The present studies focus on three key time phases in the mid-Holocene (6ka B.P.), the Last Glacial Maximum (21ka B.P.) and the mega-interstadial of last glacial period (30-40 ka B.P.), in which the climate conditions were remarkably different from the present. Lots of paleoenvironmental studies have suggested that there have been dramatic changes in the strength and extent of the Asian monsoon in response to changes not only in insolation and glaciation, but also large feedback from changes in land surface boundary conditions. According to geological data in pollen and lake records in this study, reconstructions of climate during the three can provide benchmarks for identification of the climate signals in response to changes in insolation, glacial and land surface boundary conditions for evolution of Asian palaeomonsoons, and for the validation of climate model simulations to understand the mechanisms of the climate changes.

6ka: Both the temperature field from available palaeo-GCMs (Joussaume et al., 1998) and the forest vegetation patterns from BOME model (Prentice et al., 1998) are consistent with a theory that a lower-than present winter insolation anomaly in the early and mid-Holocene (Berger, 1988) would produce a lower-than present winter temperature. Winter temperature (DJF) anomaly between 6ka and 0ka simulated by Palaeo-climate models of UGAPM (Dong et al., 1996) and CCM (Kutzback et al., 1998) shows that, 6ka winter temperature in China were 2°C colder in UGAMP model and 2-4°C in CCM model than the present. The palaeoclimate-driven biome models show the mid-Holocene forest vegetation zones was shifted southwards in China (Harrison et al., 1998; Kutzback et al., 1998). However, these are neither consistent with geological interpretations of vegetation changes (Yu et al., 1998) nor with reconstruction of winter temperature during the mid-Holocene that reconstructed a high-than-present winter temperature (Yu and Qin, 1997). These differences in changeable directions at least raise the questions: Are there regional differences in the Asian monsoon regions? What are climate mechanics behind the vegetation patterns?

Vegetation feedback in high latitudes through decrease in albedo of land surface simulated by climate-biome models (Foley et al., 1994; TEMPO, 1996) and through increase of CO2 in coupled ocean-atmosphere model (Manabe and Stouffer, 1994). These models have provided prediction at 6ka when there was a warmer-than present high latitude land in Eurasia and north America, which changeable amplitudes in temperature are much higher than changes directly driven from orbital-induced insolation. These Changes in land surface in high latitudes may have offset the direct affects of the orbital-forcing on winter temperature in the mid-low latitudes of Asian monsoon regions (e.g. Wang, 1999).

The winter monsoon patterns in the mid-Holocene from geological evidence are consistent with this interpretation. Changes in the land surface such as disappearance of icesheet in northern Europe (Bjorck, 1995), reduction of the permafrost areas in Siberia and north China (Zhou et al., 1991), and increases in taiga vegetation covering in high latitudes (Texier et al., 1997), would decrease the cold high pressure in northern Asian continent in the mid-Holocene. This can lead to a weaker winter monsoon than the present, consequently an occurrence of warm winter with a less-decreased temperature. This hypothesis is waiting for confirmation by palaeoclimate and biome models. Thus the land surface feedback has paid important role in climate changes in Asian monsoon regions.

21ka: Palaeoclimate models simulated more positive mean annual P-E in western North America and in the Mediterranean region (e.g. Dong et al., 1996; Kutzbach et al, 1998) which reflect the dominance of glacial anticyclonic conditions and a southward shift of the westerlies. Otherwise, P-E was generally less than today across most continents of the world from the tropical to the high latitudes. Indian monsoon and Pacific monsoon were weakened in the east and the south Asia, leading to less precipitation over the regions.

