Investigation of a deep ice core from the Elbrus western plateau, the Caucasus, Russia

. A 182 m ice core was recovered from a borehole drilled into bedrock on the western plateau of Mt. Elbrus (43 ◦ 20 (cid:48) 53.9 5115 in the Caucasus, Russia, in 2009. This is the ﬁrst ice core in the region that represents a paleoclimate record that is practically undisturbed by seasonal melting. Relatively high snow accumulation rates at the drilling site enabled the analysis of the intraseasonal variability in climate proxies. Borehole temperatures ranged from − 17 ◦ C at 10 m depth to − 2.4 ◦ C at 182 m. A detailed radio-echo sounding survey showed that the glacier thickness ranged from 45 m near the marginal zone of the plateau up to 255 m at the glacier center. The ice core has been analyzed for stable isotopes ( δ 18 O and δ D), major ions (K + , Na + , Ca 2 + , Mg 2 + , NH + 4 , SO 2 − 4 , NO − 3 , Cl − , F − ), succinic acid (HOOCCH 2 COOH), and tritium content.

the instrumental climatic records. Some proxies have seasonal to annual resolution, and can be combined into large networks covering continental and even global scales. Individual proxies can be calibrated with instrumental data, resulting in time series appropriate for statistical analyses and numerical modeling. Due to both the urgency of climate change, and our increased ability to synthesize paleoclimate data with future projections, it is essential to have reliable regional paleoclimate reconstructions for the last millennia (Vaughan et al., 2013). The study of chemical impurities in glacier snow and ice permits the reconstruction of our changing atmosphere from the pre-industrial era to present-day (see Legrand and Mayewski, 1997, for a review).
Ice cores from polar glaciers that result in multi-millennial records due to minimal disturbance by melt/refreeze processes are presently considered to be the best representation of past climate conditions at hemispheric scales. However, calculations based on observational data trends in the major climatic characteristics show highly pronounced regional variability. Such variability is reproduced by modern climate models and can be projected into the future (AMAP, 2011), but the reliability of the simulations depends on the amount and the quality of existent data, and some of the results such as the precipitation rate are questionable (Anisimov and Zhil'tsova, 2012).
The need for regional paleoclimate records from non-polar areas has led to the development of numerous reconstructions of annual and seasonal resolution based on instrumental climate data and paleoclimate proxies. Ice cores from low and mid-latitude high mountain glaciers can reconstruct past atmospheric conditions in areas with long human histories. A number of studies examined climate and environmental changes in various non-polar areas (Vimeux et al., 2009;Thompson, 2010) including the European Alps (Barbante et al., 2004;Preunkert and Legrand, 2013;, the continental Siberian Altai (Eichler et al., 2011), and Kamchatka (Kawamura et al., 2012;Sato et al., 2014).
Climate records located in the region of interest often best represent the climate variability from the region itself. Despite their temporal length and their continuous records, the Greenland and Antarctic ice-core data are from sites that are very remote from most inhabited areas. Therefore, the comparable paleoclimate records derived directly from glaciers in Europe and Asia are highly valuable. However, seasonal melting and water infiltration distort the climate proxies recorded in firn and ice even at high altitudes in the Andes (Ginot et al., 2010), Himalayas (Hou et al., 2013) and the low latitudes of the Arctic islands (Kotlyakov et al., 2004).
The documented conditions (Tushinskii, 1968;Mikhalenko, 2008) near the top of Mt. Elbrus suggest the possibility of a reasonably long climatic record in an ice core not affected by meltwater infiltration. Relatively high accumulation on the western plateau (Mikhalenko et al., 2005) assures high temporal resolution of the ice-core data with the possibility of seasonal variations in the analyt-ical results (Werner et al., 2000). Due to this combination of factors, we were motivated to recover ice cores from the western plateau of Mt. Elbrus to obtain natural archives that preserve environmental data associated with atmospheric chemistry, dust deposition, biomass burning, anthropogenic emission, and climate change in the Caucasus (Mikhalenko, 2010). The aim of the Elbrus drilling project is to reconstruct past climate and environmental changes for the Caucasian region from the ice core. Here, we provide an overview of the existing geographical, glaciological, meteorological, and climatological knowledge from the region, and then focus on the glaciological and glacio-chemical characterization of a new drilling site located on the western plateau of Mt. Elbrus. We use stable isotopes, glacial-chemical records and simplified thermomechanical modeling to create a chronology for the 182 m Elbrus ice core. Finally, we present the possibilities to develop the high-resolution regional paleoclimate reconstruction from this ice core.

Geographical and glaciological characteristics of the Caucasus region
The Caucasus are situated between the Black and the Caspian seas, and generally trend east-southeast, with the Greater Caucasus range often considered as the divide between Europe and Asia. The glaciers in the Caucasus cover an area of around 1121 ± 30 km 2   (Fig. 1). Glacier studies in the Caucasus began more than 100 years ago and mainly focused on glacier mapping (Pastukhov, 1893;Podozerskii, 1911) or reconstructing past glacier positions by geomorphological methods (Abich, 1874;Mushketov, 1882;Kovalev, 1961;Serebryannyi et al., 1984). Records of contemporary glaciological processes were obtained during the International Geophysical Year (IGY) in 1957-1959(Tushinskii, 1968 when the climatic conditions of the glacial zone, accumulation and ablation of the glaciers, glacier runoff, glacier ice formation zones, and snow and firn stratigraphy were investigated. These studies were mainly conducted on the southern slope of Elbrus extending from the glacier tongue to the summit (Fig. 1b), and determined that surface snowmelt did not occur above 5000 m (Troshkina, 1968). Complex studies of mass, water, and the heat balance of glaciers in the Caucasus were started during the International Hydrological Decade (1964)(1965)(1966)(1967)(1968)(1969)(1970)(1971)(1972)(1973)(1974) (Golubev et al., 1978;Dyurgerov and Popovnin, 1988;Krenke et al., 1988). A number of studies examined fluctuations in glacier dimensions and volume (Stokes et al., 2006;Kutuzov et al., 2012Kutuzov et al., , 2015Nosenko et al., 2013, Shahgedanova et al., 2014, glacier mass balance (Rototaeva and Tarasova, 2000), and regional snow chemistry (Kerimov et al., 2011). Characteristics of the mineral dust and its source were inves- tigated using records of shallow ice cores and snow pits from Elbrus (Kutuzov et al., 2013;. In addition to the glaciological studies, multiple tree-ringbased reconstructions represent mean summer air temperature, river runoff, and glacier mass balance in the region Solomina et al., 2012). The first regional lake sediment cores retrieved in 2010, 2012, and 2013 demonstrate an excellent potential for using lacustrine records to study long-term climate and glacier history variations .
Despite the substantial glacier area in the Caucasus, few suitable sites for ice-core research exist due to the relatively low elevation and considerable surface melt below 5000 m. Several shallow and intermediate depth ice cores have been recovered at the Caucasus glaciers (Golubev et al., 1988;Zagorodnov et al., 1992;Bazhev et al., 1998), but these previous studies were conducted at sites where considerable meltwater percolation smoothed isotopic and geochemical profiles. However, the vast high-elevation plateaus on the glaciers of Elbrus (5642 m), Kazbek (5033 m), and Bezengi (∼ 5000 m) (see Fig. 1b) present promising sites for obtaining ice-core records.

