1. Introduction
The climatological significance of δ18O in precipitation from the tropics and subtropics has received much less attention than that of δl8O in precipitation from the polar regions. However, pioneering work by Reference Rozanski, Araguas-Araguas and GonfiantiniRozanski and others (1992), using the past three decades of δ180 of monthly precipitation and surface air temperature available through the International Atomic Energy Agency World-Meteorological Organization global network, indicated that the slope values of δ18O — Ta varied from 0.71 for high-latitude areas to virtually zero in the tropics where a strong relationship between δ18O and precipitation amount is usually observed (Reference DansgaardDansgaard, 1964; Reference Grootes., Stuiver, Thompson and Mosley-Thompson.Grootes and others, 1989). Alternately, a "precipitation-amount effect" for δ18O and δD was also observed both on Himalayan slopes and in the southern/ central region of the Qinghai Tibetan Plateau (Reference WushikiWushiki, 1977a; Reference Wake, Stiévenard., Mikami, Matsumoto, Ohta and SwedaWake and Stiévenard, 1995). However, recent work suggests that the δ18O values of precipitation samples collected on the northern Qinghai Tibetan Plateau during the period 1991-94 are positively correlated to Ta (Reference Tandong, Thompson, Mosley-Thompson, Zhihong, Xinping and Ping-Nan.Yao and others, 1996), with slopes of linear fit ranging from 0.29‰ to 0.67‰ oC-1.
On mountainous glaciers, there is evidence of seasonal snowmelt resulting directly from the large seasonal temperature range and the intense summer radiation, and a significant part of the annual accumulation of snow is removed by melting and sublimation (e.g. Dunde ice cap in the Qilian Shan, China; Reference Thompson, Xiaoling, Mosley-Thompson and ZichuThompson and others, 1988). Thus the seasonal δ18O oscillations in precipitation are rapidly reduced during recrystallization, in the presence of percolating melt water (Reference ArnasonArnason, 1969). Simultaneously, isotopic fractionation occurs, leaving the solid phase enriched relative to the liquid phase (Reference ArnasonArnason, 1969; Reference BúasonBúason, 1972). Reference Grabczak, Niewodniczanski and Rózanski.Grabczak and others (1983), using the δ18O and δD profiles from two crevasses on temperate glaciers high in the Andes and in the Himalaya, suggested, however, that the short-and long-term isotopic variations can still be observed in the deeper parts of such glaciers.
Despite the strong temporal relationship between δ18O in precipitation and Ta on the northern Qinghai Tibetan Plateau, 30 year records of annually averaged δ18O from three different ice-coring sites in this region are not correlated significantly with the contemporaneous air-temperature records from their corresponding closest meteorological stations (150— 200km away) (Reference Tandong, Thompson, Mosley-Thompson, Zhihong, Xinping and Ping-Nan.Yao and others, 1996). Seasonal distribution of the precipitation, post-depositional modification of the original isotopic signals in precipitation, and elevation differences between the ice-coring sites and their corresponding closest meteorological stations are known to affect the δ18O — Ta relationship. Several ice cores recently drilled from the Qinghai-Tibetan Plateau for paleoclimatic reconstruction (Reference ThompsonThompson and others, 1989,Reference Thompson1997) necessitate a better understanding of the δ18O — Ta , relationships in precipitation, as well as in glacier ice in this region.
From the beginning of june 1995 to the end of june 1996, precipitation sampling was conducted at the Daxigou meteorological station (DMS; Fig. 1) at the head of the Ürümqi river, Tien Shan, northwest China, at 3545 m a.s.l. From 11 May to 16 June 1996, a series of snow pits (T in Fig. 1) were also sampled at the nearby Ürümqi glacier No. 1 (1.84 km2 in area) at 4040 m a.s.l. During May 1996, two ice cores 2 m apart were also recovered at the snow-pit sampling site (TS-1 and TS-2 in Fig. 1). These data provide a special opportunity to test the δ18O—Ta relationships in precipitation in the mid-latitude mountainous region, to identify the effect of the post-depositional processes on the ice-core δ18 O records, and to compare the ice-core δ18 O records to the contemporaneous annual surface temperature recorded at the nearby DMS which has been in continuous operation since 1959.
