Interactions between mass balance, atmospheric circulation, and recent climate change on the Djankuat Glacier, Caucasus Mountains, Russia

Authors


Abstract

[1] This paper reports recent changes in the mass balance record from the Djankuat Glacier, central greater Caucasus, Russia, and investigates possible relationships between the components of mass balance, local climate, and distant atmospheric forcing. The results clearly show that a strong warming signal has emerged in the central greater Caucasus, particularly since the 1993/1994 mass balance year, and this has led to a significant increase in the summer ablation of Djankuat. At the same time, there has been no compensating consistent increase in winter precipitation and accumulation leading to the strong net loss of mass and increase in glacier runoff. Interannual variability in ablation and accumulation is partly associated with certain major patterns of Northern Hemisphere climatic variability. The positive phase of the North Pacific (NP) teleconnection pattern forces negative geopotential height and temperature anomalies over the Caucasus in summer and results in reduced summer melt, such as in the early 1990s, when positive NP extremes resulted in a temporary decline in ablation rates. The positive phase of the NP is related to El Niño-Southern Oscillation, and it is possible that a teleconnection between the tropical Pacific sea surface temperatures and summer air temperatures in the Caucasus is bridged through the NP pattern. More recently, the NP pattern was predominantly negative, and this distant moderating forcing on summer ablation in the Caucasus was absent. Statistically significant correlations are observed between accumulation and the Scandinavian (SCA) teleconnection pattern. The frequent occurrence of the positive SCA phase at the beginning of accumulation season results in lower than average snowfall and reduced accumulation. The relationship between the North Atlantic Oscillation (NAO), Arctic Oscillation, and accumulation is weak, although positive precipitation anomalies in the winter months are associated with the negative phase of the NAO. A stronger positive correlation is observed between accumulation on Djankuat and geopotential height over the Bay of Biscay unrelated to the established modes of the Northern Hemisphere climatic variability. These results imply that the mass balance of Djankuat is sensitive to the natural variability in the climate system. Distant forcing, however, explains only 16% of the variance in the ablation record and cannot fully explain the recent increase in ablation and negative mass balance.

1. Introduction

[2] It is widely recognized that glaciers are highly sensitive to climatic change, and much research has focused on their recent (1950s onward) response to increased global temperatures in an attempt to evaluate future changes under given climate scenarios [Intergovernmental Panel on Climate Change (IPCC), 2001]. Changes in glacier mass balance, and its components, are thought to be some of the most reliable indicators of observed climatic changes [Paterson, 1994]. These are determined by winter accumulation (snowfall) and summer ablation (melting), which are, in turn, driven by anomalies in global and regional atmospheric circulation, as well as changes in local weather patterns and specific glacier characteristics (e.g., altitude, aspect, and supraglacial debris). Unlike variations in glacier extent, which can result from multiple past climate changes, glacier mass balance is driven by contemporary climate.

[3] Recently, linkages between glacier mass balance and large-scale atmospheric and oceanic circulation anomalies have received much attention [e.g., McCabe and Fountain, 1995; Hodge et al., 1998; Bitz and Battisti, 1999; McCabe et al., 2000; Francou et al., 2003]. Analysis of climate data has revealed preferred modes of atmospheric variability known as teleconnection patterns. In addition to the globally important El Niño-Southern Oscillation (ENSO), 14 teleconnection patterns have been identified by Barnston and Livezey [1987] for the Northern Hemisphere (NH). Thompson and Wallace [1998] have identified another leading mode of interannual variability in the NH extratropical circulation, the Arctic Oscillation (AO), which serves as an indicator of the strength of the polar vortex.

[4] The strength of a given teleconnection pattern in a given month is expressed as a teleconnection index, regularly estimated by the Climate Prediction Center (CPC). These have been widely used as empirical predictors in investigations of planetary-wide associations between the parameters representing the atmosphere-ocean-cryosphere system [e.g., McCabe and Fountain, 1995; Cao, 1998; Hodge et al., 1998; Bitz and Battisti, 1999; McCabe et al., 2000; Washington et al., 2000; Reichert et al., 2001; Francou et al., 2003]. It has been recently proposed that certain preferred modes of NH climatic variability may themselves be predictable [Rodwell et al., 1999; Baldwin and Dunkerton, 2001], and this provides the potential for predictability of glacier mass balance up to several years in advance.

[5] While much research on the recent dynamics of valley glaciers has been conducted across the world, particularly in Alaska [Meier and Dyurgerov, 2002; Arend et al., 2002], Scandinavia [Wallen, 1986; Holmlund et al., 1996], and the Alps [Schoner et al., 2000; Paul, 2002; Kääb et al., 2002], glacier behavior in the mountains of landlocked regions in northern Eurasia (Russia and central Asia) and their response to recent climatic variations and teleconnection patterns are less well known. This is partly due to the notion that glaciers located in cold and dry environments are slow to respond to climatic change and variability [Braithwaite and Zhang, 1999; McCabe et al., 2000; Braithwaite, 2002]. Recent observations, however, appear to indicate a strong response of such glaciers to accelerating climatic warming. For example, glaciers in the Altay Mountains and central Asia have shown a decrease in size since the early 1990s [Liu et al., 1999], and there has been widespread wastage of ice in the Ak-shirak glacier systems of the Tien Shan [Khromova et al., 2003] and in the Pamir region [Dyurgerov and Meier, 2000] since the 1980s.

