Introduction

The glaciochemical record contained in the glaciers of central Asia represents a valuable resource that can be used to document the atmospheric deposition of this region and reconstruct past regional climate. Recovering representative long-term glaciochemical records from central Asia requires understanding the processes that control the chemical content of snow and the local-to-regional scale variations Reference Wake, Mayewski and SpencerWake and others, 1990).

An assessment of the representativeness of individual snow samples is imperative for interpreting regional atmospheric deposition or reconstructing past regional climate from glaciochemical records. Chemical inputs to mountain basins can vary spatially and temporally in a complex manner. Spatial variability in processes such as wind direction (Reference Schemenauer, Summers, Wiebe and AnlaufSchemenauer and others, 1985; Reference Anlauf, Bottenheim, Brice and WiebeAnlaug and others, 1986) andsnow type (Reference Langham, Gray and MaleLangham, 1981) during storms can affect the chemical composition of the snowfall to an unknown degree. Knowledge of the source of solutes is also essential to interpret the atmospheric deposition of the region. Acidic deposition from SO₂ emissions in coalburning furnaces is a problem in northeastern and southern China Reference Zhao and SunZhao and Sun, 1986; Reference Galloway, Dianwu, Xiong and LikensGalloway and others, 1987). The solute content of snowpacks in central Asia may provide background measurements of deposition unaffected by anthropogenic emissions.

Interpretation of glaciochemical data recovered from snow pits in central Asia also necessitates understanding the processes that control the chemical content of the snowpack after deposition. Physical processes such as wind pumping (Reference Clarke and WaddingtonClarke and Waddington, 1990) and gaseous changes such as uptake and oxidation of SO₂(Reference Bales, Davies, Tranter and JonesBales, 1991) can alter the chemical composition of the snowpack after deposition. Snowpacks store both wet and dry deposition. Aeolian deposits from the arid regions of central Asia have been shown to be an important source of solutes in the snowpack (Reference Jenkins, Drever, Reider and BuchananJenkins and others, 1987; Reference Wake, Mayewski and SpencerWake and others, 1990). Dissolution of aeolian dust may change the ionic content of snow and increase the difficulty of interpreting the glaciochemical record. For example, Reference Williams and MelackWilliams and Melack (1991b) report that dissolution of particulates in snowpack meltwater was an important source of Ca²⁺, Mg²⁺ and HCO₃ ^- in a headwater basin in the Sierra Nevada, USA.

Research on the chemical composition of snow in central Asia was conducted in the eastern Tien Shan. Based on the results of previous research in central Asia (Reference Jenkins, Drever, Reider and BuchananJenkins and others, 1987; Reference WakeWake, 1989; Reference Wake, Mayewski and SpencerWake and others, 1990), we addressed three research areas: (i) sources of solutes in snow; (ii) spatial variation in the solute content of snow at the scale of headwater basins; and (iii) postdepositional alteration of the solute content of the snowpack.

Site Description

The Tien Shan is located in northwestern China (43°06′N, 86°49′E) and extends more than 2000km westward into the Soviet Union Fig. 1). The Tien Shan is bordered by large deserts on three sides: the Gobi desert to the east, the Taklimakan in the Tarim basin to the south, and the Junggar basin to the north; the nearest sea is located more than 3000 km from our site. The research area is located in the Daxigou valley, at the headwaters of the Ürümqi River. The Lanzhou Institute of Glaciology and Geocryology has maintained a glaciological research site here since 1959. We investigated four headwater basins: three with active glaciers (Glacier No. 2, west and east branches of Glacier No. 1)and one basin free of glacial ice (Fig. 1). The areas of each basin range from 1 to 2 km². A steel mill and cement plant are located on the Ürümqi River in the town of Houxia, about 50 km northeast of the Daxigou valley.

Fig. 1.

Location map of the study area. Lines around Glaciers No. 1 and No. 2 define the glacial boundary; the line around the Dry Cirque is the watershed divide. The inset map shows the location of the Tien Shan in relation to geographic and political features of northwestern China.

