Chronostratigraphy of blue ice at the Larsen Glacier in Northern Victoria Land, East Antarctica

Blue ice areas (BIAs) allow for the collection of large-sized old ice samples in a cost-effective way because deep ice outcrops and make old ice samples available close to the surface. However, most chronostratigraphy studies on blue ice are complicated due to fold and fault structures. Here, we report a simple stratigraphy of ice from the Larsen BIA, Antarctica, making the area valuable for paleoclimate studies. Ice layers defined by dust bands and ground penetration radar (GPR) surveys indicate a monotonic increase in age along the ice flow direction on the downstream side, while the upstream ice exhibits a 20 potential repetition of ages on scales of tens of meters, as shown in the complicated fold structure. Stable water isotopes (δOice and δHice) and components of the occluded air (i.e., CO2, N2O, CH4, δN-N2, δOatm (= δO-O2), δO2/N2, δAr/N2, 81Kr and 85Kr) were analyzed for surface ice and shallow ice core samples. Correlating δOice, δOatm, and CH4 records of Larsen ice with existing ice core records indicates that the gas age at shallow coring sites ranges between 9.2–23.4 ka BP and ice age for entire surface sampling sites between 5.6–24.7 ka BP. Absolute radiometric 81Kr dating for the two cores confirms the ages 25 within acceptable levels of analytical uncertainty. Our study demonstrates that BIA in northern Victoria Land may help researchers obtain high-quality records for paleoclimate and atmospheric greenhouse gas compositions through the last deglaciation.

during the last glacial period (Dansgaard et al., 1989;Steffensen et al., 2008), and bipolar seesaw climate links on millennium timescales (Blunier and Brook, 2001;Landais et al., 2015). Thus far, continuous climate records from ice cores cover the last 800 ka and may reach more than 1 Ma in the near future by new deep drilling projects in Antarctica (Fischer et al., 2013). 35 However, the use of ice cores obtained from conventional deep ice core drilling projects remains limited as many analyses require large amounts of ice, such as trace element isotope and trace gas analyses. In addition, deep drilling projects incur high economic and labour costs. In contrast, coring in blue ice areas (BIAs) has emerged as an alternative to obtain large ice samples in a cost-effective manner (Folco et al., 2006;Petrenko et al., 2006;Schaefer et al., 2006;Sinisalo et al., 2007;Korotkikh et al., 2011;Turney et al., 2013;Bauska et al., 2016;Aarons et al., 2017;Baggenstos et al., 2017;Yan et al., 2019;Fogwill et al., 40 2020).
In BIAs, old ice is exposed on the surface because the bedrock or basal topographic obstacles cause deep glacial flow upward, and surface snow is ablated by katabatic wind and/or sublimation (Bintanja, 1999;Sinisalo and Moore, 2010). Because ice layers of the same age (isochrones) are extended on the surface, we can obtain large amounts of old ice for a specific age at the surface and/or relatively shallow depths, allowing researchers to study paleoclimate which is typically prohibited by limited 45 sample sizes Bauska et al., 2016;Fogwill et al., 2020).
In the early stage of BIA research, meteorites were the focus of investigation because ice flow and ablation cause meteorites to accumulate on the surface of the blue ice (Whillans and Cassidy, 1983;Cassidy et al., 1992, Harvey, 2003. Recently, studies on BIAs have drawn attention to the possibility of identifying ice older than 800 ka because glacial-interglacial cycles changed from 40 to 100 kyr during the 0.8-1.0 Ma period, which is called the mid-Pleistocene Transition (MPT). It has been reported 50 that ice at Allan Hills BIA has ages of 90-250 ka on the surface (Spaulding et al., 2013), reaching ~2.7 Ma near the bedrock at depths of about 150-200 m (Yan et al., 2019). However, the use of blue ice has several drawbacks. In most cases, the stratigraphy of the blue ice is complicated, as shown by the fold and fault structures on the surface (Folco et al., 2006;Petrenko et al., 2006;Curzio et al., 2008;Schaefer et al., 2009;Baggenstos et al., 2017) and stratigraphy is discontinuous at deep depths et al., 2017). Variations in continental ice mass are known to be the main factor that controls δ 18 Oatm during glacial-interglacial cycles (Bender et al., 1985;Sowers et al., 1993). In addition, variations in the Inter Tropical Convergence Zone (ITCZ) were also considered controlling δ 18 Oatm on millennial and orbital timescales Landais et al., 2010;Seltzer et al., 2017;Extier et al., 2018b). Because of the long lifetime of oxygen gas in the atmosphere (~1 ka), the δ 18 Oatm varies more 70 gradually, limiting synchronization to millennial timescales. In contrast, atmospheric CH4 concentration changes rapidly because of the short lifetime of ~12 years, allowing precise dating via stratigraphic matching. Oeschger (1987) suggested the potential of 81 Kr measurements in ice core dating. However, at that time, 10 5 -10 6 kg of ice was required. In virtue of developing Atom Trace Trap Analysis (ATTA), the required ice has kept decreasing (Lu et al., 2014;Tian et al 2019;Jiang et al., 2020;Crotti et al., 2021). Buizert et al. (2014) for the first time, showed that 81 Kr age dating is feasible for blue ice in Taylor Glacier, 75 Antarctica.
BIAs cover 1.67 % of the area of the Antarctic continent and are concentrated in Victoria Land, the Transantarctic Mountains, Dronning Maud Land and the Lambert Glacier basin (Hui et al., 2014) (Fig. 1). Several chronological studies of BIAs in Antarctica have been conducted in Southern Victoria Land: Taylor Glacier (Aciego et al., 2007;Buizert et al., 2014;Baggenstos et al., 2017;Baggenstos et al., 2018;Menking et al., 2019), Allan Hills (Spaulding et al., 2013;Higgins et al., 80 2015;Yan et al., 2019), Mullins Glacier (Yau et al., 2015); and Northern Victoria Land: Frontier Glacier (Folco et al., 2006;Curzio et al., 2008;Welten et al., 2008). Studies on the Taylor Glacier and Allan Hills BIA have constrained the age of the blue ice along several transects and cores. However, chronologies of the other BIAs in Victoria Land remain insufficient for high-resolution paleoclimate studies. In addition, important paleoclimate proxies such as stable water isotopes, greenhouse gas concentrations, and isotopic ratio of the oxygen gas are not previously addressed for the area. 85  (Hui et al., 2014). Orange dots represent the BIAs where the chronology has been studied. Deep ice core locations with brown dots are labeled as follows: Talos Dome ice core (TALDICE), EPICA Dome C (EDC), WAIS Divide (WD) and Vostok. The Antarctic map was obtained from the QGIS Quantarctica package.
Our study focused on the chronostratigraphy of ice in Larsen BIA, Antarctica, which may facilitate future research in the 90 region. We describe the ice flow and structure of the ice body using dust bands and ground penetration radar (GPR) surveys, and constrained the unknown gas and ice ages by correlating δ 18 Oatm, CH4, and δ 18 Oice with existing ice core records. We also independently confirmed the ages using the radiometric 81 Kr dating method.