[6] One such area that is currently underreported is the greater Caucasus, which extends for ∼1300 km between the Caspian Sea and the Black Sea (Figures 1 and 2). In the 1970s it accommodated more than 2000 glaciers with a combined area in excess of 1400 km2 [Katalog Lednikov SSSR, 1967–1977, 1975–1977]. There is widespread evidence of glacier recession in the greater Caucasus since the end of the Little Ice Age [Volodicheva, 2002], but information about changes in glaciated area in the second half of the twentieth century is limited and somewhat inconclusive [Bedford and Barry, 1995; Khromova and Chernova, 1996]. Although the likely contribution of the Caucasus glaciers to global sea level is negligible, the regional impacts of glacier change are extremely significant, with particularly important implications for regional water resources. Runoff from Caucasus glaciers feeds directly into the Caspian Sea, and recent increases in its level after decades of decline are a matter of concern for the Caspian nations because of economic development on the coastal plain [Koronkevich, 2002].

Figure 1.

Study area.

Figure 2.

Location map of the Djankuat Glacier and Terskol weather station.

[7] In this paper, we report an extensive (1967 onward) mass balance record from the Djankuat Glacier in the greater Caucasus, Russia, and we use this to examine glacier response in relation to climate variability and change and anomalies in large-scale atmospheric circulation. The specific objectives are (1) to examine the variability in the Djankuat mass balance record and (2) to analyze the coupling between the Djankuat mass balance and climate, including the dominant patterns of NH climate variability.

2. Study Area and Climatic Setting

[8] The Djankuat Glacier (Figure 3) is a northwest facing valley glacier located at 43°12'N and 42°46'E on the northern slope of the central section of the Glavny (Main) Ridge, the most heavily glaciated area in the Caucasus. Its elevation lies between 2700 and 3900 m above mean sea level (amsl), and it delivers meltwater to the upper reaches of the Adylsu River, which eventually drains into the Caspian Sea via the Baksan and Terek rivers (Figures 1 and 2). Typical of most glaciers in the Caucasus, Djankuat is a temperate glacier, and in 2000 the surface area was estimated at 3.01 km2.

Figure 3.

Sketch map of the Djankuat Glacier.

[9] In winter, atmospheric circulation over the central Glavny Ridge is dominated by extensions of the Icelandic depression from the west and the Siberian high from the east [Lydolph, 1977]. The Siberian high is a shallow system confined to the lower levels of the troposphere (below the 500 hPa level), and its influence is primarily restricted to the lower mountains. In contrast, the higher elevations are predominantly affected by the westerly flow [Lydolph, 1977]. In summer, extensions of the Azores high prevail. Superimposed on the general circulation are the influences of the Black Sea (exerting greater influence because of its position to the southwest) and the Caspian Sea, as well as more complex orographic effects. The climate of the region is characterized by distinct seasonality in temperature, which is apparent even at high altitudes. Precipitation maxima occur in July–September in response to convective activity triggered by a combination of strong insolation and weak depressions developing on the polar front and enhanced by the orographic uplift. Winter precipitation is associated with Atlantic depressions crossing over the east European plain and depressions that develop over the Mediterranean and propagate eastward, regenerating over the Black Sea [Shahgedanova, 2002].

[10] Regular meteorological observations are not conducted on Djankuat, with the exception of air temperature measurements during the ablation season (May–September). Standard meteorological observations, however, have been conducted at the Terskol weather station regularly since 1951. The station is located 16 km northwest of the glacier (Figure 2) at an altitude of 2141 m amsl. Figure 4 shows monthly temperature and precipitation climatologies. There is a high correlation (r = 0.82) between daily mean air temperatures in May–September at Terskol and those measured at an elevation of 2650 m on the Djankuat Glacier [Boyarsky, 1978]. Therefore the seasonal and interannual variability in Terskol temperatures is representative of temperature variability on the Djankuat Glacier. Similarly, comparison of episodic precipitation measurements on the Djankuat Glacier with those at Terskol has shown that there is generally a good agreement between precipitation events and measured amounts of precipitation at the two sites [Boyarsky, 1978].

Figure 4.

Mean monthly precipitation totals (mm) and mean air temperature averages (°C) for Terskol, 1951–2001.

[11] The time series of seasonal precipitation at Terskol (1951–2001) are shown in Figure 5. Strong interannual variability in precipitation is observed in all seasons, including the accumulation season as a whole. Long-term changes in precipitation, however, have been minor, and only the October–November (ON) time series is characterized by a weak but statistically significant (at 0.05) positive linear trend, explaining 12% of the variance.

Figure 5.

Time series of seasonal precipitation totals (mm) for Terskol. Solid horizontal lines show long-term averages (1951–2001). The diagonal line in the October–November time series shows a linear trend.

[12] The seasonal mean air temperature time series for Terskol (1951–2001) are shown in Figure 6. The June–July–August (JJA) temperatures are above average during the 1950s and below average in the 1960s–1970s. The JJA Terskol temperature time series (Figure 6) exhibits a positive linear trend between 1982 and 2001, accounting for 38% of the variance.

Figure 6.

Time series of seasonal temperature averages (°C) for Terskol. Solid horizontal lines show long-term averages (1951–2001).

3. Data and Methods

3.1. Mass Balance Data

[13] Regular mass balance measurements of the Djankuat Glacier began in 1967 in the course of the International Hydrological Decade and have been conducted without interruption since then. Today, Djankuat is studied as a reference glacier to which other glaciers of the region with less dense and/or intermittent measurement programs can be compared. Mass balance measurements, reported as meters water equivalent (mwe), refer to the mass balance year, which begins in October and ends in September of the following calendar year. Two components of mass balance are measured: October–May accumulation and June–September ablation, calculated between the first week of the opening month and the last week of the closing month of each season. Sublimation is negligible at Djankuat, constituting no more than 1% of ablation [Boyarsky, 1978]. Net mass balance is calculated from the sum of accumulation and ablation. The amount of snow accumulated on Djankuat between June and September is small in comparison with the October–May period. On average, only four snowfall events are registered at Djankuat's terminus between June and September [Boyarsky, 1978], and even at higher elevations, summer snow invariably melts during the ablation season. Likewise, ablation is negligible between October and May. During this period it occurs on a few days only and very close to the glacier's terminus (not across the glacier) where its rate does not exceed 20 mm d−1. This is considerably lower than during the summer period.