Prevailing winds over most parts of China are from the north, northwest and northeast during the winter and from the south during the summer (Reference Dömros and PengDomrös and Peng, 1988). However the area of the Xinjiang province around Ürümqi is atypical in that 38% of winds in the winter are from the south whereas in the summer the prevailing winds are from the north (Domrös and Peng, 1988). Meteorological data from the Tien Shan Glaciological Station indicate that up-valley winds predominate in the Daxigou valley during the afternoon (Reference Wake, Mayewski, Ping, Qinzhao, Jiankang and ZichuWake and others, 1992).

Methods

The research was conducted in May 1990, about two weeks after maximum accumulation of the winter snowpack. The snow samples were collected using beveled PVC tubes (50 mm diameter, 500 mm long), which had been soaked in 10% HCl and then rinsed at least five times with deionized water. Vertical, contiguous cores were collected, in increments of 400 mm, from the snow-air interface to either the snow-firnor snow-ground interface. Cores were collected by placing the coring device about 100 mm behind the pit wall and then pushing it into the snowpack, parallel with the pit face. This procedure insured that the collected snow had no contact with any device (such as shovels) other than the corer. Snow was transferred from the cores into new polyethylene bags. Snow temperatures were taken every 100 mm and were the mean of three dial-stem thermometers (±1.0°). Event samples from new snow were collected in bulk containers that were opened to the atmosphere at the start of snowfall; collectors and procedureare similar to Reference Galloway, Likens, Keene and MillerGalloway and others (1982). Samples of new snow were transferred to new polyethylene bags after collection. Each snow sample was analyzed for pH, conductance (µS cm^-1), and solute concentrations (μeq L^-1 or μmοl L^-1); a subset of samples was analyzed for acid neutralizing capacity (ANC).

Sampling strategy was designed such that analysis of variance (ANOVA) statistical tests could be usedto evaluate the presence or absence of spatial variation in the chemical content of accumulated snow.Snow samples were collected from four pits in each of the four headwater basins. Pit sites within each basin were located at high, mid, and low elevations along the center line of the basin, and on a north-facing slope at mid elevation; elevations ranged from 3775 to 4300 m (Fig. 1). Four replicate snow cores were sampled at each pit; snow cores were separated by a horizontal distance of 1 m.

Snow samples were melted in the collection bags the night after collection; air temperatures ranged from 0 to 10°C. Blank samples of distilled, deionized water were run in the bags for the same amount of time and showed no significant contamination from the bags. Air-equilibrated pH, specific conductance and ANC were measured on unfiltered samples immediately after samples were melted. Subsamples werefiltered through pre-rinsed, 47 mm Gelman A/E glass fiber filters of c. 1 μm pore size into 125 ml polyethylene bottles. These samples were hand-carried back to the United States for the analysis of major inorganic ions and reactive silicate at our analytical laboratory at the University of California,Santa Barbara.

The pH measurements were made with a Ross 8104 combination electrode and Orion pH meter. For each series of measurements the electrode was calibrated with pH 7.00 and pH 4.00 reference buffers. The electrode was rinsed with an aliquot of sample and the temperature compensated pH determination made on afresh, quiescent sample after five minutes. Conductance was measured with a YEW field meter and glasselectrode with a 1.0 cell constant. Simultaneous temperature measurements were made, and conductance was standardized to 25°C using a coefficient of 2% per °C. Calcium, magnesium, sodium and potassium were analyzed with a Varian AA6 atomic absorption spectrophotometer. Chloride, nitrate and sulfate were measured by ion chromatography (Dionex Model 1010i) employing chemical ion suppression and conductivity detection. Reactive silicate (Si) was measured by the silica-molybdate method. Acid-neutralizingcapacity was measured using the Gran titration method. Detection limits for all solutes were less than 0.5 μeq L^-1, except for Ca²⁺ (1.0 µeq L^-1). Analytical precisions were less than 3.3% of mean values, except for K⁺ (6.2%), Cl^- (13.4%) and ANC (± 4 μeq L^-1). Analytical accuracy was assessed using NBS certified controls. Our charge balance control demonstrated there was no bias in our analytical performance Reference Elder, Kattelmann, Ushnurtsev, Daqing and ChichagovBerg and others, 1989). Analytical techniques and error analysis are the same as those published in Reference Williams and MelackWilliams and Melack (1991a).

