1. Introduction
Ice-core archives provide unique information about the physical and chemical properties of the atmosphere in the past. Most of the investigations leading to high-resolution chemical records for the Northern Hemisphere were performed in the dry snow zone (e.g. in Greenland) while information from other Arctic glaciers outside the dry snow zone, like Severnaya Zemlya, is sparse and often complicated by meltwater percolation.
Between 1978 and 1988, eight ice cores were drilled on the archipelago Severnaya Zemlya (Reference Kotlyakov, Arkhipov, Henderson and NagornovKotlyakov and others, 2004), in the easternmost part of the Eurasian Arctic (see Fig. 1). Chemical analysis of these ice cores was restricted to a few components only, with emphasis on Cl- (Reference ArkhipovArkhipov, 1999; Reference Kotlyakov, Arkhipov, Henderson and NagornovKotlyakov and others, 2004), as published, for instance, by Reference Vaikmyae and PunningVaikmyae and Punning (1982) for a 556 m deep ice core from Vavilov ice cap (79° N, 96° E) drilled in 1978. The Cl- record presented by these authors covers the top 400m, but sample resolution was relatively coarse except for three small intervals where the Cl- profile is given in 2 cm steps. Additional chemical components were measured on the only ice core that had previously been drilled to bedrock (761 m) on Akademii Nauk ice cap in 1986/87 (80.50° N, 94.83° E). Concentration values for SO4 2-, Cl-, Na+, Mg2+, Ca2+ and K+ down to 562 m depth for this ice core, as well as a summary of data obtained from other Arctic ice cores, are provided at http://www. pangea.de/PangaVista?query=ddga (Reference ArkhipovArkhipov, 1999). Data resolution is often quite coarse, however. For the Akademii Nauk ice core, for example, resolution ranges from 2 to 10 m of core and decreases with depth.
Map of the Arctic. The ice core was drilled on Akademii Nauk ice cap (80°31′N 94°49′ E, star) the northernmost island of Severnaya Zemlya.
In this paper, we present the first high-resolution (on average 60 mm sample length) chemical ice-core records from Severnaya Zemlya, derived from analyses of the uppermost 53 m of a 723.9m long ice core drilled during the 1999-2001 field seasons on Akademii Nauk ice cap (80°310 N, 94°49’ E; 760 ma.s.l.) (Reference FritzscheFritzsche and others, 2002), marked on Figure 1. With an ice thickness of 723.9m and a mean annual temperature of approximately -18°C, this ice cap is the thickest and coldest in Severnaya Zemlya. The mean annual temperature has been deduced from the mean annual temperature, -14.7°C, of the nearby weather station, Golomyanny (available at http://npolar.no/transeff/transport/ice/AARI_chapter3riktig.pdf (Reference Alexandrov, Radinov and SvyashchennikovAlexandrov and others, 2000)), corrected by the temperature gradient relative to weather-station data from Akademii Nauk (Reference KuhnKuhn, 2000) obtained in 1999 (Fritzsche, unpublished information).
As it is located in the easternmost part of the Eurasian Arctic, different atmospheric information can be expected from this ice-core archive than that obtained from the Greenland and other Arctic ice caps. From trajectory analysis (850 hPa level in March/April;Reference Vinogradova and EgorovVinogradova and Egorov, 1996) it is determined that Severnaya Zemlya is mainly influenced by air masses originating over Europe, northern Eurasia and the Atlantic Ocean, whereas a large part of the air masses (back trajectories on the 700 hPa level; Reference Kahl, Martinez, Kuhns, Davidson, Jaffrezo and HarrisKahl and others, 1997) reaching Summit, Greenland, pass the Canadian Arctic and enter Greenland from the west throughout the year.
Furthermore, from investigations carried out on Greenland it has been observed that episodes of increased aerosol input (e.g. due to Arctic haze events (Reference ShawShaw, 1995)) are reflected at Dye 3 (2479 m a.s.l.) with lower concentrations and for shorter durations compared to sea-level sites (Reference DavidsonDavidson and others, 1993). Thus, it is expected that Arctic haze events especially will be more pronounced at Severnaya Zemlya due to the lower altitude of this ice cap. Accordingly, Severnaya Zemlya represents an ideal location to monitor atmospheric circulation in the Siberian Arctic and to reconstruct the history of anthropogenic air pollution in this area. However, summer temperatures can exceed 0°C at this drill site, as was the case for a few days in August 1999, seen in automatic weather station data (Reference KuhnKuhn, 2000). Maximum monthly mean air temperatures for 1999 were found for July (-1.1 °C) and August (-1.4°C). Slight surface melting was observed at the end of May 1999 (Fritzsche, unpublished information). These melting processes lead to downward percolation of meltwater and thus to a redistribution of the physical and chemical core parameters. One purpose of this study is to estimate the extent to which parameters detectable in the ice are influenced by summer melting.