[14] Accumulation is estimated as a product of snow depth and snow density measured at the end of the accumulation season. Between 1967 and 1977, accumulation was measured at 150–300 nonfixed probing points located in different altitudinal zones of the glacier [Popovnin, 2000]. Accumulation was estimated for each zone as an average of these measurements with 5% of the zonal value added to compensate for accumulation in inaccessible regions. Since 1977, the snow depth measurements have been made using a system that currently includes around 500 probing points. Snow density is measured in five snow pits located in different sectors of the glacier (Figure 3). The snow depth measurements are interpolated on a digital elevation model with a 50 m × 50 m grid, and accumulation maps are constructed. Corrections are made for accumulation in the inaccessible sectors of the glacier, the proportion of which has been evaluated for each altitudinal zone. Accumulation values for 10 altitudinal zones and the glacier as a whole are derived from the constructed maps. For a detailed description of this methodology and its application, see Popovnin [2000]. A comparison of the two methods conducted between 1977 and 1986 has shown that prior to 1977, accumulation was overestimated by 1–8% [Popovnin, 2000]. As such, the pre-1977 records have been adjusted using measurements from the cross-observation period. Ablation is measured using the stake method [Østrem and Brugman, 1991] with 80–100 stakes being used each year, and these are interpolated in a way similar to accumulation. The uncertainties and errors of estimations of seasonal totals of accumulation and ablation are discussed by Popovnin [2000] and Boyarsky [1978], respectively. The errors vary between years and altitudinal zones, but for the glacier as a whole they are estimated as 3–7% of seasonal accumulation and 3–5% of seasonal ablation.

3.2. Meteorological Data and Statistical Analysis

[15] Monthly mean air temperature and precipitation data from Terskol weather station have been used to investigate the relationships between the components of mass balance and local climate. In order to assess the relationship between the components of Djankuat mass balance and the larger-scale atmospheric circulation, correlation and composite maps have been constructed, linking accumulation and ablation data on Djankuat with the gridded atmospheric reanalyses fields with 2.5° resolution (obtained from the National Center for Environmental Prediction and National Center for Atmospheric Research).

[16] Correlation coefficients between the time series of ablation and accumulation and those of the NH teleconnection indices have been calculated. The time series of mass balance components have been standardized by subtracting the mean of the time series and dividing by standard deviation. Linear trends have been removed from all data series prior to analysis in order to isolate interannual variations. The 14 teleconnection patterns identified by Barnston and Livezey [1987] and the AO [Thompson and Wallace, 1998] have been used in this study. Only those patterns that reveal correlations statistically significant at 0.05 are discussed further, namely, the North Atlantic Oscillation (NAO), Scandinavian (SCA), and North Pacific (NP) patterns (Figure 7). The AO reveals a statistically significant correlation with accumulation. The AO is a dominant mode of variability in the Northern Hemisphere extratropics and is highly correlated with precipitation in the midlatitudes [New et al., 2001], while the NAO is an important modulator of winter precipitation and snow accumulation in northwestern and southern Europe [Hurrell and van Loon, 1997; Beniston and Jungo, 2002]. The NAO is often considered as a regional manifestation of the AO [Wallace, 2000], and further discussion will be concerned with the NAO. The SCA and NP indices were estimated by the CPC by taking averages of monthly standardized amplitudes of rotated principal components of monthly mean 700 hPa geopotential height anomalies for 1950–2001 (see http://www.cpc.noaa.gov/data/teledoc/telecontents.html). The NAO index based on the sea level pressure data from Gibraltar and Stykkisholmur [Jones et al., 1997] has been used because analysis of correlation patterns between 500 hPa geopotential height and sea level pressure has revealed a stronger teleconnection with the western Mediterranean than with a more typical NAO center located over the Azores.

Figure 7.

Modes of Northern Hemisphere climate variability relevant to this study: (a) North Atlantic Oscillation (NAO) in January, (b) Scandinavian (SCA) in November, and (c) North Pacific (NP) in April. Loading weights of rotated principal components of monthly mean 700 hPa geopotential height (provided by the Climate Prediction Center) are shown. Solid contours indicate negative loading weights; dashed contours indicate positive loading weights.

4. Results

4.1. Djankuat Glacier Mass Balance (1967–2001)

[17] The time series of net mass balance and its components are shown in Figure 8, and average values of mass balance characteristics are given in Table 1. It is often stated that variability in the mass balance of glaciers located in continental interiors is influenced by changes in summer temperatures during the ablation season, while variability in the mass balance of coastal glaciers is dominated by precipitation during the accumulation season [Walters and Meier, 1989]. Standard deviations of accumulation and ablation (Table 1) estimated for the duration of the record are very similar, indicating that the mass balance regime of Djankuat is intermediate between these two extremes, possibly because of the contribution of the local sources of moisture, in particular, the Black Sea. However, variability in accumulation increased considerably after the 1986/1987 mass balance year primarily because of high winter precipitation anomalies in 1986/1987 and 1992/1993. Prior to the 1986/1987 accumulation season the variability in accumulation was considerably lower than ablation values, with standard deviations of 0.45 and 0.27, respectively. There is a tendency for accumulation and ablation to be correlated because positive anomalies in accumulation enhance albedo during the ablation season [Kaser, 1999; Francou et al., 2003]. However, no statistically significant correlation between winter accumulation and summer ablation has been found for the Djankuat Glacier.