Fig. 1. Sampling sites: DMS, T and TS stand for the Daxigou meteorological station (also the precipitation sampling site), the snow-pit sampling site and the ice-core drilling site, respectively. Inset map shows the location of the Tien Shan in relation to geographic and political features of northwest China.
2. Methodology
The study site (43.05° N, 86.49° E) is surrounded by large deserts on the north, south and east sides. The nearest sea is located > 3000 km away, so the region is dominated by classic inland climate conditions. The weather conditions in the study area are influenced by the obstacle effect of the Qinghai-Tibetan Plateau. Since the jet stream maintains its west-east orientation along the Tien Shan around the northern part of the Qinghai-Tibetan Plateau (Reference Reiter.Reiter, 1981), it carries the moisture originating in the Atlantic Ocean and/or the Mediterranean Sea to the study site (Li and Xu, 1984). The annual average surface air temperature and precipitation at the DMS are -5.3°C and 440.6 mm w.e., respectively, for the period 1959-96 (unpublished data).
Precipitation (snowfall and rainfall) samples were collected for each precipitation event during the observation period. To eliminate contamination and sample carryover, identical sampling procedures were used throughout the study (Reference Tandong, Thompson, Mosley-Thompson, Zhihong, Xinping and Ping-Nan.Yao and others, 1996). Plastic containers used for collection were dried (if necessary) and cleaned with a brush between samples. After collection, each sample was placed in a clean plastic bag, melted at about 20°C and poured into pre-cleaned high-density polyethylene bottles, and the tops were sealed in wax to avoid evaporation or diffusion. Bottled samples were transported to the Laboratory of Ice Core and Cold Regions Environment, Chinese Academy of Sciences, and kept in a cold room at -20°C until δ18 O analysis was performed using a Finnigan MAT-252 Spectrometer (precision 0.05‰). Relevant meteorological conditions such as air temperature at the beginning and end of each precipitation event, cloud type, wind speed, wind direction and humidity were recorded for each precipitation sample.
The first snow pit was dug on 11 May 1996, when little ablation of the winter snow had occurred. It was partially refilled after sampling, and sampling on successive days involved digging it out and refacing the sampling surface by at least 30 cm. The same strata were resampled on each occasion after allowing for accumulation or ablation. Snow samples were transferred into pre-cleaned polypropylene bags with plastic scoops for further processing as discussed above.
The ice-core samples were processed in the field by scraping with a clean stainless knife to obtain a contamination-free center sample. Table tops and tools were modified or covered with plastics. The TS-1 and TS-2 ice cores were cut into disks at 3 and 5 cm intervals, respectively. Samples were then transferred into polyethylene bags for further processing as discussed above.
3. Results and Discussion
3.1. The positive δ18O-Ta relationship in precipitation
The δ18O values of the precipitation samples show a broad variation; the minimum value of -38.24‰ was obtained on 24 January 1996, and the maximum value of 0.97‰ on 3 August 1995. The arithmetic average value of δ18 O for the samples collected during July and August 1995 is -7.13‰. From 9 July to 17 August 1981, precipitation samples collected at the same sampling site had δ18 O values ranging from -16.0‰ to -1.0‰, with an arithmetic mean average of -7.15‰ (Reference Watanabe, Wu, Ikegami and AgetaWatanabe and others, 1983). No significant difference is found between these arithmetic averages.