Larsen blue ice area (BIA) 95
We sampled ice from Larsen BIA, an outlet glacier in the Northern Victoria Land, East Antarctica, in austral summer of 2018/2019. The Larsen BIA is located approximately 85 km southwest of the Korean Jang Bo Go station (Fig. 2). The mean annual temperature is −27.2 ℃ (data and information were obtained from the Meteo-Climatological Observatory at MZS and Victoria Land of PNRA-www.climantartide.it), cold enough to prohibit ice melting in the summer. There were ~20 cm wide dust bands with gentle folding structures in the mid-to downstream part, while we observed severely folded dust bands (e.g. 100 S-and Z-folds) in the upstream part (Fig. 2b). Dust bands are frequently observed in BIAs in Victoria Land and can be used as isochrons (Sinisalo and Moore, 2010). To obtain ice samples with simple stratigraphy, we avoided ice coring in the area with complicated fold structures and sampled ice in the direction of ice flow identified by the Antarctic ice velocity map in QGIS Quantarctica package (Rignot et al., 2011;Mouginot et al., 2012). However, the stratigraphy of the upstream ice, where we collected surface ice may be inverted and repeated on a scale of tens of meters. Shallow ice cores were drilled along a 1 105 km long transect at intervals of 20-30 m (Fig. 2b). Most ice cores had lengths of approximately 2 m, but ice core #23 (74.9319° S, 161.6018° E) was ~10.4 m (Fig. 2c). We also collected near-surface ice samples (~500 g) along a 1.3 km transect with 20 m intervals to measure stable water isotopes of the ice (δ 18 Oice, δ 2 Hice) at depths of ~5-10 cm (hereafter, regarded as the surface ice sample). The ice core TF (74.93042° S, 161.56975° E) with a length of ~12 m on the upstream side was sampled in 2016 for a preliminary study, as described by Jang et al. (2017). To define the horizontal distance, we used an imaginary line parallel 110 to the ice flow direction. A line perpendicular to the imaginary line that crosses each sampling location was used to find the intersection point. Then, each intersection point was used to measure the horizontal distance from the most upstream sampling site (Fig. S1).

