2.1. Site selection, sampling and analyses

In the framework of the European Project for Ice Coring in Antarctica (EPICA) a comprehensive multi-year pre-site survey was carried out on Amundsenisen (Reference OerterOerter and others, 1999, 2000). Data from 13 snow pits and 4 intermediate-depth ice cores recovered during the field campaigns 1997/98 and 1999/2000 (Fig. 1) are presented here. Details on location, altitude, and accumulation rate of each sampling site can be found in Reference GöktasGöktas and others (2002). All drilling sites are located in the Atlantic sector of the East Antarctic plateau more than 500 km away from the coast at an altitude range 2600–3160 m a.s.l., i.e. well above the marine boundary layer. The annual snow accumulation (about 50–100 kg m^–2 a^–1) is rather low at all sites but varies by a factor of 2 between individual sites.

Fig. 1. Area under investigation on Amundsenisen in DML. Location of the ice-core (squares) and snow-pit (dots) sampling sites. The asterisk marks another ice core studied in this area, Camp Victoria (CV, located 130 km from DML07, and about 250 km from DML05) (Reference KruegerKarlöf and others, 2000). The present EPICA deep-drilling site is located at 75°00.10′ S, 0°04.07′ E.

The four intermediate-depth ice cores at DML03, DML05, DML07 and DML17 were drilled with an electromechanical drill, and the recovered core pieces were sealed in polyethylene (PE) bags in the field. All snow-pit samples were taken with 60 mL PE beakers, which were pushed into the snow-pit wall slightly overlapping the same snow layer, allowing for a sample resolution of 4–8 samples per year. One snow pit excavated at DML05 (SS9908) was sampled with higher resolution (8–10 samples per year) by cutting out 2 cm layers and placing the sample into 250 mL beakers. All beakers had been precleaned with ultra-pure water, until the conductivity of the water was < 0.5 μS cm^–1. After cleaning, the beakers were dried in a contamination-free vacuum oven and directly sealed in PE bags until usage in the snow pit. After sampling, the beakers were sealed for transportation. All ice-core and snow-pit samples were transported in frozen state to the cold-room facility of the Alfred Wegener Institute at Bremerhaven, Germany. There, the ice cores were cut, decontaminated using an electromechanical plane under clean-room conditions in the cold laboratory, and sealed in precleaned PE bags before ion chromatographic (IC) analysis (for details concerning sampling of the ice cores, decontamination routine, IC set-up, accuracy and detection limit see Reference GöktasGöktas (2002) and Reference GöktasGöktas and others (2002)). The uncertainty for the measured sulphate concentrations discussed in this study is approximately 3%.

In general, samples were analyzed for methanesulphonate (MS), C1^–, NO₃ ^–, SO₄ ^2–, Na⁺, NH₄ ⁺, K⁺, Mg²⁺ and Ca²⁺ by IC analysis. In addition to IC analysis, continuous flow meltwater analyses (CFAs) had been carried out already in the field on all four ice cores (Reference StenniSommer and others, 2000b), providing high-resolution profiles for electrolytical conductivity, Na⁺, NH₄ ⁺ and Ca²⁺ for annual-layer counting. The upper approximately 5 m w.e. of the ice cores (covering the interval AD 1950–97), where bad core quality compromised the CFA results, were resampled at seasonal resolution (4–10 samples a^–1) and analyzed for anions and cations using IC. Below AD 1950 the ice cores were sampled at annual resolution according to the dating established by counting annual layers in the CFA Na⁺ records, and only anion analyses were performed. For the upper 5 m w.e., annual layers were defined using the IC profiles by setting annual markers on the falling flank of the seasonal sodium peak and the rising flank of the non-sea-salt (nss)-sulphate peak, indicating the austral spring season (for definition of nss sulphate, see below). Only snow pit SS9908, which was sampled both for ion and isotope analyses in 2 cm resolution, was dated by marking the austral summer maximum in δD for each year.

In the following, nssSO₄ ^2– concentrations are presented, which were calculated by subtracting the concentration of the sea-salt-derived sulphate from the total SO₄ ^2– concentration (in ng g^–1), using Na⁺ as sea-salt reference species and the sulphate-to-sodium ratio in bulk sea water of 0.252, i.e.:

(1)

On average the sea-salt sulphate contribution was only about 10% of the total sulphate concentrations.

