Introduction

Reliable meteorological and climatological information for climate reconstruction is scarce or absent for many parts of the world, especially on much of the Qinghai-Tibetan Plateau and surrounding mountains (Domrös and Peng, Reference Domrös and Peng1988). The high-altitude ice caps on this plateau offer the potential for recovery of a well-dated archive of atmospheric history from the subtropics on glacial time-scales. Although cores have been retrieved successfully from the polar regions over the last three decades the recovery of ice cores from high-altitude sites is a rather recent development (Reference Wagenbach, Oeschger and LangwayWagenbach, 1989). The recovery of ice-core climatic records from the Tibetan Qinghai Plateau will be essential to understand the global climate system, particularly in view of the impact of this region on northern hemisphere meteorology and climate.

The Dunde ice cap (38°06′N, 96°24′E) is located in a desert environment between the highest Chinese desert, the Qaidam Basin, to the south, and the Gobi Desert to the north (Fig. 1a). A combination of glaciological and meteorological factors: glacier thickness 140 m, flat bedrock topography, low net accumulation (0.4 m of water equivalent per year) and low mean annual temperature of −7.3°C, makes the ice cap well suited to the recovery of long-term, high temporal-resolution ice-core records. Co-operative glaciological and climatological research programs were conducted on the Dunde ice cap in 1984, 1986, and 1987 by the Byrd Polar Research Center (BPRC) and the Lanzhou Institute of Glaciology and Geocryology (LIGG). In 1987, three ice cores (Figs 1b and c) were recovered to bedrock from the ice-cap summit (5325 m a.s.l.). Core D-1 (139.8 m long) was divided along the entire length in the field and shared equally between LIGG and BPRC. Core D-2 (136.6 m long) was returned frozen to the LIGG and core D-3 (138.4 m long) was returned to BPRC (Reference Thompson and Mosley-ThompsonThompson and others, 1988a). The upper 56 m of core D-3 were melted and bottled while the lower 83 m were returned frozen to BPRC.

Ice-Core Analyses

The visible stratigraphy of each core was photographically recorded immediately after drilling. Core D-1 was then cut into 3585 samples which were melted in closed containers by passive solar heating within a laboratory tent. The samples were then poured into polyethylene bottles, sealed in wax, and shipped to BPRC. Samples were divided so that microparticle concentrations (MPC), oxygen isotopie ratios (δ¹⁸O) and liquid electrical conductivity (EC) could be measured on the same sample. The concentrations of three inorganic anions and pH were measured for the entire length of D-3 while crystal sizes were measured for selected sections.

Core D-l was analyzed along the entire length for MPC, δ¹⁸O and EC. δ¹⁸O was measured at the University of Copenhagen and all other measurements were made at BPRC under Class 100 Clean Room conditions. The concentrations of particles with diameters ≥2μm per milliliter sample (MPC) were measured using a model TA-II Coulter Counter (Reference ThompsonThompson, 1977), fitted with a 100 μm aperture tube. The very high concentrations of dust necessitated diluting most of the samples 36 times, and up to 7000 times in extreme cases, to avoid problems with coincident passage through the aperture tube. Only samples for dust measurements were diluted. EC of liquid samples was measured with a Beekman model RC-16C conductivity bridge; pH was measured with a Beekman model Φ 40 pH meter.

The concentrations of seven major soluble ions, chloride (Cl⁻), sulfate

, nitrate (NO⁻ ₃), sodium (Na⁺), potassium (K⁺), magnesium (Mg²⁺) and calcium (Ca²⁺), were determined with a Dionex Model 2010i chromatograph, equipped with AS4A (for anions) and FAST SEP CATION I and II (for cations) columns. All ion chromatographic samples were passed through 0.2 μm mesh filters prior to analysis. The cations were determined in selected sections, while the anions were measured for the entire length of core D-3 (c. 1200 samples of varying length from 40 to 5 cm from the top to the bottom of the core). Some minor ions, namely ammonium, acetate, formate, oxalate and phosphate, were also detected in ion chromatographic measurements.

