Journal of Geophysical Research: Atmospheres

Recent rapid warming trend revealed from the isotopic record in Muztagata ice core, eastern Pamirs


  • Lide Tian,

    1. Laboratory of Land Surface Processes Monitoring and Research, Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing, China
    2. Also at Key Laboratory of Cryosphere and Environment, Cold and Arid Regions Environmental and Engineering Research Institute, Chinese Academy of Sciences, Lanzhou, China.
    Search for more papers by this author
  • Tandong Yao,

    1. Laboratory of Land Surface Processes Monitoring and Research, Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing, China
    2. Also at Key Laboratory of Cryosphere and Environment, Cold and Arid Regions Environmental and Engineering Research Institute, Chinese Academy of Sciences, Lanzhou, China.
    Search for more papers by this author
  • Zhen Li,

    1. Key Laboratory of Cryosphere and Environment, Cold and Arid Regions Environmental and Engineering Research Institute, Chinese Academy of Sciences, Lanzhou, China
    Search for more papers by this author
  • Kenneth MacClune,

    1. Institute of Arctic and Alpine Research, University of Colorado, Boulder, Colorado, USA
    Search for more papers by this author
  • Guangjian Wu,

    1. Laboratory of Land Surface Processes Monitoring and Research, Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing, China
    Search for more papers by this author
  • Baiqing Xu,

    1. Laboratory of Land Surface Processes Monitoring and Research, Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing, China
    2. Also at Key Laboratory of Cryosphere and Environment, Cold and Arid Regions Environmental and Engineering Research Institute, Chinese Academy of Sciences, Lanzhou, China.
    Search for more papers by this author
  • Yuefang Li,

    1. Key Laboratory of Cryosphere and Environment, Cold and Arid Regions Environmental and Engineering Research Institute, Chinese Academy of Sciences, Lanzhou, China
    Search for more papers by this author
  • Anxian Lu,

    1. Key Laboratory of Cryosphere and Environment, Cold and Arid Regions Environmental and Engineering Research Institute, Chinese Academy of Sciences, Lanzhou, China
    Search for more papers by this author
  • Yongping Shen

    1. Key Laboratory of Cryosphere and Environment, Cold and Arid Regions Environmental and Engineering Research Institute, Chinese Academy of Sciences, Lanzhou, China
    Search for more papers by this author


[1] Many have made efforts to clarify the climatic significance of stable isotopic variations in ice cores around central Asia through the study of stable isotopes in present-day precipitation. A new shallow ice core from Muztagata, in the eastern Pamirs, allows for a detailed comparison of annual δ18O variation with local meteorological data as well as with global air temperature variations. On the basis of a comparison of seasonal fluctuations of δ18O in the local precipitation, the 41.6-m ice core drilled at 7010 m provides a record of about one-half century. The annual fluctuations of δ18O in this ice core are in good agreement (correlation coefficient of 0.67) with the annual air temperature changes at the nearby meteorological station Taxkorgen, indicating that the isotopic record from this ice core is a reliable temperature trend indicator. The most important discovery from the δ18O variation of this ice core is a rapid warming trend in the 1990s, which is consistent with a general global warming trend over this time period. This recent rapid warming at higher elevations in this area has led to the quick retreat of alpine glaciers.

1. Introduction

[2] Stable isotopes in ice cores serve as an important tool for paleoclimate reconstruction. This is based on the discovery of the close relationship between isotopic components in precipitation and concurrent air temperature in high-latitude regions [Dansgaard, 1964; Rozanski et al., 1992; Jouzel et al., 1997; White et al., 1997]. Many have done work on reconstructing long-term paleoclimate from ice core isotope records on the Tibetan Plateau, such as Guliya ice core, Dunde ice core, and Dasuopu ice core [Thompson et al., 1989, 1997, 2000; Yao et al., 1992, 1997, 2002]; however, the detailed analysis of the annual stable isotope variation has been limited. Yet a detailed comparison of Greenland ice core δ18O records and annual air temperature from meteorological stations shows that annually resolved isotopic data from Greenland ice cores contain useful climatic information [White et al., 1997]. These kinds of comparisons are still scarce in the central Asia region due to the different seasonal moisture sources in this monsoon influenced area and the weak stratigraphy resulting from low winter precipitation and summer glacier surface melting. However, for the selection of ideal ice core drilling sites and for the explanation of the variation in stable isotopes in the resulting ice cores it is necessary to study the stable isotopes in present precipitation and to correlate the annual isotope record from an ice core to annual meteorological parameters.