2.2. Annual-layer counting

Reference Palais, Germani and ZielinskiOerter and others (2000) carried out a preliminary dating based on dielectric profiling (DEP) on all ice cores from Amundsenisen, using the striking volcanic horizon of Tambora, Indonesia, as a marker for the year AD 1816. Strati-graphic dating back to AD 1100 was accomplished by Reference StenniSommer and others (2000a) using a combination of annual-layer counting in the CFA records and identification of the most prominent historic volcanic horizons Tambora (1815) and unknown events at 1279, 1269, 1259 and around 1450 (Kuwae?) in the nss-conductivity profile. In this study we performed multi-parameter (MS, SO₄ ^2–, C1^–, NO₃ ^–, Na⁺, Ca²⁺, Mg²⁺ and NH₄ ⁺) stratigraphic annual-layer counting from AD 1999 to 50 and extended the annual-layer counting for ice cores B32 (at DML05) and B31 (at DML07) before AD 1950 based on the high-resolution CFA measurements of sodium (Reference StenniSommer and others, 2000b). To quantify the stochastic error of the counting procedure, we repeated the counting four times by two different persons along two arbitrarily chosen depth intervals, 105–110 m and 135–140 m. We found that counted years did not differ by more than ±5 years for these 5 m intervals. In addition, annual-layer counting was carried out four times by each person beyond AD 1257 to the end of the ice-core record, where finally AD 165 ±24 was reached. The AD 1259 event is an important interhemispheric reference horizon (Reference Delmas, Kirchner, Palais and PetitPalais and others, 1992) with a dating uncertainty of around AD 1259 ±2 years (Reference Langway, Osada, Clausen, Hammer, Shoji and MitaniLangway and others, 1988; Reference ClausenClausen and others, 1997). A possible systematic dating error (e.g. by missing years) was estimated by annual-layer counting between well-defined volcanic reference horizons in the younger part of the core (AD 1884–16, 1815–1762 and 1762–1601). Within these reference horizons, counting has been performed ten times. We found that our counting procedure led to no systematic error compared to the absolute number of years between these historic eruptions. In conclusion, the assignment of the volcanic eruptions between AD 1997 and 1601 is accurate within ±3 years, around ±5 years between AD 1601 and 1257, increasing linearly to ±24 years to the end of the core (AD 165).

In contrast to B32, annual-layer counting could not be unambiguously performed on the complete record at DML07 (B31), because of the decreasing accumulation rate from 59 to 38 kg m ^–2 a^–1 prior to AD 1000. The dating at DML07 was verified by synchronization of the volcanic sulphate deposition values with the reliably dated core B32 in combination with the densification model used by Reference StenniSommer and others (2000a).

2.3. Background sulphate and volcanic sulphate deposition

Background nss sulphate measured in Antarctic snow and aerosol is clearly dominated by a biogenic source, i.e. photo-oxidation of phytoplankton-derived dimethylsulphide (DMS), and exhibits a high year-to-year variability (Reference Legrand and DelmasLegrand, 1995). Thus it is occasionally difficult to unambiguously identify a superimposed volcanic sulphate signal. In this study, we present a sensitive detection procedure that uses a robust measure of variability and takes long-term variation in the background concentration into account (Reference Fischer, Wagenbach and KipfstuhlFischer and others, 1998b). In addition, we tried to verify the volcanic nature of identified nss-sulphate peaks considering MS/nss-sulphate ratios. In contrast to nss sulphate, MS is exclusively formed by photooxidation of phytoplankton-derived DMS. Therefore the variability in marine bioproductivity, provoking significant nss-sulphate peaks, is expected to result in simultaneously enhanced MS concentrations. However, extensive measurements of atmospheric nss sulphate and MS at coastal Antarctic stations revealed only a moderate correlation between the two compounds in the aerosol phase, most pronounced during the polar summer (Reference Legrand and WagenbachLegrand and Pasteur, 1998). Moreover, as stated by Reference Dibb and WhitlowDibb and Whitlow (1996), the possible impact of El Niños and poorly understood post-depositional processes on the MS signal complicates an assessment of biogenic nss sulphate. In summary, MS/nss-sulphate ratios only provide an additional but ambiguous criterion for detection of volcanic sulphate deposition.

