A 181.8 m ice core was recovered from a borehole drilled into
bedrock on the western plateau of Mt El'brus (43
Large-scale modes of variability such as the NAO (North Atlantic Oscillation) are known to influence European climate variability (see review in Panagiotopulos et al., 2002). However, most studies of large-scale drivers of European climate change have been focused on low-elevation instrumental records from weather stations, and there is very limited information about climate variability at high altitudes and about differences in climate variability and trends at different elevations (EDW research group, 2015). Such differences were calculated in many mountain regions (EDW research group, 2015), except for the Caucasus, due to the lack of high-elevation instrumental observations in this region.
The Caucasus is located southwards of the East European Plain. It is a high mountain region, with typical elevations of 3200–3500 m a.s.l. and with the highest point reaching 5642 m for El'brus. The main Caucasus Ridge acts as a barrier between subtropical and temperate midlatitude climates, as observed for other high mountain regions such as the Himalaya. As in other mountain regions, there is a lack of high-elevation meteorological records in the Caucasus. Moreover, existing records are relatively short: for example, reliable Caucasus precipitation measurements only started in 1966. Improved spatio-temporal coverage is required to investigate internal variability, to explore trends and spatial differences, and to evaluate the skills of atmospheric models providing atmospheric analysis products where no meteorological data are assimilated.
Measurements of the stable isotope composition of water, and annual accumulation rates in mid- to high-latitude ice cores are widely used proxies to estimate past temperature and precipitation rate changes. In many high mountain regions such as the Caucasus, and for elevations situated above the tree line, ice core data provide the only source of detailed information to document past climate changes, complementing information from specific points and retrieved from changes in glacier extent and recent glacier mass balance. For example, a study of the water stable isotope composition of several ice cores obtained in the Alps was recently conducted by Mariani et al. (2014) and the same research in Alaska was performed by Tsushima et al. (2015). The authors explored the links between the ice cores' isotopic composition, local climate, and large-scale circulation patterns. They found that in mountain regions, the isotopic composition of the ice cores was governed both by local meteorological conditions and by regional and global factors. These studies discussed the complexity of interpreting ice core records from high-altitude glaciers due to the potential bias from post-depositional processes and frequent changes in the origin of moisture sources. For instance, even in areas without any seasonal melt, accumulation is the net effect of precipitation, sublimation, and wind erosion processes and may significantly differ from precipitation. Water stable isotope records are in mid- to high latitudes physically related to condensation temperature through distillation processes (Dansgaard, 1964), but the climate signal is archived through the snowfall deposition and post-deposition processes. One important artefact lies in the intermittency of precipitation and in the covariance between condensation temperature and precipitation, which may bias the climate record towards one season, or towards one particular weather regime, challenging an interpretation in terms of annual mean temperature (Persson et al., 2011). Moreover, water stable isotopes are integrated tracers of all phase changes occurring from evaporation to mountain condensation and are also affected by non-local processes related to evaporation characteristics or shifts in initial moisture sources. Such processes have the potential to alter the validity of an interpretation of the proxy record in terms of local, annual mean, or precipitation-weighted temperature. In some regions, isotopic records are more related to hydrological cycles, recycling, or rainout (Aemisegger et al., 2014). Finally, the condensation temperature may also strongly differ from surface air temperature; depending on elevation shifts in, e.g., planetary boundary layer or convective activity (see Ekaykin and Lipenkov, 2009, for a review). While these processes make the interpretation of ice core records complex, they do open the possibility that the ice core proxy record in fact may be more sensitive to large-scale climate variability than precipitation amounts from specific points. For instance, Casado et al. (2013) have evidenced a strong fingerprint of the NAO in water stable isotope records from central western Europe and Greenland in long instrumental records based on precipitation sampling, in seasonal ice core records, and in atmospheric models including water stable isotopes. The connection of Greenland ice cores' isotopic composition with atmospheric circulation patterns was studied by Vinther et al. (2003, 2010). The strong influence of the NAO pattern on the Greenland ice cores' isotopic composition has been ascertained and the possibility to use the ice core data for the reconstruction of past NAO changes has been suggested (Vinther et al., 2003). The authors also revealed the importance of the study of the seasonally resolved ice core records rather than annual records, as there are different factors governing the formation of the isotopic composition of precipitation in warm and in cold seasons (Vinther et al., 2010).