Results

Snow depths for all sixteen pits ranged from 0.47 to 1.51 m. The mean snow water equivalance for the east branch of Glacier No. 1 was 0.40 m Reference Clarke and WaddingtonElder and others,1992). Snowpack temperatures were 0°C and free water was observed in all snow pits. Distinctive layers of dust were observed in each snow pit in the Tien Shan. These dust layers were often 20 to 30 mm in thickness and occurred about every 200 to 300 mm. Fifty-nine snowpack samples were usable; three samples were opened in transit and only two samples were taken at the high-elevation snow pit on the west lobe of Glacier No. 1 due to the steepness of the site. Seven samples of new snow were collected.

The results of chemical analyses for the 59 snowpack samples show that the average pH of the snowpackwas near neutral at 6.91 (Table 1). The specific conductance of the snowpack varied from 3.7 to 13.9 μS cm^-1 with a mean of 8.9 μS cm^-. The mean ANC concentration of 50.2 μeq L^-1 composed 61% of the total ionic charge (Σ—) of the snowpack (Fig. 2) where Σ- = ANC + SO₄ ^2- + NO₃ ^- + Cl^-, all in μeq L^-1. Sulfate had an average concentration of 16.2 μeq L^-1 and represented 20% of Σ-. The mean concentration of NH₄ ⁺ was 12.7 μeq L^-1 Fig. 2). Ammonium was calculated using an ionic charge balance equation:

all in μeq L^-1.

Fig. 2.

Mean concentrations of individual solutes (μeq L^-1) in the snowpack of the Daxigou valley (n = 59; except for ANC = 32). Anions are shown in (a) and actions in (b).

Table 1.

Mean ionic concentrations (µeq L^-1), pH and conductance (µS cm^-1)of variation, CV,for the snowpack and new snow

This equation assumes that contributions were negligible from other anions (such as organic acids) and cations (such as Al^{3 +}). The mean Ca²⁺ concentration of 52.5 μeq L^-1 represented 64% of the total cationic charge (Σ+ ) of the snowpack, where Σ + = NH₄ ⁺ + Ca²⁺ + Mg²⁺ + Na⁺ + K⁺, all in μeq L^-1. Na⁺, Mg²⁺, and K⁺ represented 20% of Σ+. Concentrations of Si ranged from 0.1 to 1.1 μmol L^-1, with a mean of 0.4 μmol L^-1. The variance of the solute content of the snowpack was evaluated using the coefficient of variation, CV,where CV is defied as the standard deviation divided by the mean, times 100%. The CV ranged from 24% to 46% for most ions (Table 1). The slightly larger CV for K⁺ and Si may in part be the result of analytical error, as these solutes were near their detection limit.

Most of the variance in the chemical content of the snowpack was due to variations among snow pits within a basin. Spatial variation of solutes within anindividual snow pit was small, but there was significant spatial variation among snow pits within a basin. The standard deviation within each pit was about 10% of the mean, illustrated here for concentrations of SO₄ ^2- and Ca²⁺ (Fig. 3). Results of the ANOVA test show that the solute concentrations of most ions within the east branch of Glacier No. 1 varied significantly at the α = 0.05 level (Table 2). Snow pits were segregated into groups of high, mid and low elevation and north-facing aspect, on the east branch of Glacier No. 1. Concentrations of most solutes were generally highest on north-facing aspects and lowest at the snow pits located on sites at low elevation, as illustrated by concentrations of SO₄ ^2- and Ca²⁺ Fig. 3).