Following Reference Tandong, Thompson, Mosley-Thompson, Zhihong, Xinping and Ping-Nan.Yao and others (1996), we plot the δ18O and contemporaneous air temperature for individual precipitation events during the period of observation in Figure 2a; the linear relationship between δ18 O and Ta for all individual precipitation events in Figure 2b, and the linear relationship between δ18O and Ta for individual summer precipitation events (May-October) in Figure 2c. These plots validate the normal assumption that δ18O is more negative when Ta is cooler and less negative (even positive) when Ta is warmer. The slope of the δ18O—Ta relationship for all precipitation events is 0.95‰°C-1, which is higher than that reported in other mid-latitude to tropical regions (Reference Rozanski, Araguas-Araguas and GonfiantiniRozanski and others, 1992; Reference Tandong, Thompson, Mosley-Thompson, Zhihong, Xinping and Ping-Nan.Yao and others, 1996) and in polar regions (Reference DansgaardDansgaard and others, 1964; Reference Jouzel and MerlivatJouzel and Merlivat, 1984; Reference Mosley-Thompson, Thompson, Grootes and GundestrupMosley-Thompson and others, 1990; Reference Peel, Bradley and JonesPeel, 1992). Since in the Tien Shan, ~90% of the precipitation occurs during summer, we exclude δ18O values of winter precipitation samples and plot the rest in Figure 2c. The strong relationship between δ18O and Ta is still apparent, albeit with a lower slope (0.92‰°C-1). These results support the suggestion by Reference Rozanski, Araguas-Araguas and GonfiantiniRozanski and others (1992) that the isotopic composition of precipitation seems to be more sensitive to temperature fluctuations in mountainous regions than in low-altitude areas.
Fig. 2. (a) The relationship between δ18O and contemporaneous air temperature (Ta) for individual precipitation events.(b) The linear relationship between δ18O and Ta for all individual precipitation events. The correlation is significant at p = 0.001. (c) The linear relationship between δ18O and Ta for summer individual precipitation events. The correlation is significant at p = 0.001.
Although the linear correlation of δ18O — Ta for the individual precipitation events (Fig. 2) is positive, variations in δ18O do exist that are not attributable to changes in Ta . Other factors, such as the different vapor sources (not significant in our study site), different transport patterns of vapor in the atmosphere, average "rain-out history" of the air mass, and differing temperature structures which control the condensation temperature (Reference Gedzelman and LawrenceGedzelman and Lawrence, 1982; Reference Covey and HaagensonCovey and Haagenson, 1984; can affect the δ18O — Ta relationships to different degrees (Reference Gedzelman and LawrenceGedzelman and Lawrence, 1982; Reference Covey and HaagensonCovey and Haagenson, 1984; Reference Fisher and AltFisher and Alt, 1985; Reference Rozanski, Araguas-Araguas and GonfiantiniRozanski and others, 1992). Moreover, in the interior of continents, e.g. at our precipitation sampling site, the re-evaporated moisture plays an important role in the atmospheric water balance (Reference Ingraham and TaylorIngraham and Taylor, 1986). Another potentially important influence on the δ18O — Ta relationship is the amount effect (Reference DansgaardDansgaard, 1964; Reference WushikiWushiki, 1977a; Reference Wake, Stiévenard., Mikami, Matsumoto, Ohta and SwedaWake and Stiévenard,1995), which is more common in tropical regions and less important in mid-latitudes (Reference Grootes., Stuiver, Thompson and Mosley-Thompson.Grootes and others, 1989; Reference Rozanski, Araguas-Araguas and GonfiantiniRozanski and others, 1992).
As suggested by Reference Jouzel, Russell, Suozzo, Koster, White and BroeckerJouzel and others (1987), a stronger δ18O — Ta relationship can be expected when individual precipitation events are aggregated into monthly averages, which minimizes the influence of synoptic-scale differences. Figure 3 shows the arithmetic monthly-averaged δ18O plotted against the monthly-averaged air temperature. The slopes of linear fit are 1.23‰° C-1 for all months (Fig. 3a) and 1.19‰°C-1 for summer months (Fig. 3b). The coefficients ( R) for the δ18O — Ta relationship of monthly averages are much higher than the R values for the individual precipitation events. Undoubtedly, the substantial reduction in the number of data points (degrees of freedom) results in higher errors for the slopes. As suggested by Reference Rozanski, Araguas-Araguas and GonfiantiniRozanski and others (1992). the long-term δ18O — Tn relationship for a given location is believed to be the most appropriate for paleoclimatic reconstruction, because it covers a wide range of different "climate".