Lakes from eastern and southern China are consistent with the modelling results. However, very wet conditions from Tibet and inland Xinjiang (Yu et al., 2000) can not find any clues consistent with these modelling experiments. To argue the positive water balance of high lake level is resulted from more rainfalls and less evaporation, we compared with recent research products on glacier changes from Tibet. There occurred synchronous changes in water balance by lake and ice balance by glaciers in LGM Tibet and Xinjiang. When there were presence of high lake level/ large lake areas, glaciers in Tibet and Xinjiang were increased in their maximum volumes (Li et al., 1998). This implies that the high lake level in the region was not due to mountainous glaciers melt water. Pollen-based reconstruction show the very cold conditions at LGM can not provide evidence there was enough good thermal condition to melt large ice or snow at LGM.

Thus we provide interpretations for the LGM wet conditions in western China:
(1) Quaternary ice sheets of the Northern Hemisphere developed its maximum extend and consequently existence of persisting strong glacial anticyclone, leading to the southward displacement of the westerly.
(2) The presence of ice sheets in high northern latitudes in the western Eurasian, and the presence of much large areas of permafrost than today from Russian Far East to northern China, would favor a stronger cold high pressure in northern Asian continent, leading to a stronger winter monsoon. This could barred the westerly that was difficult to penetrated into higher latitudes than the displacement of the present.
(3) Much lower sea level than today in LGM and large land areas exposed in modern continental shelf in the East Asia, eastern, produced a weakened land-sea contrast and reduced summer monsoons of the South Asia and East Asia. This could be resulted in an around-year dominance of the westerly in the mid-latitudes across Eurasia continents. These suggest the wetter conditions over Tibet and Xinjiang linked with southern Europe, Near East and Central Asia was major resulted from increased precipitation by moisture transportation from dominance of the westerly. Continental cooling during the LGM produced evaporation conditions much lower than today. The significant decrease in evaporation may help to explain the wetter conditions over the Tibetan Plateau and Xinjiang inland.

30-40 ka: The period of 30-40 ka B.P. corresponds to the later phase of the marine oxygen isotope of Stage 3. According to reconstruction from the deep sea cores (mbrie et al., 1984), Antarctic Vostok ice core (Lorius et al., 1991) and Greenland GRP ice core (Fabre et al., 1995), the global temperature was generally lower at the Stage 3 than the last interglacial period (Stage 5) and also lower than the post-glacial Holocene (Stage 1), but slightly higher than the early last glacial (Stage 4) and late last glacial (Stage 2). However, our recent palaeoclimate studies showed that Tibetan Plateau climate during the period of 30-40 ka B.P. was exceptionally warm and humid, which the increased temperature were roughly equivalent to the last interglacial of Stage 5.

The lake-level and pollen records consistently indicated that during that period the climate of western China was exceptionally warm and humid that had reached the interglacial climate regime. It was estimated that the annual mean temperature at that time was 2-4°C higher than the present by isotope?18O variations in Guliya ice cap (Yao et al., 1994; 1997). There occurred numerous large freshwater lakes and their water surfaces were 30-200 m above present level. There were also large freshwater lakes occurred in the today's desert regions, such as Qaidam Basin, Tengger Desert and Badain Jaren Desert. The conifer forest in the Tibet extended ca. 400-800 km beyond their present western limits. The precipitation was estimated ca 400 mm more than today.

These evidences strongly infer a very strong summer monsoon event in the late Stage 3. We provided a hypothesis to interpret the changes in climate during the 30-40 ka. As the precessional anomaly was much stronger in the low latitudes than that in high-latitude region during the late Stage 3, the Tibetan Plateau heat low would be formed by the strong heating effect of the Plateau not only induced the ndian southwest monsoon to the inner Plateau, but also induced the extension of southeast airflow from the Pacific Ocean towards the northwest inland, thus leading to a relatively wet climate in the arid region. Meanwhile, the evaporation of the tropic and subtropical ndian Ocean surface was strengthened due to increased heat input, resulting in enhanced the cross-equatorial airflow moving towards the Northern Hemisphere and much intensified ndian southwest monsoon, which carried large amount of moisture to ndia and Tibet Plateau.

We are developing the palaeoclimate modelling to test these hypotheses.