Fig. 3.

Mean concentrations of each of four snow pits located at high, mid and low elevations, anda north-facing aspect, on the east branch of Glacier No. 1. Lines are one standard deviation from themean value in the snow pit. Sulfate is shown in (a) and Ca²⁺ in (b). The probability, p, of within-basin variance, derived from a one-way analysis of variance test, indicates that concentrations of SO₄ ^2- and Ca²⁺ varied significantly (α = 0.05) among sites within the basin, df = degrees of freedom.

Table 2.

Probability values for one-way analysis of variance tests on the solute content of the snowpack

However, there was no significant variation among basins for most solutes, using a one-way ANOVA and α = 0.05. For example, there was little variation in the mean concentrations of Ca²⁺ amongthe east and west branches of Glacier No. 1, Glacier No. 2 and the Dry Cirque Fig. 4). Mean concentrations of Ca²⁺ ranged from about 50 to 55 μeq L^-1. The standard deviations among basins were high relative to the standard deviations within an individual snow pit, about 30% of the mean. The probability value of 0.59 indicates that there was no significant variation in concentrations of Ca²⁺ among basins.

Fig. 4.

Mean concentrations of Ca²⁺ in the snow pack from each of the four basins. Lines are one standard deviation from the mean value for each basin. The probability, p, for among-basin variance, derived from a one-way analysis of variance test, indicates that concentrations of Ca²⁺ did not vary significantly (α = 0.05) among basins, df = degrees of freedom.

Information on sources of solutes can also be obtained from the results of the ANOVA tests. The ANOVAresults show that the various ions tend to cluster into two groups, suggesting a common origin for each of these two groups. Na⁺, K⁺ and Cl^- differed significantly amongbasins and ANC, H⁺ , Ca²⁺, Mg²⁺ and NO₃ ^- did not vary significantly among basins (Table 2). Τhe significance of the two ionic groups was reversed when elevation and aspect were compared, with Na⁺, K⁺ and Cl^- not significantly different and ANC, H⁺, Ca²⁺, Mg²⁺ and NO₃ ^- significantly different. Sulfate concentrations did not follow this pattern; SO₄ ^2- concentrations were significantly different, both among basins and by elevation and aspect.

Sources of solutes were also investigated by looking at associations among ions. A parametric correlation analysis for all ions in the snowpack indicates that ANC was highly correlated with Ca²⁺ (r = 0.82, n = 32) and inversely correlated with H⁺ (r = -0.63, n = 32). Specific conductance was negatively correlated with H⁺ (r = -0.76, n = 59). Na⁺ was highly correlated with Cl^- (r = 0.90, n = 59) (Fig. 5) and with K⁺ (r = 0.75, n = 59) and not with any other ions. Themolar ratio of 0.98 for Na : Cl was slightly higher than the sea-salt ratio of 0.85 (Reference Henriksen, Drablos and TollanHenriksen, 1980), indicating that the snowpack was slightly enriched in Na⁺ . Analysis of linear regression shows that SO₄ ^2- was significantly correlated with Ca²⁺ (SO₄ ^2- = 1.74 + 0.27 Ca²⁺, r² = 0.70, n = 59) Fig. 5). Adding Mg²⁺ in a multiple regression increased the r² to 0.80; addingNa⁺ did not increase the r². Multiple regression analysis also shows that Ca²⁺ was significantly correlated with ANC and SO₄ ^2- (Ca²⁺ = 2.14 + 0.47ANC + 1.60SO₄ ^2-, r² = 0.82, n = 32).

Fig. 5.

Comparison of snowpack concentrations of Na⁺ and Cl^- in (a) and SO₄ ^2- and Ca²⁺ in (b); r² is the correlation coefficient and n is the number of samples. Both Na⁺ and Cl^- and SO₄ ^2- and Ca²⁺ were significantly correlated at the p < 0.001 level.