Fig. 3. The linear relationship between monthly-averaged δ18O and monthly-averaged air temperature Ta ; (a) annual; (b) summer. Both are significant at p = 0.001. No precipitation samples were collected during November-December 1995 and February-March 1996.
3.2. The ice core δ18O — Ta relationship
The ice cores were dated using a variety of different seasonal indicators (e.g. stable isotopes, major anions and cations) from comparison between the TS-1 and TS-2 ice-core records, and comparison with the mass-balance data of the Ürúmqi glacier No.l (Chinese Academy of Sciences, 1982-95). The final dating results are 22 years for the TS-1 ice core and 24 years for the TS-2 ice core (see Fig. 4).
Fig. 4. δ18O profiles for the TS-l and TS-2 ice cores, compared with corresponding annual and summer surface air temperature as measured at the DMS. The four coarse solid lines show the smoothing trend using Gaussian weighting coefficients, approximately equal to 5year moving average, and the gray dashed lines indicate annual ice layers.
By tritium analysis of the ice samples from a crevasse wall at the head of Khumbu Glacier, Mount Everest, Reference Miller, Leventhal and LibbyMiller and others (1965) concluded that two strata are deposited in a single accumulation year. One segment represents accumulation during the summer monsoon, and the other accumulation during the months of prevailing winter storms. Reference WushikiWushiki (1977b) also suggested such a conclusion based on the deuterium-content profile of ice cliffs on Kongma Glacier, Khumbu. This stratigraphic characteristic is also apparent in our δ18O profiles of the TS-1 ice core (Fig. 4), which is supposed to be formed at the beginning and end of each ablation season, when the firnification process is controlled by the thaw-freeze cycles.
Figure 4 shows the δ18O profiles of the TS-1 and TS-2 ice cores, together with the annual and summer air temperature recorded at the nearby DMS. The δ18O values of the ice-core samples range from - 12.57‰ to -7.74‰, with an arthimetic average of -10.1 ‰. Compared to the precipitation samples, the δ18O amplitude in ice cores decreases significantly. However, the distinct annual to seasonal variability of δ18O values can still be preserved in a fixed annual ice layer. Factors that may contribute to the preservation of the depositional variability are the presence of numerous ice layers in the snowpack, acting as physical obstacles against meltwater percolation, and thick superimposed ice layers that may prevent further smoothing of the δ18O seasonal signal in underlying ice layers.
Although the TS-1 and TS-2 ice cores were cut at different intervals, similar δ18O profile features, especially for the smoothing lines, are apparent. The ice-core δ18O smoothing lines also follow the air-temperature records, particularly for the summer (May-October). As ~90% of the precipitation falls in the summer-half of the year, and δ18O is only recorded during precipitation, it is reasonable to expect that the δ18O from an annual ice-core layer records the summer average air temperature of the corresponding year. For further δ18O — Ta comparison, we calculate the mean annual δ18O value for each year by averaging all sample δ18O values between two adjacent annual δ18O peaks, and plot the annually averaged δ18O against the contemporaneous annual, summer and winter surface temperature (Fig. 5). Apparently, both the TS-1 and TS-2 ice-core δ18O records might provide a proxy of the summer air temperature, because the linear slopes against the summer temperature are not only higher than that against the annual or winter temperature, but also very similar (0.38 For the TS-1 ice core, and 0.40 for the TS-2 ice core). We notice that the significance of the regression coefficients is far from satisfactory. However, they at least show some progress compared to the results of Reference Tandong, Thompson, Mosley-Thompson, Zhihong, Xinping and Ping-Nan.Yao and others (1996) where the R2 for annually averaged δ18O from the ice cores vs surface air temperatures at the closest meteorological station is only in the range 0.000-0.011. Reference ThompsonThompson and others (1993) demonstrated that the 5 year running mean of the annual δ18O averages from the Dunde ice cap (one of Yao and others' three ice cores) was strongly correlated (R2 = 0.25, significance 99.9%) with the 5 year running mean of the Northern Hemisphere annual temperatures from 1895 to 1985. However, such work is not expected here due to the limitation of our short ice-core records. We propose to drill another relatively long ice core from Ürúmqi glacier No. 1, and wish to press this question further.