Ground penetrating radar (GPR) survey
Approximately 17 km of GPR data were collected for two days in January 2019 (Fig. S1). A MALÅ ProEx impulse radar system with a 50 MHz unshielded antenna for a larger penetration depth was used for data acquisition. Records were taken at a sampling frequency of 559.5 MHz, time interval 0.1 s and stacked 4 times. The survey was conducted at a speed of 2.5-3 km h −1 to minimize noise caused by frictional vibrations between the antenna and the surface of the glacier. The position of 125 traces was recorded using a single-frequency code-phase GPS with an accuracy of < 3 m. Data processing was performed in ReflexW v.9.5 in the order of dewow, DC filter, band-pass filter, time-zero drift, energy decay correction, background removal, static correction, migration, and stack. As the constant radar signal velocity "v" in the glacier was 0.17 m ns −1 (Borgorodsky, 1985;Reynolds, 1985), and the frequency "f" of the radar system was 50 MHz, the GPR wavelength "λ" was about 3.4 m (λ = v f −1 ). 130 https://doi.org/10.5194/tc-2021-294 Preprint. Discussion started: 5 October 2021 c Author(s) 2021. CC BY 4.0 License.

Stable water isotope measurement
Stable water isotopes (δ 18 Oice, δ 2 Hice) were measured simultaneously at the Korea Polar Research Institute (KOPRI) using the Cavity Ring-Down Spectroscopy (CRDS) method using a Picarro L2130-i with a vaporizer (A0211). Surface ice samples (~5-10 cm depth) and ice core samples (10-30 cm and 190-200 cm depths of ice cores) were used for the analyses. The average horizontal interval of the surface ice samples and ice cores was approximately 10.3 m. Ice from the ~10 m long core 135 #23 was used for measurements at 20 cm intervals. Ice samples were kept in Whirl-Pak and melted at room temperature. The melted sample was then injected into a 2 ml vial via a disposable syringe with a 0.45 μm filter. One batch of measurements consisted of five duplicate samples (10 samples in total) and a working standard. The first and second aliquots of the sample were measured 12 times each, and the working standard was measured 20 times. To remove the memory effects from the previous samples, the last six measurements were used. The precision was evaluated by measuring the working standard 140 repeatedly (n = 67); 1 sigma (standard deviation) was 0.07 ‰ for δ 18 Oice and 0.90 ‰ for δ 2 Hice. The working standards and samples were calibrated against the international standards for water isotopes, VSMOW2 (Vienna Standard Mean Ocean Water 2), SLAP2 (Standard Light Antarctic Precipitation 2) and GISP (Greenland Ice Sheet Precipitation).