Figure 8.

Time series of accumulation, ablation, net mass balance, and cumulative mass balance (meters water equivalent) of the Djankuat Glacier. Horizontal solid lines refer to zero mass balance and mean values of ablation and accumulation.

Table 1. Average Statistics of Mass Balance Components of the Djankuat Glacier, 1966/1967–2000/2001a
ParameterMeanStandard Deviation
  • a

    Unit mwe is meters water equivalent.

Net mass balance, mwe−0.160.64
Accumulation, mwe2.410.41
Ablation, mwe2.570.47

[18] While strong interannual variability characterizes the time series of accumulation, ablation, and net mass balance (Figure 8), no time series exhibits a linear trend (significant at 0.05) between 1966/1967 and 2000/2001. Since 1993, however, the net mass balance has been predominantly negative with the exception of the low positive values of 0.04 and 0.27 mwe in 1994/1995 and 1996/1997 (Figure 8). Since 1997/1998, negative values have been observed for 4 consecutive mass balance years (Figure 8). Prior to this the net mass balance was negative for more than 1 year on only three occasions (in 1971–1973, 1984–1986, and 1990–1992). Moreover, prior to 1993 (with the exception of the early 1970s when negative values of mass balance reached a low of 1.14 mwe), negative values of mass balance were generally quite small.

[19] The overall mass changes are best illustrated by the cumulative mass balance trend (Figure 8). The glacier's mass was declining steadily between 1970 and 1986; however, the highest accumulation values on record that occurred in 1986/1987 and 1992/1993 (4 and 3.18 mwe, respectively) allowed for the partial recovery of the glacier's cumulative mass balance trend. This brief recovery period ended in the 1993/1994 mass balance year and was followed by a negative trend, which rapidly accelerated after 1996/1997 (Figure 8). This has resulted in increased runoff from the glacier, which is calculated as a product of specific ablation and glacier area. On the basis of this estimation, ∼9.1 × 106 m3 of water per annum discharged into the basin of the Terek River between the 1992/1993 and 2000/2001 mass balance years. This only constitutes 0.1% of the annual Terek River discharge, but a similar response from over a thousand glaciers in the greater Caucasus, discharging into the Terek, could have a profound impact on inputs to the Caspian Sea.

[20] Inspection of the ablation time series (Figure 8) shows that melting has increased considerably since the 1986/1987 mass balance year with positive trend (significant at 0.01) explaining 46% of variance in the time series between 1986/1987 and 2000/2001. Analysis of ablation rates at different altitudinal zones shows that melting of the lower part of the glacier (2800–2900 m) has been increasing since the early 1970s, despite the continuous expansion and thickening of supraglacial debris cover leading to declining rates of submoraine melting in this area [Popovnin and Rozova, 2002]. Ablation in the middle part of the glacier has been increasing since 1987, and since 1991, acceleration of the positive trend in ablation has been observed across the whole glacier, explaining over 60% of the variance in the time series. At the same time, accumulation rates have remained stable, with the notable exception of the 1986/1987 and 1992/1993 mass balance years. These results indicate that the decline in net mass balance since 1993/1994 has been largely due to increasing summer melt.

[21] The two-sample t test, assuming normal distribution and unequal variances, has been used to test the hypothesis that the change in ablation and the associated change in mass balance observed in the late 1980s to early 1990s are statistically significant. Visual inspection and analysis of linear trends in the time series (Figure 8) have shown that ablation has been increasing since the summer of 1988 with a step increase in the summer of 1994. The test has been applied to two pairs of samples: (1) 1967/1968–1986/1987 and 1988/1989–2000/2001 and (2) 1967/1968–1992/1993 and 1993/1994–2000/2001. The change in ablation that occurred since the summer of 1994 passes the test at 0.02 confidence level, while the associated change in net mass balance is significant at 0.06 (Table 2). This implies that there is a statistically significant change in Djankuat mass balance and ablation after the 1993/1994 mass balance year. The fact that there has been no statistically significant change before and after the 1988/1989 mass balance year may be attributed to the impacts of atmospheric circulation (section 5) and the impact of the 1991 eruption of Mount Pinatubo on temperature and precipitation. Global mean air temperature declined by up to 0.5°C at the surface in 1991–1993 following the eruption [Parker et al., 1996], resulting in the abrupt decrease in ablation recorded in mass balance series worldwide [Abdalati and Steffen, 1997; Dyurgerov and Meier, 2000]. It can be seen that below average ablation values were registered on Djankuat in 1991–1993 (Figure 8).

Table 2. Statistical Significance of Assumed Changes in Ablation and Mass Balance of the Djankuat Glaciera
ParameterBefore “Change”After “Change”P Value
YearsNmσYearsNmσ
  • a

    Means and standard deviations of the before- and after-change samples are m and σ; the number of years in the sample is N.

Ablation1967/1968212.480.441988/1989132.720.500.17
Mass balance1986/198721−0.120.672000/200113−0.230.620.64
Ablation1967/1968262.450.411993/199482.954.60.02
Mass balance1992/199326−0.060.662000/20018−0.50.510.06

4.2. Relationship Between Glacier Mass Balance and Local Climate

[22] Winter accumulation (October–April) on the Djankuat Glacier results largely from snowfall, and only ∼10% of winter accumulation is contributed to by avalanches and snowdrift [Boyarsky, 1978]. There is a positive correlation of 0.68 (significant at 0.01) between accumulation on Djankuat and October–April precipitation at Terskol. The highest correlation coefficients between precipitation in individual months and seasonal accumulation occur in December and January (DJ) (0.50 and 0.51, respectively, significant at 0.01). This indicates that precipitation in DJ dominates the interannual variability in winter accumulation. Although interannual variability in DJ precipitation at Terskol is considerable, there is no long-term linear trend between 1951 and 2000 (Figure 5). A corollary of this is that the recently observed increase in ON precipitation (Figure 5) has been insufficient to sustain a growth in accumulation.