Numerous studies have been carried out to identify volcanic events in ice-core time series (see above), and different peak detection algorithms have been applied (e.g. Reference Delmas, Kirchner, Palais and PetitDelmas and others, 1992). To provide a robust measure of variability in the presence of volcanic peaks and long-term variations in the atmospheric background sulphate concentrations, we use the running median (RM_i ═ median (x _{i–(n–1)/2}, ... ,x _i+(n–1)/2)) to calculate background concentrations, and the median of absolute deviation (MAD_i ═ median(|x _i(n–1)/2 – RM _i |),...,(|x _i+(n–1)/2 – RM _i |)) as robust measure of variability, where n is the window width of the running filter and x_i the ion concentration of sample i (Reference Fischer, Wagenbach and KipfstuhlFischer and others, 1998b). Peaks were detected if nss-sulphate concentrations exceeded the running threshold value y _r defined by:

(2)

The parameter k ═ 4, which adjusts the threshold, and the window width n ═ 30 were selected empirically to be most suited for peak detection in our nss-sulphate records (Fig. 2). Finally, of the 39 nss-sulphate peaks coinciding in ice cores B31 and B32 over the period AD 540–1997, 24 were tentatively assigned to known volcanic events, 6 to unknown (i.e. documented but not named) events and 8 to unidentified events (i.e. not documented and either seen for the first time in an ice-core record or seen previously in other ice-core records). Before AD 540 there was one coinciding unidentified event detected in both cores. Note that only those peaks coinciding in both cores are accepted to be volcanic horizons (Table 1). Further down in the interval covered only by B32 at DML05, two known and six unidentified events occurred.

Fig. 2. Identification of volcanically derived peaks in the nss-sulphate records. The nss-sulphate record (grey line) and y_r (black line) vs assigned time-scale of the ice core at DML05. The numbers refer to the volcanic events listed in Table 1.

Table 1. Volcanic eruption chronology and volcanic sulphate depositions, AD 1997–165

We note, however, that such an assignment of high sulphate concentrations to prominent eruptions listed by Simkin and Siebert (1993) implies a linkage of explosivity and sulphur emissions into the stratosphere that does not necessarily hold. Moreover, the threshold value y _r is by definition just an outlier detection criterion and by no means a selective volcanic sulphate detector. In order to isolate volcanic from background sulphate peaks, we further considered the MS/nss-sulphate ratios. Our results from DML05 and DML07 revealed that during the impact of volcanic eruptions (detected by using y _r) the MS/nss-sulphate ratio was 0.062±0.03 (mean±σ) and 0.039±0.027, respectively, while during the remaining periods the mean MS/nss-sulphate ratio was 0.122±0.05 for DML05 and 0.064±0.048 for DML07. Taking an upper threshold for volcanic MS/nss-sulphate ratios of 0.07, in DML05 only six nss-sulphate peaks identified by y _r showed ratios above this threshold (Table 1), indicating that these events are probably biogenic. Unfortunately, MS/nss-sulphate ratios found in DML07 appeared too variable to identify a meaningful threshold value to define a volcanic impact. Most probably, the impact of post-depositional losses is more severe at this low-accumulation site (Reference GöktasGöktas and others, 2002).

The total volcanic sulphate deposition D(volc-SO₄ ^2–) related to each volcanic eruption detected on the ice sheet was calculated by integration of the part of a volcanic nss-sulphate peak above the reduced running mean RRM_i centered on the peak (where all detected volcanic peaks had been removed from the dataset), i.e.:

(3)

where Δz _i is the length of each sample i in water equivalents and where the sum includes all samples of a peak affected by nss-sulphate concentrations above the threshold level. The deposition value is mainly dependent on the chosen background-level RRM_i, hence, the natural variability of non-volcanic sulphate, while the measurement accuracy of x_i and Δz _i is of secondary importance. Therefore, the following approach was taken to determine the uncertainty for volcanic nss-sulphate deposition values. In addition to the running mean (window width n ═ 30), the corresponding running standard deviation (σ _i ) was computed for the volcanic sulphate reduced dataset. The uncertainty of the deposition values is given by:

(4)

where 95% of all background concentration values lie within twice the standard deviation σ_i. The relative error for volcanic sulphate depositions higher than 20 kg km^–2 was about ±15%, for depositions between 10 and 20 kg km^–2 about ±50%, and for values below 10 kg km^–2 about ±75% and in single cases even higher. The uncertainty of the calculated volcanic sulphate deposition values of our snow-pit data was on average ±82%. Here the uncertainty of RRM_i is much higher, because of the small number of years available to determine non-volcanic background levels. For example, at DML19 no volcanic sulphate deposition could be calculated at all. Based on _i of the background sulphate deposition flux, it can be concluded that for Amundsenisen there is a 65% likelihood of detecting a volcanic event with a sulphate deposition flux of about 1– 2 kg km^–2 a^–1 in firn cores or 10 kg km^–2 a^–1 in snow pits.

Finally, we emphasize that, besides enabling the detection of smaller volcanic peaks in the presence of high background variability, a more sensitive peak detection algorithm, even in combination with MS/nss-sulphate ratios, still carries a risk of including non volcanic nss-sulphate peaks. In summary, a totally unambiguous detection of volcanic peaks in ice-core records and a final assignment to individual documented eruptions is only possible when specific volcanic tephra can be found in the ice as well (e.g. Reference ZielinskiZielinski and others, 1997b; Reference Basile, Petit, Touron, Grousset and BarkovBasile and others, 2001). Nevertheless, the detection of smaller volcanic eruptions enabled by using y _r as the identification criterion should significantly improve the dating of further Antarctic ice cores. In addition, a more reliable quantification of volcanic sulphate deposition on DML, which is intrinsically linked to the volcanic sulphate burden of the atmosphere, is possible.

3.1. Spatial variability of two recent volcanic signals

To derive a measure of the atmospheric sulphur aerosol load caused by individual volcano eruptions from an ice core, we have to show that the sulphate deposition at the core location is at least of regional representativeness and that local differences in deposition and wind reworking of the snow cover have only minor effects. Hence, we first scrutinized the spatial variability of two recent volcanic signals within the area under investigation in 13 snow pits, i.e. the eruptions of Pinatubo on 15 June 1991 and Cerro Hudson on 8 August 1991, to assess the spatial and temporal consistency of the imprint. In contrast to Pinatubo, the Cerro Hudson event had mainly a tropospheric impact, and a marked regional variability of the volcanic sulphate deposition can be anticipated for Antarctica. The Pinatubo eruption was among the strongest recorded in the past century, assigned a VEI of 6, while that of Cerro Hudson was assigned a VEI of 5 (Simkin and Siebert, 1993). All snow pits covered up to 14 years back in time ending at the year 1999 (for DML25) or 1997 (for the remaining sites).

The nss-sulphate records of these 13 snow pits are plotted vs depth in Figure 3. In all snow pits nss-sulphate concentrations increased at the end of AD 1991, except at DML13 and DML19 where the signal rose in late AD 1992 or even 1993. The reason for this discrepancy is not obvious. Dating ambiguities, low sampling resolution (±2 cm) and post-depositional processes such as wind erosion could be put forward as plausible explanations. Typically, elevated nss-sulphate concentrations persisting up to 3 years after major volcanic eruptions were observed in other ice cores (Reference Legrand and DelmasLangway and others, 1994; Reference McCormick, Thompson and TrepteLegrand and Wagenbach, 1999). Because of the limited resolution of our sampling and the high standard deviation of the background level, the Pinatubo and Cerro Hudson signals could not be separated in our records. This is in contrast to results from the South Pole, where the imprints of the two eruptions are well separated (Reference Cole-Dai, Mosley-Thompson and ThompsonCole-Dai and others, 1997b).

Fig. 3. Records of nss-sulphate concentrations from snow-pit samples. The increase of concentrations is assigned to the eruptions of Pinatubo (15 June 1991) and Cerro Hudson (8 August 1991). Asterisks mark the year 1992, vertical lines the turn of the year.