The interpretation and discussion of the data are focused on the following aspects:
On what time-scale can undisturbed atmospheric information be deduced? Do variations in the chemical records reflect seasonal cycles or are they a result of the melting processes? This aspect is closely related to the dating of the ice. A method for deriving a depth-to-age relationship for this ice core will be presented.
Are maxima in the electrical conductivity record linked to volcanic events, as is the case in the dry snow zone (Reference HerronHerron, 1982)?
What is the long-term anthropogenic influence on the SO4 2- and NO3 - content in an ice core due to strong summer percolation?
2. Methods of Sample Preparation and Chemical Analysis
At the Alfred Wegener Institute (AWI), Bremerhaven, Germany, high-resolution measurements of electrical conductivity (dielectric profiling: Reference Wilhelms, Kipfstuhl, Miller, Heinloth and FirestoneWilhelms and others, 1998; Reference WilhelmsWilhelms, 2000) and density (g-absorption method: Reference WilhelmsWilhelms, 2000) were carried out. The stratigraphy of the ice was recorded using a line-scanning camera (Reference Wilhelms, Fritzsche, Kipfstuhl, Miller, Schwager and ThorsteinssonWilhelms and others, 1997).
A total of 450 ice samples were prepared for chemical analysis. To this end, rectangular ice strips with a crosssection of 30 mm × 40 mm and a length of about 500 mm were cut from the inner part of the ice core. To decontaminate the ice, a ~2mm thick layer was removed from the whole surface under a clean bench using a contamination- free electromechanical plane (Reference Fischer, Wagenbach and KipfstuhlFischer and others, 1998a). Aliquots with a mean depth resolution of 60 mm (corresponding to 8-9 samples per year) were then prepared using a contamination-free circular saw, and sealed in pre-cleaned polyethylene (PE) bags for storing in a cold room to await measurement.
Melted decontaminated ice samples were analyzed for anions (methanesulphonate (MSA-), Cl-, NO3 -, SO4 2-) and cations (Na+, NH4 +, Mg2+, Ca2+) using a Dionex IC20 ion chromatograph. Blank samples were measured routinely by processing ultrapure water ice (Milli-Q water; specific resistance >18.2 MO cm) together with the core samples.
The PE bags and tools needed for further work were precleaned by rinsing them with ultrapure water until the remnant conductivity of the water was <0.8 μScm-1 for the PE bags and the PE vials for the ion chromatograph, <0.5 μScm-1 for the PE bottles in which the melted aliquots were stored, and <1 μScm-1 for the tools needed to handle the samples in the clean laboratory.
In Table 1 the lower concentration limits (which are derived from comparison of the blank-samples concentrations with the response function of the system) and the upper concentration limits (defined by external calibration standards) are summarized for the different ions. The corresponding concentration errors (lower and upper limit) are also listed. The values in parentheses correspond to around half of the samples, which have been measured at a different time and with a modified system. Note that the error decreases with increasing ion concentration due to the decreasing impact of the blank sample values. The number of samples with a concentration error >10% shows that the majority of the detected ion concentrations are well above the lower concentration limit except for Ca2+, NH4 + and MSA-. The discrepancy is caused by a slight contamination of the Ca2+ samples during the preparation process in the laboratory and an intensive variation in laboratory background concentrations for NH4 +. In the case of MSA-, a poorer baseline definition of the system is the reason.
Summary of the concentration limits and concentration errors derived for the various ions. The numbers of samples with a concentration error larger then 10% are listed. Values in parentheses correspond to around 50% of the samples which have been measured at a different time and with a slightly different system
3. Results and Discussion
3.1. The reflection of atmospheric seasonality in the ion records
The concentration of various chemical components which are distributed in the Arctic atmosphere varies during the year due to a change in the meteorological circulation pattern, mainly associated with the seasonal migration of the polar front (Reference Rahn, Borys, Shaw, Roddy and O’ConnorRahn and others, 1977). Consequently, a seasonal oscillation in the ion records of ice cores drilled in the dry snow zone (e.g. in central Greenland) can be interpreted as a seasonal variation in atmospheric concentration and then used to distinguish individual years (Reference SteffensenSteffensen, 1988; Reference BeerBeer and others, 1991; Reference Fischer, Wagenbach and KipfstuhlFischer and others, 1998a). At Akademii Nauk ice cap, however, melting and refreezing processes occur and the snowpack stratigraphy is disturbed. Further, ion fractionation processes may be expected, owing to differences in elution strength of individual ion species.