Greenhouse gas (GHG) measurement
The CH4 concentrations in ice cores (#306, #23, #120, and #201) were analyzed at Seoul National University (SNU) using 145 a melt-refreeze technique with a flame ionization detector gas chromatograph (FID-GC) at 10 cm intervals. The process of CH4 measurement was described in detail by Yang (2019). Briefly, we trimmed the outermost side and the cracks of the ice samples by approximately 2 mm with a clean band saw to eliminate contamination by ambient air. The mass of the ice used for measurement was about 35-55 g. Then, we placed the ice inside a custom-made flask and evacuated the flask. We used a 740.6 ppb standard air sample from the National Oceanic and Atmospheric Administration (NOAA) Global Monitoring 150 Division (GMD) on the WMO X2004A scale to establish a daily calibration line (Dlugokencky et al., 2005). The NOAA standard air was injected into four flasks containing bubble-free ice samples, which served as the control group. The CH4 mixing ratio of the control group was measured, and the average offset to the standard value of 740.6 ppb was assumed to be a systematic error; thus, the average offset (~2-20 ppb) was subtracted from each measurement result. Trapped gas in the ice was liberated by immersing the flask in a hot water, and then refreezed the melt water. In addition, as CH4 is more soluble than 155 the major air components (N2, O2, and Ar), the CH4 concentration of the 1 st extracted air is lower than the original value.
Therefore, 2 nd gas extraction was conducted to correct the solubility effect.
The same ice cores (#306, #23, #120, and #201) were used for CO2 measurements. Generally, CO2 should be measured using a dry extraction system rather than wet extraction because of its high solubility in water and the possibility of CO2 production by carbonate-acid reactions in ice melt (Delmas et al., 1980). Therefore, for our measurement, air was collected 160 using a needle-crusher dry extraction instrument and detected by FID-GC at SNU. The ice was pre-treated in the same manner as that for CH4, as described above, while the required amount of ice was 15-20 g. We placed the ice in the needle crush chamber, which was cooled to −35 ℃ during ice preparation. The CO2 results could be affected by adsorption and desorption by the instrument and the sample tube (Ahn et al., 2009). Therefore, before crushing the ice, NOAA standard air was released through the chamber and collected into the sample tube. The CO2 mixing ratio was measured, and the average offset with 165 respect to NOAA standard air was used to subtract the value of the CO2 result from each ice sample measurement. The CO2 mixing ratio of the standard air we used for making a daily calibration line was 285.66 ppm or 293.32 ppm, which is from NOAA GMD in WMO X2019 calibration scale (Hall et al., 2021). Trapped air in the ice was collected in a sample tube cooled by a helium closed cycle refrigerator (He-CCR) after crossing the water trap. The entire process and principle are described in detail by Shin (2014). 170 We also used ice samples from 35 cores at depths of 190-200 cm (hereafter, regarded as horizontal measurement) and ice core #23 to measure CH4, CO2, and N2O concentrations, along with other gas isotopes (see below), by a wet extraction method at the National Institute of Polar Research (NIPR) in Japan. For CO2 and CH4, FID-GC was used, and for N2O, an electron capture detector (ECD) GC was used. To determine the concentrations of CH4 and CO2, a calibration line was established using TU-2008 (Tohoku University) scale standard air; for N2O, the TU-2006 scale standard air was used. More information 175 on the standard air we used (named STD 1, 3, 5) is given by Oyabu et al. (2020). Three standard air and a quadratic calibration line were used to determine GHG concentration in the NIPR, while single standard air and a linear calibration line were used to determine the GHG concentration in SNU. Different calibration scales of standard air (TU and NOAA/WMO) and different calibration methods may contribute to making an inter-laboratory offset of GHG concentrations. CO2 was measured again at SNU by the dry extraction method at 190-195 cm depth for the horizontal measurement (Table S4). The ice preparation process 180 is briefly explained in Sect. 2.5.
The isotope results are valued with respect to the modern atmosphere. As mentioned previously, we measured 35 ice cores at depths of 190-200 cm to evaluate how ancient air compositions change horizontally and also measured gas isotopes using ice 185 core #23 at several depths with a range of 0-10 m to compare with the results from the neighbouring cores at depths of 190-200 cm. Because gas loss fractionation by molecular diffusion during storage could have affected the isotope results (Ikeda-Fukazawa et al., 2005), we trimmed the surface and the cracks of the ice approximately 3-5 mm and then removed some unclear ice surface with a ceramic knife. The weight of the ice was 55-80 g after trimming the surfaces. The ice inside the vessel was evacuated for ~120 min. To extract the gas, each vessel was gradually immersed in a hot water container. 190 Simultaneously, the released air was collected into a sample tube, which was cooled by a He-CCR. After homogenizing the sample tube for a night, the gas was split into two aliquots, one for isotope analysis (1.5 ml) and the other one for greenhouse gas measurement (5 ml). The measurement process is described in more detail by Oyabu et al. (2020).
Obtaining a true δ 18 O value in the past atmosphere from ice cores requires gravitational, thermal, and gas loss fractionation corrections. The gravitational factor is proportional to the difference in the mass number between isotopes (Craig et al., 1988;Severinghaus et al.,1998). Therefore, δ 18 O ( 18 O/ 16 O) was affected twice as much as δ 15 N ( 15 N/ 14 N) by gravity. Hence, each gas isotope was gravity-corrected using Eq. (1): In the same principle, δO2/N2 is affected by a factor of 4 than δ 15 N, thus it was corrected using Eq.
(2) and used for assessing the gas loss fractionation: 200 Along with the gravitational correction, thermal fractionation should also be considered because temperature gradients in the firn column affect the distribution of the isotopes (Severinghaus et al., 1998;Goujon et al., 2003). Thermal fractionation is typically small in Antarctic ice cores due to the relatively gradual nature of surface climate change. Comparing δ 40 Ar/4 ( 40 Ar/ 36 Ar) and δ 15 N together allow discrimination of the contribution of thermal and gravity fractionation (Severinghaus et 205 al., 1998). However, for our samples, thermal fractionation could not be considered because 36 Ar interfered with the 18 O2.
δO2/N2,gravcorr of around −30 ‰ indicates that the ice is poorly preserved, and has experienced considerable gas loss either during drilling or storage (Landais et al., 2003). Following Capron et al. (2010), we did not correct for gas loss fractionation in our samples because the δO2/N2,gravcorr values were significantly greater than −30 ‰, except for one measurement for sample #301 (Table S1, S2). We did not use the results of #301 when constraining the gas age using δ 18 Oatm. 210

81 Kr dating
For the 81 Kr measurement of ice core #23 and TF, 5.3 kg (depth: 711-1040 cm) and 5.4 kg (depth: 798-1192.5 cm) of ice were used, respectively. Air for the 81 Kr measurement was extracted at SNU by the instrument provided by the University of Science and Technology of China (USTC). The ice was kept in a tank with an O-ring lid and evacuated using a dry scroll pump with a water trap. Then, the ice was melted by immersing the tank in hot water. The released gas was collected in containers 215 and shipped to USTC for Kr purification and 81 Kr analysis using Atom Trap Trace Analysis (ATTA). The extraction procedure was described in detail by Tian et al. (2019), and the principle of ATTA was described by Jiang et al. (2012).
The anthropogenic 85 Kr is measured simultaneously with 81 Kr to quantify any contamination with modern air. The ice sample details and krypton dating results are reported in Table 1. For both samples, the measured 85 Kr activity is below the detection limit, so no correction for contamination with modern air is necessary. For the calculation of the 81 Kr-ages, the changes in the 220 past atmospheric 81 Kr abundance due to variation of the cosmic ray flux on the Earth (Zappala et al., 2020) are taken into account. The age uncertainty calculation is based on the statistical error of the atom counting.