The solute content of new snow was at times different than that of the snowpack (Table 1). The average pH of new snow was 6.39 and had an average conductance of 4.7 μS cm^-1. Correlation analysis shows that there was a small inverse correlation betweεn conductance and H⁺ (r = -0.22). Average concentrations in new snow of Ca²⁺ and ANC were 21.6 μeq L^-1 and 15.6 μeq L^-1, 41% and 33% of Σ+ and Σ- in new snow, respectively. In contrast to the snowpack, ANC was not significantly correlatεd with H⁺ (r = 0.07). Ammonium concentrations in new snow were 6.2 μeq L^-1, calculated by the samr procedure as for NH₄ ⁺ concentrations in the snowpack. Using a two-tailed t-test, the solute content of the snowpack and new snow differed significantly at the α = 0.05 confidence level for the following parameters. Total ionic charge of the snowpack (n = 59) was 164 μeq L^-1 and total ionic charge of the new snow (n = 7) was 65 μeq L^-1. Conductance, pH, ANC and Ca²⁺ were significantly higher in the snowpack than in new snow. In contrast, SO₄ ^2- concentrations did not differ significantly between the snowpack and new snow (p = 0.40). Furthermore, SO₄ ^2- concentrations accounted for 37% of the Σ- of new snow and for only 20% of the Σ- in the snowpack.

Discussion

Base cations in the atmosphere neutralize some of the acidity in snowfall caused by H₂SO₄ from anthropogenic sources. Deposition of SO₄ ^2- in association with basic cations may have contributed to the relatively high pH of 6.34 for new snow. Therefore, the measured concentration of H⁺ in snowfall and in the snowpack reflects the acidity after neturalization by atmospheric bases and is not a measure of the original acidity. Our results are similar to those reported by other investigators on atmospheric acidity in China. Reference Galloway, Dianwu, Xiong and LikensGalloway and others (1987) and Reference Zhao, Xiong, Rohde and HerreraZhao and Xiong (1988) report that the pH of rainfall is much higher in northern China than in southern China due to the neutralization of H⁺ by the base cations Ca²⁺ and NH₄ ⁺. Reference Huebert, Wang and LuHuebert and others (1988) report high levels of Ca²⁺ aerosols and also report that the ratio of ([Ca²⁺] + [NH₄ ⁺])/([SO₄ ^2-] + [NO₃ ^-]) was always > 1 in aerosols and vapor collected on teflon/nylon filter packs in northeastern China. Buffering of H⁺ from atmospheric sources is consistent with available data, which shows that much of the Ca²⁺ in the atmosphere of China was originally in the form of CaCO₃ (see Reference Galloway, Dianwu, Xiong and LikensGalloway and others (1987) for a summary of data sources).

Sulfate concentrations in snow from the Tien Shan are about three-fold more than concentrations from other remote areas of the world. Mean SO₄ ^2- concentrations in snow from three areas in the central Himalayas varied from 4.8 to 7.7 μeq L^-1 (Reference Wake, Mayewski and SpencerWake and others, 1990). Sulfate concentrations in precipitation from remote areas of the soutlern hemisphere are around 4 μeq L^-1 (Reference Galloway, Dianwu, Xiong and LikensGalloway and others, 1987), similar to the SO₄ ^2- concentrations inthe Himalayas. The elevated concentrations of SO₄ ^2- in the Tien Shan could be from a variety of sources. Crustal dust in the form of CaSO₄ is a potential source of sulfate, and is consistent with the high correlation of Ca²⁺ with SO₄ ^2-. Reference Wake, Mayewski, Ping, Qinzhao, Jiankang and ZichuWake and others (1992) report that anthropogenic emissions from the city of Ürümqi are a major sourceof the SO₄ ^2- in snow at the Bogda Shan, located about 100 km from the Daxigou valley. Upslope winds that bring air from the steel mill in Houxia to the Daxigou valley are a potential source of anthropogenic SO₄ ^2-. Regional surveys would help to discriminate the various sources of SO₄ ^2- in the Tien Shan.