Map showing the region around El'brus (black rectangle in the world map in the lower right corner), with shading indicating elevation (m a.s.l.). Drilling sites are indicated as red filled circles, GNIP stations as green filled circles, and meteorological stations as blue dots. Stations situated to the south of the main Caucasus Ridge according to the precipitation cycle pattern are shown using a blue dot surrounded by a white circle and the stations situated to the north are displayed surrounded by a black circle (see text for details). The brown dotted line shows the border between two types of precipitation seasonal cycles. The numbers of the various stations refer to Table 1, where detailed descriptions are found.
We will now briefly review earlier studies performed on climate variability
in the Caucasus area, which have already explored the relationships between
regional climate, glacier expansion, and large-scale modes of variability:
the NAO, AO (Arctic Oscillation), and NCP (North
Sea–Caspian Pattern). For example, Shahgedanova et al. (2005) monitored the
mass balance of the Djankuat glacier, situated at an altitude between 2700
and 3900 m a.s.l. While no significant correlation was identified between
the accumulation rate and the winter NAO index, the years of high
accumulation systematically occurred during winters with a very negative NAO
index. Brunetti and Kutiel (2011) explored the influence of the NCP mode on
climate in Europe and around the Mediterranean region. They evidenced a
negative correlation coefficient of
Here, we take advantage of the new El'brus deep ice cores (Mikhalenko et al., 2015) and produce the first analysis of water stable isotope and accumulation records. Section 2 introduces the data and methods, with a description of the ice core analyses and age scale, an overview of regional meteorological information, and the source of information for indices of modes of variability. Section 3 presents the results of the comparison and statistical analyses of the relationships between regional climate parameters (temperature and precipitation), El'brus ice core records, and modes of variability. In Sect. 4, we summarize our key findings and the next steps envisaged to strengthen the climatic interpretation of the Caucasus ice core records.
Here, we report on results from the new, deepest ice core from Mt El'brus and compare them with results from shallow ice cores.
Deep drilling was performed on the western plateau
(43
In order to update the ice core records towards the present day and enable a comparison of the measurements with local meteorological monitoring data, surface drilling operations were repeated at the same place in 2012 (11.5 m long core) and in 2013 (20.5 m long core). Results are also compared here with previously published isotopic composition data measured along the 22 m shallow ice core drilled at the same place in 2004 which covered the period from 1998 until 2004 (Mikhalenko et al., 2005).
Vertical profile of
In 2014, drilling operations were also successful at the Maili Plateau (Mt Kazbek), at an altitude of 4500 m a.s.l. 200 km eastwards from El'brus (Fig. 1), delivering a 20 m ice core. The Kazbek core is shown for purposes of comparison only. A detailed description of it will be published elsewhere.
For the upper and the lower parts of the deep core (0–106 m and 158–181.8 m) and for the shallow firn cores drilled in 2012 and 2013, sampling was performed using classic cutting–melting procedures. For the other depth intervals, melted samples were extracted from the continuous-flow analysis system of LGGE (Laboratoire de Glaciologie et Geophysique de l'Environnement, Grenoble, France), automatically sub-sampled, frozen, and stored in vials for subsequent isotopic analysis. The description of the CFA (continuous-flow analysis) system will be published elsewhere.
The sampling resolution was 15 cm for the upper 16 m of the deep core (see the sketch of the sampling resolution in Fig. 2c). It was then increased to 5 cm in order to achieve better resolution from 16 to 70 m depth and in the bottom part of the core (158–182 m depth). To ensure 15–20 samples per year, the sampling resolution was increased to 4 cm in the depth range from 70 to 106 m, similar to the sampling resolution of the CFA system (3.7 cm).
Samples from the shallow cores drilled in 2012 and 2013 were cut with a resolution of 10 and 5 cm respectively.