Development of WD2014 timescale for TALDICE
For the TALDICE ice core we use a timescale that is synchronized to the WAIS Divide (WD) WD2014 chronology (Buizert et al., 2015;Sigl et al., 2016) in the following way. The ice age scales are synchronized using volcanic deposits identified in sulfur/sulfate data from the WD and TALDICE cores (Buizert et al., 2018;Severi et al., 2012). Next, the TALDICE gas-age 230 ice-age difference (∆age) is established empirically by matching abrupt changes in atmospheric CH4 to the WD core; at each match point this provides one discrete ∆age constraint. A dynamical firn densification model based on Herron-Langway firn physics (Herron & Langway, 1980) is then used to interpolate between these empirical ∆age constraints, in order to obtain a gas age scale for all depths (Buizert et al., 2021).

Ground penetrating radar (GPR) survey
In the GPR survey, we identified ice layers (or isochrones) in the transect parallel to the ice flow direction (Fig. 3). The dips of the ice layers range from 1° to 6° with a decreasing trend from the upstream to the downstream direction. The ice layers of radargram were not clearly visible at a depth of < 10 m because of the direct wave signal. We did not observe any stratigraphic folding structure in the ice layers that made age inversion along the ice flow direction in the mid-to downstream areas. 240 Therefore, we expect monotonic and continuous age changes along the ice flow direction. However, as shown in the dust bands with S-and Z-folds in the upstream area (Fig. 2b), the upstream stratigraphy might be repeated on a scale of tens of meters.
The basal topography is well defined from the GPR data, and we observed an ice thickness variation of 200-400 m (Fig. 3b).
The subsurface ice layer in the upstream area (0-800 m from the most upstream side) was not well recognized from the GPR profile (Fig. 3c). It is possible that noise caused by crevasses, cavities, or cracks could obscure the signals. In addition, 245 accurate data acquisition might have been hindered by antenna tremors or low battery power at a severely cold temperature.

Stable water isotopes
Stable water isotopes in ice record surface temperatures in the past at the snow deposition site (Jouzel et al., 1997). The δ 18 Oice and δ 2 Hice records of Larsen BIA are presented in Table S5, S6, and S7. The horizontal δ 2 Hice result has a distinct local minimum around a horizontal distance of 600-800 m, indicating a transient cold event (Fig. 4b). The δ 2 Hice of the transient 255 cold event plunges by approximately 70 ‰, heading to the upstream ice. The most downstream sample, ice core #200, has a δ 2 Hice of −369 ‰. Conversely, the most upstream sample, surface ice #81W, has a δ 2 Hice of −245 ‰. Ice core #23 was located in the middle of the transect, and the vertical δ 2 Hice profile shows a range of −353 ‰ to −291 ‰. It appears that the water isotope values from Larsen BIA are more scattered with a wider range than other published ice core records. The highly variable δ 18 Oice has also been reported in the Taylor Glacier (Baggenstos et al., 2018;Menking et al., 2019). Severe scattering 260 might indicate that the accumulation zone (original source of ice) of Larsen BIA might have experienced more variability in temperature and/or the vapour source (variability of atmospheric conditions) than other sites.
To match the two δ 2 Hice profiles (measurement of ice core #23 and horizontal measurement of near-surface ice), the depth of ice core #23 was converted to the horizontal distances by pinpointing the deepest result of #23 to the result of ice core #104 and conducting linear interpolation. The two δ 2 Hice profiles have r 2 value of 0.85 (p < 0.001) (Fig. 4b). The similarity of δ 2 Hice 265 between the vertical and horizontal measurements demonstrate that stable water isotopes were not altered while the glacier flowed to the surface and also showed that ice stratigraphy was not disturbed. As the ~10 m long #23 core covers ages that correspond to a surface distance of ~117 m, we calculated the average dip of the ice layer of 4.96°, which is comparable with the average dip derived from the GPR profile (Fig. 3d). This depth/distance relationship is also supported by the comparison of gas isotope values (δ 18 Oatm and δ 15 N-N2) from ice core #23 to the records from the horizontal measurement at a depth of 270 1.95 m (Fig. A2).

CH4 and CO2 mixing ratios
The measurement results of CH4 and CO2 concentrations for the vertical cores are presented in Fig. A1. The shallow ice cores (#306, #120, #201) show that greenhouse gases are significantly altered for the top 2 m. CH4 records of #23 also fluctuate significantly at 0-4.6 m depth but settle down at > 4.6 m. CO2 records of #23, in contrast, gradually decrease and become 280 steady at a depth of > 4.6 m. A comparison of the results from NIPR to SNU with ice core #23 shows that the difference in the concentration decreases significantly at depths of > 4 m; huge difference (30-140 ppb) of CH4 at 0-4 m depths, but decreased considerably (5-10 ppb) at a depth of > 4 m; CO2 difference was 10-20 ppm at < 4 m, but decreased to 2-10 ppm at a depth of > 4 m (Table S2 and S3). Altered CH4 proximity to the surface is also by reported by Petrenko et al. (2006) and Baggenstos et al. (2017). However, the CO2 mixing ratio of the TF core, reported by Jang et al. (2017), shows that CO2 is scattered even 285 in the deeper part of the ice core. We speculate that this scattering of CO2 can be due to complicated ice stratigraphy because we observed many folding structures identified by dust bands near the TF core site (Fig. 2b).