Geographical and glaciological characteristics of Elbrus
Elbrus, the highest summit of the Caucasus, has two peaks at its highest elevations with both an eastern (5621 m a.s.l.) and western (5642 m a.s.l.) summit where the whole complex is covered by glaciers with a total area of 120 km 2 (Zolotarev and Khar'kovets, 2012) (Fig. 1). Elbrus is an active volcano but only minor fumarole activity is currently observed (Laverov et al., 2005). Glaciers on Elbrus are situated in the altitudinal range of 2800 to 5642 m. Stratigraphic records display several ice formation zones on Mt. Elbrus (Bazhev and Bazheva, 1964;Psareva, 1964;Troshkina, 1968). The coldest conditions occur above 5200 m a.s.l., where the mean summer air temperature does not exceed 0 • C, while the Elbrus glaciers between 4700 and 4900 m a.s.l. have limited surface melt. Ice lenses up to 30 cm thick alternate with firn horizons in the uppermost snow and firn at 5050 m a.s.l. (Mikhalenko, 2008). Snow accumulation measurements from 1985 and 1988 demonstrate total snow accumulation of 400-600 mm w.e. a −1 with considerable wind-driven snow erosion at the col of Elbrus (5300 m a.s.l.; Fig. 1c). The snow/firn temperature measured at a depth of 6 m was −14 • C at the col, indicating absence of meltwater runoff from this zone.
Long-term (1983 to present) mass-balance records of Garabashi Glacier show negative values for the period since 1994. Extremely high summer temperatures and glacier melting accompany this negative trend. Garabashi Glacier surface elevation has thinned by 3.2 m over the last decade near the equilibrium line .
A 76 m long ice core was recovered in the accumula-  (Zagorodnov et al., 1992). The firn in this ice core completely transformed into ice as a result of meltwater refreezing at 23-24 m depth. Thus, the geochemical profiles obtained from the ice core were smoothed by meltwater percolation and could not be used for paleoclimate and environmental reconstruction. The next ice core drilled in this region was recovered on the western plateau of Elbrus, at 5115 m a.s.l. (Fig. 1). The plateau area is ∼ 0.5 km 2 and is bordered to the south and southeast by two lava ridges, and by a vertical wall of Mt. Elbrus to the east. The first ice-core drilling campaign during 4-6 July 2004 resulted in a 21.4 m ice core with associated borehole temperature and glacier thickness measurements (Mikhalenko et al., 2005). The 10 m depth temperature of −17 • C indicated that any meltwater refreezes at only a few centimeters below the surface and thus isotopic and soluble ions profiles are preserved. Ice-core records of this first shallow ice core indicated good preserved seasonal stable isotopic (δ 18 O and δD) oscillations and mean annual accumulation rates of approximately 1200 mm w.e.