[3] The high-resolution isotopic record from a recently drilled ice core taken from the 7010-m level of Muztagata Peak in the eastern Pamirs makes it possible to associate the annual isotopic record to local annual air temperature variation.

2. Research Area and Ice Core Drilling

[4] Muztagata Peak, which is shaped like a big dome, is located in the eastern Pamirs, West China (Figure 1) and has a summit altitude of 7546 m. On the west, above 5500 m, the mountain is covered by a glacier, and in the valleys the glaciers flow as low as to 4300 m. In the summer of 2003, ice cores were extracted at 7010 m a.s.l. on a rather gentle slope of the Muztagata glacier (75°06′E, 38°17′N). These ice cores were transported to a low-temperature laboratory in frozen condition.

Figure 1.

Location of the Muztagata Mountain and the ice core drilling site.

[5] An ice core 41.6-m in depth was cut to 819 snow and ice samples in 5-cm intervals. These samples were melted at room temperature before analyzing. Oxygen isotopes were measured for all the samples using MAT-252 and Delta-Plus mass spectrometers. The precision of the δ18O measurement was better than ±0.2‰.

[6] Figure 2 provides the δ18O profile as well as the density of ice and ice temperature with depth. The δ18O in this ice core shows a large span of over 20‰ from −26.72‰ at 33.71 m depth to −4.95‰ for snow near the surface. The average was −16.55‰. The seasonal variations of δ18O are apparent throughout the profile. Note that the annual layers become thinner below 35 m where the ice density is its greatest. This is due to the ice layer being near the bottom of the glacier (the depth of the glacier is 52.6 m). The gradual increase of ice density due to the densification of firn layers shows that the glacier has not overturned at the drilling site. The extremely low air temperature at over 7000 m limits the possible melting of surface snow. Recurring strong winds on the glacier do remove part of the fresh snow; however, the in situ observation shows that a hard layer forms on the snow surface with these winds which prevents further erosion of the fresh snow surface. Clear stratigraphy had been found in a 2-m-depth snow pit at the drilling site. In addition, four δ18O profiles along the glacier at different altitudes also show that the seasonal variation of δ18O was well preserved in the snow layers of the Muztagata glacier (Figure 3).

Figure 2.

The δ18O of Muztagata ice core with depth as well as the firn density and ice temperature profiles.

Figure 3.

The δ18O variations from snow pits dug at different altitudes of the Muztagata glacier. Surface snow at 6800 m had suffered from apparent wind erosion. Snowmelting was found at the bottom of the 6300 m snow pit. Nevertheless, seasonal variations of δ18O were clearly recorded in the four snow pits.

3. Regional δ18O Variations and Ice Core Dating

[7] In this paper, the variations in δ18O of the ice core will be discussed with an emphasis on local and global climate change. The spatial and temporal variations of δ18O in precipitation and affected parameters will be reviewed before discussing the ice core δ18O variations.