The methods for the isotopic measurements have been partially discussed in
Mikhalenko et al. (2015). Water stable isotope ratios (
Moreover, 600 samples from the depth interval from 23 to 35 m were measured
in the Laboratory of Isotope Hydrology of the IAEA (Vienna, Austria). The two
records are highly correlated (
We also stress the close overlap of the upper part of the profiles of the water stable isotope records versus depth from the different cores drilled in 2009, 2012, and 2013 (Fig. S2a). Based on this close agreement within the different shallow firn cores, we decided to calculate a stack record for the period from 1914 until 2013, which is used for dating hereafter.
In the depth interval from 100 to 106 m depth, we also have an overlap of samples obtained with a classic cutting method and the CFA method described above, without any significant difference (Fig. S2c), again allowing us to combine the two records into one stack record.
The chronology is based on the identification of annual layers. These are
prominent in
Hereafter, we focus our analysis on 100 years, from 1914 until 2013,
which corresponds to the total of 140 m of the ice thickness studied here
(the 15 m covered by the shallow cores plus the 126 m covered by the deep
ice core). This period has been chosen because at this depth, the age scale
is well defined by the time horizon found slightly below (Katmai 1912),
resulting in a relatively small dating uncertainty of
For warm and cold season allocation, we used a method adapted slightly from
Vinther et al. (2010). The original method requires ascribing an equal
accumulation rate for the warm and cold season of each year. Basically, we
used the same approach as there is an obvious seasonal cycle of
Figure 3 illustrates the identification of seasons using the isotopic composition seasonal cycle. In the meteorological data we used the period from November to April for the cold season and May to October for the warm season.
Illustration of the scheme used to identify warm and cold half-years
(respectively indicated by the light red and light blue shaded areas) based
on the deviation of the mean
Average seasonal cycle of temperature (black dots and line) and
precipitation (grey bars) calculated over 1966–1990 period,
There are some gaps in the isotopic composition data that resulted from technical
problems during the drilling operations and the process of analysis. The
drilling problems are described in (Mikhalenko et al., 2015). The biggest gap
appears at the depths of 31.3 and 32.1 m. A piece of the core was lost during
the drilling operations. This part is covered by the bottom part of the 2004
core where the sampling resolution was 50 cm. It is evident that two seasons
(one warm and one cold) are partially missing. We did not use these values
for the correlation analysis because of the large uncertainty of the seasonal
value calculations in this case. In case of a missing sample we considered
its isotopic value to be the average of the two neighbouring samples. For
a detailed description of the raw isotopic data and annual layer allocation
for the upper 106 m of the core, please refer to Mikhalenko et al. (2015).
Mean annual and seasonal values of
Annual variations in
Annual variations in deuterium excess in the warm season (red line), in the cold season (blue line), and mean annual values (green line). Thick lines show the 10-year smoothed values, and the thin ones display the raw values.
The annual accumulation rate is calculated as the thickness of the seasonal layer, multiplied by the layer density using the density profile from Mikhalenko et al. (2015), and corrected for layer thinning using the Nye model (Nye, 1963; Dansgaard and Johnsen, 1969), with the following parameters: accumulation rate – 1.583 m ice equivalent; pore close-off depth – 55 m (Mikhalenko et al., 2015).
Description of meteorological and instrumental data used in the paper.
We calculated the potential influence of diffusion on the stable isotopes
record according to the Johnsen et al. (2000) model. We used the following
parameters for the calculation. Our calculation showed that the seasonal
amplitude of
We used the daily meteorological data (precipitation rate and mean daily temperature) from several weather stations around the drilling site (see map in Fig. 1 and Table 1) for comparison with the ice core data. We also investigated records of precipitation isotopic composition based on monthly sampling, performed at three stations to the south of the Caucasus within the WMO-IAEA GNIP program (Table 1).
For comparison we used the NCEP/NCAR reanalysis temperature data (Kalnay et al., 1996) for the 500 mbar level, which corresponds to the drilling site altitude. Two different models were used to calculate back trajectories: FLEXPART (FLEXible PARTicle dispersion model; Forster et al., 2007; Stohl and Thompson, 1999) and HYSPLIT (Hybrid Single Particle Lagrangian Integrated Trajectory Model; Draxler and Hess, 1998; Stein et al., 2015; Rolph, 2016). The LMDZiso model was used to estimate the precipitation isotopic composition at the drilling site (Risi et al., 2010).
