The Skallen terminus ice

Figure 3a shows the complete sequence of ice and debris strata observed in the Skallen terminus ice. It also indicates the height profiles of δ¹⁸O and some ion (Cl⁻, Na⁺ and Ca²⁺) concentrations. The horizontal axis of this figure represents the height measured upward from the base of the ice sheet; this convention is used for the remainder of the paper.

Fig. 3.

(a) Profiles of the debris distribution, δ¹⁸O values and Cl⁻, Na⁺ and Ca²⁺ concentrations of the Skallen terminal ice. White, light gray, heavy gray, and black marks indicate zero, low, medium and high levels of debris respectively. (b, c) Same as (a), but for the Skarvsnes and Hamna terminal ice respectively.

The debris-rich layer extends from 0.84 to 1.80 m in height. The Na⁺ and Ca²⁺ profiles show a similar trend, namely that they are much higher within this layer than in precipitated ice from EDML (Reference Satow, Watanabe, Shoji and MotoyamaSatow and others 1999; Reference Suzuki, Iizuka, Matsuoka, Furukawa, Kamiyama and WatanabeSuzuki and others 2002) even during the cold climatic stages (Reference WatanabeWatanabe and others, 2003). Looking at the data in more detail, we find that both Na⁺ and Ca²⁺ are strongly correlated with the debris concentration (r ² = 0.78). These results suggest that the primary source of Na⁺ and Ca²⁺ is dissolution from subglacial bedrock debris. The SO₄ ²⁻, K⁺ and Mg²⁺ profiles are all similar to the Na⁺ and Ca²⁺ profiles.

δ¹⁸O ranges from −52‰ to −20‰. The δ¹⁸O values change abruptly on crossing the upper border of the debris-rich layer at 1.84 m. The upper part of the ice core (1.84–4.81 m) has the lowest average value of δ¹⁸O, −51.7 ± 0.6‰ (Table 1). The lower layers (0–1.84 m) of the sample have high δ¹⁸O values, with an upper limit of −20‰ and an average of 35.7 ± 10.5‰ (Table 1). Even if the Skallen terminus ice experienced some fractionation due to refreezing at the inland bedrock, the δ¹⁸O value should change by no more than several per mil (e.g. Reference Souchez and JouzelSouchez and Jouzel, 1984). Thus, the difference between the upper and lower profiles can be attributed mainly to differences in the δ¹⁸O of the precipitation. The Cl⁻ concentration profile is similar to the δ¹⁸O profile, their correlation coefficient being relatively high (r ² = 0.80). This result suggests that the primary source of Cl⁻ is not debris dissolution but sea salt from the atmosphere. In general, the continental crust seldom contains chlorine. The upper (1.84–4.81 m) and lower (0–1.84 m) parts of the sample have average Cl⁻ concentrations of 1.67 ± 1.00 and 7.06 ± 2.93 μmol L⁻¹ respectively (Table 1).

Table 1.

Average δ¹⁸O values (‰) and average concentrations of seven ions (μmol L⁻¹) in the Skallen, Skarvsnes and Hamna terminal ice sections

In the upper part of the Skallen terminus, the white glacier ice and the debris-laden basal ice have similar δ¹⁸O values and Cl⁻ concentrations (Table 1). We therefore integrate these data in the following discussion.

The Skarvsnes terminus ice

Figure 3b shows a complete stratigraph of the Skarvsnes terminus ice, along with the height profiles of δ¹⁸O and some ion (Cl⁻, Na⁺, Ca²⁺) concentrations.

There are some thin (several cm or less), debris-rich layers in the Skarvsnes terminus ice, although the total amount of debris present is less than that observed in the Skallen terminus. The Ca²⁺ concentration is much higher than in precipitated ice from EDML (Reference Satow, Watanabe, Shoji and MotoyamaSatow and others 1999; Reference Suzuki, Iizuka, Matsuoka, Furukawa, Kamiyama and WatanabeSuzuki and others, 2002; Reference WatanabeWatanabe and others, 2003), so the primary source of Ca²⁺ in these layers is probably dissolution from subglacial bedrock debris.

