In the deeper part of the ice core, between 628–635 m, we observe some discontinuities, with rapid and simultaneous changes in water isotopes and methane at the same depth. These will be discussed in more detail later, but they represent likely depths of flow disturbance or folding, as has been observed in other ice core records, including those of the LIG in Greenland (Chappellaz et al., 1997; NEEM Community Members, 2013; Yau et al., 2016). We also deduce that some disturbance may exist in a region between about 605 and 615 m depth. From 600 m downwards we therefore carefully examine individual data points (using paired values of CH4 and δ18Oatm matched against reference data) to reconstruct discrete ages for particular depths. This allows us to assess which sections are in order with well-constrained ages and which are disturbed in the deeper ice. We then use Paleochrono to derive a continuous age model to 628 m, making manual adjustments to the final age scale to avoid assigning spurious ages to data in the disturbed section.

Methane measurements are a particularly powerful way of aligning the gas ages of different ice cores because they exhibit large (from tens to 200 ppb) and abrupt changes of concentration across millennial-scale Dansgaard–Oeschger events that recur throughout the last glacial period (e.g. Epifanio et al., 2020). Using high-resolution continuous analysis it has also been shown that centennial and faster variability down to below 10 ppb amplitude is well reproduced between cores (Lee et al., 2020; Mitchell et al., 2013; Rhodes et al., 2017). As methane is well mixed in the Antarctic troposphere, not just the pattern but the absolute values should match with reference datasets within uncertainty. Our main dataset, from continuous flow analysis (CFA), is good at showing the high-resolution variability but has a large and unknown uncertainty in absolute values. We therefore supplement it with some discrete analyses (Sect. 3.2) that constrain the concentration tightly at key sections of ice.

We measured methane (CH4) continuously during the continuous flow analysis (CFA) campaign (Grieman et al., 2021). Briefly, the core was melted at a mean rate of 3.2 cm min-1 and the air was separated from residual water flow using a 3M Liqui-Cel MM-0.5x1 Series membrane contactor. The dried air was then directed to a Picarro G2301 CRDS for CH4 analysis. While the methane Picarro calibration could not be checked against external certified standards, comparison of our data produced by CFA with analysis of discrete samples analysed in Bern (Sect. 4.2), as well as comparison of our CFA data with reference data across the Holocene and glacial, suggests that the CFA methane reproduces the variability in methane at centennial scales. However, the absolute values are offset (mainly low) by an amount that varied by a few percentage points over the campaign, but the offset was typically below 10 %. This offset arises partly from dissolution of a small percentage of gas into the meltwater stream, as has been observed previously using CFA to measure methane (Rhodes et al., 2015). Continuous analyses started at 244 m depth and continued in all sections where the ice was of suitable quality to 649.4 m. A short section from 144.0–161.3 m was also analysed continuously for methane with an improved measurement setup which is discussed in the companion paper (Hoffmann et al., 2022).

Two significant issues affected the measurements. Firstly, a section of data between 534 and 545 m was affected by a leak of lab air at the membrane contactor. The absolute values in this section of ice are therefore substantially higher than the palaeoatmosphere, but the pattern of variability can still partly be used for wiggle matching after correction using discrete analyses (next section).

A second issue is that there were increasing numbers of breaks and cracks in the ice with depth, particularly below 450 m. Badly cracked sections were removed before the ice was placed on the melter, and breaks across the core were smoothed with a cleaned file to ensure that the contact between ice sections was as close as possible. With these precautions, such occurrences do not affect the ice phase chemical measurements and most do not affect CH4 either. Nonetheless some of the remaining cracks and transitions between different bags provide an opportunity for the ingress of lab air as the ice melts, leading to spikes in methane concentration. Major short peaks and troughs were identified manually and removed from the dataset. Above 500 m, ∼ 25 spikes that were at least a factor 2 higher or lower than the mean of the dataset were removed. Below 500 m, the data became much noisier and ∼ 100 deviations from the dataset were manually removed. Even after removal of the obvious spike artefacts, the data remain more noisy than the data that are unaffected by such artefacts, suggesting that positive artefacts arising from inclusion of modern air remain in the dataset. This makes it trickier to clearly align data with a reference dataset in the deeper ice. The dataset below 244 m is shown in Fig. 3.