[8] The primary control on the spatial distribution of the δ18O of precipitation in the northern part of Southeast Asia (approximately 35°N) is air temperature [Arguas-Arguas et al., 1998]. Ice cores taken from the southern Tibetan Plateau do not have good correlation between annual air temperature and variations in δ18O primarily due to the influence of the Indian monsoon on the isotopes of the precipitation [Tian et al., 2001, 2003]. In contrast, the Muztagata glacier is much farther north and is not influenced by the Indian monsoon. Figure 1 illustrates the locations of stations where there are observations of stable isotopes in precipitation around the Muztagata ice core drilling site, and Table 1 lists the observational correlation between δ18O of precipitation and air temperature. The results from Hetian, Urumqi, Kabul, and Delingha are based on the monthly averaged δ18O and monthly averaged air temperature. Measurements from the Urumqi river basin were made over a 1-year observation from the summer of 1996 [Yao et al., 1999]. The results from Altay are based on the observation of individual precipitation events over 1 year from June 2000 to July 2001. The positive correlations listed in Table 1 indicate that the δ18O values in the research area are related to temperature changes, with high δ18O values found in summer precipitation and low δ18O values found in winter precipitation. This seasonality provides a basis for dating the Muztagata ice core.

Table 1. Temperature Dependence of δ18O of Precipitation on Temperature at Stations Surrounding Muztagata Ice Core Drilling Site
StationsAltitude, mRegression ResultNote
Hetian1375δ18O = 0.68T − 15.35 (r = 0.86)GNIP
Urumqi918δ18O = 0.45T − 15.97 (r = 0.86)GNIP
Kabul1860δ18O = 0.46T − 10.18 (r = 0.76)GNIP
Delingha2981δ18O = 0.66T − 14.27 (r = 0.88)Tian et al. [2003]
Urumqi river basin (Daxigou)4200δ18O = 0.73T − 11.97 (r = 0.95)Yao et al. [1999]
Altay737δ18O = 0.62 T − 15.75 (r = 0.72) 

[9] Using the seasonal fluctuations of δ18O along the profile, the 41.6-m ice core was dated to 48 years before today (Figure 4). We must also remain alert to the processes that can lead to uncertainty and make age dating difficult, these include erosion and replacement of snow layers by wind, temporal fluctuation of δ18O even in the same season, smoothing by surface snow melting, and even reversal due to movement by the glacier. The very high elevation and low temperatures prevent surface snow melting year-round. No interruption was found in this ice core, which is confirmed by the gradual increase in snow density. In situ observation found that strong winds could blow away some part of fresh snow, which could possibly remove a part of annual layers, but there was no topographical evidence of overturning of snow layers due to wind. Sublimation will also have some effects on the limited surface firn, but the modification of the isotopic composition of firm by sublimation has little influence on the quality of the ice core record [Stichler et al., 2001].

Figure 4.

Muztagata ice core dating result based on seasonal variations of δ18O.

[10] Thus we cannot absolutely exclude the possibility of the loss of annual layers by strong wind erosion. An example of a possible wind erosion event is the slight drop in δ18O at 32.6m depth in an otherwise broad peak (Figure 4). However, the depth over which the larger peak was measured, over 2 m, is far greater than would be expected for a single annual layer. Thus we used the slight drop as demarking the annual layer and divided it into the 2 years, 1968 and 1969. Also, intraseasonal δ18O variation can occur due to temperature variations, which can potentially make dating difficult. Such values can be cross referenced to temperature data from Taxkorgen. We see such fluctuations between 26.4 and 28 m depth with a slight drop at 26.8–27.4 m which we assign as the year 1973. We also assign the several fluctuations at 2.3–3.8 m depth as the year 1998, as there was a sharp decrease of air temperature in 1999 at Taxkorgen.

[11] The peak in β activity found between 36.85 m and 37.89 m of the ice core (shown in Figure 4) we identify as the 1963 peak associated with nuclear bomb testing. This peak is widely found in northern hemispheric ice cores [Clausen and Hammer, 1988]. We estimate that the dating uncertainty is within 1 year for the near past half century.