[23] In contrast to winter accumulation, summer ablation is a more complex process driven by several factors, which determine the net radiation at the glacier surface. These factors include incoming long-wave radiation, which is a function of cloud cover [Ohmura, 2001], albedo [Oerlemans and Hoogendoorn, 1989; van de Wal et al., 1992; Klok and Oerlemans, 2004], and sensible heat flux [Braithwaite, 1981]. The latter is particularly important at the edges of a glacier where it may be enhanced by local advection of warm air from the surrounding rocks [Wendler, 1974]. Precipitation is another control over ablation on alpine glaciers, through its effect on surface albedo and through the effect of cloud cover on spectral distribution of radiation [Jonsell et al., 2003]. It has been shown that in low-latitude glaciers a precipitation deficit and the associated low albedo of the aged snow cover can drive glacier recession [Kaser, 1999; Francou et al., 2003].

[24] With regard to Djankuat, there are strong positive correlations of 0.61 and 0.77 between ablation and JJA mean and mean maximum air temperatures, respectively, and therefore the recently observed increase in ablation is consistent with the increase in JJA air temperatures (Figure 6). As mentioned in section 3.1, significant snowfalls at the terminus of the Djankuat Glacier occur, on average, only four times during the ablation season. After each snowfall event, albedo increases by 40–50% and can remain above its seasonal value for the next 2–3 days, but this only results in a small reduction in overall ablation values for the season [Boyarsky, 1978]. We note a weak negative correlation of −0.34 (significant at 0.05) between April–September precipitation and ablation, but it is likely that this reflects the fact that hotter months are also drier months rather than any precipitation-ablation feedback.

4.3. Relationship Between Glacier Mass Balance and Large-Scale Atmospheric Circulation Anomalies

4.3.1. Accumulation Season

[25] Our correlation analysis indicates that three teleconnection patterns (monthly indices) exhibit a statistically significant relationship (at 0.05) with seasonal accumulation on Djankuat. These are the November indices for SCA and NAO with correlation coefficients of −0.40 and 0.38, respectively. These correlation coefficients are lower than those between teleconnection indices and components of mass balance in the climatically sensitive regions such as northwestern Pacific [Hodge et al., 1998; Bitz and Battisti, 1999] and Scandinavia [Washington et al., 2000], where statistically significant coefficients range between 0.30 and 0.6. Seasonal (October–April) accumulation only is measured on Djankuat, and the lack of accumulation data at monthly resolution may have an effect on the strength of the relationship between accumulation and atmospheric dynamics.

[26] In addition to the correlation analysis we have also examined links between anomalies in large-scale atmospheric circulation and accumulation by constructing 500 hPa geopotential height composite anomaly maps for the months with strongly above and below average precipitation. Monthly precipitation at Terskol (starting in 1951) has been used as a proxy for accumulation because monthly totals of accumulation are not available and because precipitation at Terskol exhibits strong correlation with accumulation on Djankuat (section 4.2). The low- or high-precipitation months have been defined as not more than one standard deviation below or above the 1951–2001 monthly average for each month of the accumulation season. The only exception is for low January precipitation whereby an arbitrary threshold of 15 mm was used because the standard deviation exceeds the monthly average. The 500 hPa geopotential height has been used because it is close to the Caucasus summit elevations at ∼5000 m above sea level.

[27] Figure 9 shows the 500 hPa geopotential height anomaly maps for low-precipitation months (all months exhibit similar patterns; the largest anomalies occur in November when SCA pattern is most prominent, and this month only is shown for brevity). Negative precipitation anomalies are associated with the above average geopotential heights over Scandinavia and northwestern Russia and the below average geopotential heights over western Europe, which is consistent with the positive phase of the SCA pattern [Barnston and Livezey, 1987].

Figure 9.

The 500 hPa geopotential height anomaly map (departures from the mean 1948–2001 500 hPa fields) for low-precipitation months in November. Contour intervals are 2 decameters (dam). The main features are the above average geopotential heights over Scandinavia and northwestern Russia and the below average geopotential heights over western Europe which are consistent with the positive phase of the SCA pattern.

[28] The largest perturbations in the 500 hPa height fields, associated with the high-precipitation months (Figure 10), occur in the North Atlantic and over the eastern Europe-Black Sea region. In November a high-pressure ridge that is normally present over the Azores is displaced northeastward from its climatological position extending from the Iberian peninsula to the British Isles. The above average geopotential heights occur over the Iberian peninsula, and below average heights occur over central and eastern Europe, extending as a weak negative anomaly toward the Norwegian Sea (Figure 10a). Although all high-precipitation months except Novembers of 1956 and 1979 are characterized by positive NAO index values [Jones et al., 1997], the high-precipitation composite pattern does not resemble that typically associated with the positive NAO phase, in which centers occur over the Azores and over the Denmark Strait near Iceland. The positive correlation between accumulation and NAO should be understood not through the simultaneous strengthening of the Icelandic low and the Azores high that characterizes a standard positive NAO phase but through the presence of a positive geopotential height anomaly centered over the Bay of Biscay. This anomaly initiates a wave train of anomalies extending across Eurasia with an enhanced trough extending from the Barents Sea to the eastern Mediterranean (Figure 10a). The wave-like pattern is present through the depth of the troposphere with an equivalent barotropic structure typical of the quasi-stationary Rossby waves [Charney and Eliassen, 1949]. Anomaly patterns for other winter months (only January is shown for brevity) (Figure 10b) are different from November. They exhibit a strong positive geopotential height departure over the subpolar North Atlantic and a negative anomaly over the Azores that is consistent with the negative phase of the NAO. In spite of the difference between the November and the winter month composite patterns the main feature in all high-precipitation months is a strong negative geopotential height anomaly extending from the Barents Sea to the Mediterranean Sea.