To validate and control that the absolute levels of our continuous CH4 record are consistent within uncertainty with the absolute values in reference data, we obtained some well-calibrated discrete measurements (Fig. 3), particularly in the deep ice and in the section impacted by the air leak (Sect. 3.1). Ten discrete samples were therefore measured at the University of Bern between 533–546 m, and a further 25 samples were measured between 600 and 650 m depth. Details of the method have been published elsewhere (Schmidely et al., 2021). Concentrations ranged between 413 and 644 ppb, with an estimated precision (1σ) of 7 ppb (Table S1 in the Supplement). Note that the discrete data presented here have been corrected by -18 ppb (Schmidely et al., 2021) to align them with previously published CH4 records. These offsets are potentially due to different remnant solubility of CH4 in meltwater using different melt extraction methods in different labs. Taking the uncertainty of the correction into account, the total uncertainty is estimated at 12 ppb (Schmidely et al., 2021), while that of the reference data is estimated at 10 ppb (Loulergue et al., 2008). Combining these uncertainties suggests that when comparing absolute values of methane (discrete data) with reference datasets we should allow an uncertainty of 16 ppb (much higher offsets are possible for the data derived by CFA, and there we mainly look for similar patterns to those in the reference data). The discrete data measured in Bern are displayed along with the continuous data in Figs. 3 and S1. A number of discrete measurements were also made between 84 and 144 m at Oregon State University which are described in the companion paper (Hoffmann et al., 2022).

The isotopic ratio of oxygen in air provides a good additional constraint because it is well mixed globally and varies in line with precession, providing opportunities for aligning measurements with calculated orbital targets as well as with measurements from other ice cores (Extier et al., 2018). CH4 and δ18Oatm have previously been used powerfully in tandem to untangle disturbed ice chronologies in the LIG (Chappellaz et al., 1997; Yau et al., 2016).

In this work, 27 samples were analysed for δ18Oatm at the Laboratoire des Sciences du Climat et de l'Environnement (LSCE). Two samples were in the depth range 160–170 m, and five samples were between 435 and 471 m. The remaining samples were in the depth range 602–635 m. Data were corrected for firn fractionation and gas loss (Extier et al., 2018) and are shown in Table S1; data below 600 m are shown on a depth scale in Fig. S1. Uncertainty on each value is estimated at ±0.03 ‰. Combining this with the similar uncertainty in data points in the reference dataset suggests that we should allow an uncertainty of 0.04 ‰ when comparing our data with the reference.

The flux/concentration of 10Be in ice shows a pattern related to variations in the magnetic field of the Sun and, on longer timescales, Earth. The pattern of these variations can be matched between ice cores, and with 14C variations in other archives such as tree rings, in order to synchronise records (e.g. Adolphi and Muscheler, 2016). A particularly clear and prominent pattern is seen across the Laschamps event, a weakening of Earth's magnetic field that occurred around 41 ka (e.g. Raisbeck et al., 2017). Because this section of ice is in the last glacial period, its synchronisation in the ice phase should allow for a particularly useful and unambiguous estimate of the offset between ice age and gas age (Δage) in the glacial period.

Seventy samples from between 509 and 520 m depth were spiked with a known amount of 9Be, processed in Lund and analysed for 10Be by accelerator mass spectrometry at ETH Zurich. Measured 10Be / 9Be ratios were normalised to the ETH Zurich in-house standards S2007N and S2010N with nominal 10Be / 9Be ratios of 28.1 × 10-12 and 3.3 × 10-12 (Christl et al., 2013). Data and associated uncertainties are presented in Supplement Table S2.

When synchronising ice cores from different sites, it is important to use only parameters for which there is a sound reason to assume that both cores share synchronous variability. This is the case, for example, with volcanic eruption spikes, with 10Be, and with well-mixed atmospheric gases, such as methane. It is not safe to make such an assumption for water isotopes, which are site-dependent because climatic changes may vary asynchronously in different parts of Antarctica. While methane synchronisation (see above) and a relatively small Δage compared to inland sites (due to the higher accumulation rate) allows us to make a reasonable estimate of the ice age along our core, it would be advantageous to have further ties in the ice phase. It has been argued previously that variations in the components of terrestrial dust (such as Ca) can be assumed to be synchronous across Antarctica (Baggenstos et al., 2018; Mulvaney et al., 2000). This is because their concentrations are strongly controlled by events at a common source in South America and in a common part of the transport pathway towards Antarctica, with only a minor part of the variability likely to be dependent on the final stages of transport to each ice core site.