Fig. 1. Fig. 1(A). Location of the Dunde ice cap, Qaidam Basin, and Loess Plateau; (B) topography of the ice cap, position of the survey network, firn and equilibrium lines, and sites where ice thickness was determined; (C) location of drill sites (D1, D2, and D3), detailed surface topography, inferred flow lines and ice velocities.

Samples of 200–250 g were cut from both the D-1 and D-3 cores and pumped through ion-exchange filters to measure total β radioactivity in a Tennelec LB 1000 Series α/β Counting System. Total β radioactivity is routinely used to identify time-stratigraphic horizons associated with known atmospheric thermonuclear tests (Reference Crozaz, Langway and PicciottoCrozaz and others, 1966). When isolated in an ice core, these horizons (Reference Picciotto and WilgainPicciotto and Wilgain, 1963) allow verification and calibration of seasonally varying parameters which are used to date the core.

Fig. 2. Summit area of the Dunde ice cap showing (A) circular and strain network, (B) a contour map of surface topography, (C) a contour map of bedrock topography, (D) an isopach map of the ice thickness, and (E) typical echograms obtained from the summit area of the Dunde ice cap showing the complete wave train including the surface wave (A) and the reflected wave (B).

Geophysical Studies

A short-pulse ice-radar sounder was used on the Dunde ice cap to determine ice thicknesses. The system used has been described elsewhere (Reference Jezek and ThompsonJezek and Thompson, 1982; Reference ThompsonThompson, 1989). Examples of radar data collected from the summit of the Dunde ice cap are illustrated in Figure 2. The complete wave train includes the surface wave (A) which travels along the upper surface of the air–snow interface, arrivals scattered from features within the ice, and a strong echo (B) which is interpreted as the reflection of the ice–rock boundary. Very few of the waves interpreted as reflections from the ice–rock interface appear to be distorted, and there are no strong arrivals later than the bedrock reflections, suggesting that the bedrock surface is relatively smooth. The ice-cap summit topography was extensively surveyed and radar soundings were made at 42 sites (Reference ThompsonThompson, 1989). At the drill sites for cores D-l and D-3 the measured depths between the surface and the bedrock were 139.6 and 138.4 m, respectively, while the average short-pulse ice-radar thicknesses were 142.4 and 140.3 m, respectively, providing a radar accuracy within 2%. The radar data and the surface survey data were used to produce a contour map of the surface topography, a contour map of bedrock topography, and an isopach map of the ice thickness (Figs. 2b, c, d). The three deep core sites (Fig. 2b) are located near the summit dome of the ice cap and the bedrock surface is gently sloping to the southeast (Fig. 2c).

Temperature profiles were measured at 5 and 10 m intervals in each of the three bore holes using a 150 m thermistor cable designed with two thermistors at each point to allow cross-checks of temperature measurements at each depth. The resulting core 1 bore-hole temperature profile is presented in Figure 3a. The 10 m surface and basal bore-hole temperatures were −7.3 and −4.7°C respectively, indicating that the Dunde ice cap is “polar.”

Late Glacial Stage Ice

Late in the 1986 field season ice merging along the margin of the Dunde ice cap contained an abrupt contact between overlying clean ice and underlying dirty ice. MPC and δ^l8O analyses of two shallow cores drilled through this transition indicated that the underlying ice is probably of Wisconsin/Würm Late Glacial Stage (LGS) age (Reference Thompson and Mosley-ThompsonThompson and others, 1988a). The results from core D-1 drilled at the summit confirm that a strong temporal relationship exists between the large increase in dust below 129.2 m and δ¹⁸O depletion (Fig. 4) in the ice. The LGS ice contains major dust events where concentrations are 4–8 times average Holocene concentrations. Concurrently, mean δ¹⁸O decreases by 2‰. Just as observed in Greenland cores from Camp Century (Reference ThompsonThompson, 1977) and Dye 3 (Reference Hammer, Clausen, Dansgaard, Neftel, Kristinsdottir and JohnsonHammer and others, 1985), the transition from high to low-dust concentrations is very abrupt (Fig. 4b) across the LGS/Holocene boundary. In Dunde ice cores this transition in the dust concentration is contained in a 30 cm section of ice representing approximately 40 years.