In order to compare sulphate deposition from sites with significantly different accumulation rates, we have to consider the systematic change in the relative contribution of wet and dry to total deposition. In a first-order approximation the concentration C_firn in firn and the total deposition flux J _total of a conservative compound can be described as a composite of an average annual concentration C _snow in fresh snow and an annual total dry deposition flux J _dry (comprising dry deposition, fog deposition and scavenging by drifting snow). J _dry is uniformly diluted by the snow accumulation rate (accu), while J _total is proportional to the snow accumulation rate (Reference Fischer, Wagenbach and KipfstuhlFischer and others, 1998a):

(5a)

(5b)

According to Equation (5a), the contribution of dry deposition and the impact of snow accumulation rate on volcanic sulphate concentrations in firn can be assessed. A linear regression of the mean volcanic sulphate concentrations ((volc-SO₄ ^2–) in ngg^–1) from 1992 to 1993 vs the inverse accumulation rate (accu^–1 in m² a kg^–1) for each snow pit results in:

(6)

Note that the values for DML19 and DML20 are not included because of highly uncertain RRM_i. This relationship indicates that only 17% of the spatial variability in volcanogenic sulphate concentrations in the firn can be explained by the change in snow accumulation. Following the simplistic approach that annual mean volcanic sulphate concentrations in fresh snow are independent of the snow accumulation rate, about half the sulphate aerosol from the Pinatubo and Cerro Hudson eruptions was removed by dry deposition on Amundsenisen (J _dry ═ 2:0 ±1.5 kg a^–1 km^–2). In addition, a corresponding linear regression revealed that only 11% of the variation of D(volc-SO₄ ^2–), i.e. J _total(volcSO₄ ^2–) integrated over the duration of the volcanic eruption impact, can be explained by the snow accumulation rate in this area (Fig. 4). Although clearly not ideal, the total volcanic sulphate deposition D(volc-SO₄ ^2–) appears to be a more suitable, i.e. less accumulation-dependent, measure of the volcanogenic sulphuric acid impact in the area of investigation than the corresponding firn concentration C _firn(volc-SO₄ ^2–).

Fig. 4. Spatial distribution of volcanic sulphate deposition on Amundsenisen. The total volcanic sulphate deposition of the years 1992 and 1993 is plotted vs the snow accumulation rate.

Accordingly, the volcanic chronology in this area can be cross-checked by comparing the calculated volcanic sulphate deposition at different sites. All calculated volc-SO₄ ^2– deposition values were within the range between 12 kg km^–2 (at DML18) and 21.7 kg km^–2 (at DML17), except for the snow pits at DML20 (6.9 kg km^–2) and DML19, where no volcanic sulphate deposition could be determined. The latter cases could be explained by local effects (e.g. wind erosion removing a part of the snow deposited during 1992 and 1993). Excluding DML19 and DML20 from our further evaluation, for the remaining sampling sites the average sulphate deposition caused by the Pinatubo and Cerro Hudson eruptions is 15.8 ± 3.4 kg km^–2 on Amundsenisen. For the South Pole, a site of similar altitude (2835 m) and only slightly higher mean accumulation rate (80 kg m^–2 a^–1 (Reference Delmas, Kirchner, Palais and PetitDelmas and others, 1992)), a comparable sulphate deposition of 13.8–14.5 kg km^–2 was determined for the sum of these two volcanic eruptions (Reference Cole-Dai, Mosley-Thompson and ThompsonCole-Dai and others, 1997b). The roughly uniform geographical distribution of the sulphate deposition caused on Amundsenisen by the Pinatubo and Cerro Hudson eruptions supports the contention that any eruption of similar size should be well and comparably documented in the snowpack of the area under investigation. However, even for an eruption such as that of Pinatubo, an inherent error in volcanic sulphate deposition of at least ±22% has to be taken into account at this site.

3.2.1. The time period AD 1800–1997

In this period, six volcanic signals at DML03, nine at DML05, ten at DML07 and DML17 were identified using y _r as the detection criterion (Table 1). For the same time period, seven peaks were observed by Reference KruegerKarlöf and others (2000) in the ice core from Camp Victoria (CV).