Climatology of the Caucasus and Elbrus
The summer atmospheric circulation pattern in the Caucasus is dominated by the subtropical high pressure to the west and the Asian depression in the east. In the winter, circulation is affected by the western extension of the Siberian High (Volodicheva, 2002). The Caucasus are located in the southern section of the vast Russian Plain and are therefore buffeted by the unobstructed passage of cold air masses from the north. High elevation ridges in the southern Caucasus deflect air flowing from the west and southwest. The influence of the free atmosphere on the Elbrus glacier regime is significantly larger than local orographic effects as the glacier accumulation area lies above main ridges.
Most of the annual precipitation occurs in the western and southern sections of the Caucasus, reaching 3240 mm a −1 at Achishkho weather station (1880 m). Precipitation ranges between 2000 and 2500 mm a −1 at 2500 m a.s.l. in the west and declines to 800-1150 mm a −1 in the east on the northern slope of the Caucasus. Precipitation ranges from 3000-3200 mm a −1 in the west to 1000 mm a −1 in the east for the southern macroslope. The proportion of winter precipitation (October-April) also declines eastward from more than 50 to 35-40 % for the northern Greater Caucasus and from 60-70 to 50-55 % for the southern slope (Rototaeva et al., 2006). The proportion of solid precipitation increases with altitude and reaches 100 % above 4000-4200 m. The altitude of the glacier equilibrium line (ELA), tends to increase from 2500-2700 m in the Belaya, Laba, and Mzymta river basins in the west to 3700-3950 m in the Samur and Kusurchay basins in the eastern sector of the northern macroslope of the Caucasus.
Mean summer (May-September) air temperature at the ELA ranges from 6-7 • C in the west to 1-2 • C in the east. The ELA is much higher on the glaciers of the northern macroslope, especially in the central Caucasus, where the ELA on the northern slope of Elbrus is 1000 m higher than on Svanetia glaciers 80 km southward. The number of highelevation meteorological stations is very limited in the Caucasus. Figure 2 shows the mean monthly air temperature and precipitation at the Klukhor Pass, Teberda, and Terskol meteorological stations in the western and central Caucasus (Fig. 1, Table 1).
Air temperatures at these stations are in good agreement and correlate well with lowland stations (r = 0.7-0.9, p < 0.01), indicating the homogeneity of the temperature regime for the investigated area (Solomina et al., 2012). Vari-The Cryosphere, 9, 2253-2270, 2015 www.the-cryosphere.net/9/2253/2015/ ations in mean annual and monthly temperatures for the Klukhor Pass station for the period of observation (see Ta-ble 1) do not display a statistically significant trend. A positive trend for mean annual temperature (r = 0.33, p < 0.05) and a slight positive trend for summer temperature occur at the Teberda station. Temperature records from the Terskol station located 7 km southward from the Elbrus glaciers show a negative mean annual temperature trend for the whole period of observation (r = −0.35, p < 0.05) (Solomina et al., 2012) but mean summer (May-September) temperatures increased from 11.5 • C in the 1987-2001 period to 12.0 • C over the last decade. Winter precipitation increased by 20 % over the same period, while summer precipitation did not show any change . The first meteorological measurements were taken on the Elbrus glaciers in 1934-1935 by an expedition of the USSR Academy of Sciences (Baranov and Pokrovskaya, 1936). Air temperatures, pressure, humidity, wind regime, and incoming solar radiation have been measured at four sites from Terskol at 2214 m a.s.l. to the col of Elbrus at 5300 m. A permanent meteorological station was established near Priyut-9 on the southern slope of the Garabashi Glacier at 4200 m a.s.l. in 1934. 1949-1952 data demonstrate a mean annual air temperature of −9.2 • C. The temperature of the coldest month (January) was −17.1 • C, while the July temperature was −0.5 • C. The minimum air temperature of −36.1 • C was measured on 30 January 1950, with a maximum of 10.7 • C on 1 August 1950. An annual precipitation rate of 1128 mm was observed for 1949-1952. The summer months (April-October) contribute 75 % of the total precipitation, while the winter months (November-March) account for only 25 % (Matyukhin, 1960). The maximum wind speed at Priyut-11 station of 56 m s −1 was measured in January 1952.
During the IGY (1957)(1958)(1959) the permanent all-year meteorological station was established on the glacier base on the southern slope of the Elbrus at 3700 m a.s.l. Meteorological records from this site include diurnal air pressure and temperature, precipitation, humidity, cloudiness, wind regime, and snow cover thickness (Tushinskii, 1968). Heat balance, air temperatures and wind speed were recorded during occasional observations in the col of Mt. Elbrus (5300 m). The first accumulation and ablation measurements on the southern slope of Mt. Elbrus were done during the IGY and in 1961-1962(Bazhev and Bazheva, 1964.

The western Elbrus plateau glacier archive
In the following section we will present recent meteorological, glaciological, and glacio-chemical investigations conducted on the western Elbrus glacier plateau with the aim of obtaining knowledge about the suitability of this site to obtain relevant atmospheric ice-core records.

On-site meteorological measurements
An automatic weather station (AWS) from AANDERAA Data Instruments was installed on the western Elbrus plateau at 5115 m a.s.l. at the drill site in 2007. The AWS was working between 30 July 2007 and 11 January 2008, but disappeared afterwards under unascertained circumstances. Here we discuss records until 12 October 2007, comprising the period with uninterrupted, consistent data. Air temperature, wind speed and direction, humidity, air pressure, radiation balance, and snow cover thickness have been measured with a time resolution of 1 h. According to AWS records, mean daily air temperatures were negative during the period of observations. Hourly averaged temperatures were also negative, while the maximum un-averaged air temperatures were recorded on eight occasions and ranged from 0.1 to 3.1 • C. Mean hourly averaged wind speed on the drilling site was 2.9 m s −1 throughout the entire period of observation. Wind gusts up to 21.4 m s −1 were observed when fonts passed the station while the mean daily maximum wind speed was 6.7 m s −1 in August-September 2007. Our data did not cover the whole year but according to measurements from 1961 to 1962, the average wind speed was approximately 30 % higher in the winter on the southern slope of Elbrus (Tushinskii, 1968). A combination of high snow accumulation and the relatively low average wind speed from the prevailing westerlies allows us to assume that most of the precipitation did not move far from its depositional site and was not scoured by wind.
AWS records were compared with measurements from the mountain meteorological station Klukhor Pass (2037 m a.s.l.; 50 km westward) and the lowland Mineralnye Vody station (316 m a.s.l.; 120 km northeastward) (Table 1) as well as with the 20th Century Reanalysis V2 data provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, (http: //www.esrl.noaa.gov/psd/) (Fig. 3a, b). A temperature lapse rate of 0.6 • per 100 m elevation was observed during the summer months. In winter, however, the lapse rate decreases due to temperature inversions at the Mineralnye Vody station. There is a good agreement between the temporal variations of mean daily air temperature measured by the AWS at the drill site, and evident in the data of 20th Century Reanalysis and other meteorological stations (r > 0.85). Therefore the temperature variations at the West Elbrus plateau are consistent with the regional temperature regime.
In an AWS DAVIS Vantage Pro 2 including air temperature, humidity, and wind speed at 0.5 and 2.0 m, with 15 min resolution (Figs. 1,4). Along with the estimation of eddy flux of heat and moisture, we measured the fluxes of total, scattered, and reflected radiation. Meteorological conditions during the observation period encompassing the maximum insolation at the summer solstice were close to mean annual parameters. Downward short-wave radiation varied between 1 and 1.2 kW m −2 which is 73-88 % of the solar constant at the outer boundary of the atmosphere and 78-93 % of total insolation at 43 • N latitude during that time of year. Albedo has a dominant role in the short-wave radiation balance. Mean albedo values of 0.66 were measured at the plateau in June 2013. Initial measurements of radiation balance were conducted in Elbrus in  and showed that downward short wave radiation ranged from 1.1 kW m −2 at an elevation of 3750 m a.s.l. up to 1.2 kW m −2 at 5300 m a.s.l. (Tushinskii, 1968). Despite the negative air temperatures, the radiation www.the-cryosphere.net/9/2253/2015/ balance was positive except for during the night. The mean value of the radiation balance encompassing both short-wave and long-wave radiation was 150 W m −2 , affecting surface melt and snow recrystallization.