δ¹⁸O ranges from −48‰ to −37‰, changing dramatically at a height of 0.39 m. Above this point (0.39–3.89 m), the average δ¹⁸O value is −46.4 ± 1.4‰. Below this point (0–0.39 m), the average δ¹⁸O value is 42.5 ± 2.5‰ (Table 1). The upper part (0.39–3.89 m) and the lower part (0–0.39 m) have average Cl⁻ concentrations of 1.71 ± 1.25 and 6.99 ± 5.29 μmol L⁻¹ respectively (Table 1), comparable to those observed in precipitated ice from inland EDML. As with the Skallen terminus ice, we can conclude that the Cl⁻ in the Skarvsnes terminus ice originates mainly from atmospheric sea salt.

The upper (0.39–3.89 m) and lower (0–0.39 m) parts of the sample have average Na⁺ concentrations of 3.36 ± 2.08 and 13.87 ± 10.12 μmol L⁻¹ respectively (Table 1). On average, assuming that all Cl⁻ ions came from sea salt, we expect non-sea-salt Na to account for 46.9% and 52.7% of the Na⁺ in the upper and lower parts respectively, of the Skarvsnes terminus ice. Meteoric nssNa in precipitation contributes at most 30% to the overall Na concentration, even during the LGM when terrestrial sodium was most prevalent (Reference RöthlisbergerRöthlisberger and others, 2002). Thus, contributions of nssNa⁺ higher than 30% indicate that the Skarvsnes terminus ice contains ions from both atmospheric sea salt and subglacial debris dissolution.

The white glacier ice and debris-laden basal ice have similar δ¹⁸O values and Cl⁻ concentrations in the upper part of the Skarvsnes terminus (Table 1). Thus, as with the Skallen ice, we integrate these data in the following discussion.

The Hamna terminus ice

Figure 3c shows a complete stratigraphy of the Hamna terminus ice and indicates the height profiles of δ¹⁸O and some ion (Cl⁻, Na⁺, Ca²⁺) concentrations. As with the Hamna terminus ice, the chemical characteristics of this sample are reported elsewhere (Reference Iizuka, Igarashi, Shiraiwa, Naruse, Yamada and WatanabeIizuka and others, 2000, Reference Iizuka, Satake, Shiraiwa and Naruse2001). In summary, the Hamna terminus ice can be divided into upper (1.30–7.50 m) and lower (0–1.30 m) parts with different characteristics. The primary source of Na⁺ and Ca²⁺ is dissolution from subglacial bedrock debris. As with the two other sites, the primary source of Cl⁻ is atmospheric sea salt. The average δ¹⁸O values and Cl⁻ concentrations in ice from the upper part, including white glacier ice, are −45.8 ± 0.7‰ and 0.59 ± 0.39 μmol L⁻¹ respectively (Table 1).

Origin of the upper part of the Skallen, Skarvsnes and Hamna terminus samples

Here we focus on the upper parts of the Skallen, Skarvsnes and Hamna terminus ice sections. Their average δ¹⁸O values are −51.7‰, −46.4‰ and −45.8‰ respectively. These are very low values compared with precipitation from the marginal region of the Sôya drainage basin, where seasonal δ¹⁸O values range from −33‰ to −10‰ at present (Reference KatoKato, 1979). This indicates that the basal ice originated from precipitation on the inland regions of the Shirase and Sôya drainage basins.