Circulation of the atmosphere sufficiently influences isotopic composition of the ice cores (Casado et al., 2013, and references therein). Atmospheric circulation is quantitatively characterized by circulation indices. In this research we used three indices: NAO, AO, and NCP; these are widely used to characterize European climate (Jones et al., 2003; Thompson and Wallace, 2001; Brunetti et al., 2011, and references therein). Time span and references for the indices are presented in Table 1.
NAO characterizes the type of circulation in Europe, the strength of Azores
maximum, and the Icelandic minimum. The positive values of the NAO index
correspond to the lower than usual value of the atmospheric pressure in
Iceland and the higher than usual value of atmospheric pressure in the
Azores. The negative index corresponds to the less prominent centres of
action in the Northern Hemisphere. Usually this index is calculated as a
difference of atmospheric pressure measured at Reykjavík and Lisbon,
Ponta Delgada or Gibraltar. Here we used data from Vinther et al. (2003) and
The AO is also a characteristic of the Northern Hemisphere circulation. It is
used to analyse climatic variability with periods longer than 10 years. It is
calculated as EOF (empirical orthogonal function) of 500 hPa surface. Negative values correspond to high
pressure at the North Pole and the cooling of Europe, while positive values
correspond to low pressure at the North Pole and the drying of the Mediterranean (Thompson
and Wallace, 2001). We used AO data from NOAA
(
The NCP index is less widely used although it has been proved that it is convenient to use it in Mediterranean climate studies
(Kutiel et al., 1996; Brunetti et al., 2011). The index is calculated as a
normalized difference of geopotential heights between the Caspian and
northern seas. Positive values correspond to stronger meridional circulation
in Europe and lower summer temperatures, while negative values reflect the
strengthening of zonal circulation and higher summer temperatures in Europe
(Brunetti et al., 2011). We used NCP data from NOAA
(
The main peculiarity of the drilling site is its location on the border between subtropical and temperate climatic zones (Volodicheva, 2002). Back-trajectory calculations show that the drilling site is characterized by remarkable seasonal differences in the locations of moisture sources. In winter, the origin of air masses varies from the Mediterranean to the North Atlantic. In summer, local moisture sources from the surrounding continents or from the Black Sea are predominant (see Fig. S1 for examples).
Meteorological data depict large regional variations in the seasonal cycle of precipitation. To the south of the Caucasus, there is no distinct seasonal cycle (Fig. 4a), showing the climatology for the Klukhorskiy Pereval station. In fact, the Klukhorskiy Pereval station is situated north of the main ridge, but in terms of the seasonal cycle of precipitation it undoubtedly belongs to the southern group. However, we are nevertheless using this station as an example because of the uninterrupted record of temperature and precipitation for the 1966–1990 period. By contrast, the north of the Caucasus is marked by a distinct seasonality in precipitation amounts, which are at a maximum in summer and a minimum in winter (Fig. 4b), showing the climatology for the Mineral'nyye Vody station. More examples of the Caucasus weather stations' climatologies are given in Mikhalenko et al. (2015). Moreover, the annual precipitation rate to the south of the Caucasus is much higher than to the north. For example, the typical annual precipitation rate to the north of the Caucasus at an altitude close to sea level is 500 mm per year, while to the south of the Caucasus at the same altitude it is about 1500 mm. The amount of precipitation in the region is affected by the altitude and the distance from the sea shore.
The seasonal changes in temperature appear uniform throughout the region
surrounding the Caucasus, with the warmest conditions observed in summer and
the coldest observed in winter. The seasonal amplitude depends on the
distance from the sea and the mean annual temperature depends on the
altitude. The average regional lapse rate was calculated using the available
meteorological data. We used the data from all the stations for the
calculation. The lapse rate is lowest in December–February (2.3
Monthly
Normalized regional temperature record based on meteorological data,
with respect to the reference period 1966–1990, expressed as annual
anomalies (
Based on the lapse rate, we calculated the temperature at the drilling site
taking into account its seasonal variability shown on the Fig. S3. This
record was used for the estimation of the
Correlation coefficients between meteorological data and
indices of large-scale modes of variability (statistically significant
coefficients at
We also compared the data from meteorological stations with the NCEP
reanalysis (Kalnay et al., 1996) outputs (not shown) for the 500 mbar level.