Ground-based survey: surface topography and radar sounding
Detailed measurements of ice thickness were carried out in 2005 and 2007 using the monopulse ice-penetrating radar VIRL with a central frequency of 20 MHz (Vasilenko et al., 2002(Vasilenko et al., , 2003. VIRL ice-penetrating radar consists of a transmitter, receiver, and digital recording system with GPS. In order to synchronize the transmitter and receiver, we used a special radio channel with a repetition rate of 20 MHz. We modified the advanced VIRL-6 radar with an optical channel in 2007 (Berikashvili et al., 2006) allowing simultaneous recording and controlling. The time interval ranged between 1 and 99 s for obtaining both radar and navigation data as well as for performing the hardware and program stacking (from 1 to 6192 times) of wave traces. The average radio wave velocity (RWV) in firn and ice can calculate ice thickness from measured time delays of radar signals reflected from the bedrock. RWV depends on firn/ice density and temperature. We did not measure RWV (V ) at the western Elbrus plateau, but calculated the result as a function of glacier depth (z) through measured ice-core density ρ d (z) and borehole temperature T (z) profiles: where c = 300 m µs −1 -radio wave velocity in air; ε (ρ d , T ) is the dielectric permeability of snow, firn, and ice as a function of density ρ d (z) and temperature T (z) (Macheret, 2006). ε (ρ d ) was calculated for two component dielectric mixture of ice and air (Looyenga, 1965): where ρ i = 917 kg m −3 is the density of glacier ice. ε i (T ) was calculated from Mätzler and Wegmüller (1987): The average RWV of 180 m µs −1 was calculated for the 181.8 m ice thickness at the drilling site, taking into account the depth variations from ρ d and T and the measured time delay of radio signal to ice thickness at each point. We combined two data sets, from 2005 and 2007, to construct an ice thickness map. In total, we measured the glacier depth at more than 10 000 sounding points along 6.5 km profiles, with an estimated accuracy of ice thickness measurements of 3 % (Lavrentiev et al., 2010). The maximum depth was 255 ± 8 m at the central part of the plateau, with minimum values of about 60 m near the edge. Radar records and digital elevation model ASTER GDEM averaged for 2000-2009 map the bedrock topography (Fig. 4). ASTER GDEM with an error of ±20 m (ASTER GDEM Validation Team, 2009) is in a good agreement with the 1959 northern Caucasus topographic map and the 1997 digital orthophotomap of Elbrus (Zolotarev and Khar'kovets, 2000).

Methods
Due to the promising glacier archive conditions obtained from the shallow ice coring in 2004 (see Sect. 2.2), a fulldepth ice-core drilling was completed on the western plateau from 27 August to 6 September 2009 (Mikhalenko, 2010). Bedrock was reached at the depth of 181.80 m. Drilling was done in a dry borehole using the lightweight electromechanical drilling system developed by Geotech Co. Ltd., Nagoya, Japan. Technical details of the drill are described in Takeuchi et al. (2004). The recovered ice cores were subjected to stratigraphic observations, packed into plastic sleeves, and stored in a snow pit with temperatures of −10 • C. Ice-core drilling was accompanied by borehole temperature measurements (using thermistor chains which were left for 3 days in the borehole and calibrated before and after the study with an error of ±0.1 • C), and snow pit sampling conducted 30 m to the south of the drilling site. The ice core was shipped in a frozen condition to the cold laboratory of the Lomonosov Moscow State University where detailed stratigraphic descriptions, including photographing each piece of the core, and bulk density measurements, were conducted.
In addition to the 2009 deep ice core, a subsequent 12 m ice core was extracted in June 2012 at the same site to expand the existing ice-core sample set from 2009 to 2012. The 2012 ice core was also used for the dust study of Kutuzov et al. (2013). Finally, in 27-30 June 2013, a 20.36 m ice core was recovered at the same location.
Stratigraphic descriptions of the ice core were carried out using transmitted-light illumination, resulting in 1 mm resolution details of the depths and thickness of individual horizons. The density of firn and ice were measured on 457 individual samples. Figure 5 shows the bulk density distribution with depth. The sharp random outliers from the general profile, especially with the lower values, could result from uncertainties in estimations of samples lengths. The uncertainty increases for the denser and smaller samples.
Ionic species such as ammonium (NH + 4 ) succinate (HOOCCH 2 COO − , also denoted succinic acid) were investigated along the uppermost 157 m of the Elbrus core (122 m w.e.) with the aim of aiding in the sampling of ice core for alpine firn and ice (Legrand et al., 2007a). Pieces of firn and ice were decontaminated in a clean air bench located in a cold room using a pre-cleaned electric plane tool. with a sample resolution decreasing from 10 cm at the top to 2 cm at 157 m depth. For cations (Na + , K + , Mg 2+ , Ca 2+ , and NH + 4 ), a Dionex ICS-1000 chromatograph equipped with a CS12 separator column was employed. For anions, a Dionex 600 equipped with an AS11 separator column was used with an eluent mixture of H 2 O, NaOH at 2.5, 100 mM, and CH 3 OH. A gradient pump system allows determining inorganic species (F − , Cl − , NO − 3 , and SO 2− 4 ) as well as short-chain monocarboxylates (denoted MonoAc − ) and dicarboxylates (denoted DiAc 2− ). For all investigated species, ion chromatography and ice-core decontamination blanks were found to be insignificant with respect to respective levels found in the ice-core samples.
As will be discussed in greater detail in Sect. 3.3.5, the search for volcanic horizons in the Elbrus ice cores requires examining the acidity (or alkalinity) of samples by evaluating the ionic balance between anions and cations where concentrations are expressed in micro-equivalents per liter, µEq L −1 ): Within the present study, we focus (Sect. 3.3.4) on the NH + 4 and succinic acid profiles, in order to (1) define a criterion which allows the separation of winter and summer snow deposition and (2) to apply this criterion to the first 157 m of the Elbrus ice core, in order to establish a depth-age relationship based on both annual layer counting and the NH + 4 and succinic acid depth profiles.
The shallow 2012 and 2013 ice cores and the deep (down to 106.7 m) 2009 ice core were analyzed for deuteriumhydrogen (D / H) and oxygen ( 18 O/ 16 O) isotope ratios using a Picarro L1102-i instrument in the Climate and Environmental Research Laboratory (CERL), Arctic and Antarctic Research Institute, St. Petersburg, Russia. The instrument was calibrated on a regular basis with the isotopic standards V-SMOW, GISP, and SLAP provided by the International Atomic Energy Agency (IAEA) to estimate the precision of the measurements and to minimize the memory effect associated with continuous measurements. The reproducibility of the measurements was ±0.07 ‰ for oxygen isotopes (δ 18 O) and ±0.3 ‰ for deuterium (δD). The CERL laboratory work standard SPB was measured after every five samples. The δ 18 O and δD values are expressed in ‰ units relative to the V-SMOW value. This value is 4-5 times higher than the average heat flux density for the Earth's surface and higher than the mean value for central Caucasus, and may be associated with a heat magma chamber of the Elbrus volcano. Figure 6 also shows the temperature profile measured in the 19 m borehole in 2013 and temperature records obtained in 2004 after the 22 m depth shallow ice-core drilling on the western plateau (Mikhalenko et al., 2005). The good match between records is indicative of the stable temperature regime on the western Elbrus plateau for the last decade. Using the altitudinal temperature gradient estimated in Sect. 3.1 based on western plateau AWS temperature data close to the 2009 drill site and the low-elevation station Mineralnye Vody, we estimate that the annual mean air temperature at the drill site is approximately −19 • C. This value is close to the mean annual air temperature of −19.4 • C, calculated using the general relationship with the ice temperature at the bottom of the active layer (Zagorodnov et al., 2006) and only slightly enhanced compared to the 10 m firn temperatures.

