 2004) or for the early Holocene (Savina and Khotinsky 1984) in the former Soviet Union. However, since the Boreal period, and completely since the Atlantic when the vegetation zones became established in Siberia (Khotinsky 1977), past pollen and vegetation assemblages have modern analogs. This fact allowed us to apply this method in the present study.
Pollen-based studies of reconstructed vegetation and climates in Central Asia during the Holocene are still limited. Our pollen-reconstructed paleovegetation and paleoclimates of the Holocene in general agreed with previous findings of other paleoecologists for the Altai-Sayan mountains, although our studies revealed some contradictory conclusions as well (table 5). Our reconstructions of paleovegetation and climates of all time slices (10 000, 8000, 5300, and 3200 BP) corresponded well with those of Blyakharchuk et al (2004Blyakharchuk et al ( , 2007. We reconstructed the 5300 BP climate to be wetter and warmer, as other authors did within the study area at 5000-6000 BP (Wu and Lin 1988), at 5400 BP (Herzschuh et al 2004), at 5300 BP (Yamskikh et al 1981), and at 5200 BP (Savina and Koshkarova 1981). However, Grunert et al (2000) found from archeological evidence that the climate about 5000 BP near Lake Uvs Nuur, Mongolia, was drier (table 3). The 3200 BP climate we reconstructed as cooler and drier corresponded well to the reconstructions of Herzschuh et al (2004) at 3100 BP and Savina and Koshkarova (1981) at 3200 BP, although, Wu and Lin (1988) reconstructed the 3000 BP climate as cooler and wetter.
The Holocene vegetation distribution across the Altai-Sayan region was different from the current vegetation distribution: kappa statistics varied between 0 and up to 0.27. The vegetation distribution predicted for the 21st century is also dissimilar to that under the contemporary climate, resulting in kappa statistics less than 0.4. Just as Monserud  (1993) concluded that the vegetation of the mid-Holocene and the future vegetation in Siberia would be dissimilar, no analogous patterns in the vegetation distribution were found over the Altai-Sayan mountains in the past (the Holocene) and the future (the 21st century), because all kappa statistics paired between the Holocene time slices and 2020, 2050 and 2080 are less than 0.4, which means a 'very poor' agreement. Note, that both the kappa statistics (table 4) being around null or negative, and the vegetation maps ( figure 4) show 'no agreement' between the mid-Holocene and 2050. This result may shed doubt on the assumption that the mid-Holocene may serve as an analog of warming by the current mid-century (Borzenkova and Zubakov 1984, Budyko 1986, Avenarius et al 1987. We may expect no analogous vegetation between a warmed climate in the near future and that which occurred in the past. In the past, with slow climate change, the time interval for the vegetation change could have been as long as hundreds to thousands of years. In the 21st century, with rapid climate change caused by unprecedented rates of increasing CO 2 concentration, the time interval might be as little as decades, as predicted from global circulation models (GCM) runs. Recent experiments suggest that high CO 2 levels can spur faster photosynthesis and growth (Tangley 2001), increase the total net primary production (DeLucia et al 1999), increase the fecundity of forest trees, and change dispersal and recruitment patterns (LaDeau and Clark 2001) which may facilitate quick adaptation to a changing environment, although GCM runs also suggest unprecedented rates of warming and decreased moisture in the near future greenhouse world with doubled or tripled CO 2 concentrations. Global vegetation model runs suggest significant shifts in vegetation distribution (Smith and Shugart 1993, Cramer et al 2001, Brovkin et al 2006, Tchebakova et al 2009 with chances that some plant assemblages and species will become extinct in a rapidly changing climate. Still, plants may adapt in this changing world by migration and adaptation. Although migration rates of Boreal tree species, as estimated from paleoecological evidence, were only as 300-500 m yr −1 (King and Herstrom 1997), species with broad climatic niches could adjust to a rapidly warming climate while species with a restricted range of suitable habitats and limited dispersal are likely to disappear first (Solomon and Leemans 1990). Man's role in transporting seed throughout the world using contemporary technical means substantially increase migration rates. Environmental changes can also alter the behavior of long-distance dispersal: warming can promote wind-driven movements of plant genotypes and populations in Boreal forests (Kuparinen et al 2004).While extirpation and immigration are the main processes at the margins of the forest distribution, within the forest zone, natural selection and gene flow are primary processes favoring tree adaptation to climate change (Davis andShaw 2001, Rehfeldt et al 2004). Evolutionary processes of adjusting to predicted climate change would take a time at least one order of magnitude greater than the time it takes for the climate itself to change. Estimates for Pinus sylvestris in Siberia suggest that 5-10 generations (about 150 years) are required for the evolutionary process to follow a predicted warming. Genotypes would be reorganized within tree distributions, and tree boundaries would follow a changing climate. The forest adjustment to climate change would occur, but it would require a long time due to the large amount of change predicted by the end of the current century (Rehfeldt et al 2004).
Our predictions of climate and past vegetation are subject to many potential sources of error: sparse fossil pollen sites and their complete absence in some parts of the study area such as eastern Tuva and Mongolia, which therefore necessitates data extrapolation; sparse climate data (especially in Mongolia) and related poor interpolation of climate surfaces; erroneous or inadequate pollen analysis (including radiocarbon sample dating) and related 'biomization' of the pollen; and inaccurate climatic limits of orobiomes in the bioclimatic model and a related degree of disagreement between modeled and actual vegetation. Nevertheless, for the study area, the main conclusions regarding pollen-reconstructed vegetation and the climate of the major time slices during the Holocene showed agreement with many previously held interpretations by other paleoecologists.