[12] We will not discuss the annual accumulation variations here because we do not know the change of the rate of erosion of snow to actual annual snowfall. Despite the increase of firn density with depth, there is no corresponding increase in the number of years per meter of ice up to 35 m in depth. This is possibly due to a decreasing rate of snowfall or increasing of snow erosion by wind or a combination of these factors in recent years. Below 35 m, the annual layers become much thinner, partly due to the firn densification process but primarily due to the proximity to bedrock.

4. The δ18O Variations in Muztagata Ice Core and Recent Rapid Warming

[13] The annual variation of δ18O from 1955 to 2002 as reconstructed from the Muztagata ice core is presented in Figure 5. As a comparison, the annual air temperature variation is also given at Taxkorgen Meteorological Station (75°14′E, 37°47′N, 3100 m), the nearest meteorological station to the ice core drilling site. The annual fluctuations of δ18O in the ice core record are highly correlated with the instrumentally measured annual air temperature at the meteorological station, despite the very different elevations of 3100 m versus 7000 m, and despite the sporadic precipitation events versus the continuous air temperature measurement. The δ18O captured the cooling event in 1986, which resulted from an extremely low average January air temperature of −24.2°C (the average January air temperature is −11.8°C). This cold event is also widely recorded at other local meteorological stations. This cold period corresponds to low δ18O values in the ice core record of −21.5‰ and an annual δ18O value of −18.8‰. The lowest annual δ18O record from this ice core is from 1965, with a value of −20.4‰. The annual air temperature at the meteorological station in 1965 is also extremely low. The slight discrepancy between the annual variations of δ18O in ice core and local air temperature can mostly be attributed to the difference in elevation, errors associated with comparing intermittent precipitation events to a continuous air temperature record, seasonal shifts in precipitation, and loss of snow layers on the glacier surface.

Figure 5.

The annual variations of δ18O in Muztagata ice core with the annual air temperature variations at Taxkorgen Meteorological station.

[14] A regression analysis between the δ18O of the Muztagata ice core and annual air temperature at Taxkorgen reveals a close relationship: δ18O = 1.50T + 22.00 (r = 0.67). The good agreement indicates that the δ18O variations in the ice core from 7010 m of Muztagata are correlated to the local air temperature change.

[15] The most important finding from the δ18O variation in Muztagata ice core is an increasing trend in the value of δ18O over the past decade resulting from a change in the average air temperature over that time period. The increase of δ18O in both summer and winter precipitation in recent decades, as shown in Figure 2, excludes the possibility that the increasing trend is caused by a shift from winter-dominated precipitation to summer-dominated precipitation.

[16] In order to discuss the amplitude of warming at this site, it is necessary to derive a δ18O-air temperature relationship curve. A suitable isotope/temperature relationship should be built to reconstruct the paleoclimate from the ice core record. Globally, there is a great deal of variability in the relationship of δ18O with temperature. It has been found that the slope of the δ18O-temperature relationship curve is substantially higher at high latitudes than that found at low latitudes, varying from a very low value (near zero in the tropics) to as high as 0.9‰/°C in Antarctica [Rozanski et al., 1993]. However, in the high latitudes of northern hemisphere, the slope of the δ18O-temperature relationship curve usually varies between 0.6 and 0.7‰/°C. The slope of 0.69‰/°C, derived from the data gathered from middle and high northern latitude coastal stations [Dansgaard, 1964], has been frequently used in isotope-aided paleoclimatic reconstructions. This relationship, derived from long-term annual measurements, yields a slope of 0.67‰/°C in southern and western Greenland [Johnsen et al., 1989]. Work done for all IAEA/WMO network stations produced a lower slope value of 0.58‰/°C [Rozanski et al., 1993]. However, there are some doubts as to whether a spatial relationship accurately reflects the conditions found at a single location such as an ice core [Jouzel et al., 1997]. Long-term changes in isotopic composition of precipitation over mid and high-latitude regions during the last 3 decades have closely followed the long-term changes in surface air temperature with a corresponding average δ18O-temperature ratio around 0.6‰/°C [Rozanski et al., 1992]. A nearly 10-year observation of δ18O of precipitation at Delingha, on the northern side of the Tibetan Plateau by Yao et al. [1996] has yielded a temporal slope of 0.66‰/°C [Tian et al., 2003], which is very close to the spatially derived slope for high-latitude regions [Dansgaard, 1964]. Thus we use the range of 0.6 to 0.7‰/°C δ18O-temperature relationship to transform the δ18O notation from the Muztagata ice core into degrees of Celsius (Figure 6a).