Fig. 3(A). Temperature profiles measured at 5 m intervals in the core 1 bore hole on the summit of the Dunde ice cap in 1987. The temperature profile indicates that the ice cap is “polar” throughout with a bottom temperature of −4.7°C. (B). 30 m density profile for Dunde ice cap. Measured values are indicated by dots, while the smooth curve is a third-order best fit to the measured values. The depth (X)-density (ρ) relationship is expressed as ρ = A + BX + CX ² + DX ³ where A = 473.558, B = 55.233, C = −2.418, and D = 0.0346. For this fit r ² = 0.557 which is statistically significant at the 99.9% confidence level.

Figure 4b illustrates MPC and δ¹⁸O in the lower 20 m of D-1 and similarly Cl⁻, NO⁻ ₃ and δ¹⁸O in D-3. The more negative δ¹⁸O ratios in the lower 10 m support the presence of LGS ice. However, unlike the rapid transition in MPC and anion concentrations, the ¹⁸O enrichment in precipitation appears more gradual suggesting that Holocene conditions were attained over several centuries.

Chloride, nitrate and sulfate concentrations in core D-3 (Fig. 4) show large variations with depth, as well as infrequent, but extremely high concentrations. There are significantly positive correlations between concentrations of chloride, sulfate, sodium, calcium and magnesium. Throughout D-3 alkaline cations (Ca²⁺, Mg²⁺ and Na⁺) and Cl⁻, δ¹⁸O and HCO⁻ ₃ (estimated with pH) constitute the majority of the dissolved species, indicating that the ice core is highly alkaline.

Throughout core D-3 the pH of melt-water samples remains essentially neutral (average pH = 7.3) with occasional values ranging from 5.7 to 9.2. For the few sections analyzed, cation and anion concentrations vary proportionally. For example, Na⁺ and Cl⁻ concentrations maintain a molarity ratio near unity for all sample sections for which both cations and anions were measured, indicating that these two ions share a common dominant source.

The most distinctive feature of the anion profiles is the marked concentration decrease in the lower 12 m of the core (Fig. 4), where average concentrations of Cl⁻, NO⁻ ₃ andδ¹⁸O are 30, 50, and 32% lower than their respective averages for the entire core. The evidence of co-variation between major anions and major cations suggests that cation concentrations are also lower in the bottom 12 m of the ice core. Indeed, one-meter averages of Ca²⁺ concentration decrease from 1.6 ppm at mid-core (64 m) to 1.3 ppm in LGS ice (130 m). A corresponding decrease in EC below 129 m was also observed in core D-1 (not shown). This decrease of the ionic concentrations in core D-3 is temporally correlative with the increase in MPC below 129m of core D-1, with a similarly abrupt transition.

Fig. 4(A). Averages over 1 m depth intervals (roughly 25 samples per section in core D-1) for total particles (diameters ≥2.0 μm) per ml sample, δ¹⁸O, Cl⁻, NO⁻ ₃, and δ¹⁸O. The time-scale is indicated at 25 m intervals. (B). Averages every 0.5m interval (roughly 17 samples per section in core D-1) for the lower 20 m of the cores illustrate the rapid transition in the concentrations of dust and dissolved species and the more gradual δ¹⁸O transition.

No new ions were found below the LGS/Holocene transition. This suggests that the dominant sources of the soluble aerosol were essentially the same during both the LGS and Holocene. Later discussion demonstrates that the soluble aerosol is supplied by the deserts and surface mineral deposits surrounding the ice cap, and the higher concentrations in the Holocene are the result of the increased strength of the predominant local sources, due to the progressively drier climatic conditions. This is different from the polar areas, where LGS concentrations of soluble species (Cl⁻, δ¹⁸O) are generally higher than those in the Holocene (Reference Herron, Langway, Langway, Oeschger and DansgaardHerron and Langway, 1985; Reference Legrand and DelmasLegrand and Delmas, 1988). The decrease of soluble aerosol from LGS to Holocene in the polar cores is attributed to a decrease of the meridional atmospheric circulation and increased precipitation, while reduced levels of insoluble dust are attributed to a decrease of both atmospheric circulation and continental aridity.