Figure 4 plots δ¹⁸O values against Cl⁻ concentrations for the three terminal ice sections and present surface snow. Also shown are data representing this relationship during the corresponding climatic period of the Dome Fuji (DF) ice core. The DF ice core was drilled at the summit of the Shirase drainage basin, and covers several glacial and interglacial periods (e.g. Reference WatanabeWatanabe and others, 2003). In general, cold-stage ice has lower δ¹⁸O values and higher Cl⁻ concentrations than warm-stage ice (Figs 4 and 5). Typically, Cl⁻ concentrations are >3 μmol L⁻¹ in the cold stages (marine isotope stages (MIS) 2 and 4) but <2 μmol L⁻¹ in the warm stages (MIS 1, 5a, 5c and 5e) (Figs 4 and 5). At present, Cl⁻ concentrations vary little over the inland region above 2500 m a.s.l. (e.g. Reference Suzuki, Iizuka, Matsuoka, Furukawa, Kamiyama and WatanabeSuzuki and others, 2002) where δ¹⁸O values are less than −40‰. Figure 4 indicates that all three terminal ice sections have δ¹⁸O values less than −40‰ and Cl⁻ concentrations of ∼2 μmol L⁻¹, suggesting that the three terminal ices originate from inland precipitation during a warm climatic period (e.g. MIS 1, 5a, 5c and 5e).

Fig. 4.

(a) The relationship between δ¹⁸O values and Cl⁻ concentrations for ice around EDML. (b) A close-up of the same plot for low δ¹⁸O values and Cl⁻ concentrations. The average values of the upper parts of the Skallen, Skarvsnes and Hamna terminal ice sections are shown by open squares. The average values in snow from the surface to 2 m depth obtained from nine altitudes are shown by black diamonds. An approximate fit is shown by the black curve. The average values of past climatic stages (MIS 1, 2, 3, 4, 5a, 5c and 5e) are shown by open circles. The surface snow and past ice data are taken from Reference HigashiHigashi (1997) and Reference WatanabeWatanabe and others (2003) respectively.

Fig. 5.

δ¹⁸O values and Cl⁻ concentrations in the most recent 150 ka BP of the DF ice core, as reported by Reference WatanabeWatanabe and others (2003). The gray shaded areas of each component correspond to the upper part of the terminal ice sections.

The above suggestion is supported by the average Na⁺ concentrations in white glacier ice of the three termini (Table 1), which are <2 μmol L⁻¹. The main source of Na⁺ is also sea salt (1.18 for the sea-salt ratio of Cl⁻/Na⁺), and the Na⁺ concentrations of the Sôya and Shirase drainage basins above 2500 m a.s.l. fall below 2 μmol L⁻¹ (e.g. Reference Suzuki, Iizuka, Matsuoka, Furukawa, Kamiyama and WatanabeSuzuki and others, 2002) only during warm climatic periods such as MIS 1, 5a, 5c and 5e (Reference WatanabeWatanabe and others, 2003). Thus, the Na⁺ in this ice also originates from inland precipitation during a warm climatic period.

The average δ¹⁸O values observed in the MIS 1, 5a, 5c and 5e sections are also within ±3‰ of the present Dome Fuji δ¹⁸O value (Figs 4 and 5). The relationship between δ¹⁸O and surface elevation (Reference Satow and WatanabeSatow and Watanabe, 1992) is

where E is elevation (m). If we assume that the present spatial distribution of δ¹⁸O in EDML has not changed since previous climatic periods (MIS 1, 5a, 5c and 5e), then the upper parts of the Skallen, Skarvsnes and Hamna terminus ice sections could have originated from inland precipitation around 3423 ± 253, 2976 ± 253 and 2926 ± 253 m a.s.l. respectively. Note that the precipitation originates from higher elevations in the southern sample area (Skallen), and from lower elevations in the northern sample area (Hamna).

Since the present head of the Sôya drainage basin is ∼3000 m a.s.l. (Fig. 1), precipitation in the Skallen terminal ice section must have originated from the neighboring Shirase drainage basin which lies at a higher elevation. As a reference, the amplitude of the altitude change does not exceed 150 m on the Antarctic Plateau over the last 420 ka BP (Reference Ritz, Rommelaere and DumasRitz and others, 2001). This result implies that the ice divide between the Shirase and Sôya drainage basins was different in the past, and that ice presently located in the terminal part of the Sôya drainage basin originated from different past precipitation. On the other hand, the precipitation in the Skarvsnes and Hamna terminal ice section must have originated from the top of the Sôya drainage basin. This result implies that the past ice divide between the Shirase and Sôya drainage basins was located between the Skallen and Skarvsnes regions.

