Chronological characteristics for snow accumulation on Styx Glacier in northern Victoria Land, Antarctica

Abstract

Under the potential to reconstruct the past climatic and atmospheric conditions from a deep ice core in the coastal Antarctic site (Styx Glacier), an 8.84 m long firn core (73°50.975′ S, 163°41.640′ E; 1623 m a.s.l.) was initially studied to propose a reliable age scale for the local estimation of snow accumulation rate. The seasonal variations of δ¹⁸O, methanesulfonic acid (MSA) and non-sea-salt sulfate (nssSO₄^2–) were used for the firn core dating and revealed 25 annual peaks (from 1990 to 2014) with volcanic sulfate signal. The observed declining trend in annual accumulation rate with a mean value of 146 ± 60 kg m^–2 a^–1 is likely to be linked to the changes of sea-ice extent in the Ross Sea region. Moreover, the temporal variation of the annual mean δ¹⁸O, an annual flux of MSA and nssSO₄^2– also likely to be under the influence of ice-covered and open water area. This study suggests a potential to recover past changes in an oceanic environment and will be useful for the interpretation of the long ice core drilled at the same site.

1. Introduction

Ice core chronology is an initial and essential step for the interpretation of the past records of climate, environment and atmospheric circulation from polar ice cores (Legrand and Mayewski, 1997; Traversi and others, 2004; Sinclair and others, 2010; Furukawa and others, 2017). A precise chronology provides an accurate quantification of snow accumulation rate in sub-annual and annual scales which is an important measure to surface mass-balance studies (Frezzotti and others, 2007; Rignot and others, 2011). Associated to the large spatial variability in Antarctic snowfall (Anschütz and others, 2011), the local estimation of snow accumulation rate is vital for the coastal Antarctic mass balance (Krinner and others, 2007), assessment of climate models, satellite measurements (Stenni and others, 2000; Goursaud and others, 2017) and the response of climate and environmental changes on polar ice sheet (Stenni and others, 2000; Sinclair and others, 2010; Kwak and others, 2015).

Age dating of firn cores has commonly been done by counting the seasonal signals of its stable water isotopic compositions (δ¹⁸O and δD) and ionic species (e.g. Udisti, 1996; Wagenbach and others, 1998; Kreutz and others, 1999; Udisti and others, 1999; Delmotte and others, 2000; Stenni and others, 2000; Rhodes and others, 2012; Markle and others, 2012; Caiazzo and others, 2016; Goursaud and others, 2019). In coastal Antarctica, distinguishable seasonal signal of isotopic composition is trapped in snow layers and can be interpreted based on the conventional correlation of δ¹⁸O (or δD) and air temperature (e.g. Dansgaard, 1964; Jouzel and others, 1997), coupled with the seasonal changes of sea-ice extent (SIE) (Bromwich and Weaver, 1983; Noone and Simmonds, 2004; Rhodes and others, 2012; Tuohy and others, 2015; Holloway and others, 2016). Moreover, the peak values of CH₃SO₃^– (methanesulfonic acid [MSA]) and non-sea-salt sulfate (nssSO₄^2–) illustrate noticeable peaks during spring and summer in the northern Victoria Land (e.g. Udisti and others, 1998; Rhodes and others, 2012; Becagli and others, 2016).

Due to the short-term and sparse instrumental data (e.g. Stenni and others, 2000; Tuohy and others, 2015, Jones and others, 2016), trying to capture the high-spatial variability of the vast Antarctic continent (Masson-Delmotte and others, 2008), more proxy records with sufficient resolution are still necessary to be reconstructed (Goursaud and others, 2019). Particularly at coastal sites, there is a need to improve the interpretation of past environmental and climate records (Dansgaard, 1964; Stenni and others, 2000, 2017; Krinner and others, 2007; Jouzel and Masson-Delmotte, 2010; Tuohy and others, 2015; Goursaud and others, 2019). Accordingly, the northern Victoria Land is characterized by less available data and large variability in temperature (Yang and others, 2018). The combination of the high-altitude Transantarctic Mountains and the low-altitude coastal areas means that this site is a potential location for the study of air mass transport at both the regional and global scale (Udisti and others, 1998). Moreover, a strong influence from marine sources also enhances the particular interest to investigate the variability of SIE (Drewry and others, 1982; Han and others, 2015; Kwak and others, 2015; Thomas and others, 2019). In addition, the previous studies showed a relatively higher accumulation that ranged from 203 to 226 kg m^–2 a^–1 in Styx Glacier (Stenni and others, 2000; Kwak and others, 2015), thus, the site is assumed to provide better annual signals of proxy records with negligible effects of post-deposition noise (Maggi and others, 1998; Frezzotti and others, 2007; Goursaud and others, 2019).