Snow lysimeter experiments carried out by Reference Johannessen and HenriksenJohannessen and Henriksen (1978) show that the first 30% of water to melt contains 44-76% of the total amount of all examined chemical components (including sea salt, mineral dust, sulphate and nitrogen components). Furthermore, SO4 2-and NO3 - are eluted from snow more rapidly than, for instance, Cl-or Na+ (Reference Goto-Azuma, Enomoto, Takahashi, Kobayashi, Kameda and WatanabeGoto-Azuma and others, 1993, Reference Goto-Azuma1995). Investigations by Reference Davies, Vincent and BrimblecombeDavies and others (1982) show a stronger mobility within the snow for Cl-than for Na+. Additionally, the degree of elution depends on the duration of different melt/freeze cycles, the melting rate and the initial distribution of the impurities in the snowpack as well as the location of the ions in the crystal lattice or on the surface of the ice grains (Reference Mulvaney, Wolff and OatesMulvaney and others, 1988; Reference Davis, Petersen and BalesDavis and others, 1995).
At Akademii Nauk, the influence of the melting processes on the behaviour of the ion signal, the conductivity and the density of the ice can be deduced from the visual ice-core stratigraphy (Fig. 2). As expected for ice drilled in the percolation snow zone (Reference BensonBenson, 1961), the occurrence of periodic melting events often leads to the penetration of pure firn by vertical pipe-like percolation channels (darker sectors with lower density) and ice lenses and layers (white strata; density near 0.91 gcm-3) (Reference Schutt, Fritzsche, Meyer, Miller, Wilhelms and HubbertenSchutt and others, 2003). Furthermore, these ice layers seem to correlate with high signals in electrical conductivity, where maximum values can often be found in or at the edge of an ice layer. As the conductivities of the different ions are reflected in the electrical conductivity of the ice, this supports the assumption that the ion profile is strongly affected by melting.
Visual ice-core stratigraphy plotted against the core depth in metres. The electrical conductivity (grey diamonds) and the ice density (grey line) are shown, as well as the Na+ record (black line). Single years as deduced from counting Na+ cycles are marked by dashed grey horizontal lines.
For that reason, it is not possible to link the ionic ice-core records to the seasonal cycles of the chemical composition of the atmosphere. However, the ion profile can be considered as an annually varying melting/refreezing signal, from which a counted depth-to-age relation can be deduced.
3.2. Dating the ice core by counting sodium cycles
The chemical profile and the stratigraphic record of the ice core show periodic features as illustrated in Figure 2. Assuming the annual snow accumulation is affected by melting processes in a similar manner each year, it is possible to use these cycles to mark individual years. Thus, it can be assumed that ice layers formed from firn penetrated by meltwater operate as barriers for further penetration (Reference Goto-Azuma, Enomoto, Takahashi, Kobayashi, Kameda and WatanabeGoto-Azuma and others, 1993; Reference Schutt, Fritzsche, Meyer, Miller, Wilhelms and HubbertenSchutt and others, 2003) and therefore lead to an accumulation of chemical components above them (see, e.g., the Na+ peak between 2.6 and 2.8 m in Fig. 2). From counting these ion peaks, the number of years in a certain depth range can be obtained.
As an example, in Figure 3 the top few metres of the Na+ concentration profile are plotted against the core depth. Black lines are used to indicate the boundaries between different years according to the recurring pattern in the Na+ record. The Na+ record was chosen, firstly, because the mean Cl- to Na+ ratio (2.00 ± 0.63) is close to that of sea water (1.84). This indicates that essentially all of the Na+ detected at Severnaya Zemlya is of sea-salt origin. Thus, compared to the other chemical components, the signal is influenced by one source only and no anthropogenic contribution needs to be considered. Secondly, Na+ has a lower mobility than NO3 - and SO4 2-, and is therefore least affected by elution processes (see section 3.1).
Na+ profile plotted against core depth in mw.e. (grey line). Black lines indicate the boundaries between different years. Not that no seasonal atmospheric information can be deduced with this method. The Chernobyl horizon (derived from the 137Cs record) is depicted as a grey rectangle.