Borehole temperatures
The measured temperature profile shows that basal melting can occur due to ice pressure at the deepest part of the glacier. Potential bottom melting has been estimated using a mathematical model of temperature regimes (Salamatin et al., 2001). Our modeling results demonstrate that basal melting occurs under ice thicknesses of more than 220 m, but that its value does not exceed 10 mm w.e. a −1 .

Bulk density and ice-core stratigraphy
The bulk density profile suggests a change in densification around the critical densities (Maeno and Ebinuma, 1983) of 550 and 840 kg m −3 , and no visible change at 730 kg m −3 , which is consistent with other analyses of density profiles (Hörhold et al., 2011;Ligtenberg et al., 2011). However, the slight decreasing trend in density at depths below the maximum values at ∼ 80 m (Fig. 5), close to the critical density across the whole depth interval, is unlikely to be a systematic error in measurements and needs further investigation. Future research should account for the ice flow characterization and the possible effects related to the "intervening depth interval" (the alternation of layers which have already reached the close-off density, with layers that are still permeable) due to seasonal (Bender et al., 1997) or wind-induced (due to seasonal differences in wind speeds) snow density variability at high accumulation sites. Unlike polar ice cores where the "intervening depth interval" is just a fraction of the whole length of the ice core (Bender et al., 1997), the measured bulk density in the Elbrus ice core spans a wide interval between 800 and 915 kg m −3 towards the bottom of the glacier (Fig. 5). Comparing Elbrus's density profile with the results from the Salamatin et al. (2009) densification model demonstrates that there is an increase in the accumulation rate over the past several years. The minimum deviation between simulated and measured ice-core density profiles occurred when the accumulation history was assumed to be similar to the long-term precipitation changes observed at meteorological stations .
According to the morphogenetic classification of stratigraphic features (Arkhipov et al., 2001), two distinct types of layering were observed in the core: firn layers which have not been affected by melting, and ice layers formed by the refreezing of meltwater in the surface snow. The thickness of the infiltration ice layers, which do not form every year, does not exceed 10 mm. Ice formation occurs in cold, dry conditions, as already concluded on the basis of borehole and air temperatures at the drill site (see Sects. 3.1, 3.3.2). The pore close-off depth occurs at around 55 m, where the air bubbles separate from the surrounding ice matrix. This depth coincides with a measured bulk density of around 840 kg m −3 . This density is consistent with the presence of ice layers, as these layers increase the close-off density value above what is expected in ice in which no melting occurs (i.e., 830 kg m −3 ).

Seasonal ice-core stratigraphy of stable water isotopes
The seasonal cycle of the isotopic composition is detectable over the entire measured part of the core (Fig. 7). Mean seasonal values of δD are −200 ‰ for the winter and −25 ‰ for the summer. Values of δ 18 O are about −5 to −10 ‰ in summer and −30 ‰ in winter. Using isotope values to determine annual cycles over 106.7 m of the ice core suggests that this 106.7 m covers 86 years (AD 1924(AD -2009). The mean accumulation rate for this period, based on this dating and taking the firn density and layer thinning into account, was 1455 mm w.e. Figure 7 shows results of isotopic measurements of four different ice cores obtained from the western Elbrus plateau. While 2009, 2012, and 2013 cores were obtained from almost the same location; the 2004 core was drilled 120 m to the southwest. Good agreement in isotopic variations of all cores suggests a relatively homogeneous snow deposition on the plateau. We used the isotope diffusion model of Johnsen et al. (2000) to estimate the preservation of the isotopic signal with diffusive smoothing. Although the drilling site is located in a relatively warm place (−17 • C), the high snow accumulation rate does not favor strong diffusion, since any firn layer rapidly reaches the pore close-off depth. The maximum "diffusion length" at the pore close-off depth is estimated as 5 cm in ice equivalent (IE). The effective diffusion length may be even smaller if we take into account the ice lenses in the firn that prevent the vertical exchange of the water molecules.
Such a diffusion length means that all oscillations shorter than 13 cm IE, will be completely erased due to the diffusion, while oscillations between 13 and 70 cm IE will survive but will be damped to some extent, and the cycles longer than 70 cm (e.g., the annual cycle) IE will not be affected by the diffusion. Thus, if during a single snowfall results in a 35 cm snow layer (that corresponds to 13 cm IE), the isotopic signal of this layer will survive during the diffusion processes and will be seen in the ice core.
Diffusion occurs below the pore close-off depth, but the in ice is much slower than in firn. The final diffusion length solely depends on the time and temperature of the firn-ice thickness. Even if we take a maximum possible temperature (−2.4 • C) and an age estimate of a few hundred years, the additional diffusion in ice will still be very small. This combination leads us to an important conclusion; we may expect to obtain seasonal cycles in the isotope profile down to the very bottom of the core, and our ability to detect the annual cycle in the core depend on the sampling resolution, as well as on such basal processes such as layer folding and mixing.