Ice-crystal sizes were measured for selected sections of D-3 in ice below 50 m (Fig. 5) to explore the possibility of a relationship with temperature (Reference Petit, Duval and LoriusPetit and others, 1987) and dust (Reference Koerner and FisherKoerner and Fisher, 1979) as reported in polar ice cores. Crystal growth rate increases with temperature (Reference Duval and LoriusDuval and Lorius, 1980). Between 52 and 89 m, a growth rate of 0.2 mm² a⁻¹ was calculated, which, if applied to the entire core, should result in a crystal size of 300 mm² at 128.5 m. However, measured crystal size decreases substantially below 128 m where a minimum of 11.4 mm² was measured, and changes sharply at the LGS/Holocene transition. Similarly, LGS ice in cores from Antarctica (Reference Petit, Duval and LoriusPetit and others, 1987) and Greenland (Reference Herron, Langway, Langway, Oeschger and DansgaardHerron and Langway, 1985) is characterized by substantially smaller crystals and abrupt crystal size changes at the LGS/Holocene transition.

All these analyses lead us to conclude that the lower 10 m in the Dunde ice cap represent ice deposited during the last glacial stage. The high dust concentrations correlate closely with δ¹⁸O depletion (temperature proxy) as found in all polar cores extending below the LGS/Holocene transition. In fact, the association of high dust content and more negative δ¹⁸O has been used routinely for the last decade as the principal characteristic of LGS ice in polar ice cores, e.g. Camp Century and Dye 3, Greenland (Thompson and Mosley-Thompson, Reference Thompson and Mosley-Thompson1981a; Reference Dansgaard, Clausen, Gundestrup, Johnsen, Rygner, Langway, Oeschger and DansgaardDansgaard and others, 1985); Ellesmere Island and Devon Island (Reference PatersonPaterson and others, 1977; Reference KoernerKoerner, 1977, Reference Koerner, Fisher and PatersonKoerner and others, 1987); and, in Antarctica, Vostok (De Angelis and others, Reference De Angelis, Barkov and Petrov1987); Byrd Station (Reference Johnsen, Dansgaard, Clausen and LangwayJohnsen and others, 1972; Thompson and Mosley-Thompson, Reference Thompson and Mosley-Thompson1981a) and Dome C (Reference Thompson, Mosley-Thompson and PetitThompson and others, 1981b). In addition to the same δ¹⁸O and MPC characteristics which are found in glacial stage ice in polar cores, the decrease in ice-crystal size and the abrupt nature of concentration changes in both ions and insoluble dust at the glacial/interglacial transition further support the LGS age interpretation of the lower 10 m of each core.

Time-Scale Construction

Total β radioactivity horizons from the upper parts of the 1987 D-1 and D-3 cores, as well as from the 1986 summit cores, provide time stratigraphie markers (Fig. 6) for verification of seasonal variability of MPC and δ¹⁸O. As demonstrated in Figure 6, the 1963 radioactivity horizon is pronounced and provides an average accumulation rate of ∼400 m of H2O, which is twice the 1980–86 rate of ∼200 m (H₂O equivalent) determined from snow-pit studies (Reference Thompson, Mosley-Thompson, Wu and XieThompson and others, 1988b); however, it is similar to the measured values of ∼350 m (H₂0 equivalent) from 31 summit stakes for the 1986–87 accumulation year.