Figure 10.

The 500 hPa geopotential height anomaly maps for high-precipitation months in (a) November and (b) January (departures from the mean 1948–2001 November and January 500 hPa fields). Contour intervals are 2 dam. The main features are a wave train of geopotential height anomalies extending across Eurasia from the Bay of Biscay (Figure 10a) and geopotential height anomalies over the Azores and the subpolar North Atlantic consistent with the negative phase of the NAO (Figure 10b).

[29] Analysis of the North Atlantic storm track variability over Europe [Rogers, 1997] has shown that storm track (and consequently precipitation) variability is not uniquely associated with the NAO but with the variability in sea level pressure in the far northeastern Atlantic and the Bay of Biscay. Our results support this conclusion. For example, the highest accumulation on record occurred in the 1986/1987 mass balance year, resulting in the highest annual mass balance on record (Figure 8). While monthly precipitation totals exceeded the 1951–2001 average by factors of 2.1 and 4.8, monthly values of the NAO index [Jones et al., 1997] changed from strongly positive in December (3.42) to strongly negative in January (−2.12), showing that high precipitation anomalies occur during both NAO extremes. In both months the 500 mbar geopotential height fields were characterized by an enhanced trough extending from the Barents and Kara seas to the Mediterranean, a strong positive anomaly in zonal wind in the middle and upper troposphere directly over the Caucasus Mountains (not shown), and associated near-surface cyclogenesis.

4.3.2. Ablation Season

[30] Correlation of the Djankuat ablation record with 500 hPa geopotential height fields (not shown) has revealed that the May–October ablation exhibits a strong link with the circulation over the Caucasus: A strong (weak) melt is linked to anticyclonic (cyclonic) circulation over the region and the warmer/colder troposphere associated with it. This relationship is not driven by the Atlantic-based teleconnection patterns as there is no statistical association between the summer melt at Djankuat and the Euro-Atlantic teleconnection indices. However, ablation at Djankuat appears to be linked to the NP teleconnection pattern, which is active between March and April [Rogers, 1981; Barnston and Livezey, 1987]. The positive phase of NP features below average geopotential heights over northeastern Siberia and Alaska and above average geopotential heights over the subtropical North Pacific (Figure 7) (CPC, see http://www.cpc.noaa.gov/data/teledoc/telecontents.html). There is a statistically significant (at 0.05) correlation between June–September ablation at Djankuat and July and April NP indices with correlation coefficients of −0.38 and −0.44, respectively; that is, low (high) ablation is associated with the positive (negative) phase of the NP pattern (either concurrent with or leading ablation). The NP index for April has been correlated with JJA 500 hPA geopotential height and 500 hPa temperature (Figure 11). Although correlation coefficients do not exceed −0.50, it is clear that negative (positive) geopotential height and temperature anomalies associated with positive (negative) NP index extend eastward from the North Pacific with centers being positioned over the central United States and over the Caucasus. To confirm the stability of this relationship, correlation analysis has been repeated for the periods of 1968–1980 and 1980–2001 separately (not shown). Negative correlations of the April index of NP with 500 hPa geopotential height and temperature in the central United States and the Caucasus have been found to be similar to those obtained for the whole period with higher (−0.7) correlations in 1968–1980 (a period free of major volcanic eruptions whose impact is similar to that of NP).

Figure 11.

Correlation coefficients (multiplied by 10) between the NP index (April) and (a) June–July–August (JJA) 500 hPa geopotential height and (b) JJA temperature at 500 hPa for the period 1968–2001. Areas of correlations statistically significant at 0.05 are shaded.

[31] There is evidence of quasi-decadal variability in the NP index with a mostly negative phase in the 1950s, a mostly positive phase in the late 1980s to early 1990s, and positive extremes in 1992 and 1993, followed by predominantly negative values and a negative extreme in 1998 (CPC, see http://www.cpc.noaa.gov/data/teledoc/np.html). The above average summer temperatures in the central greater Caucasus (Terskol) in the 1950s and the late 1990s (Figure 6) were associated with the negative phases of the NP. Lower temperatures and reduced ablation rates at Djankuat were associated with the positive NP phase of the early 1990s. Thus ablation at Djankuat was among the lowest on record in 1982 and 1992–1993 (positive NP extremes) and among the highest in 1998 (negative NP extreme) (Figure 8).

5. Discussion

[32] The most prominent result emerging from these analyses is a significant increase in the summer melt on the Djankuat Glacier in the last 2 decades. In particular, during the last decade, accelerating ablation has been observed across the whole glacier, including the highest elevations. Ablation at Djankuat is strongly correlated with summer temperature. Previous research by Diaz and Bradley [1997] focusing on changes in mean annual temperature at five high-altitude sites in the Caucasus Mountains between 1930 and 1990 (regular observations at these sites were discontinued in the early 1990s) has shown that climate was not becoming warmer in the Caucasus Mountains, unlike in the Alps. In contrast to these results the Terskol temperature records, which were not considered by Diaz and Bradley [1997], have shown that a strong warming signal has emerged in the central greater Caucasus in the last 20 years and particularly during the last 10 years. At the same time, there has been no compensating consistent increase in accumulation on Djankuat, which has resulted in the net loss of mass and increase in the runoff to the Caspian Sea. Accumulation is closely correlated with precipitation during the accumulation season, in particular, in December and January. The Terskol precipitation records do not show any consistent change in winter precipitation, and the observed increase in ON precipitation (Figure 5) has not been sufficient to sustain a growth in accumulation.