Seasonal ice-core stratigraphy of chemical parameters and ice-core dating based on annual layer counting
We attempted to date the core by counting annual layers based on chemical ice-core stratigraphical records, as we previously successfully applied such layer counting to midlatitude Alpine ice cores using the NH + 4 signal (Preunkert et al., 2000). Since NH + 4 experiences strong maximum emissions in phase with strengthened summer upward transport of air masses, a particularly well-pronounced seasonal cycle is expected, such as observed at the Col du Dôme Alpine site (Preunkert et al., 2000;Fagerli et al., 2007). However, it appears that the NH + 4 seasonal cycle at Elbrus is less pronounced than in the Alps. Whereas recent summer NH + 4 levels are comparable at both sites, recent winter concentrations at Elbrus are significantly higher than at Col du Dôme.
The first study regarding the seasonality of Elbrus snow accumulation was conducted by Kutuzov et al. (2013) along a short firn core spanning the years 2012-2009. Based on the dust layer stratigraphy of absolute-dated dust events and the stable isotope record the authors showed that the annual deposition at Elbrus has a mean δ 18 O signature of −15 ‰. The isotopic signature consists of nearly equal deposition amounts from the warm season (45 % of total accumulation), where δ 18 O values vary between −5.5 and −10 ‰, and from the cold season (55 % of total accumulation), for which values vary between −17 and −27 ‰.
The concentration distribution of NH + 4 values was inspected in recent firn layers (0-12 m w.e.), and the 50 % concentration limit of 100 ppb was a first approach to separate  4 (a, b), succinic acid (c, d), and δ 18 O (e, f) for different sections of the Elbrus ice core. Red sections demonstrate winter samples based on the following criteria: succinic acid less than 5 ppb, NH + 4 less than 100 ppb for recent years, and less than 50 ppb prior to 1950. Green sections meet the following winter-background criterion: succinic acid less than 3 ppb, NH + 4 less than 50 ppb for recent years, and less than 20 ppb prior to 1950. Black bars in ionic plots refer to the winter criteria. The black bars in the δ 18 O plots refer to the respective mean values. snow deposition arriving from summer and winter precipitation at Elbrus. However this criterion is not conserved in time as the NH + 4 sources are mainly anthropogenic in origin, with an expected NH + 4 trend in summer as well as in winter over the last 100 years. Therefore, a second criterion was used to confirm our winter snow selection. This criterion used succinic acid, a light dicarboxylic acid with an observed strong summer maximum; and a winter level is almost non-existent in the current atmosphere in Europe (Legrand et al., 2007b). The very low winter levels are related to the absence of source as this species is mainly photochemically produced from biogenic precursors. The concentration distribution of succinate values was inspected in recent firn layers (0-12 m w.e.), and the 50 % concentration limit of 5 ppb separates snow deposition arriving from summer and winter precipitation at Elbrus. Winter snow and ice layers were identified when both ammonium and succinate criteria were fulfilled for more than two successive samples. Figure 8a, shows the result of this data dissection over the uppermost 12 m w.e. along with the δ 18 O record (Fig. 8e). The mean δ 18 O level of selected winter data is −19.6 ‰, and as demonstrated in Fig. 8a and c, the winter season selected from ammonium and succinate concentrations matches with winter sections deduced from the δ 18 O profile quite well. However, when examining the δ 18 O variability compared to the major ions, it appears that sometimes the spring season or even the beginning of the summer season may be included. For dating by annual layer counting, this shortcoming is not critical; however, if the data set is inspected at seasonal resolution this definition of the spring season might be a handicap. In this case a stronger criteria (NH + 4 < 50 ppb and succinate < 3 ppb) may be applied in addition to ensure that only deposition corresponding to winter precipitation and associated atmospheric background conditions are selected within the winter period. The mean δ 18 O level of winter data selected using this criteria is −21.1 ‰, whereas seen in Fig. 8a, c, and e, this selection may omit some winter seasons.
Examination of NH + 4 and succinate minima below a depth of 12 m contrasts with results from the European Alps, where in Elbrus the NH + 4 winter level decreases significantly from near the surface to around 70 m w.e. depth (see Fig. 8). Therefore, the NH + 4 winter and background criteria were adjusted using a winter ( from below a core depth of 62 m w.e. In contrast, the succinate winter levels did not change and the 5 ppb criterion applied in recent times was also applied in deeper layers. Figure 8b, d, and f showed a comparison of NH + 4 , succinate with the δ 18 O record between 75 and 78 m w.e. (i.e., from AD 1931(i.e., from AD to 1928. The winter criteria match well with recent winter deposition as deduced from the stable isotope content, although the stable isotope record tends to already be a bit smoothed compared to the uppermost firn layers. As observed for the uppermost core section, we cannot exclude that the winter criteria include parts of an intermediate season, whereas the background criteria select only deposition arriving from the coldest precipitation. Figure 9 shows the result of the dating of the Elbrus ice core. In addition to model calculations detailed in Sect. 3.4, the depth-age scale obtained by annual layer counting using the NH + 4 and succinate criteria is reported down to 122 m w.e. Annual layer counting was achieved as described above down to 85 m w.e. Below 85 m w.e., winter levels became rather thin due to annual layer thinning but also likely due to upstream effects as commonly encountered on such small-scale glaciers (Preunkert et al., 2000). Therefore, below 85 m w.e. ice-core layers, in which fewer than three samples had reached the winter criteria were considered as winter seasons, and from 113 to 122 m w.e. winter layers were also assigned when only one of the two criteria was fulfilled for at least one sample, while the other sample showed only a relative minimum that sometimes exceeded the fixed threshold. This lack of sharp minima could be either due to the fact that winter sections become smaller than our sampling resolution of 2-3 cm applied to core depths below 90 m w.e., and/or could be the result of an incomplete precipitation preservation due to wind erosion upstream of the borehole as ob-served on other small-scale Alpine glacier sites (e.g., Preunkert et al., 2000). In this latter case a systematic lack of winter snow accumulation would occur in the deeper ice-core layers.