The main component used for such synchronisation to date has been non-sea-salt calcium (nssCa), calculated using marine and terrestrial ratios of Ca and Na, e.g. Röthlisberger et al., 2002). However, after an initial attempt, we observed that while nssCa at Skytrain Ice Rise shows a good coherence with that of other sites (EDC, EDML) until a depth of about 500 m (30 ka), it diverges below that. Other terrestrial markers such as Al and nssMg (calculated as Mg - 0.12 ⋅ Na and both measured by ICP-MS during the CFA campaign, Grieman et al., 2021) do not mirror the Skytrain nssCa signal and do appear to follow nssCa at other East Antarctic sites (see Sect. 5.3). It appears that an additional source of Ca-rich material, not seen in other Antarctic cores and presumably due to local sources, is present at this site in the earlier part of the last glacial. The reasons for this will be explored elsewhere. However, the solution for us is to use the terrestrial markers that appear free from this extra source but that are coherent with nssCa records at other sites. The limits of detection of Al and Mg are 3.3 and 1.3 ppb, respectively. We concentrate on alignments from nssMg because a majority of Al values in the Holocene and Marine Isotope Stage 5 fall below the detection limit; in the glacial the Al values support our conclusions with nssMg.

In order to use the more detailed variability that can be traced during the Holocene, we compared our methane data to the high-resolution Roosevelt Island methane record between 2–7 ka. Between 7–68 ka we used the WAIS Divide record (Buizert et al., 2015; Rhodes et al., 2017). Between 68 and 156 ka, we used the Southern Hemisphere methane spline generated from the EDC ice core (Köhler et al., 2017). To investigate possible matches with older ice we used the EDC data themselves (Loulergue et al., 2008). As previously explained, the Roosevelt and WAIS Divide data are on the WD2014 age scale, but we eventually used a conversion table (based on Buizert et al., 2018) to place all matches onto a common AICC2012 age scale.

A composite EDC-Vostok record of δ18Oatm (Extier et al., 2018) was used for comparison to Skytrain ice core δ18Oatm.

The clear pattern of the 10Be record across the Laschamps event has been shown to be closely replicated at several sites in Greenland and Antarctica (Raisbeck et al., 2017). For the synchronisation, we used the normalised stack that was recently created based on three Greenland and three Antarctic records (Adolphi et al., 2018).

As the reference dataset for terrestrial deposition we used the nssCa record from EDML (Fischer et al., 2007), because of its greater proximity to Skytrain in the Atlantic sector of Antarctica, with further validation using the record from EDC (Wolff et al., 2010).

First, we note that the discrete methane data (Fig. 3) confirm that the methane concentrations in the section from 534–545 m are much too high. In this section of ice we therefore use the values from the discrete data to match with reference data.

In Table S3, we list the methane tie points that we used in this section. The very clear match between our record and the reference data is ideally seen in the past 15 kyr (460 m), where there are few spikes in the methane record due to air ingress into cracks (Fig. 4). However, the pattern of Dansgaard–Oeschger events remains clear right down to 100 ka and is shown in Fig. 5, along with the tie points used. We note that the comparisons in Fig. 4 suggest that the Skytrain data might be up to 10 % too low in concentration (but with a variable offset along the core) compared to the reference data; this results from the dissolution of gas in the melt stream (as discussed in Sect. 3.1) and the difficulty of accurately calibrating data from the continuous melter due to the absence of an external certified standard. In Fig. 5 we show the full methane record on the eventual age scale, compared to reference data. It is clear that some spikes due to air ingress across cracks remain in the dataset beyond about 60 ka, but the pattern for matching is still apparent to at least 100 ka. The match between Skytrain and reference methane between 80 and 100 ka is less secure than it is in shallower ice, because ice with high concentration outliers and/or missing data is common as a result of extensive cracking. This makes it hard to match absolute values of methane and forces us to rely on the pattern with depth. Nonetheless, the methane ties we have made result in a good match in this part of the core between nssMg and reference nssCa (Fig. 7 and Sect. 5.3), supporting our choices. The section beyond 100 ka will be discussed in Sect. 6.

In Fig. 6 we show the Skytrain 10Be concentration from 509–520 m, aligned with the reference dataset. The common shape across the wider event as well as the presence of individual peaks and troughs is clear. We chose five tie points in the range 39.9–42.0 ka (Table S4).

Skytrain nssMg was compared to nssCa from EDML (Fischer et al., 2007) (Fig. 7). The two records show strong similarities, as does Skytrain Al (not shown) where it exceeds the detection limit; comparison with EDC nssCa (Wolff et al., 2010) shows a comparably good match. We chose a few obvious tie points (Table S4) concentrating on regions with clear variability and trying to fill the gaps where fewer ice tie points existed. We discuss the ice below 100 ka in Sect. 6.