Figure 7 illustrates the annual MPC and δ¹⁸O cycles used to date the upper 70 m of D-1. The diamonds mark the annual layers which thin with depth. Due to the low accumulation rate, coupled with rapid thinning with depth, layer separations below 70 m are too small to be resolved by analyzing seasonally varying parameters, as the average sampling interval is 3.7 cm. However, the enhanced dust deposition during the dry season produces an additional annual stratigraphie marker visible deeper in the ice. Between 70 and 117 m, these visible dust layers were used for dating. The measured layer thicknesses from the visible dust layers as a function of depth are shown in Figure 8. Counting the annual δ¹⁸O and MPC peaks down to 70 m, and the visible dust layers below 70 m, provides an age of about 4550 year B.P. at 117 m depth.

Fig. 5. Ice-crystal sizes measured for selected sections of core D-3, illustrated along with the δ¹⁸O record. Light micrographs illustrate ice crystals for sections of typical Holocene ice, ice containing a major dust layer and LGS ice near the ice-cap bottom.

If annual layers could be identified down to the stratigraphie transition at 129.2 m depth (125.8 m ice equivalent), the corresponding age could be determined by simply counting the number of layers. Unfortunately, annual layers are too thin to be discerned below about 117 m depth. However, a lower boundary to the age can be estimated by assuming no additional layer thinning below the last measured value, which is about 0.0037 m a⁻¹. The resulting value is about 7000 year B P.

The LGS/Holocene transition has been dated at 10 750 year BP. in the Camp Century ice core from northern Greenland (Reference Hammer, Clausen and TauberHammer and others, 1986), and we seek to determine whether the thinning of measured annual layers is consistent with a similar age at the 129.2 m stratigraphie transition. A more refined estimate of the age of this horizon can be made by extrapolating the layer-thickness data. The measured variation of layer thickness with depth could be fitted with an nth-order polynomial or an appropriate set of orthogonal functions. However, because the behavior of such least-squares, best-fit functions is not well constrained beyond the data region, the resulting extrapolation would not be very useful. Also, the scatter in the data is related to secular variations in the accumulation rate and, to a lesser degree, subsequent adjustments in the velocity field. Thus, extrapolating a least-squares fit to the data places unrealistic emphasis on the temporal variations that occur later in time than the interval of interest.

Instead, a more physical approach is taken and the nature of the flow to the bore hole is considered. Measurements of the detailed surface topography and surface velocities (1.8 m a⁻¹) around the bore hole indicate that the topographic summit is also the center of flow, or flow divide; and that the D-1 bore hole is about 17 m southeast of the summit (Fig. 1c). Various studies (Reference RaymondRaymond, 1983; Reference BolzanBolzan, 1985; Reference ReehReeh, 1988) suggest that layers tend to thin more rapidly with depth near a flow divide than at distances many ice thicknesses away. Here the D-1 bore hole is much less than one ice thickness (about 136 m ice equivalent) from the flow divide and is well within the near-divide zone, which seems to be on the order of several ice thicknesses wide. Based on the analysis of a deep temperature profile in East Antarctica (Reference BolzanBolzan, 1985), a simple analytical representation for the annual layer thickness as a function of depth in the near-divide zone was found to be:

where Η is the ice thickness, b is the accumulation rate in ma⁻¹, and z is the ice-equivalent depth. To compute the time interval corresponding to a given depth interval, the inverse of the annual layer thickness (i.e. the number of years per meter) is integrated over the depth interval.

The accumulation rate (b) is approximately the average value over the interval in question. Here only the average value over the past 25 years is known, and is assumed to be an appropriate value for the Early Holocene. The exponent Ρ is a priori unknown and is determined to be 1.612 by requiring the age to be 4550 years at 117 m depth, as given by the annual layer thickness data. With these values for b and p, we find the age of the stratigraphie transition at 129.2 m depth (125.8 m ice equivalent) to be about 11 950 year B.P.; within 10% of the assumed age (the calculated layer thicknesses as a function of depth are also shown in Figure 7). This result also supports the identification of the LGS/Holocene horizon on the basis of the ice-core stratigraphy.