Figure 6.

A comparison of the air temperature anomalies derived from Muztagata ice core (a) with that of Taxkorgen (b) and the northern hemisphere (c). The shadowed area in 6a shows the possible span of temperature change derived from the δ18O/T slope of 0.6 to 0.7. The decadal variations of temperature anomaly are given as horizon lines.

[17] The annual warming trend in the isotope-derived temperature record from Muztagata ice core is apparent. For the annual fluctuation, the Muztagata ice core isotope record is consistent (with a correlation coefficient of 0.67) with the local meteorological station record (Figure 6b). The warming trend derived from the Muztagata ice core isotope record is also in step with the trend in the instrumental record of warming in the northern hemisphere (Jones et al. [1999], updated from (Figure 6c). Both records have common warming features. For instance, both records show the recent warming trend to have begun in the mid-1970s with an abrupt warming at the beginning of 1990s. The result of a regression analysis between the Muztagata ice core isotope record and the northern hemisphere air temperature anomalies gives a close relationship with a correlation coefficient of r = 0.75. This implies that the recent warming shown in the Muztagata ice core isotope record is probably a result of the global warming trend of the past decades.

[18] Looking at longer time periods, Thompson et al. [2003] had compared the millennial δ18O variations from three ice cores on the Tibetan Plateau and found all three δ18O ice histories show a consistent trend of enrichment since 1800, suggesting a large spatial-scale warming. Our more recent record from Muztagata reveals that this warming trend has persisted and has become more intense in recent decades. There are, however, exceptions, such as in Malan ice core in the center of the Tibetan Plateau, where δ18O shows no warming trend in the last 30 years [Wang et al., 2003]. The spatial difference among ice cores on the Tibetan Plateau, probably reveal the difference in the patterns of climate change in different parts of the plateau.

[19] For the decadal variations of temperature in the past 4 decades from 1960s to 1990s (Figure 6), the three records show the warming trend of the 1990s, although there are only slight differences for the 3 decades from of the 1960s to the 1980s.

[20] The warming magnitude derived from the Muztagata ice core record is much larger than that of the other two records shown in Figure 6. The regression results for Taxkorgen meteorological station show a warming of 0.18°C per decade and those for the northern hemisphere show a warming of 0.14°C per decade. However, the warming at Muztagata as derived from the ice core isotope record is around 2.0∼2.4°C per decade.

[21] One possible explanation for the relatively large magnitude of warming found at the Muztagata ice core site is that high elevation regions are more sensitive to a warming climate. A detailed statistical analysis of surface mean air temperature variations which looked at the dependency of warming on elevation in and near the Tibetan Plateau region found that the rate of warming increases with altitude. The study found that the warming rates at approximately 700 m, 2500 m, and 4300 m, a.s.l., are about 0.05, 0.15, and 0.25°C/decade, respectively [Liu and Chen, 2000]. The sharp warming is also found in ice core records from Wyoming in the United States from which an increase in air temperature of about 3.5°C since the middle of 1960s was reconstructed [Naftz et al., 2002]. Though the magnitude of the warming is substantial, it is not unheard of. The warming trend in some areas of northeastern North America is substantially greater than the global average with annual values of 2–3°/decade in the 1990s (NASA-GISS,

[22] Nevertheless, other processes, which are presently not well understood, add to the uncertainty of the magnitude of warming derived from the isotope derivations. For example, air temperatures at extremely high elevations can often vary from the standard lapse rate model to be quite different from those expected as measured at a lower elevation meteorology stations. Another example is that trends in precipitation at extremely high elevations might differ than those at lower elevations, although we found no apparent annual precipitation trend at the surrounding meteorological stations. These and other processes may affect the magnitude of the isotope derived air temperature. Though the magnitude of the warming is not well constrained, the pattern of warming is consistent with both local and northern hemisphere instrumental records as is the pattern of a larger degree of warming with altitude.