Sea salts contributed only a small amount of the chemical content in snow in the Tien Shan. Both the ANOVA results and correlation analysis indicate that Na⁺, K⁺ and Cl^-have a common source. The molar ratio of Na : Cl has been used to identify marine contributions to the chemical content of snow in central Asia by Reference Wake, Mayewski and SpencerWake and others (1990), who conclude that average Na : Cl ratios of 0.98 to 1.2 from the Biafo, Sentik and Yala Glaciers, and the Mount Everest regions are indicative of a marine source for these ions. The Na : Cl of 0.96 that we report for the Tien Shan issimilar to the Na : Cl ratio of these other regions in central Asia, and only slightly higher than the Na : Cl ratio of 0.85 found in sea salt. However, Reference Wake, Mayewski and SpencerWake and others (1990) conclude that Na⁺ and Cl^- in the Tien Shan represent an input of Na⁺ and Cl^- from evaporite deposits of areas to the north of the Tien Shan. We can place an upward limit on the contribution of marine air masses to the chemical content of snow in the Tien Shan by assuming that all Cl^- is from sea salt. The maximum contribution of marine air masses to the chemical content ofsnow is about 12%, based on the assumption that all Cl^- is from sea salt.

Terrestrial dust may be the primary source of NO₃ ^- in the snowpack. Reference Lyons, Mayewski, Spencer and TwicklerLyons and others (1990) report that crustally derived NO₃ ^- from soils appears to be the main source of NO₃ ^- in snow samples from central Asia. Reference Wake, Mayewski and SpencerWake and others (1990) also report that terrestrial dust is the primary source of NO₃ ^- in snow from central Asia. Terrestrial dust as a source of NO₃ in snow in the Tien Shan is supported by our ANOVA results. Our ANOVA tests showed that NO₃ ^- was grouped with Ca²⁺ and ANC (Table 2). The primary source of Ca²⁺ and ANC in the snowpack may be from aeolian dust. The grouping of NO₃ ^- with Ca²⁺ and ANC suggests that NO₃ ^- has a similar source.

It is unknown why there is significant spatial variation within basins, but not among basins, for most solutes. Loss of ions in meltwater may be one reason, and is consistent with higher elevations and north-facing aspects having higher concentrations for most ions. However, Reference Wake, Mayewski and SpencerWake and others (1990) haveevaluated meltwater infiltration as low for Glacier No. 1, at elevations ranging from 3450 to 4100 m.An increased rate of post-depositional alterations at higher elevations as compared to lower elevations may account for the significant within-basin differences of the chemical content of snow. The parametric correlation results and ANOVA results both suggest that solutes can be divided into two main groups, one group composed of Na⁺, K⁺ and Cl^-, and a second group compsed of Η⁺, ANC, NO₃ ^-, Ca²⁺ and Mg²⁺. Previousfield research in the Sierra Nevada has shown that melt-freeze cycles can increase concentrations of ANC, Ca²⁺ and Mg²⁺ and buffer concentrations of H⁺ in snowpack meltwater Reference Williams and MelackWilliams and Melack, 1991b). If the primary source in the Tien Shan for ANC, Ca²⁺, Mg²⁺ and NO₃ ^- is particulates from aeolian deposition, increased melt-freeze actions at higher elevations and north-facing aspects may account for the significant variation among sites of these ions. More research is necessary to determine the factors that cause significant spatial variation within headwater basins of the eastern Tien Shan.