δ 15 N-N2 and δ 18 Oatm
As N2 and O2 are the dominant gas species in atmospheric air, we assumed that the δ 15 N-N2 and δ 18 Oatm were not significantly altered at a depth of 1.95 m. We tested the assumption by comparing the vertical distribution of the ice core #23 with horizontal 290 distribution at depths of 1.95 m in nearby shallow cores. The depths of ice core #23 were converted to horizontal distances at a depth of 1.95 m (Fig. A2) using Eq. (3): Horizontal distance in meter (at 1.95 m depth) = {|(depth of ice core #23) − 1.95|/tan(4.96°)} + 663 (3) We added 663 m because ice core #23 was located 663 m from the most upstream sampling site. The average dip of the ice layer (4.96°) was calculated by matching the δ 2 Hice of ice core #23 with the horizontal records from the neighbouring cores. 295 The δ 15 N-N2 and δ 18 Oatm values obtained from the horizontal record at 1.95 m are comparable to those from ice core #23 except those from the very shallow depths of < 0.5 m (Fig. A2), confirming that the gas isotope ratios are generally reliable at a depth of 1.95 m. We estimated the uncertainty of δ 15 N-N2 and δ 18 Oatm values as ±0.05 ‰ because the offsets between #23 and the horizontal record are approximately 0.05 ‰. Petrenko et al. (2006) also reported that as long as the ice is not affected by surface ice melting, nitrogen and oxygen gas isotopes are not altered even at a depth of 0.3 to 0.4 m at the western Greenland 300 ice. No significant gaps, discontinuities, or anomalies were found within the results from horizontal measurements of δ 18 Oatm, which indicate no significant stratigraphic disturbance (Fig. 4a).

Glacial termination identification in Larsen ice
The increase in δ 2 Hice and decrease in δ 18 Oatm from the downstream to upstream ice shows that the atmospheric conditions changed to a warmer climate (Fig. 4). In particular, the δ 18 Oatm values of the Larsen ice (1.126 ‰ to −0.075 ‰) reveal a typical 305 glacial termination period. The maximum δ 18 Oatm value of > 1 ‰ implies that the ice is younger than 800 ka because the glacial climate conditions became more extreme after the MPT (Lisiecki and Raymo, 2005;Elderfield et al., 2012;Chalk et al., 2017); and it is likely that the maximum δ 18 Oatm values were < 1.0 ‰ during the pre-MPT glacial periods (Yan et al., 2019). The EPICA Dome C (EDC) record shows that Termination Ⅰ, Ⅱ, Ⅳ, Ⅴ and Ⅶ are the only terminations that have both values of negative and > 1.0 ‰ during the last 800 ka (Landais et al., 2013;Extier et al., 2018a) (Fig. 5). Hence, we exclude Termination 310 Ⅲ and Ⅵ for the Larsen BIA ages. Among the candidates, Termination Ⅴ should also be excluded because the maximum δ 18 Oatm value in the EDC record is ~1.4 ‰, which is significantly higher than that of the Larsen ice. The δ 2 Hice decrease in the middle of the glacial termination in the Larsen BIA (Fig. 4b) is similar to that during the Antarctic Cold Reversal (ACR, 14.6-12.7 ka BP), which is a distinct feature only during Termination Ⅰ among the candidate terminations (i.e., Termination I, II, IV and VII) as we see in the EDC δ 2 Hice record (Fig. 6). To confirm the age of the Larsen ice, we compared the δ 18 Oatm-CH4 315 relationship with the ice from core #23 at depths of 4.6-10 m with the EDC and WAIS Divide (Fig. 7). We found that the Larsen δ 18 Oatm-CH4 distribution was well matched only at Termination Ⅰ. It appears that δ 18 Oatm of Larsen has a 0.05 ‰ offset with that of the WAIS Divide. However, this offset is not odd, as we can also see a ~0.05 ‰ offset between core #23 and horizontal measurements at a depth of > 1.95 m; the offset may also come from age difference (Fig. A2). The δ 18 Oatm-CO2 and CO2-CH4 relationships do not clearly match with those of Termination Ⅰ in the EDC and WAIS Divide records. The Larsen 320 ice showed a higher CO2 concentration of 10-20 ppm. Probably, it is altered naturally and/or contaminated, even at depths of 4.6-10 m (Fig. A3, Fig. A4). Finally, we confirmed the ages of the Larsen ice with 81 Kr dating, indicating 9-41 and 14-43 ka for ice from the TF and #23 cores, respectively (Table 1), and conclude that Larsen ice covers the Last Glacial Termination (LGT, T1).    Extier et al. (2018a). CH4 record of EDC is from Bazin et al. (2013d). The area within the black box is magnified to compare the Larsen #23 record with those from WAIS Divide T1 and EDC T1.