Deducing the age of the upper Skallen terminal ice by principal component analysis

The upper part of the Skallen terminal ice originates from inland precipitation on the Shirase drainage basin. Here we deduce the age of the upper part of the Skallen terminal ice by a principal component analysis (PCA) of its chemical data, δ¹⁸O values and Cl⁻, NO₃ ⁻, SO₄ ²⁻, Na⁺, Mg²⁺ and Ca²⁺ concentrations. Time series of these values have already been derived from the DF ice core over the most recent 150 ka BP (Reference WatanabeWatanabe and others, 2003). The six ions are the most common species in the DF core (Reference IizukaIizuka and others, 2008). Note that the K⁺ concentration in the DF ice core is so small (Reference WatanabeWatanabe and others, 2003) that it is occasionally absent from the ion chromatogram, and so was left out of the PCA.

After smoothing the DF series by a 500 year running mean, we were left with 284 data points. Along with 12 independent chemical analyses of white glacial ice from the Skallen terminus, the smoothed chemical vectors from the DF core are then projected onto a new basis of seven principal components.

Figure 6 plots the second principal component against the first for each vector in the dataset. Each climatic stage of the DF ice core is given a different symbol, as is the Skallen ice. We calculated average values for the seven components and subtracted these from the data to give deviations around zero. The eigenvalues (and contributions) of the first and second components are 5.87 (83.9%) and 0.59 (8.5%) respectively; together they explain 92.4% of the total variance. Figure 6 suggests that the first component is closely related to air temperature, because it is positive during MIS 1 and 5e and negative during MIS 2 and 4. The meaning of the second component is not obvious, but it probably depends on the ratio of Ca²⁺ (terrestrial material) and Na⁺ (sea salt). The MIS 1 and 5e periods are interglacial, with low Ca²⁺ and Na⁺ concentrations. The MIS 2, 3 and 4 periods are relatively cold glacial, with high Ca²⁺ and Na⁺ concentrations. MIS 5a–5d are relatively warm glacial, with low Ca²⁺ but high Na⁺ concentrations.

Fig. 6.

The first two principal components of chemical data from the white glacier ice of the Skallen terminal ice and each climatic stage of the DF ice core.

Several climatic stages can be distinguished clearly by their position in Figure 6, and each stage forms a distinct cluster indicating consistent chemical characteristics and δ¹⁸O values. The Skallen terminal ice is positioned near the MIS 5e cluster but far from the MIS 1, 5a and 5c clusters. This analysis strongly suggests that the upper part of the Skallen terminal ice originates in MIS 5e precipitation, and rules out MIS 1, 5a and 5c.

If we take into account a simple flow model (Reference NyeNye, 1963) of an ice sheet 3000 m thick with 50 mm of annual accumulation (precipitation giving rise to the Skarvsnes and Hamna terminal ice), the thicknesses of annual layers located at one-half and three-quarters of the total height are 25 and 13 mm respectively. These layers are disturbed as they flow to the terminus by various dynamic processes (e.g. folding, boudinage, and faulting near the bedrock) (Reference PatersonPaterson, 1994). The marginal ice sections are several meters thick, with a sampling resolution of about 100 mm, corresponding to a period of several decades and a resolution of several years. This terminal period corresponds to a single datum of the DF ice core (Fig. 5). The relatively low resolution is sufficient to transform the annual layer concentrations into a well-homogenized series, with smooth stochastic peaks. Thus, estimations from the ice-flow model allow us to relate the δ¹⁸O values and Cl⁻ concentrations of the marginal ice samples to recent climatic stages in the DF ice core as described above.