The imprint of the eruptions from Pinatubo and Cerro Hudson (unresolved peak 1) is described in detail in section 3.1. Aerosol measurements at the coastal stations Neumayer and Dumont d’Urville revealed that about 30–40% of the volcanic sulphate in the atmosphere between 1991 and 1993 can be attributed to the weaker but closer-located eruption of Cerro Hudson (Reference McCormick, Thompson and TrepteLegrand and Wagenbach, 1999). This is also supported by firn-core samples from the South Pole, where the sulphate signals caused by the two eruptions are well separated (Reference Cole-Dai, Mosley-Thompson and ThompsonCole-Dai and others, 1997b). The Pinatubo eruption injected about 18±2 Mt (10¹² g) SO₂ into the middle stratosphere (Reference Langway, Clausen and HammerKrueger and others, 1995), while only about 2 Mt SO₂ were injected, mainly below 14 km altitude, by Cerro Hudson (Reference Doiron, Bluth, Schnetzler, Krueger and WalterDoiron and others, 1991). The relatively high sulphate deposition caused in coastal Antarctica by the Cerro Hudson eruption is most probably due to an effective SO₂/H₂SO₄ advection through the upper troposphere and lower stratosphere, while sulphate deposition from the stronger Pinatubo eruption can be mainly accounted for by long-range transport via the stratosphere (Reference McCormick, Thompson and TrepteLegrand and Wagenbach, 1999). Thus, an indeterminate but most probably significant part of the observed sulphate deposition was caused by the eruption of Cerro Hudson. As a consequence, the sum of sulphate deposition of both eruptions archived in DML is not representative for a corresponding stratospheric aerosol loading.

We assign peak 2 to the eruption of Reference Oerter, Graf, Wilhelms, Minikin and MillerEl Chichón in 1982. The calculated volcanic sulphate deposition varied between 2.2 and 3.4 kg km^–2 (Table 1). Interestingly, this eruption was not reported in studies on other Antarctic ice cores (Table 2), except Law Dome cores (Reference Palmer, Morgan, Curran, van Ommen and MayewskiPalmer and others, 2001). Satellite images revealed that the eruption veil of El Chichón was nearly exclusively distributed in the Northern Hemisphere (Reference Wilson, Ambraseys, Bradley and WalkerStrong, 1984). An extensive snow-pit study across central Greenland revealed a total volcanic sulphate deposition from El Chichón of about 12–16 kg km^–2 (Reference ZielinskiZielinski and others, 1997a).

Concurrently with peak 3 (deposited in AD 1969), in all four ice cores maxima in the MS concentration were detected. Although the MS peaks were very high, the nss-sulphate concentration peaks could hardly be detected even with our sensitive algorithm. In comparison, in none of the other glaciochemical studies of Antarctic ice cores was a significant nss-sulphate peak found, while electrical conductivity measurement (ECM) and DEP records from CV (Reference KruegerKarlöf and others, 2000) and Mizuho (Reference Newhall and SelfMoore and others, 1991) revealed a distinct peak in the corresponding year. We conclude that high biogenic MS and nss-sulphate concentrations, rather than a volcanic event, most probably caused the corresponding ECM and DEP peaks.

The eruption of Agung in AD 1963 (peak 4) is well archived in the glaciochemical records of the ice cores from sites DML07 and DML17, and by using k ═ 3 it was also possible to identify this peak at DML03. Although this eruption was not detected at CV, it is generally well represented in all Antarctic ice cores listed in Table 2 as well as at Dome C (Reference Legrand and PasteurLegrand and Delmas, 1987), Law Dome (Reference Palmer, Morgan, Curran, van Ommen and MayewskiPalmer and others, 2001) and Mizuho (Reference Newhall and SelfMoore and others, 1991). Reference Palais, Germani and ZielinskiOerter and others (2000) discerned this event by ECM and DEP in 7 out of 12 intermediate-depth ice cores from Amundsenisen. In the GISP2 record from Summit, Greenland, the imprint of the Agung eruption is most probably negligible and additionally masked by the Sheveluch (Kamchatka, AD 1964) eruption (Reference ZielinskiZielinski, 1995).

We assigned peak 5 to the eruption of Cerro Azul in AD 1932, which was also detected at CV. The calculated volcanic sulphate deposition was quite low for an eruption classified with a VEI of 5, most probably because of unusually low sulphur loading of the eruption veil. This is supported by the absence of a corresponding volcanic signal in all other Antarctic ice cores listed above. As might be expected, this low-sulphur-loading eruption, which occurred in the mid-latitudes of the Southern Hemisphere, was not detected in the GISP2 record (Reference ZielinskiZielinski, 1995).