Despite the difference in absolute values on a daily scale when compared with
the AWS data (the difference is random and varies from
We then investigated long-term trends in the meteorological records. Mean annual temperatures show a significant increase during the last 2 decades. We also observe higher than average values of mean decadal temperature in 1930–1940. The beginning of the observations in the region, i.e. the period from 1881 until 1900, was as cold as the 1990s. It is evident that the last 20 years in the warm season were the warmest for the whole observation period (Fig. 8), while in the cold season the recent warming is not unprecedented. For example, cold seasons in the 1960s–1970s were even warmer (Fig. 8). Multi-decadal patterns of temperature variations also differ in the late 19th century, where negative anomalies are identified in cold season temperature (Fig. 8) but not in warm season temperature (Fig. 8). On the other hand, in cold season temperatures we can observe lower temperatures at the end of the 19th century that might be due to the impact of volcanic eruptions (Stoffel et al., 2015). We also noted the high temperature values in the 1910s–1920s, which are not completely understood. We did not find any trends in the precipitation rate for any of the groups of stations (Fig. S4).
A significant anti-correlation is observed between temperature and the NAO index, both in the cold and warm seasons (Table 2, the information about the time series used for the correlation analysis can be found in Table 1). Stronger anti-correlations are identified between temperature and the NCP index, especially in the cold season, as also reported by Brunetti et al. (2011). Relationships with indices of large-scale modes of variability are systematically weaker for precipitation, with contradictory results for the south/north Caucasus stack; they appear significant for the NCP in both seasons (Table 2).
GNIP data are only available at low-elevation stations. They show a rather
uniform distribution of the isotopic composition of precipitation in the
region during summer and a gradual depletion of
GNIP records are too short and intermittent (1–2 years with gaps) to
investigate the variability and relationships with the local temperature on
an interannual scale. We therefore restrict the discussion of GNIP data to
seasonal variations. The
Climate variability as a driver for glacier variations in the Caucasus has
recently been explored by several authors. Elizbarashvili et al. (2013) found
an increased frequency of extremely hot months during the 20th century,
especially over eastern Georgia, whereas the number of extremely cold months
decreased faster in the eastern than in the western region. In addition, the
highest rates for positive trends in annual mean air temperature can be
observed in the Caucasus Mountains. Shahgedanova et al. (2014) evidenced
significant glacier recession on the northern slopes of the Caucasus,
consistent with increasing air temperature of the ablation season. They
report that the most recent decade (2001–2010) was 0.7–0.8
The comparison of the four cores obtained on the western plateau of El'brus shows similar variations during overlap periods (see Fig. S2). We therefore calculate a stack record for each season, based on the average value of individual ice cores for the overlapping seasons. The inter-core disagreement is almost negligible (Fig. S2) and can be explained by a different sampling resolution.
We note that the shallow ice core from the Maili plateau of Kazbek shows the
same mean values of
The interannual variability in isotopic composition is about twice larger in
the cold season than in the warm season for
The
Mean characteristics of the El'brus ice core records, calculated for the period from 1914 to 2013.
No significant (
Our deuterium excess record (Fig. 2b) does not depict any robust seasonal
variation. Moreover, the distribution of deuterium excess as a function of
We compared the ice core data with the regional meteorological data and the large-scale modes of variability. The result of the correlation analysis is summarized in Table 4. Multiannual variations in the parameters are shown in Fig. 9 for the cold season and in Fig. 10 for the warm season.
Comparison of the ice core record with instrumental regional climate
information for the cold season:
We found no significant correlation between the ice core
Obviously, the above inferences strongly depend on the uncertainties of the
timescale used. If one concedes that the error of the timescale could be
significantly greater than
Same as Fig. 9 but for the warm season.