Past migrations of the ice divide between the Shirase and Sôya drainage basins

In order to deduce the past position of the ice divide between the Shirase and Sôya drainage basins, we calculate the past elevation of the Antarctic ice sheet using a large-scale ice-sheet model. SICOPOLIS (SImulation COde for POLythermal Ice Sheets) is a commonly used 3-D dynamic/thermodynamic model developed by Reference GreveGreve (1995, Reference Greve1997) and available as free software (http://sicopolis.greveweb.net). We conducted a paleoclimatic simulation from 422 ka BP until the present, adopting a spatial resolution of 40 km. Greve (http://hdl.handle.net/2115/34433) provides details of the simulation set-up.

The main outputs used in this study are snapshot elevation maps of the Antarctic ice sheet at 140, 130, 84, 66, 45, 20, 9 and 0 ka BP, corresponding to typical MIS 6, 5e, 5a, 4, 3, 2, 1 and present conditions. These ages cover fluctuations of the ice sheet during the most recent glacial–interglacial period. Figure 7 shows some of these elevation maps. We compared the 0 ka BP elevation map calculated by SICOPOLIS with the present elevation distributions (Reference HigashiHigashi, 1997), and found flexion points on the 2400 and 2600 m contours of EDML in both distributions (Figs 7 and 8). The 2400 and 2600 m flexion points are located on the right and left banks of the Sôya drainage ice divide respectively. We confirmed that the 2400 and 2600 m flexion points at 140, 130, 84, 66, 45, 20 and 9 ka BP are also located on the right and left banks of the Sôya drainage ice divide. Hence, these two features can serve as useful reference points to determine past fluctuations of the ice divide. We measured changes in the latitude and longitude of the 2400 and 2600 m flexion points with respect to their present locations in each snapshot.

Fig. 7.

Some snapshots of past ice elevations of the Antarctic ice sheet (66, 45, 20 and 0 ka BP) calculated by SICOPOLIS. The arrow between the MIS 2 and present snapshots shows an example of the moving of flexion points.

Fig. 8.

A scaled-up map of the Sôya drainage basin from Figure 1. The locations of the 2400 and 2600 m flexion points at 20 ka BP are indicated by black circles. The two dotted lines are the reconstructed ice divides at 20 ka BP obtained by processing parallel shifts from the present ice divides.

Since SICOPOLIS does not take into account the dynamics of an ice shelf, the past elevation distributions of the marginal part have an undetermined margin of error. By choosing our reference points from the inland part of the Sôya drainage basin, we avoided this source of error.

Table 2 describes the migration of the 2400 and 2600 m flexion points. At 140, 84, 66, 45 and 20 ka BP (MIS 2, 3, 4 and 6), during the glacial period, these points were situated to the northeast of their present locations. For example, the 2400 and 2600 m points at 20 ka BP (MIS 2) are located at 0.12° N, 0.37° E and 0.14° N, 0.46° E with respect to their present positions (two-way arrow in Fig. 7). In particular, the rapid eastward migration (by 0.46° E) of the 2600 m point means that the 20 ka BP ice divide was located in the present center of the Sôya drainage basin (Fig. 8).

Table 2.

Past migrations of the 2400 and 2600 m flexion points relative to their present positions

This is how precipitated ice can have flowed from physical coordinates now located in the Shirase drainage basin to the terminus of the present Sôya drainage basin. Before the eastward migration of the ice divide, the west side of the region now comprising the Sôya drainage basin consisted of inland Shirase drainage ice (black enclosure in Fig. 8). After the ice divide migrated to its present position, ice originally located in the Shirase drainage basin became part of the Sôya drainage basin and flowed toward the Sôya coast. The northeastward migration also explains why only the Skallen region, located close to the central estuary of the past Shirase drainage basin, stores ice from an ice divide farther upstream. The Skallen and Skarvsnes regions are located farther from the divide, so their compositions are less affected by its past migration.