Post-depositional alterations of the solute content of snowpacks are indicated by the significant differences between solute concentrations in the snowpack and new snow. Seasonal differences in deposition may account for our observed differences in the solute content of the snowpack and new snow. However, the similarity in concentrations of SO₄ ^2- in the snowpack and in new snow indicates that seasonal differences in the solute content of snowfall were not the primary reason thatconcentrations of Ca²⁺ and ANC were significantly different in new snow and the snowpack. Furthermore, an additional source of anions after deposition is indicated by the decrease in the ratio of SO₄ ^2-: Σ- in the snowpack relative to new snow. Dissolution of dust in thesnowpack may be responsible for the increase in solute content of the snowpack relative to new snow. Reference Wake, Mayewski, Ping, Qinzhao, Jiankang and ZichuWake and others (1992) also report that the dissolution of dust increases the solute content of the snowpack in the Tien Shan. Moreover, dust from western China has been shown to change the chemical content of snow in areas as distant as the Canadian Arctic (Reference WelchWelch and others, 1991).

The in situ production of ANC and Ca²⁺ may be from weathering of carbonates contained in the dust. Dissolution of CaCO₃ in Saharan dust has been shown to increase significantly thepH and Ca²⁺ content of rain in Corsica (Reference Loÿe–Pilot, Martin and MorelliLoÿe-Pilot and others, 1986). Weathering of CaCO₃ to Ca²⁺ and HCO₃ ^- in meltwater can explain the high correlation between concentrations of ANC and Ca²⁺ in the snowpack, and the inverse relationship between ANC and H⁺, Furthermore, dissolution of carbonates can explain the increasein the ratio of ANC:Σ- and Ca:Σ+ in the snowpack relative to new snow. The low concentrations of Si in the snowpack indicate that weathering of silicate minerals was not an important source of solutes.

Our results indicate that aeolian dust deposited as dry deposition may be an important component of the chemical content of snow in the eastern Tien Shan. However, no data exist on the relative contributions of wet and dry deposition to the chemical content of snow in northwestern China. Additional research on the relative contributions of wet and dry depositon to the chemical content of snow in central Asia would assist in interpreting the glaciochemical record.

Conclusions

Sulfate concentrations in snow from the eastern Tien Shan are about three-fold greater than SO₄ ^2- concentrations from other remote areas of the world. Regional surveys are needed to determine if these elevated concentrations of SO₄ ^2- are from regional gradients in natural concentrations, local anthropogenic sources, regional anthropogenic sources, or some combination of these sources.

Efficient sampling strategies to measure the chemical content of snow are suggested by the results ofstatistical tests for spatial variation. There was significant spatial variation within basins but not among basins for most solutes. Spatial variation within a snow pit was generally less than 10% of mean concentrations. Where a snow pit is located within a basin appears to be more important than which basin is sampled. Fewer replications within a snow pit and more snow pits within a basin will result in more representative data.

Aeolian dust is a major component of the chemical content of snowpacks in central Asia. The large amount of dust presents new challenges for interpretation of the glaciochemical record that may be unique to central Asia. To interpret correctly the glaciochemical record of the Tien Shan, it is importantto include post-depositional changes. It may also be important to evaluate post-depositional changes for snow and ice at higher elevations as well. Reference ColbeckColbeck (1989) has shown that solar heating can inducesub-surface melting in high-elevation snowpacks during periods when the surface is entirely at sub-freezing temperatures. Furthermore, small amounts of dust (c. 10 parts per million by weight) can increase absorption of incident solar radiation in a snowpack by several per cent (Reference WarrenWarren, 1982), increasing the free water content of the snowpack. This wet snow may provide liquid water for post-depositional alterations of the chemical content of the snowpack even in snowpacks that do not become isothermal at 0°C. Reconstruction of past climates and evaluation of regional deposition from glaciochemical records in central Asia must account for post-depositional changes in the solute content of snow and ice caused by aeolian deposition.

Acknowledgements

We thank the Lanzhou Institute of Glaciology and Geocryology for logistical support. Kelly Elder, Rick Kattelmann and Frank Setaro supplied field and laboratory assistance. Comments from two anonymous reviewers improved the manuscript. Funding was provided by a NASA EOS-IDS contract and an NSF pre-doctoral fellowship.