We also did not find any statistically significant correlations when we
compared 3-, 5-, and 7-year running means of these parameters. This result
implies that the isotopic composition at El'brus is controlled by both local
and regional factors such as changes in moisture sources. The possibilities
for accurate reconstructions of past temperatures are therefore limited. For
a more accurate investigation of the
Correlation coefficients between ice core data, meteorological data,
and indices of large-scale modes of variability (statistically significant
coefficients at
Our results are comparable to those obtained in the Alps by Mariani et
al. (2014) for the Fiescherhorn glacier where the authors found significant,
though weak, correlation between temperature and
Other research carried out in the Alps by Bohleber et al. (2013) revealed a significant correlation between modified local temperature and the ice core isotopic composition on a decadal scale. The authors also report that there are some periods of correlation absence. The main finding is that for the periods of less than 25 years the difference between the modified dataset according to the authors' method and original dataset temperature is crucial, but for longer periods the two temperature datasets are close to each other. That conclusion implies that the isotopic composition reflects the local temperature in the high mountain regions to a limited extent. It seems to be impossible to calculate the modified temperature for the Caucasus region according to the methods described by Bohleber et al. (2013) because of the relatively short and sparse original datasets.
The seasonal accumulation rate is seasonal layer thickness corrected for
densification using the density profile from Mikhalenko et al. (2015) and for
the layer thinning due to glacier flow using the Nye model (Nye, 1963;
Dansgaard and Johnsen, 1969). It is linked to the precipitation rate at the
stations situated south of the Caucasus in both seasons (
The calculation of the seasonal cycle of precipitation isotopic composition using the LMDZiso model (Risi et al., 2010) does not correspond to the results obtained from the ice core in absolute values or in amplitude (Fig. S5). This can be explained by a complicated relief of the region that strongly influences the isotopic composition, but it is not taken into account in the model. Also, in summer, El'brus is in a local convective precipitation system that is not included in the model.
We did not find any statistically significant correlations between ice core
data and large-scale modes of variability when using the mean annual values.
We present the results of calculations in the Table 4. We report a weak
though significant (
For the cold season, the ice core
We explored the links between the ice core parameters (
No significant correlation was identified between deuterium excess and indices of large-scale modes of variability. So far, no regional or large-scale climate signal could be identified in El'brus deuterium excess. Further investigations using back trajectories and diagnoses of moisture source and evaporation characteristics will be needed to explore further the drivers of this second-order isotopic parameter.
We found no persistent link between ice cores
Our ice core records depict large decadal variations in
Based on regional meteorological information and trajectory analyses, the
main moisture source is situated not far from the drilling site in the warm
season, and consists of evaporation from the Black Sea and continental
evapotranspiration. Changes in regional temperature during the warm season
may affect the initial vapour isotopic composition as well as the atmospheric
distillation processes, including convective activity, in a complex way. This
may explain the significant, albeit non-persistent, correlation between
summer
Our data can be used in atmospheric models equipped with water stable isotopes, for instance to assess their ability to resolve NAO–water isotope relationships (Langebroek et al., 2011; Casado et al., 2013). The accumulation rate at the drilling site is significantly correlated with the precipitation rate and gives information about precipitation variability before the beginning of meteorological observations.
Data for this paper can be found in the Supplement.
The authors declare that they have no conflict of interest.
The research was supported by the RFBR grants 14-05-31102 mol_a and 17-05-00771 a. The analytical procedure ensuring a high accuracy of isotope data obtained at CERL was made possible with financial support from the Russian Science Foundation, grant 14-27-00030. The study of dust layers was conducted with the support of RFBR grant 14-05-00137. The measurement of the samples in IAEA was conducted according to research contracts 16184/R0 and 16795. This research work was conducted within the framework of the International Associated Laboratory (LIA) “Climate and Environments from Ice Archives” 2012–2016, linking several Russian and French laboratories and institutes. We thank Obbe Tuinenburg and Jean-Louis Bonne for the back-trajectory calculations. We thank Alice Lagnado for improving the English. We are grateful to four anonymous reviewers and the editor Shugui Hou for their comments, which helped to improve the paper. Edited by: S. Hou Reviewed by: four anonymous referees