The SICOPOLIS software also outputs the surface velocity of the ice sheet. Using these data, we can calculate a lower limit for the age of the terminal ice. We began tracking the flow from the inferred origin of Skallen marginal ice, at 74.6° S, 47.0° E (3400 m a.s.l.). The surface velocity at this position is about 5 m a⁻¹. Moving along the flowline toward the Skallen region, we obtain velocities of >10 m a⁻¹ at 70.8° S, 47.3° E, and >100 m a⁻¹ at 70.1° S, 42.1° E. Integrating along this path, we find that the total flow time from the origin to the Skallen region is about 98 ka. As the surface velocity is higher than the particle velocity within the ice sheet, we can take 98 ka as a lower limit. This calculation supports the PCA result, which implies that the upper part of the Skallen terminal ice was precipitated during MIS 5e (124–134 ka BP). This calculation rules out the MIS 1 and MIS 5a stages.

Flow history and formation mechanism of the Skallen terminal ice

The above results and discussion allow us to propose a detailed flow history of the Skallen terminal ice in the Sôya drainage basin. Figures 1 and 7 describe the history of the upper ice, which could originate in precipitation during a warm period of MIS 5e over the inland area of the Shirase drainage basin at about 3423 m a.s.l. (black circle in Fig. 1). The precipitated ice then flowed over the present ice-divide line between the Shirase and Sôya drainage basins (black arrow in Fig. 1) at a time when the ice divide was to the northeast of its present location (MIS 2, 3 or 4) (black enclosure in Fig. 8). Finally, the ice flowed through the Sôya drainage basin to the Skallen region after the ice divide moved to its present position (MIS 1; black arrow in Fig. 8).

This flow history implies that the Shirase drainage basin of the glacial period, one of the fastest discharges of the Antarctic ice sheet, is narrower than it once was. Further analysis of terminal ice taken from other parts of the Sôya and Harald drainage basins will allow a more precise evaluation of mass balance and ice dynamics in Shirase Glacier.

Sometime during the aforementioned flow history, the upper part of the basal ice layer (1.84–3.48 m) would have touched the bedrock, although the white glacier ice (3.48–4.81 m) did not. The δ¹⁸O of the lower part of the basal ice (0–0.56 m) is at most −20‰, corresponding to the annual average in local precipitation at the Sôya coast. Thus, the lower part of the Skallen basal ice consists of on-site precipitation. Two mechanisms may have caused the inland ice (the upper part) to overlap the local ice (the lower part) near the bedrock. First, the lower part was already dead ice when the Skallen terminal ice passed over it. In this case, the inland ice could have obducted on the dead ice by thrusting. Second, the lower part was formed from local summer meltwater, which froze at the base of the ice sheet and adhered to the underside of the inland ice.

The δ¹⁸O of the boundary between the upper and lower parts of the basal ice (0.56–1.84 m) is between −48‰ and −28‰. These values suggest that the boundary layer was formed by mixing inland ice and local ice. The boundary layer has a high Ca²⁺ concentration and a significant amount of debris, which may have been entrained by ice forming at the interface between ice sheet and bed. Figure 2e shows a transmissive photograph of the boundary layer ice, which consists of bubbly ice and contains dispersed (Reference LawsonLawson, 1979) or clotted (Reference KnightKnight, 1997) debris facies. Reference KnightKnight (1997) suggested that such debris types have traveled long and far since their entrainment in the basal ice, which if true would imply that the boundary layer formed inland somewhere other than the present Skallen region. On the other hand, if the basal ice formed by freezing meltwater, it should consist of bubble-free ice (e.g. Reference KnightKnight, 1997; Reference Iizuka, Satake, Shiraiwa and NaruseIizuka and others, 2001). We conclude that the lower part of the basal ice already existed as dead ice of local precipitation when the Skallen terminal ice advanced through its original location. The upper part was then obducted on the dead ice by thrusting.