To confirm this approach, external reference horizons have been considered. A maximum of the 137Cs activity at 21.36-22.68 m depth was identified as the 1963 layer corresponding to fallout from the atmospheric nuclear bomb tests in 1961-62 (Reference PinglotPinglot and others, 2003). In addition, on the shallow core SZ 2000-1, a second 137Cs peak was identified at 10.19-10.89 m depth (converted to the 1999 surface using the mean annual net mass balance derived from SZ 2000-1), reflecting the Chernobyl accident in 1986 (for details see Reference FritzscheFritzsche and others, 2002).
The age obtained by counting Na+ cycles agrees, within 2 years, with the Chernobyl horizon in 1986 (Fig. 3, grey shaded area), which itself may be shifted to deeper layers by meltwater percolation (Reference PinglotPinglot and others, 2003). Furthermore, it has to be considered that, especially due to the summer melting, strong horizontal variations in the surface structure of the firn are likely (snow-pit studies; Ruth, unpublished information). Therefore, an error in the depth assignment of this reference layer, caused by the matching of the shallow core with the main core, cannot be excluded.
Following this procedure, an age discrepancy of 4years is observed for the nuclear-bomb peak in 1963. A discrepancy of 3 years is found for a strong peak in the electrical conductivity record at 25.86 m depth. This peak is assumed to be the result of the eruption of the Bezymianny (Kamchatka) volcano in 1956 (see also section 3.4). For these two horizons it has to be considered that the percentage of firn ice is <30%, sometimes <10%, for this depth range (personal communication from R. Schutt). Thus, it is likely that the accumulation for some years, and therefore the corresponding Na+ cycles, are missing completely.
The good agreement between the number of Na+ cycles and the various reference horizons indicates the method is useful for dating this ice core. For the whole depth range, 94 years could be counted, placing 1906 at 53 m depth, with a maximum dating error of-4 to +2 years to 26 m depth and probably with a larger error below this depth, where no more reference horizons are available.
A similar approach has been used to date various other ice cores subjected to summer melting: for instance, the top 42.5 m of a 84m deep ice core from Snofjellafonna, Svalbard (Reference Goto-AzumaGoto-Azuma and others, 1995), where Na+ cycles were counted with a dating accuracy of ±2 or ±3years above the 1963 bomb peak horizon, and possibly less accuracy below this depth, since no additional reference horizons could be documented in the deeper part of this core. Counting of seasonal cycles in the chemical profiles was also used to date a 122 m ice core from Lomonosovfonna, Svalbard (Reference Kekonen, Moore, Mulvaney, Isaksson, Pohjola and van deWalKekonen and others, 2002), where, in addition to the 1963 radioactive layer, the Laki (Iceland) eruption could be detected as a supplementary reference.
3.3. Mean net accumulation for the 20th century
As shown above, a continuous depth-to-age relationship can be established by counting annual Na+ melt cycles. With this result, it is possible to determine the thickness of each annual layer (Fig. 4, grey triangles). However, annual layer thickness is not necessarily equal to the snow accumulation for the corresponding years, since meltwater transport in and out of these layers has occurred. Therefore, the annual mean net accumulation is estimated as a mean value for a core section for which the loss of mass compared to the total snow amount can be assumed to be negligible (approximately 10 years).
Annual layer thickness derived from counting Na+ cycles (grey triangles). The annual net accumulation as derived from a 10 year running mean is indicated by a black line. The mean annual net mass balance is shown as a thick dashed black line. The mean annual net mass balances deduced from the external reference horizons are depicted as thin dashed black lines. Note that the Chernobyl horizon was detected on a shallow ice core drilled close to the main core.
A mean net accumulation of 0.46 m w.e. a-1 (Fig. 4, dashed thick black line) is deduced for Akademii Nauk from a 10year running mean (black line, Fig. 4) covering approximately the last 100 years bp. This value fits well with the observations of Reference Zagorodnov, Klementyev, Nikiforov, Nikolaev, Savatyugin and SasunkevichZagorodnov and others (1990), who deduced a mean annual net mass balance of 0.430.44 m w.e. a-1 from structural stratigraphic methods. Other results for the accumulation at Akademii Nauk ranging from 0.23 to 0.29 m w.e. a-1, as summarized by Reference Kotlyakov, Arkhipov, Henderson and NagornovKotlyakov and others (2004), could not be confirmed.
Note that variations in the temporal evolution of the net accumulation between around 1950 and 1980 mainly reflect the dating error and thus do not represent the natural variation in precipitation. As mentioned above, the definition of individual years in the record, as well as the transfer of the Chernobyl horizon from the shallow core to the main core, may also affect the dating error. This is why the mean annual net mass balances independently calculated from the external reference horizons (Fig. 4, thin dashed lines) differ from the values achieved from averaging the layer thicknesses.