Peaks 7–11 are well represented in our ice cores and in all the ice cores from Antarctica listed in Table 2. However, peaks 7 and 9, assigned to the eruption of Tarawera (in AD 1886) and Coseguina (in AD 1835) were not found at CV. At least one reason for the absence of peak 7 could be super-imposition by the strong volcanic eruption of Krakatau in AD 1883 in combination with a less sensitive peak-detection algorithm. Assuming that the Tarawera eruption did not inject a large amount of sulphur into the stratosphere, an originally faint deposition signal at CV could also be eroded at this site. The mean sulphate deposition caused by the Krakatau eruption (peak 8) was 9.6 kg km^–2 on Amundsenisen, which agrees well with values determined in ice cores from the South Pole, Byrd, Siple Station and Plateau Remote (Table 2). A striking horizon for all Antarctic and Greenland ice cores derives from the eruption of Tambora in AD 1815 and that of an unknown volcano in AD 1808, preserved in Amundsenisen with outstanding volcanic sulphate depositions (Table 2). The 1808 event is most peculiar, because it is also a prominent signal in Greenland ice cores (Table 2). As yet it is not clear whether a single huge Equatorial eruption or, as suggested by Reference Newhall and SelfMoore and others (1991), two minor eruptions that may have occurred around this time at high northern and southern latitudes caused these imprints in polar ice. The latter possibility seems more realistic, because it is hardly conceivable that a volcanic eruption comparable to the most prominent events of the last few decades was not recognized at the beginning of the 19th century.

3.2.2. The time period AD 165–1800

As mentioned above, a detailed discussion of individual volcanic events is inappropriate for the period beyond AD 1800, because assignment of the detected peaks is highly uncertain and equivocal. Nevertheless, the sequence compiled in Table 1 represents the most accurately dated volcanic eruption chronology available for Antarctica to date for the period AD 1200–165.

Our record reveals a period of 450 years exhibiting very low volcanic activity (AD 700–1150). Supporting the findings of Reference Cole-Dai, Mosley-Thompson, Wight and ThompsonCole-Dai and others (2000), this quiescent period is followed by a series of violent volcanic eruptions between AD 1150 and 1300 marked by peaks 22–29 (Fig. 2; Table 1). During the quiescent period, only about 86±20 kg km^–2 (mean flux: 0.19 kg km^–2 a^–1) of volcanic sulphate was deposited, while from AD 1150 to 1300 a total of 188 ± 30 kg km^–2 (mean flux: 1.25 kg km^–2 a^–1) was deposited in the area under investigation. This is also significantly different from the long-term mean volcanic sulphate deposition flux of around 0.42 kg km^–2 a^–1 derived from the ice core at DML05 between AD 165 and 1997. Note that for each period the background sulphate deposition flux (i.e. nss sulphate minus volcanic sulphate deposition flux) is about 3.5±0.2 kg km^–2 a^–1 and thus much higher than the mean volcanic sulphate flux within the corresponding period.

The striking peak 17 is assigned to the eruption of Huaynaputina in AD 1600. Beyond AD 1601 a more accurate dating of the volcanic eruptions is generally performed by the applied annual-layer counting. The years mentioned are the dates assigned in the ice core at DML05 and are listed in Table 1 with the corresponding uncertainties. The strong volcanic horizon at AD 1453±5 was detected in many ice cores between AD 1450 and 1460 (Reference Delmas, Kirchner, Palais and PetitDelmas and others 1992; Reference Legrand and DelmasLangway and others 1994; Reference Cole-Dai, Mosley-Thompson and ThompsonCole-Dai and others, 1997a) and has been tentatively assigned to the eruption of Kuwae (Reference ZielinskiZielinski, 2000). Of the remaining signals, only peaks 28, 40 and 48 were clearly due to a reported volcanic event, namely the violent eruptions of Tarawera, Rabaul and Taupo, respectively. The AD 1257 event has been tentatively assigned to an eruption of El Chichón (Reference Delmas, Kirchner, Palais and PetitPalais and others, 1992). Note that the outstanding signal around AD 540 is now thought to be caused by an eruption of Krakatau (personal communication from G.A. Zielinski, 2003).