[23] Although instrumental air temperature data at extremely high elevations are scarce in the research region, several studies show that the recent warming at high elevation areas of central Asia results in a higher rate of melting and the rapid retreat of alpine glaciers. Pu et al. [2004] had studied Tibetan Plateau glacier movement during the last century and found the periodic movement of glaciers. From 1940 to 1960s, glaciers were in rapid retreat, while during 1970s–1980s, glaciers were stable or advancing slightly. Strong and rapid retreat has occurred since the end of 1980s with the retreat speed accelerating over time. Another example is the Number 1 Glacier at Urumqi River Basin of East Tianshan. This glacier is also in rapid retreat. In the 42 years of observation from 1959 to 2000, the negative mass balance of the Number 1 Glacier has been enhanced. Eighty-five percent of glacial mass loss has occurred in the last 20 years while 51% of total loss has occurred in just the last 10 years [Jiao et al., 2004]. Although it varies with region, the retreat of glaciers in central Asia has been greatest in the past 10 years [Yao et al., 2004]. The accelerated retreat of high alpine glaciers in this region is consistent with the isotopic variation trend in Muztagata ice core.

5. Conclusions

[24] A 41.6m ice core drilled at 7010m of Muztagata Mountain, eastern Pamirs in the summer of 2003, provides a high-resolution temperature record. The stratigraphy of this ice core is well preserved by the extremely cold environment. The ice core was dated by the clear seasonal variation of δ18O to a record of nearly a half century.

[25] The detailed annual δ18O in ice core record allowed us to compare it with the local meteorological station air temperature data. The annual variation of δ18O in this ice core is consistent with the local air temperature record from the Taxkorgen meteorological station. The annual correlation regression between the δ18O in the ice core and air temperature at Taxkorgen gives a relation of δ18O = 1.50T + 22.00, with a correlation of R2 = 0.45. This result indicates that a δ18O record from ice cores taken at high-elevation locations, such as Muztagata, can provide a reliable high-resolution air temperature record.

[26] The most apparent feature of this ice core record is the rapid warming during the 1990s as revealed by the abrupt increase in the value of δ18O. The annual fluctuations of temperature derived from the Muztagata ice core isotope record are in step with the local meteorological annual temperature variations and the instrumental record of northern hemisphere temperature, revealing that the local warming trend is related to the global warming trend of recent years. The magnitude of warming at Muztagata, however, is much larger then that revealed by the other two records. The regression result shows that the decadal warming trend is around 2.0∼2.4°C per decade from the decadally averaged temperature at Muztagata, while only 0.18°C per decade for Taxkorgen meteorological station and 0.14°C per decade for the northern hemisphere. A possible reason for this might be that the warming trend is greater at higher elevation, though other nontemperature related processes are also a potential reason. In order to substantiate the magnitude of temperature change, continuous direct observations of air temperature at this elevation are necessary. We now make these efforts as part of our ongoing program to monitor the air temperature change at high glacier surfaces.

[27] The recent rapid warming in the Tianshan mountains has led to the rapid retreat of glaciers. This retreat intensified in the 1980s and 1990s, which is consistent with the Muztagata ice core warming trend.


[28] This work is supported by the National Basic Research Program of China (grant 2005CB422002), the Innovation Program of Chinese Academy of Sciences (KZCX3-SW-339), and National Natural Science Foundation of China (grant 40121101 and grant 40571039). We also thank Sun Wenzhen and Wang Yu for sample analysis in the laboratory.