As a part of the ongoing research at Jang Bogo Korean research station focused on reconstructing past climate with a focus on changes in SIE, two cores (210.5 and 8.84 m long) were obtained from the Styx Glacier during 2014–15 austral summer. The 210.5 m long ice core (Styx-M) is approximately estimated to cover the time period back to >2 ka (Han and others, 2015; Yang and others, 2018). To interpret the long ice core (Styx-M) and for accurate dating, the isotopic and ionic composition of the less consolidated snow layers in the firn core (Styx-B) are important. In this study, the chronological characteristics of 8.84 m long firn core (Styx-B) was discussed by evaluating the seasonal variations of the isotopes (δ¹⁸O and δD) and major ionic species (Na⁺, K⁺, Mg²⁺, Ca²⁺, MSA, Cl^– and SO₄^2–). The correlation and principal component analyses were applied to characterize the record of isotopes and ionic species. The seasonal signals of the δ¹⁸O and δD, nssSO₄²⁻ and MSA were detected as already done in earlier studies (e.g. Udisti, 1996; Udisti and others, 1998; Stenni and others, 2000; Kwak and others, 2015; Caiazzo and others, 2016) and the annual mean snow accumulation rates were estimated. The mean accumulation rate was compared with the estimation by firn densification models (Herron and Langway, 1980; Morris, 2018). More reliably interpreting this core is made possible by comparison with instrumental data and the known climate events in recent years. This study will extend the previous records in this site (Udisti and others, 1998; Stenni and others, 2000; Traversi and others, 2004) and will be useful for the interpretation of the long ice core drilled at same site to represent the historical changes of oceanic (e.g. sea ice) and atmospheric environment in this region.

2. Materials and methods

2.1. Study area

The drilling site (73°50.975′ S, 163°41.640′ E; 1623 m a.s.l.) is situated on the Styx Glacier (150 km² plateau area) in the northern Victoria Land, East Antarctica, which is 85 km north of the Jang Bogo Station and ~60 km from the western coast of the Ross Sea (Fig. 1). During the Korean ice core drilling program (2014–15), the firn core (Styx-B) was drilled from the adjacent position (at a distance of ~100 m) of the longer ice core (Styx-M) by Korea Polar Research Institute (KOPRI). The annual mean temperature was −32.5°C, based on the 15 m depth borehole temperature measurement (Yang and others, 2018) and a horizontal ice flow was estimated to be ~0.9 m a^–1, and an ice thickness was determined to be 550 m by using a ground-penetrating radar (Hur, 2013).

Fig. 1.

The location of drilling site on Styx Glacier with South Korean Jang Bogo Research Station, Italian Mario Zucchelli Station and the nearest AWS (Lola) in Northern Victoria Land, East Antarctica.

[Figure omitted. See PDF]

Air mass mainly originates from the low-pressure center locally formed over the Ross Sea (Drewry and others, 1982; Scarchilli and others, 2011; Sinclair and others, 2012) with the prevailing southerly and southwesterly wind in this region (Udisti and others, 1999). During the period of 1980–2010, the sources from oceanic/West Antarctic (57.4%) and continental/East Antarctic (42.6%) were suggested as the dominant source to Ross Sea region (Markle and others, 2012). Stenni and others (2000) indicated that this region is a flat area with well-preserved snow layers, accompanying by a minimal effect from katabatic wind/existence of modest snowdrifts and absence of melting of ice layers (Udisti and others, 1998). Moreover, the accumulation rate is relatively high in this region (Stenni and others, 2000; Han and others, 2015) with respect to other inland Antarctic sites (e.g. Wagnon and others, 1999; Traversi and others, 2000). These features support the correct detection of the seasonal signals, because the high thickness of the annual snow layer prevents the uncertain stratum of snow that may result from the diffusion of isotopic components (Hoshina and others, 2016).

2.2. Sampling and analysis

The firn core with an average diameter of 0.09 m (ranged from 0.092 to 0.096 m) was collected in 18 runs by a drilling system by Geo Tecs Co., Ltd, Japan (Han and others, 2015). The average length for each run was 0.51 m (ranged between 0.25 and 0.84 m). During sampling, the density was estimated by measuring the mass and dimensions of the firn core subsections. The samples were kept frozen in polyethylene plastic bags and transported to KOPRI in Korea. The samples were stored under −20°C temperature in a cold room prior to the isotopic and ionic analysis.