Gas and ice ages of Larsen ice
The CH4 was severely altered at a depth of 1.95 m and the available CH4 record was only from ice core #23 (> 4.6 m). Therefore, we developed the tentative gas age by correlating the horizontal δ 18 Oatm value to the WAIS Divide record on the 340 WD2014 timescale. Then, fine correlation was conducted with CH4 record from the core #23, which corresponds to 13.3-14.1 ka at depths of 4.6-10.4 m (Fig. 8c). We take advantage of the high-resolution records of CH4 from both Larsen core #23 and WAIS Divide ice, which greatly improves the precision of age construction. The following paragraph describes this process in detail.
Spline curves were created for both WAIS Divide and Larsen δ 18 Oatm records to reduce artefact from insufficient sampling 345 resolution and/or ice quality at a 1.95 m depth of the Larsen ice (Fig. 8). The spline curve was drawn after interpolating the original records at 5 m horizontal distance intervals. Three points were chosen, which divide the upstream part of the Larsen ice into three equal parts to tie with those in the WAIS Divide record (pink dots in Fig. 8a and 8b). For the downstream side of the Larsen ice, we chose a local maximum and a minimum as the tie points (purple dots in Fig. 8a and 8b) because the slope of the δ 18 Oatm spline curve is small. When the horizontal distance was greater than 1280 m, the age was constrained by 350 extrapolation using the age/distance relationship at 18-22 ka. Based on the tentative horizontal gas ages and the depth/distance relationship of core #23 (obtained from Fig. 4b), we determined the gas age of core #23 (~10 m vertical ice core). A small offset (53.6 ± 38.5 yr) existed between the CH4 record of core #23 (> 4.6 m) and the WAIS Divide record (offset between the red and dark blue lines in Fig. 8d). To eliminate this gap, four tie-points (orange dots) were chosen and interpolated (light blue line). The gas age, corresponding to 13.3-14.1 ka was obtained more accurately by using the CH4 correlation. As a result, the 355 gas age for the Larsen ice at 1.95 m depth was estimated to be 9.2-23.4 ka BP on WAIS Divide chronology in 2014 (WD2014).
The ages on AICC2012 scale were estimated to be 9.4-23.4 ka by using the depth and the WD2014 age of EDC (Bazin et al., 2013a, Buizert et al., 2021.   (Severinghaus et al., 2015). Six tie-points were used to correlate each other. (c) Comparison of synchronized Larsen δ 18 Oatm with EDC (Landais et al., 2013) and WAIS Divide records. (d) Comparison of CH4 record from Larsen #23 (> 4.6 m depth) with WAIS Divide (Rhodes et al., 2017). Tentative gas age determined by δ 18 Oatm correlation with WAIS Divide is tuned by correlating CH4 record using four tie-points. WD2014 timescale for WAIS Divide and EDC is from Sigl et al. (2016) and Buizert et al. (2021), respectively.