We emphasize that this calculation provides only a rough estimate of the age deep in the core. We have ignored factors such as temporal variations in accumulation rate, ice-sheet shape, and the position of the flow divide, as the subsequent ice-sheet response to these changes is poorly understood; and rarely is sufficient information available to specify these secular changes. Nevertheless, it is interesting to use the above results to estimate the time-scale deeper in the core. To do this we recalculate the time-scale by fixing the age at the LGS/Holocene transition to be at 10 750 year B.P. The resulting ages are indicated in Figure 4(a, b) and in Figure 9. These results yield a preliminary age well in excess of 100 000 years.

Fig. 6. β radioactivity results from three cores (two drilled in 1987 and one in 1986) at the summit provide an accurate determination of the present accumulation rate. From 1963 to 1987, the average annual accumulation is 390 mm of water equivalent and from 1963 to 1986 it is 405 mm water equivalent. Decadal time lines were determined by counting annual MPC and δ¹⁸O layers (Fig. 7).

Fig. 7. Three representative 5 m ice-core sections illustrate the annual variations in total particles (diameter ≥2.0 μm) per ml of sample and δ¹⁸O which were used to date the upper 70 m of core D-1. The diamonds indicate the annual layers. The average thickness (λ) of the annual layers is shown; annual layer thinning is evident (see Fig. 6).

Fig. 8. The curve represents the calculated annual layer thicknesses (in cm of ice) as a function of depth in core D-1. The squares illustrate the annual layer thicknesses based upon the visible dust layers. Data points A, Β and C were calculated from the annual MPC and oxygen-isotope layers for sections illustrated in Figure 7. The insert shows the thicknesses in detail from 90 m to the bottom of the core. The curve represents the calculated variation of layer thickness with depth for the analytical representation described in the text with the parameters p = 1.612 and b = 0.428 ma⁻¹.

Of particular significance for the age of the bottom few meters would be the occurrence of basal melting, which could remove many thousands of years of ice from the bottom if a very small amount of melting were sustained for a prolonged period of time. While the current basal temperature is about −4.7°C (Reference Thompson and Mosley-ThompsonThompson and others, 1988a; Fig. 3a), a decrease in the accumulation rate, an increase in ice thickness, or an increase in the mean annual surface temperature (currently −7.3°C) would tend to raise the basal temperature. Estimates of the changes in temperature and ice thickness necessary to raise the basal temperature to the pressure-melting point are obtained by numerically integrating the heat equation. A surface temperature 1°C above the present value requires the ice to thicken by over 250 m to initiate basal melting. For a surface temperature rise of 4°C, the ice must be only about 30 m thicker. However, no previously warmer period is evident in the isotopie record for the entire Holocene period. Additionally, the thicknesses required to initiate basal melting by a small surface warming are several times larger than would be possible, even with a doubling of the accumulation rate over an extended period of time. This suggests that conditions resulting in basal melting are unlikely.

Discussion

The Late Glacial Stage record of climatic and environmental conditions on the Tibetan Plateau now begins to emerge (Fig. 9) from ice cores on the Dunde ice cap. The high alkalinity and NaCl concentrations reveal that the soluble dust deposited on the ice cap originates from the surrounding salt and calcareous deposits, with the Qaidam Basin probably contributing a significantly large portion of the soluble dust. The presence of similar cations and anions in both Holocene and LGS ice suggests that these sources have remained the same over the last ∼40 000 years. However, the reduced concentrations of dissolved species in LGS ice may reflect higher precipitation rates during the last glaciation. There is evidence (Reference Chen and BowlerChen and Bowler, 1985) suggesting that a large part of the Qaidam Basin was covered by fresh-water lakes during the LGS. Desiccation of the lakes began in the latter part of the LGS in response to reduced precipitation and/or precipitation/evaporation ratios, resulting in the formation of more extensive halite and calcareous deposits. As the Holocene is approached, the increasing ion concentrations and electrical conductivity probably reflects both decreased precipitation and increased dry surface area in the Qaidam Basin.