Flow disturbances affecting LIG ice have been seen previously, though until now this has been observed mainly in Greenland. To confirm the age of ice with difficult stratigraphy, and even to re-order disordered layers, previous authors have used a combination of methane and δ18Oatm (Chappellaz et al., 1997; NEEM Community Members, 2013; Yau et al., 2016). Provided data are sufficiently precise, the two-dimensional field of these parameters can define an age for a given layer that is close to unique within the plausible range. In Fig. 8a we show the reference data for CH4 and δ18Oatm.

By plotting the two-dimensional distribution of values (Fig. 8b) one can see how the data clearly differentiate samples of different ages – this is particularly true in the section from about 120–140 ka (section that goes clockwise in increasing age coloured yellow). While the δ18Oatm data were used mainly in combination with CH4 to assess the ages of ice around the LIG, δ18Oatm was also measured in two Skytrain ice core samples from the Holocene and two from the Last Glacial Maximum: these were not used to construct the age scale, but the values were entirely consistent with the modelled ages. Three samples were also measured between 435 and 456 m. These three values of δ18Oatm, along with the less precise CH4 data obtained from the continuous measurements, were used to assign ages (Table S1) more precisely between 11 and 15 ka in a section in which δ18Oatm is increasing rapidly with age (Fig. 8).

Twenty Skytrain ice core samples were analysed for δ18Oatm between 600 and 635 m depth, covering the period that the discussion above would lead us to expect is older than 100 ka. In all but two cases discrete methane measurements were made (in Bern) on an adjacent sample (a few centimetres away from the δ18Oatm sample).

We now examine the data at depths for which we have both δ18Oatm and methane measurements. We start with the data from 603–618 m (Fig. 9a). The data point at 603.1 m can be assigned an age of ∼ 106 ka, as we had already deduced above from the shape and amplitude of the methane peak alone. While the point at 606.4 m matches best with ∼ 118 ka, the three data points deeper than that (609–618 m) are only compatible with younger ages, between 106 and 117 ka. We cannot untangle this section, but there is apparently some degree of disturbance at least between 605 and 615 m.

We have no reason to doubt that the ice is in good order until 605 m, but we acknowledge that the section we date as 95–107 ka (Figs. 5, 7) relies on the pattern of methane and on a single CH4/δ18Oatm data point. This point, dated at 106 ka, firmly defines the lower end of this section, with values that do not occur again as a pair until 57 ka.

Turning now to Fig. 9b, the data from 615.3 to 627.3 m plot in chronological sequence with respect to the reference data between about 110–126 ka. Most of these points are not consistent with δ18Oatm and methane values at any other ages in the range 60–180 ka. Crucially the two data points at 623.2 and 624.7 m with very negative δ18Oatm and CH4>600 ppb are not compatible with any other age in the past 200 kyr other than the LIG at around 122 ka and a short period in the Holocene at 7 ka. These data points are also incompatible with any mixtures of ice from other depths. Because the data point at 615.3 m is compatible with a range of ages, we choose a conservative range of depths from 617 m (just above the clear match at 618.3 m) to 628 m where we are very confident that we have a sequence of ice from the last interglacial, covering the period 126 to 117 ka. Although it lies within the uncertainty of the values at 627.3 m, the data point at 630.3 m (shown in red) is also only consistent with the last interglacial but does not show the expected increase in age with depth and could show a reversal in age. As this is already in the section that appears disturbed in methane and δ18Oice, we consider this data point and the ice around it as subject to disturbance.

Finally, we examine the data from 630 to 635 m (Fig. 10). The point at 631.6, sitting close to clearly disturbed ice with rapidly changing values of CH4 and δ18Oice, has values not seen in the reference data and is probably a mixture of interglacial and glacial ice. The other data have values consistent with ages that would occur in the middle of MIS6 (140–180 ka) or alternatively could originate from ice that is much older (from an earlier glacial cycle). Because there are a number of age solutions within the uncertainty of the measurements, we do not attempt to assign ages to these data points.

Combining the observation that no ice has methane values that fit in the age ranges 127–129 or ∼ 140 ka and the positive identification of ice with unique combinations of CH4 and δ18Oatm, we conclude the following: a.

there is probably a flow disturbance at the top of the last interglacial section, with ice from ∼ 106–117 ka repeated;