Dating based on annual layer counting of the chemical stratigraphy agrees fairly well with the AD 1963 tritium time horizon located at the core depth of 50.7 m w.e. and which is dated at AD 1965 using the ammonium stratigraphy; Fig. 10a. The layer counting results fit well with the dating achieved to 106.7 m based on the seasonal stratigraphy of the stable isotope profile. Whereas stable isotopes predict the year AD 1924 at a core depth of 106.7 m, the chemical stratigraphy leads to an estimate of the year AD 1926 at this depth.
To anchor the depth-age relationship with further absolute time horizons, we inspected the sulfate profile to identify volcanic horizons such as found in other northern hemispheric ice cores between AD 1912 (Katmai) and AD 1783 (Laki) in Greenland Clausen et al., 1997) and at Colle Gnifetti (Bohleber 2008). However, since Elbrus is a volcanic crater, it is sometimes difficult to attribute a peak either to a well-known global eruption or to a local event. Furthermore, numerous sulfate peaks in the Elbrus ice core originate from terrestrial input as suggested by the presence of concomitant calcium peaks. The Katmai eruption in AD 1912 could be clearly identified at 87.7 m w.e. (dated at AD 1911 using the ammonium stratigraphy), with several neighboring samples showing relatively high sulfate levels (up to 1200 ppb, i.e., 25 µEq L −1 ) compared to sulfate peaks generally present in summer layers in the early 20th century. Furthermore, in contrast to neighboring summer sulfate peaks located at 87.2, 87.4, 88.0, and 89.3 m w.e. that are alkaline (see Fig. 10b the 87.7 m w.e. sulfate peak reaches 8 µEq L −1 . Furthermore, samples located at the top part of the 87.7 m w.e. sulfate peak remain neutral in spite of a large presence of calcium. Figure 11 demonstrates that the 1-year uncertainty in dating this horizon is in excellent agreement with our annual counting. We were still able to easily examine annual counting down to 113 m w.e. resulting in a date of AD 1860, but below this depth the dating becomes more uncertain (see the green line in Fig. 9). Below 88 m w.e., seven significant potential volcano horizons are suspected based on their ionic balance and sulfate levels (not shown), from which at least one horizon is of local origin as suggested by small stones up to 1-2 mm in the corresponding ice layer. Nevertheless, a series of three narrow ionic spikes occur at 118-120 m w.e. (dated at ∼ AD 1840-1833), where two of the spikes are characterized by an increase of sulfate and acidity (up to 7.8 µEq L −1 , not shown) that may be related to the well-known eruptions observed in Greenland that are dated to AD 1840 ± 2 years. One of these eruptions may possibly be due to the Coseguina eruption in AD 1835 .
We calculated the depth-age relationship from the depths and thicknesses of the counted annual layers (Fig. 9). Despite high variability in the annual layers' thickness, the data demonstrate layer thinning with depth related to the ice flow. Figure 11. Vertical transect of the western Elbrus plateau glacier along a reference flow line. Predicted ice-particle paths (lines with arrows) and isochrones are shown.
Applying the thickness-age relationship developed by Nye (Dansgaard and Johnsen, 1969) to the actual annual layer data ( Fig. 9) provides the mean accumulation over the whole studied time period in the ice core resulting in 1.583 m of ice equivalent per year. The Nye curve corresponds to the depthage relationship from Nye's model with a best-fit (constant over time) accumulation rate and the glacier thickness at the drilling site (Fig. 9). The green line is the depth-age relationship as suggested by Salamatin's model  with the same best-fit accumulation rate and drill site basal and surface descriptions as assumed when applying Nye's model.
Dating based on annual layer counting of the chemical stratigraphy agrees well with the AD 1963 time horizon that is located at the core depth of 50.7 m w.e. (dated at 1965 using the ammonium stratigraphy, Fig. 10a). In addition it fits very well with the dating achieved so far (i.e., core down to 106.7 m) on the base of the seasonal stratigraphy of the stable isotope profile. Whereas stable isotopes predict the year 1924 at a core depth of 106.7 m, the chemical stratigraphy leads to an estimation of the year 1926 in this depth.
To anchor the depth age relation with further absolute time horizons, a first inspection of the sulfate profile was made with the view to identify volcanic horizons as found in other northern hemispheric ice cores between 1912 (Katmai) and 1783 (Laki eruption) in Greenland Clausen et al., 1997) and at Colle Gnifetti (Bohleber, 2008). However since the Elbrus is an in active volcanic crater, it is sometimes difficult to attribute a peak either to a wellknown global eruption or to a local event. Furthermore, numerous sulfate peaks in the Elbrus ice core originate from terrestrial inputs as suggested by the presence of concomitant calcium peaks. So far, the Katmai eruption in 1912 could be clearly identified at 87.7 m w.e. (dated at 1911 using the ammonium stratigraphy) with several neighboring samples showing relatively high sulfate levels (up to 1200 ppb, i.e., 25 µEq L −1 ) compared to those seen in sulfate peaks generally present in summer layers of the early 20th century. Furthermore, as seen in Fig. 10b, in contrast to neighboring summer sulfate peaks located at 87. that are alkaline (see Fig. 10b), the acidity of samples of the 87.7 m w.e. sulfate peak reaches 8 µEq L −1 at the bottom part of the sulfate peak. Furthermore, samples located at the top part of the 87.7 m w.e. sulfate peak remain neutral in spite of a large presence of calcium (similar to those seen in neighboring summer sulfate peaks). As seen in Fig. 9 it appears that within 1-year uncertainty, this horizon is in excellent agreement with our annual counting. Below 88 m w.e., we were still able to easily proceed annual counting down to 113 (1860); whereas further down, the dating become more uncertain (see the green line in Fig. 9). Below 88 m w.e., seven significant potential volcano horizons can be suspected on the basis of the ionic balance and sulfate levels (not shown), from which however at least one are of local origin (as suggested by small stones with size of up to 1-2 mm were found in the corresponding layer). Nevertheless, a series of three narrow spikes was located at 118-120 m w.e. (dated at around 1840-1833) among which two that are characterized by an increase of sulfate and acidity (up to 7.8 µEq L −1 , not shown) may be related to the wellknown eruptions observed in Greenland in a time distance of 2 years around 1840 (one of them being possibly due to the Coseguina eruption in 1835) .