Fig. 5. The linear relationship between annually averaged δ18O and contemporaneous annual, summer and winter surface temperature.
3.3. Depositional and post-depositional modification of ice-core δ18O records
The relationship between the ice-core δ18O records and contemporaneous air temperature is less significant than that for the precipitation samples due to depositional and post-depositional processes.
As suggested by Reference Tandong, Thompson, Mosley-Thompson, Zhihong, Xinping and Ping-Nan.Yao and others (1996), the air-temperature data reflect an equal weighting of monthly temperatures, while the ice-core δ18O record is skewed toward wet-season precipitation. It is unrealistic to expect the δ18O from an annual layer in ice cores to record the annual average air temperature of the corresponding year. This fact may account for the different linear-fit slopes and correlation significance between contemporaneous annual, summer and winter surface temperature records and the annually averaged ice-core δ18O records.
The isotopic signal in snowpack may be further modified by post-depositional processes. In polar snow and firn the isotopic homogenization connects to recrystallization of the grains via the vapor phase, e.g. storms and barometric-pressure changes cause vertical air movements, particularly in the upper firn, where mass exchange between the strata is further accentuated by high temperature gradients (Reference Dansgaard, Johnsen, Clausen and GundestrupDansgaard and others, 1973). Such processes can, however, also happen in mountainous glacier snow and firn, especially during the winter-half of the year. Diffusion in the vapor phase also causes considerable interstratificial mass exchange (Reference Dansgaard, Johnsen, Clausen and GundestrupDansgaard and others, 1973). In low-accumulation areas, the seasonal δ18O oscillations are simply missing due to redistribution by snowdrifting or lack of winter (summer). snow, which introduces small-scale and local depositional noise (Reference Fisher, Koerner, Patersun, Dansgaard, Gundestrup and ReehFisher and others, 1983). As shown in Figures 6 and 7, the Tl snow pit was sampled at the ice-core drilling site on 11 May 1996 when no snowmelt was observed in the snow-pit stratigraphic profile, and the minimum δ18O value in the snow pit is -23.29‰, which is much higher than the δ18 O values observed in the winter precipitation samples (e.g. -35.85‰ on 10 January 1996, and -38.24‰ on 24January 1996). Therefore, the effect of wind-scouring has also been observed on an ice cap, Ellesmere Island, Canada, where the annual mean δ18O at an ice divide was 2.5‰ less negative than 1.2 km downslope ( Reference Fisher, Koerner, Patersun, Dansgaard, Gundestrup and ReehFisher and others, 1983).
Fig. 6. The successive stratigraphic profiles of snow pits T1 to T5 that were collected during the early-summer melt period.
Fig. 7. The successive δ18O profiles of snow pits T1 to T5 that were collected during the early-summer melt period.
The effect of snowmelt-percolation processes on the ice-core δ18O records may be studied using a set of successive snow-pit stratigraphic and δ18O profiles (Figs 6 and 7). Because the surface of the snow pits changed with time, we adjust the bottoms of all the snow pits to the same 95 cm depth. The Tl snow pit was sampled at intervals of exactly 10 cm. Each δ18O value presents the average of its corresponding depth range, which covers different stratigraphic snow and firn layers. The minimum δ18O value of-23.29‰ is present in the middle of the Tl snow pit, which was accumulated during the winter season. The relatively high δ18O values in the upper and bottom layers correspond to the accumulation deposited during the spring and previous autumn seasons. It is apparent that the δ18O profile of the Tl snow pit resembles the temperature history during the accumulation period.