3.4. Specific events
Major volcanic eruptions release large quantities of impurities into the atmosphere which often lead to significantly higher concentrations of non-sea-salt (nss) SO4 2- and, in some instances, also nssCl-. Thus, maxima in the electrical conductivity of the ice onto which these impurities are deposited are recorded (Reference HerronHerron, 1982). The non-sea-salt portion of an ion species is calculated as the difference between the total concentration and the sea-water contribution as represented by the Na+ concentration.
In the Academii Nauk ice core, three extraordinarily high conductivity peaks at depths of around 20, 26 and 44 m (black diamonds; see Fig. 5a-c), which were initially suspected to be of volcanic origin, were analyzed for their chemical composition. As can be seen from Figure 5, higher ion concentrations are detectable not only for volcanogenic components but for all ion species in these peaks. Thus maxima in electrical conductivity cannot be linked directly to volcanic eruptions, as is normally assumed for ice cores drilled in the dry snow zone. Instead, these peaks seem to be the result of ion fractionation processes caused by melting/ refreezing cycles (see section 3.1). This assumption is also supported by the locations of the ion maxima within ice layers (20 and 44 m depth) or between two ice layers (26 m depth; see Fig. 5).
Ion profiles for three extraordinarily high peaks in the electrical conductivity record (black diamonds) at around (a) 20 m, (b) 26m and (c) 44m depth. Concentrations of the different ions (black lines), corresponding core stratigraphy and the density signal (black line) are shown on the left.
Similar observations have been made using other cores affected by summer melting. In the case of the core from Snofjellafonna, Spitsbergen (Reference Goto-AzumaGoto-Azuma and others, 1995), no clear volcanic signal could be detected, as extremely high concentrations of SO4 2- (4178.8 ngg-1) dated to around 1940 were accompanied by strong maxima in Na+, Cl- and NO3 -. The same pattern occurred in an ice core from Penny Ice Cap, Baffin Island, Canada (Reference Grumet, Wake, Zielinski, Fisher, Koerner and JacobsGrumet and others, 1998), where evidence for a considerable peak in the nssSO4 2- (around 1914, almost 800ngg-1) associated with the eruption of Katla, Iceland (1918), could only be substantiated by the presence of basaltic glass, as the dust-related species showed high concentration values, too.
In the case of Akademii Nauk, the conductivity maximum detected at 26 m depth can be attributed to the eruption of Bezymianny in 1956 (volcanic explosivity index (VEI) = 5; Reference Simkin and SiebertSimkin and Siebert, 1994). This is done by comparison of the total peak load ratio for the event relative to the background. The total peak load ratio compares the peak area (peak load), calculated for a specific ion [Ion]i with the total peak load (sum of the peak loads contributed by every single ion species; for definitions see footnotes in Table 2). This measure is independent of peak width or height. As a background value, the mean total peak load ratio will be considered. This is the average of the total peak load ratios contributed by three or four smaller peaks located directly above or below the corresponding conductivity peak. The three or four smaller peaks are chosen to be representative of the contemporary conditions in atmospheric concentrations and meltwater percolation in the snowpack.
Total peak load and total peak load ratio calculated for nssSO4 2_ and nssCl_ for three extraordinarily high maxima in the electrical conductivity record at depths of approximately 20, 26 and 44 m. As a background value, the mean total peak load ratio, which was computed from the average of the total peak load ratios contributed by peaks above and below the corresponding maximum in the electrical conductivity record, is listed (for details see text)
As can be seen from Table 2, the percentage of nssSO4 2- in the total peak load ratio at 26 m (28.6%) is more than twice the corresponding background value (12.3%). Much smaller contributions are observed for the 20 and 44m maxima. Additionally, a nssCl- contribution of 11.1% is observed for the conductivity maximum at 26 m. This value is exceptionally high compared to the corresponding mean total peak load ratio of 3.9%. In the case of the maximum at 44 m depth, the total nssCl- peak load ratio is much lower compared to the background value, and for the 20 m horizon a negative nssCl- contribution is found. Accordingly, the relative contribution of nssSO4 2- and nssCl- to the total peak load at 26m is higher than expected from background conditions, supporting a volcanic origin of this horizon.