Modeled ice flow and ice-core dating
Mountain glaciers in present volcanic craters have different thermodynamic properties than other mountain glaciers. The limited ice flux over the crater rims forms flat glacier surfaces with low surface ice velocity. The intense volcanic heat flux may result in basal melting and the associated removal of the oldest basal layers. A simplified thermomechanically coupled model for simulating ice flow along a fixed flow tube and heat transfer in ice caps filling volcanic craters was developed by . The model description and ice-flow and heat-transfer equations are described in detail in . The model takes into account surface and bedrock topography and snow-firn densification parameters (see Sect. 3.3.2), the distribution of the basal melt rate (see Sect. 3.3.1), and normalizes the results by the present-day accumulation rate. We calculated the depthage relationship for the western Elbrus plateau using the recent accumulation rate of 1200 mm w.e. The ice melt rate at the glacier bedrock is negligible and comprises less than 10 mm w.e a −1 (see Sect. 3.3.1) in the deepest glacier sections. Figure 11 shows the cross section of the western Elbrus plateau along a reference flow line. Predicted ice flow paths are shown by lines with arrows while isochrones are designated by numbers which specify the ice age in years.

Conclusions
Paleoclimatological records for southern and eastern Europe are based on geomorphological, palynological, limnological, and dendrochronological data. Ice-core records have not been taken into account as a source of paleoclimatological and environmental information for this area due to rapid glacier mass exchange rate and significant surface melt, often resulting in smoother isotopic and chemical profiles in glacier records. However, the deep Elbrus ice core drilled in 2009 at 5115 m a.s.l. provides new evidence for significant regional-scale multiproxy climatic implications. The negative ice temperature of the glacier at the drilling site results in an undisturbed incoming climate signal. The considerable snow accumulation rate of 1455 mm w.e. coupled with highresolution ice-core sampling allows us to separate snow seasonal climate signals from summer and winter precipitation. Annual layers were differentiated on the basis of seasonal oscillations of NH + 4 , succinic acid, and δ 18 O. Annual layer counting was confirmed by the well-known reference horizons of the AD 1963 nuclear tests and the AD 1912 Katmai volcanic eruption. Annual layer counting extends down to 85 mw.e. Ice flow models show that the basal ice age at the maximum glacier depth of 255 m is more than 600 years BP. The 2009 drilling site was situated downstream from this maximum depth location and the resulting basal ice age does not exceed 350-400 years BP. An essential difference between reported depth-age scale constructed from annual layer counting versus the age scale created from flow models requires the inspection of the model algorithm and the development of a reliable ice flow model.
The combination of the different glacio-chemical features of the western Elbrus plateau detailed in this study demonstrates that this high elevation glacier archive offers the possibility to extract relevant atmospheric information from longterm ice-core records in the Caucasus. Ongoing research is therefore dedicated to reconstructing several key aspects of the changing atmosphere of this central European region, in particular for various aerosol components such as sulfate, ammonium, terrigenous matter, and carbonaceous compounds or fractions and species related to the nitrogen cycle. The comparison of Elbrus ice core with ice-core records from Alpine glaciers (Col du Dôme and Colle Gnifetti) will allow us to estimate the tendency of climatic changes over Europe for the last few centuries, and to obtain high-resolution multiproxy reconstructions of atmospheric chemistry, air temperature and precipitation oscillations, black carbon pollution, and atmospheric circulation change.
the RSF (project 14-37-00038) in his contribution to the paper. The ongoing laboratory analyses at LGGE and logistics were supported by the EU FP7 IP PEGASOS (FP7-ENV-2010/265148), the French ANR program PAPRIKA (ANR-09-CEP-005-02), the CNRS-DFG bilateral project entitled "Secondary organic aerosol production in the lower free troposphere over western Europe", and the LEFE-CHAT program ESCCARGO. Stable water isotopic analyses were supported by the RFBR grant 14-05-31102 (A. Kozachek, A. Ekaykin, and V. Lipenkov, AARI) and IAEA research contracts 16184/R0 (Stable water isotopes in the cryosphere of the Northern Eurasia), and 16795 (Paleo-Climate Isotope Record from European Mt. Elbrus Ice Core). This research work was conducted in the framework of the International Associated Laboratory (LIA) "Climate and Environments from Ice Archives" 2012-2016, linking several Russian and French laboratories and institutes. Two anonymous reviewers and the editor are acknowledged for their valuable comments and suggestions to improve the manuscript. The authors thank Natalie Kehrwald for her help in tuning the language.
Edited by: O. Eisen