[33] Low-frequency distant modes of atmospheric variability have been identified in many studies as an important factor controlling glacial accumulation and melt [McCabe and Fountain, 1995; Cao, 1998; Hodge et al., 1998; Bitz and Battisti, 1999; McCabe et al., 2000; Washington et al., 2000; Reichert et al., 2001; Francou et al., 2003]. Distant forcing that often originates on the ocean-atmosphere interface in the tropics and propagates through the atmosphere as quasi-stationary Rossby waves may exert stronger influence on variability in mass balance of middle-latitude and subpolar glaciers than atmospheric forcing in the immediate glacial environment. Thus glaciers of the northwestern Pacific respond strongly to the interannual variability in atmospheric circulation associated with ENSO [Hodge et al., 1998; Bitz and Battisti, 1999]. Similarly, circulation anomalies over the Pacific exert a strong influence on the mass balance of Svalbard glaciers, despite the fact that the NAO is widely regarded as the most important regional mode of atmospheric variability [Washington et al., 2000].

[34] Distant forcing has a moderate influence on glacial mass balance in the central greater Caucasus. The highest (although not strong) correlation is between ablation at the Djankuat Glacier and atmospheric circulation over the North Pacific quantified by the NP teleconnection index, showing that low (high) ablation at Djankuat is associated with the positive (negative) phase of the NP. The NP pattern, which is active from March through July, controls the position and intensity of the North Pacific storm track [Rogers, 1981], which, in turn, is an important modulator of glacier dynamics in the northwestern Pacific. Enhanced anticyclonic circulation over western North America and enhanced cyclonic circulation over the central United States [Bell and Janowiak, 1995] are associated with the strong positive phases of the NP. Our analysis tentatively suggests that the influence of the NP pattern extends much farther east, reaching the Caucasus Mountains (Figure 11), where negative (positive) geopotential height and temperature anomalies, associated with strong positive (negative) phases of the NP, result in suppressed (enhanced) glacial melt. The correlations between geopotential height over the Caucasus region and the NP index are moderate and as such may not imply causation; however, they are as strong as those over the central United States where strong positive precipitation anomalies result from strongly positive NP [Bell and Janowiak, 1995]. Inspection of the geopotential height anomaly maps constructed for those months of April when the NP index was one standard deviation above the mean (not shown) has revealed the presence of wave trains of anomalies extending from the North Pacific to the Caspian Sea. These wave trains, which are best defined in the middle troposphere (500 hPa), provide a possible mechanism of the NP impact on ablation in the Caucasus.

[35] Bell and Janowiak [1995] linked a prolonged positive phase of NP that occurred during the 1992 and 1993 spring seasons and associated floods in the American Midwest to the extremely prolonged period (early 1991 to late 1993) of ENSO conditions, arguing that positive NP mode represents one of the main responses of the extratropical atmosphere to ENSO forcing. While no direct link has been found between ablation on Djankuat and ENSO, the relationship between ablation and the NP pattern suggests that tropical signal may indirectly impact glacial melt in the Caucasus. Thus, in the early 1990s the prolonged ENSO events and strengthened NP pattern might have moderated the local temperature increase and ablation. By contrast, during the last decade the NP pattern was predominantly negative, and this distant moderating forcing was absent.

[36] The NP pattern is the only pattern significantly correlated to the ablation record, and it explains 16% of the variance in the Djankuat ablation record. This suggests that distant forcing provides a limited explanation of the Djankuat Glacier climate dynamics, including the recent (the 1980s onward) warming trend and negative mass balance of the Djankuat Glacier, and tentatively points toward an increased role of anthropogenic climate warming which began to exceed natural variations during the last 2 decades [IPCC, 2001].

[37] Statistically significant correlations are observed between accumulation and the SCA and NAO/AO patterns. The SCA pattern has a strong influence on the variability in autumn and winter precipitation in Europe [Qian et al., 2000] and Eurasian snow cover [Clark et al., 1999]. The positive phase of the SCA pattern is associated with the presence of a blocking anticyclone over northwestern Russia or eastern Scandinavia and upper level troughs over western Europe and central Asia (Figure 7) [Barnston and Livezey, 1987]. The blocking events that are most prominent in the late autumn to early winter present a barrier to the westerly flow over Scandinavia and European Russia, resulting in lower than average precipitation in central and eastern Europe. At the same time, the displacement of the Atlantic jet stream south of its climatological position results in above average precipitation at the peripheries of the high-pressure center, causing, as in the case of the extreme positive SCA phase in the autumn of 2000, anomalous precipitation in western Europe and floods and mud slides in the southern Alps [Lawrimore et al., 2001]. Correlation coefficients between precipitation measured at Terskol and the SCA index in November and December of −0.51 and −0.56 (significant at 0.01), respectively, indicate that negative precipitation anomalies associated with the positive phase of the SCA extend to the northern Caucasus and frequent occurrence of the positive SCA pattern at the beginning of the accumulation season results in lower than average snowfall. Thus the lowest accumulation values on record observed in 1983–1985 (Figure 8) were associated with a frequent occurrence of the positive SCA phase.