According to Reference Simkin and SiebertSimkin and Siebert (1994), no strong volcanic events (with VEI > 3) are registered in historical records for the time-span corresponding to the maximum at 20 m depth (1971 +2/—4 years). For the 44 m maximum, three large volcanic eruptions are recorded in the years 1924 (Raikoke, Kurile Islands, Russia; VEI = 4), 1918 (Katla, Iceland; VEI = 4) and 1912 (Katmai, Alaska, USA; VEI = 6) (Reference Simkin and SiebertSimkin and Siebert, 1994). Dating by counting Na+ cycles leads to an age of 1924 +2/—4 years for this horizon, and the eruption of Raikoke in 1924 therefore best fits with this event. However, this result contradicts the study of Reference Schutt, Fritzsche, Meyer, Miller, Wilhelms and HubbertenSchutt and others (2003), who assigned the 44 m maximum in the electrical conductivity record to the eruption of Katmai in 1912. This assumption was based on the strength of the event (VEI = 6). Furthermore, the construction of a continuous layer thickness curve for the whole deep ice core by means of the δ18O and conductivity signal supports Katmai as the source, where further potential volcanoes (e.g. Laki, 1783; VEI = 4; Reference Simkin and SiebertSimkin and Siebert, 1994) have been considered. Note however, that sulphur emission is not a primary parameter in the definition of the VEI (Reference Simkin and SiebertSimkin and Siebert, 1994). Thus, highly explosive eruptions do not necessarily imply high sulphur output and vice versa.
At this stage it is not possible to decide unambiguously which volcanic eruption causes the high conductivity or whether it is a result of melting/refreezing processes. Therefore, microscopic analysis of microparticles to identify the potential presence of volcanic glasses might be helpful. The chemical analysis should be continued to identify additional volcanoes as external reference horizons and to achieve a continuous depth-to-age relation for the whole core. A synthesis of these results with the approach from Reference Schutt, Fritzsche, Meyer, Miller, Wilhelms and HubbertenSchutt and others (2003) may lead to a consistent depth scale for this ice core.
3.5. Long-term variations
As shown above, the seasonal information in the ion signal of the ice from Akademii Nauk reflects primarily seasonal melting/refreezing processes rather than seasonal variations in atmospheric aerosol concentrations. Owing to the large fraction of refrozen ice per annual layer, it is not even certain that all the ions deposited in a specific year are conserved in the respective layer. They may elute to layers below. However, averages of ion concentration on longer timescales where melting/refreezing processes are in balance (i.e. within a few years) should be representative of changes in the atmospheric load, at least for conservative species like sulphate and sodium, where no post-depositional losses via the gas phase occur. Accordingly, we calculated water- weighted, 5 year running averages for non-sea-salt sulphate, nitrate and sodium concentrations in snow over the time- span 1906-99. As clearly illustrated by the grey dots in Figure 6a and b, the anthropogenically influenced species, SO4 2- and NO3 -, are subject to a substantial increase starting in the 1960s and also to short-term events caused by the specific events discussed above. Because those horizons may be attributed to singular volcanic events, we removed the affected samples (about five samples per horizon) from the dataset and calculated corrected 5 year averages (open circles in Fig. 6). In the following the corrected long-term variations are discussed.
Long-term variations in the (a) nss SO4 2_, (b) nssNO3 _ and (c) nssNa+ concentrations detected at Akademii Nauk derived from water-weighted 5 year running averages. The grey circles represent the raw data; open circles depict the result after removing samples suspected to be influenced by volcanic input (around five samples for each horizon at approximately 20, 26 and 44 m; see Fig. 4). Open squares indicate total (a) Eurasian SO2 (Reference MylonaMylona, 1996) and (b) European NOx emissions (Reference HovHov and others, 1987).
3.5.1. Sulphate
As depicted in Figure 6a, background nssSO4 2- concentrations decreased by about 100ngg-1 at the beginning of the 20th century, then gradually increased by about 100ngg-1 in the following 30 years, until an intermittent decline around 1940. In the mid-1960s, however, sulphate concentrations show an extraordinary rise to background concentrations higher than 400 ngg-1, unprecedented in the decades before. For comparison, we calculated running averages for Na+, which as a sea-salt tracer is of natural origin only (see section 3.2). The latter shows no long-term increase after 1960 (Fig. 6c). This implies that the trends in sulphate (and nitrate; see below) do not reflect an increase of the load related to melting/refreezing processes as is the case for the discrete events, but can be attributed to a long-term anthropogenic influence on the atmospheric sulphur load. Note also that after maximum concentrations were reached in the mid-1970s, nssSO4 2- concentrations declined steadily to values below 200 ngg-1 in the 1990s, reflecting a substantial reduction in the anthropogenic sulphur load during recent decades.