365
The ice ages of the Larsen ice were determined based on the δ 18 Oice correlation with the TALDICE record (Fig. 9). TALDICE was chosen because the coring site is close to the Larsen BIA (~240 km apart) and the direction of the TALDICE site from the Larsen BIA is similar to the ice flow direction of the Larsen ice (Rignot et al., 2011;Mouginot et al., 2012). It is likely that the trend of surface temperature changes in the snow accumulation zone of Larsen BIA is comparable to that of the TALDICE site. We constructed spline curves of the horizontal measurement of the surface ice. As described above, we interpolated the 370 original δ 18 Oice records at 5 m interval in order to avoid bias in the δ 18 Oice record. The correlation was established by visually choosing a similar inflection point. Then, in order to validate the ice age, we estimated the ice age at 1.95 m depth (Appendix C for more information) and calculated the Δage; Δage for Larsen ice is defined as (ice age) − (gas age) at a depth of 1.95 m.
It should be noted that the ice is older than the gas at the corresponding depth because the gas is isolated when the firn completely transforms into ice (Schwander and Stauffer, 1984). 375 First, eight tie-points were selected and linearly interpolated. In this case, Δage showed a minimum value at around 17.5 ka ( Fig. A6), which is near to the time period of the Last Glacial Maximum (LGM). Since it is known that Δage increases when the climate is colder (Schwander et al., 1997), we assumed that either the ice age is undervalued or the gas age is overvalued.
Because the δ 18 Oatm value of the Larsen ice is comparable with the EDC and WAIS Divide records, the gas age seems to be reasonable and not likely to be overvalued (Fig. 8). Hence, we concluded that the ice age should be revised. In addition, the 380 age/distance relationship of ice shows a fast age increase near ~1000 m (Fig. A7), which is not supported by the average dip of the ice layer deduced from the GPR profile (Fig. 3). As the dip decreases to the downstream ice, an abrupt increase in the slope of the age/distance relationship is not expected. However, this is the case only when we assume no abrupt change in the snow accumulation rate.
After excluding one tie-point (indicated by the red dotted line in Fig. 9), the Δage of Larsen ice shows a maximum value at 385 approximately 17.5 ka, which appears more appropriate (Fig. A6). For records with horizontal distances less than 107 m and greater than 1310 m, extrapolation was conducted to estimate the ice age using the age/distance relationship at 7-11 ka and 18-24 ka, respectively. In conclusion, the ice age was estimated to be 5.6-24.7 ka (WD2014) for the surface ice. However, as noted above, local age inversion might have occurred in the upstream ice areas. Hence, the age constraint for upstream ice should be considered cautiously. The ice age calculated for a 1.95 m depth also seems reliable because the water isotope record 390 is comparable with the surface ice result when plotted with the ice age (Fig. 9c).   (Stenni et al., 2010, Bazin et al., 2013c. The ice age of the surface ice from Larsen was estimated by correlating 7 inflection points of the spline curve (indicated by black dotted lines) with TALDICE. (c) Comparison of near-395 surface δ 18 Oice with those in 1.95 m depth at Larsen BIA. (d) δ 18 Oice record of TALDICE shown in more detail (Stenni et al., 2010, Bazin et al., 2013c).
More precise ages may be obtained if additional ice is available for a higher sampling resolution. However, our data clearly show that the studied Larsen ice ages cover the Last Glacial Maximum, Last Glacial Termination, and early Holocene, and we may obtain reliable records for surface temperature at the original snow accumulation sites and atmospheric greenhouse gas 400 concentrations, although the latter is not guaranteed for depths < 4-10 m, depending on the gas species. Assuming that the average age/depth relationship of ice core #23 (~156 yr m −1 ) is maintained throughout the ice, the age may reach ~50 ka near the bedrock (~240 m of ice). Thus, ice may be obtained at least ~50 ka by shallow coring at the Larsen BIA.

Conclusions
Based on the dust bands, GPR profile, and no significant anomaly of chemical results (δ 18 Oatm, δ 2 Hice, and CH4) we conclude 405 that the ages of the downstream ice in Larsen BIA monotonically increase along the ice flow direction. The correlation of δ 2 Hice values, which are from ice core #23 and the surface ice samples, also shows that the stratigraphy of ice is not significantly disturbed and does not cause an age inversion at the site. CH4 is well-preserved in ice core #23 at a depth of > 4.6 m, and gas isotopes (δ 18 Oatm, δ 15 N-N2) are well-preserved at 1.95 m depth. Horizontal δ 18 Oatm values at a depth of 1.95 m in the Larsen BIA reveal a very typical glacial termination. Along with the stable water isotopes, correlation of CH4 concentration, and 410 δ 18 Oatm with existing ice core records, the gas ages of the studied Larsen ice cover 9.2-23.4 ka. In addition, the 81 Kr ages from #23 and TF cores also support the age. We provide high-precision ages for ice and fossil air trapped in blue ice at the Larsen Glacier. Because well-constrained ages and high quality of proxies are essential for paleoclimate study, our work may help future studies for blue ice in the Northern Victoria Land. Especially, since large amount of ice with the same age is exposed at the surface in the BIA, this study may lead to further research on paleoclimate that was difficult because of the limited amount 415 of ice samples.  Figure A2. Nitrogen and oxygen gas isotope record of ice core #23 and horizontal measurement. Horizontal measurement is conducted using ice cores at 1.95 m depth. Depth of ice core #23 was converted to horizontal distance at 1.95 m depth by using Eq. (3). The uncertainty of measured δ 18 Oatm at a depth of 1.95 m is assumed to be ± 0.05 ‰, which was deduced by the offset between the dark-red and dark-blue

Appendix C: Ice age estimation for the Larsen Glacier
To estimate the ice age at a depth of 1.95 m, the age/depth relationship of each ice core sample must be known. The relation is deduced by a simple calculation using Eq. (A1) and (A2); α, β, γ: ice age at each location; D: horizontal distance between two surface ice samples (at a depth of 20 cm); θ is the average dip of the ice layer; H + 0.2 m: depth where the age is the same as β. The average dip was estimated from the GPR profile (yellow box in Fig. 3c). The average dip of the ice layer located 455 near core #23 was 4.96°, as inferred from Fig. 4b. Then, the ice age at a depth of 1.95 m was calculated using Eq. (A3). Refer  Bindschadler, R., Vornberger, P., Fleming, A., Fox, A., Mullins, J., Binnie, D., Paulsen, S. J., Granneman, B., and Gorodetzky, 535