Fig. 9. The 1000 year averages of dust concentrations (diameters ≥2.0 μm), δ¹⁸O, Cl⁻, NO⁻ ₃ and δ¹⁸O, are illustrated for the last 40 000 years. The LGS–Holocene transition is characterized by an abrupt decrease in insoluble dust, an increase in dissolved species and a more gradual warming inferred from δ¹⁸O. The average values represent all samples in the 40 000 year record and not the individual 1000 year averages. There are 5 m of ice below the 40 000 year model cut-off shown here (Reference ThompsonThompson and others, 1989).

The enhanced dust deposition in LGS ice likely results from increased wind strength in response to increased baroclinicity (steeper isobars and isotherms) due to the expanded continental ice sheet in the northern hemisphere. The high winds from the deserts of western Asia deposited loess hundreds of feet thick across central and eastern China, extending to the Pacific Basin (Reference Braaten and CahillBraaten and Cahill, 1986) and most likely to the North American continent. Under present-day conditions, large quantities of mineral dust are transported to the North Pacific each spring (Reference BetzerBetzer and others, 1988). During the LGS, very large and discrete dust events blanketed the Dunde ice cap and most likely much of the extensively snow-covered and glacierized plateau, significantly altering the radiation balance. Growing evidence indicates that under the current climate regime albedo changes due to variation in Eurasian snow cover affects global climate. These include strong influences on the Asian summer monsoon system and the El Nino–Southern Oscillation phenomenon (Reference Barnett, Dümenil, Schlese and RoecknerBarnett and others, 1988). During the LGS the large short-term variations in albedo on the Tibetan Plateau due to large, multiple, discrete, high-altitude dust events may have increased the variability in the monsoonal system and, hence, of the global climate system. The decrease in dust deposition at the LGS/Holocene transition was rapid (within ∼40 years) reflecting an abrupt and substantial change in the prevailing climatic conditions which severely limited the capability of the atmosphere to transport dust.

The δ¹⁸O record (Fig. 9) reveals an apparent warming between ∼38 000 and ∼30 000 year B.P. which is associated with less dust deposition. At about 30 000 year B.P., full glacial conditions were re-established and dust deposition increased substantially and abruptly. Between 30 000 to 10 000 year B.P., concentrations of Cl⁻ and δ¹⁸O increased gradually as conditions became drier and the areal extent of salt and loess deposits increased. About 10 750 year B.P. a marked increase in these species occured as drying of the fresh-water lakes was completed. Simultaneously, the insoluble dust decreased sharply (Figs 4b and 9). The δ¹⁸O profile suggests a gradual warming while the aerosol (soluble and insoluble) data indicate rapid environmental changes including reduced wind intensity.

A most striking feature of the δ¹⁸O record (Fig. 4a) is the extreme warmth of the last 60 years with the warmest decades being the 1940s, 1950s, and 1980s. Using the δ¹⁸O record as a proxy of temperature, the last 60 years constitute one of the warmest periods in the entire record, equalling levels of the Holocene maximum 6000–8000 year B.P. (Fig. 4). Model results of Hansen and others (Reference Hansen1988) suggest that the central part of the Asian continent may be one of the first places to exhibit an unambiguous signal of the anticipated “greenhouse warming”. Certainly, the Dunde ice-core (D-1) results suggest that the recent warming on the Tibetan Plateau has been substantial. Recent radiosonde data from southern India (Reference Flohn and KapalaFlohn and Kapala, 1989) show that, in fact, the average tropospheric temperature has increased nearly 1° since 1965.

Acknowledgements

This research was supported by the U.S. NSF Office of Climate Dynamics and the Division of Polar Programs (ATM-8519794), the National Geographic Society (3323-86), The Ohio State University, and Academia Sinica of China. We thank the many individuals who contributed to the success of this program, especially the 50 Chinese and American participants. Initial support was provided by the U.S. National Academy of Sciences’ Committee for Scholarly Communication with the People’s Republic of China. We thank the Fred C. Hansen Company for assistance in the modification of a tree-ring incremental measuring device for use in ice-core dating, S. Smith for illustrations and K. Doddroe for typing. This paper is Contribution 711 of the Byrd Polar Research Center, The Ohio State University.