Owing to the significant variation in background concentrations at the beginning of the last century, it is difficult to define a representative early-industrial to pre-industrial level. Taking into account that around 1910 we find a significant number of smaller sulphate peaks in the record, which might also be related to less pronounced volcanic depositions expected for this time interval, we estimate average early- industrial sulphate concentrations in Figure 6a to be 100- 120ngg-1. Thus, industrial emissions after 1965 lead to a four-fold increase in sulphate concentrations in only a few years, implying dramatic ecological consequences for the High Arctic. As a matter of fact, in the raw data this rise takes place within only 1-2 years and indicates rapid increases in additional emission sources relevant for Severnaya Zemlya.
Comparison of the Severnaya Zemlya with other Northern Hemisphere sulphate ice-core records shows no such dramatic and sudden rise in Greenland records (Reference Mayewski, Lyons, Spencer, Twickler, Buck and WhitlowMayewski and others, 1990; Reference Legrand and MayewskiLegrand and Mayewski, 1997; Reference Fischer, Wagenbach and KipfstuhlFischer and others, 1998b) or in records from Svalbard (e.g. Reference Goto-AzumaGoto-Azuma, 1995; Reference Simoes and ZagorodnovSimoes and Zagarodnov, 2001) or Penny Ice Cap (Reference Grumet, Wake, Zielinski, Fisher, Koerner and JacobsGrumet and others, 1998; Reference Goto-Azuma and KoernerGoto-Azuma and Koerner, 2001). This points to proximal sources being more important for the Akademii Nauk ice core. The records from Svalbard also indicate an increase in the second half of the 20th century. However, sulphate concentrations in these cores increase by a factor of only two in the course of the 20th century, and for Snofjellafonna (Reference Goto-AzumaGoto-Azuma and others, 1995) the increase is more gradual and starts at around 1950. A sudden rise in SO4 2- concentrations at the end of the 1960s in that record is counterbalanced by a pronounced minimum in layers above, pointing to a melting/refreezing feature and not to a comparable change in the atmospheric load. The Greenland records (Reference Mayewski, Lyons, Spencer, Twickler, Buck and WhitlowMayewski and others, 1990; Reference Fischer, Wagenbach and KipfstuhlFischer and others, 1998b) reveal a more gradual increase in sulphate concentration over the last century than any of the lower-altitude Arctic ice cores. This gradual rise has been attributed to the increase of total Eurasian sulphur dioxide emissions (Reference MylonaMylona, 1996) in the 20th century (Reference Fischer, Wagenbach and KipfstuhlFischer and others, 1998b). The sudden rise at Severnaya Zemlya is not in line with a gradual increase in total Eurasian sources, as clearly illustrated in Figure 6a (squares). While Greenland is remote from all industrial sources and, with its high altitude, is more representative of tropospheric background conditions, the proximity of Akademii Nauk to the Norilsk mining complex in Siberia and to mining locations on the Kola peninsula, and its much lower altitude, points to more regional emission sources. Thus, the sudden rise in sulphate concentrations around 1966 +2/-4 years may point to an increase in emission rates from these sites. Indeed, a major shift from the use of low-sulphur nickel ore of local origin to ore from Norilsk with extremely high sulphur content is reported for the Kola region during the mid-1960s (personal communication from A. Ryaboshapko). No information on the temporal development of sulphuremission is availablefor the Norilsk region for that time. Nowadays, Norilsk is the most important point source of sulphur in the former USSR, with annual emissions of 1.2 and 1.1 × 109 kgSa-1 in 1985 and 1990, respectively (Reference Ryaboshapko, Brukhanov, Gromov, Proshina and AfinogenovaRyaboshapko and others, 1996). Model results on sulphur dioxide and sulphate transport in the Arctic by Reference ChristensenChristensen (1997) support an important contribution to inner Arctic sulphur levels by the nickel- and copper-producing industries on the Kola peninsula and in Norilsk. Also, back trajectory studies by Reference Vinogradova and EgorovVinogradova and Egorov (1996) clearly show the dominant influence of air masses originating over central Siberia for the aerosol transport onto Severnaya Zemlya. Furthermore, elemental tracer studies on the aerosol composition at the coastal sites Point Barrow, Alaska (Reference Rahn, Loewenthal and StonehouseRahn and Loewenthal, 1986), and Svalbard (Reference Pacyna and SturgesPacyna, 1991) reveal a dominant influence of anthropogenic emission from sources located in the central part of the former USSR during the early 1980s, not only for the Eurasian but also for the North American sector of the Arctic. In summary, we conclude that a primary influence of regional Siberian industrial emissions on the sulphate levels in Severnaya Zemlya is most likely.