The T2 snow pit was dug 3 days later and sampled at its stratigraphic layers. The measured snowpack temperatures were below-2.5° C. Disregarding the upper 15 cm new snow layer, the δ18O values of the T2 snow pit swing around those of the Tl snow pit. During the period of the T2 and T3 sampling dates, no precipitation event occurred. Though the measured temperatures in the upper layers of the T3 snow pit exceed 0° C at noon, the snowpack below 70 cm remained negative, and several percolation ice layers (or ice lenses) had formed. However, the δ18O profile of the T3 snow pit preserved the original isotopic characteristics.
Thick superimposed ice layers formed at the bottoms of T4 and T5 snow pits, and the δ18O amplitudes decreased dramatically compared with that of the former snow pits, especially in the superimposed ice layers where the δ18O values range only from -14.84‰ to 12.57‰. These values are close to the smallest δ18O (-12.57‰) as observed in the ice cores. Because of the rapid formation of the superimposed ice layer, snowmelt cannot penetrate into the underlying annual ice layers. Thus the modification of the snowpack δ18O variations occurs only within a certain annual layer.
Reference KoernerKoerner (1997) concluded that early summer melt will cause melting of the very negative (cold) — δ winter/spring snow which lies near the surface. The meltwater then percolates down to refreeze within, or at the base of, the current annual snowpack. Further melt may cause runoff of less negative (warmer) −δ snow deposited during the early winter/fall period. A very negative (cold) −δ snowpack remains (Fig. 7). However, the remaining very negative (cold) −δ snowpack does not necessarily correspond to the winter precipitation with minimum δ values, but rather to a mixture of snow deposited during the winter half of the year. In addition, similar melting and percolation processes can be expected to happen year by year. Isotopic enrichment by partial melting of the snow cover can be roughly estimated by assuming a Rayleigh-type removal of meltwater from the system (Reference Moser, Stichler, Fritz and FonteMoser and Stichler, 1980). Nevertheless, such isotopic enrichment should influence the ice-core δ18O — Ta records, in such a way that the final isotopic composition of the glacier ice is shifted in the same direction (Reference Grabczak, Niewodniczanski and Rózanski.Grabczak and others, 1983), or even enhance the climatic significance of ice-core δ18O records, because more enrichment in heavy isotopic content accompanies higher ablation, that is, generally due to higher air temperatures. This guarantees the climatic significance of ice-core δ18O records from areas of very heavy melt, especially on a long time-scale.
4. Conclusions
Our data provide the first observations of the relationship between the contemporaneous temperature and δ18O in precipitation and ice-core samples from the Tien Shan. Although post-depositional processes certainly decrease the significance or the relationship between δ18O of the precipitation samples and contemporaneous temperature, a positive δ18O — Ta relationship is still observed in ice cores collected from the nearby Ürümqi glacier No. 1. The presence of numerous ice layers in the snowpack and rapid formation of the superimposed ice layer guarantee preservation of the seasonal δ18O variation in ice cores despite the dramatic decrease in amplitude. Other factors, such as the seasonal distribution of the annual precipitation, isotope diffusion in the snow and firn layers, snowpack redistribution caused by snowdrifting, and the heavy-isotope-content enrichment caused by the partial melting of the snow and firn layers, have a potential but perhaps less important influence on the ice-core δ18O record. Nevertheless, the climatic significance of mountainous ice-core δ18O records can be preserved in temperate glacier ice, which might demonstrate that the ice-core climate records can be obtained from less than ideal environments, thus expanding the number of sites in the world where such climatic records could be collected.
Acknowledgements
This research was supported by the State Committee of Science and Technology of China (Climbing Project 95-YU-40), the Chinese Academy of Sciences (CAS) (the Great Project of CAS KZ-95I-A1-402,204 and 205) and the National Natural Science Foundation of China (49571020, 49871022), and is part of the cooperative project between Lanzhou Institute of Glaciology and Geocryology, CAS and the Climate Change Research Center, University of New Hampshire. We thank C. P. Wake, J. E. Dibb and S. Whitlow for their valuable suggestions. The manuscript was improved by comments from J. Jouzel and an anonymous reviewer.