[38] Although the NAO is the most prominent mode of climatic variability in the Euro-Atlantic region, affecting a variety of climate-related parameters [Marshall et al., 2001], there is a lack of a strong and consistent association between NAO, regional precipitation, and accumulation on the Djankuat Glacier. Positive correlation between Jones et al.'s [1997] NAO index and accumulation is unexpected as the positive phase of the NAO is associated with below average precipitation in southern Europe [Hurrell and van Loon, 1997; New et al., 2001]. Indeed, correlation of the NAO indices with regional precipitation using the Climate Research Unit (CRU) global precipitation analysis [New et al., 2000] has shown a weak negative correlation with precipitation in the Black Sea-northern Caucasus region between November and February. The CRU precipitation analysis does not resolve precipitation in the high-altitude zone of the Caucasus. However, correlation of monthly precipitation records from Terskol with the NAO indices has also shown weak but significant (at 0.05) negative correlations between NAO and precipitation in October, December, January, and February. This relationship is not found in November, March, and April. Composite analysis supports this conclusion, showing that strong precipitation measured at Terskol in the winter months is associated predominantly with a negative NAO phase (Figure 10b).

[39] There are three possible explanations for the inconsistent relationship between NAO, precipitation, and accumulation on the Djankuat Glacier. First, the northern macroslope of the greater Caucasus is located between the zones of enhanced NAO-precipitation links located approximately at 60°N (positively correlated NAO and precipitation) and 40°N (negatively correlated NAO and precipitation) [Hurrell and van Loon, 1997]. As such, regional circulation anomalies and local orographic effects on precipitation may be highly variable and more important than distant forcing. This is consistent with results of Schmidli et al. [2002], who have shown that the relationship between the NAO and winter precipitation in the northern Alps is weak and intermittent. Second, the relationship between precipitation and NAO may change with altitude. Indeed, analyses by Beniston [2000] and Beniston and Jungo [2002] have revealed an altitudinal dependency of local climatic response to the NAO in the Alps and an amplified response of precipitation and snow accumulation at high-altitude sites to NAO. Similarly, correlation of accumulation in the 10 altitudinal zones of Djankuat with Jones et al.'s [1997] NAO index revealed stronger positive links between NAO and accumulation in the upper part of Djankuat (above 3200 m) with a correlation coefficient of 0.50 between October–April accumulation totals and the NAO index in November. Third, strong negative geopotential height anomaly extending from the Barents Sea to southeastern Europe (Figure 10a) is the main cause of extreme late autumn-winter precipitation in the central greater Caucasus. The persistence and coincidence of this anomalous pattern with prolonged and heavy precipitation in December 1986 and January 1987 and the highest accumulation on record were notable. However, this pattern is not uniquely associated with the NAO and in the late autumn is more likely to be a part of a wavelike pattern propagating from the Bay of Biscay eastward. In this context a positive correlation between Jones et al.'s [1997] NAO index in November and accumulation on Djankuat is likely to reflect an association between snowfall and the anomalous circulation over the easternmost Atlantic rather than a true NAO signal.

6. Conclusions

[40] The mass balance record for the Djankuat Glacier located in the greater Caucasus, Russia, has been analyzed and links between the components of mass balance, local climate, and distant atmospheric forcing have been examined. Our analyses suggest the following conclusions:

[41] 1. Accumulation and ablation on the Djankuat Glacier are strongly correlated with local precipitation and temperature (measured at a nearby Terskol weather station).

[42] 2. The temperature record at Terskol has shown that a strong warming signal has emerged in the last 2 decades, leading to a significant increase in the summer melt. In particular, during the last decade, there has been an acceleration in the increase of ablation losses observed across the whole glacier, including the highest elevations.

[43] 3. There has been no compensating consistent increase in accumulation to offset the increase in ablation, and cumulative mass balance has shown a strong negative trend since the 1993/1994 mass balance year. Although the t test supports the hypothesis of a significant step change before and after 1993/1994, it is suggested that a shift in glacier regime to a new mode occurred earlier, in 1986/1987, but was masked by the impacts of Mount Pinatubo eruption and a possible impact of the NP teleconnection pattern.

[44] 4. There is a weak but statistically significant correlation between accumulation on Djankuat and the SCA teleconnection pattern. Frequent occurrence of the positive SCA phase at the beginning of the accumulation season results in lower than average snowfall, and the lowest accumulation on record observed during 1983–1985 is explained by the recurring positive SCA.

[45] 5. The relationship between the NAO and accumulation is weak, although positive precipitation anomalies in the winter months are associated with the negative phase of the NAO. A stronger positive correlation is observed between accumulation on Djankuat and atmospheric circulation anomalies over the Bay of Biscay unrelated to the established modes of the Northern Hemisphere climatic variability.

[46] 6. There is a link between the positive (negative) phase of the NP and negative (positive) geopotential height and temperature anomalies over the Caucasus and reduced (increased) glacial melt. Thus positive NP extremes may have resulted in temporary decline in ablation in the early 1990s. The positive phase of the NP is related to ENSO, and it is possible that teleconnection between the tropical Pacific sea surface temperature and summer temperatures in the Caucasus is bridged through the NP pattern.

[47] 7. The observed trends in the Djankuat Glacier ablation and mass balance provide compelling evidence that the glacier is responding to a recent warming trend, which may lie outside of the natural variability expressed by the main NH circulation modes.

[48] Traditionally, it has been argued that glaciers located in continental interiors are slower to respond to climatic fluctuations. This study has shown that such glaciers may respond to the observed climatic warming and may be more sensitive than those located in coastal regions because the temperature-related increase in ablation is not compensated for by positive trends in accumulation. Indeed, our preliminary analyses of satellite imagery (1985–2000) indicate the widespread decline of other glaciers in the greater Caucasus (C. R. Stokes et al., manuscript in preparation, 2005).

Ancillary

Advertisement