Assuming that the hypothesis of regional industrial sources in Siberia being the dominant influence on sulphate deposition on Severnaya Zemlya is correct, the measured reduction in sulphate snow concentrations over recent decades would also imply a decrease of emissions from these sites, possibly related to an economic decline of the companies in Norilsk and the Kola peninsula.
3.5.2. Nitrate
In Figure 6b the nitrate background concentrations show a very similar temporal evolution to the sulphate levels. A systematic decrease of NO3 - from about 70 ngg-1 in 1910 to about 50ngg-1 in 1940 is followed by 20ngg-1 higher concentrations around 1950. The major increase, however, is from a local minimum of 55 ngg-1 in the early 1960s to >160ngg-1 in the 1970s. Thus, nitrate concentrations have risen by a factor of three since early-industrial levels at the start of the last century. As was the case for the sulphate raw data, major parts of this increase occurred in a 1-2yearstep at 1966 +2/-4 years.
Again, no such step increase in NO3 - concentrations can be found in the Greenland records (Reference Mayewski, Lyons, Spencer, Twickler, Buck and WhitlowMayewski and others, 1990; Reference Legrand and MayewskiLegrand and Mayewski, 1997; Reference Fischer, Wagenbach and KipfstuhlFischer and others, 1998b);however, it is indicated by records from smaller ice caps in Svalbard and the Canadian Arctic (Reference Goto-Azuma and KoernerGoto-Azuma, 2001; Reference Kekonen, Moore, Mulvaney, Isaksson, Pohjola and van deWalKekonen and others, 2002). The ice core from Lomonosovfonna, Svalbard, (Reference Kekonen, Moore, Mulvaney, Isaksson, Pohjola and van deWalKekonen and others, 2002) shows an increase to about 200 ngg-1 both in the late 1950s and during the period 1970-85. The first of these increases is not clearly found on Severnaya Zemlya. The record from Snofjellafonna shows an outstanding maximum in nitrate concentrations around 1968, at least partly related to melt processes (see above). Total European NOx emissions (Reference HovHov and others, 1987) as depicted in Figure 6b rose gradually starting in the 1940s and can therefore not explain the step increase in nitrate concentrations on Severnaya Zemlya. Even more striking is the nitrate decline by 70-100ngg-1 synchronous with the decrease of sulphate concentrations in the last two decades. Such a significant nitrate decrease is not found in the Greenland records. Accordingly, this result rules out a comparable anthropogenic source mix for both regions and strongly points to regional NOx emissions by industrial combustion processes occurring in the Siberian Arctic, leading to drastic changes in nitrate concentrations on Severnaya Zemlya.
4. Conclusions and Outlook
This study has shown that even if the chemical signal in the ice core from Akademii Nauk is disturbed due to annually occurring melt processes, information about the aerosol composition of the past atmosphere can still be deduced. Long-term variations in anthropogenic sulphate and nitrate precursors could be determined from the ice-core records, where the sulphate trend can probably be ascribed to the nickel- and copper-producing industries at Norilsk or the Kola peninsula, as the principal inner Arctic sources.
In contrast, the cyclic behaviour observed in the ion records cannot be assigned to the seasonal variation in atmospheric aerosol concentrations. Rather, these variations reflect a melting/refreezing signal from which, however, a depth-to-age relation can be derived. Thus, it was possible to achieve continuous dating of the first 53 m of the ice core by counting cycles in the Na+ record. Using this approach, the depth interval investigated in this study reflects a time-span of around 100 years, and a mean annual net mass balance through the 20th century of 0.46 m w.e. a-1 was deduced for Severnaya Zemlya.
Peaks in electrical conductivity in ice cores drilled in the percolation snow zone are not necessarily linked to volcanic events. An exceptionally high maximum at around 26m depth could be attributed to the eruption of Bezymianny in 1956, but the ionic composition of two peaks at approximately 20 and 44 m depth did not give unambiguous evidence for a volcanic input. Thus, it has to be expected that only a few reference horizons will be available to date a deep ice core.
To improve insight into the melt processes, more detailed investigations of the core stratigraphy may be helpful. With such information it could also be possible to estimate the time-scale for which an atmospheric signal can be deduced from the chemical composition of the ice, i.e. the number of annual layers which are penetrated by the meltwater of one season.
Acknowledgements
We thank B. Twarloh for help with the chemical analysis of the ice samples. We thank R. Schutt for useful comments on the manuscript. This project is funded by the German Ministry of Education and Research (BMBF research project 03PL 027A). It is a joint venture of the Alfred Wegener Institute for Polar and Marine Research, Germany, and the Arctic and Antarctic Research Institute and the Mining Institute, both in St Petersburg, Russia.