Inside the cold room, the external thin layers were removed and cut with an average length of 0.04 m for isotopic analysis. The subsampled ice samples were melted at room temperature (~20°C) in a clean room and filtered by a 0.45-μm PVDF syringe filter (Merck Millipore, USA) before being decanted into 2-mL glass vials (dried for over 72 h at 50°C). Isotopic measurements were performed on 227 samples by a cavity ring-down spectroscopy (L1102-i, L2130-i, L2140-i, Picarro Inc., USA) and calibrated by International Atomic Energy Agency standards including Vienna Standard Mean Ocean Water (VSMOW), Greenland Ice Sheet Precipitation and Standard Light Antarctic Precipitation. Moreover, an in-house reference standard (with the isotopic composition of −34.6 ± 0.07‰ for δ¹⁸O and −272.4 ± 0.6‰ for δD, respectively), prepared by Antarctic snowmelt was measured in every ten samples to monitor the operation of the analyzer. Isotopic ratios were expressed by the comparison with the VSMOW standard in delta notation (δ¹⁸O and δD) with the unit of per mil (‰) as described in the following equation:1\[\delta = \left({\displaystyle{{R_{{\rm sample}}-R_{{\rm VSMOW}}} \over {R_{{\rm VSMOW}}}}} \right)\;\times \;1000\]where R represents the ratio of heavy to light isotopes (¹⁸O/¹⁶O and D/H) in the sample and VSMOW standard. Analytical reproducibility was better than 0.1 and 1‰ for δ¹⁸O and δD, respectively. The deuterium excess (d-excess), which represents the deviation of the distribution of isotopic composition from the global meteoric waterline (GMWL), was estimated by using the equation: d-excess = δD − 8 × δ¹⁸O (Dansgaard, 1964).

Ionic species (Na⁺, K⁺, Mg²⁺, Ca²⁺, MSA, Cl^– and SO₄^2–) were simultaneously measured with an average depth resolution of 0.045 m (in 197 samples) by the two-channel ion chromatography system (ICS-200 and ICS-2100; Thermo Fisher Scientific Inc., USA) at KOPRI. Anions (MSA, Cl^– and SO₄^2–) were measured using a Dionex model ICS-2000 with an IonPac AS15 column and KOH eluent (6–55 mM). For the cation analysis (Na⁺, Ca²⁺, Mg²⁺ and K⁺), a Dionex model ICS-2100 with an IonPac CS12A column and MSA eluent (20 mM) were used. The analytical detection limit, reproducibility and accuracy were 0.01–0.26 μg L^–1, 0.4–17.4% and 4.5–12.0% for cations and 0.02–0.26 μg L^–1, 0.1–27.6% and 1.3–5.6% for anions, respectively (Hong and others, 2012). The non-sea-salt (nss) proportion of the ions including K⁺, Mg²⁺, Ca²⁺, Cl^– and SO₄^2– were calculated to disentangle the input from the sea salt (ss) aerosol by Eqn (2). The theoretical ratio of certain ions and Na⁺ in sea water were used (Pilson, 2013). In this equation, Na⁺ was considered to be solely supplied from a sea spray (Kuramoto and others, 2011):2\[{\rm nss}\lsqb {\rm X} \rsqb = {\rm tot}\lsqb {\rm X} \rsqb -\lpar {{\rm X/N}{\rm a}^ + } \rpar _{{\rm sw}} \times {\rm N}{\rm a}^ + \]where X is the ion species and (X/Na⁺)_sw is the ratio of the certain ion and Na⁺ in sea water. Similarly, the enrichment factor of SO₄^2– was estimated by dividing the ratio (SO₄^2–/Na⁺)_sample in the sample to the theoretical ratio in sea water (SO₄^2–/Na⁺)_sw (Pilson, 2013).

2.3. Age–depth determination

Principal component analysis (PCA), the multivariate statistical technique, is applicable to classify the large dataset (e.g. Lee and others, 2008). Several studies have applied PCA to measure proxies for past atmospheric conditions from ice core records (e.g. Knüsel and others, 2005) and snow pits (Kwak and others, 2015). It was utilized to emphasize the dominant characteristics of isotopic and ionic values in same depth resolution of the firn core (n = 197) using JMP^® statistical software. The inter-correlated variation, which represents the same atmospheric sources and common transport path (Knüsel and others, 2005; Kwak and others, 2015), is shown (Fig. 2) and assist the visual interpretation of the temporal signals by detecting the seasonal tendency.

Fig. 2.

Annual layer counting of the isotopic and ionic species with PC 1 and PC 2 scores, starting from the year of drilling (2014). The Cl^–, Mg²⁺ and K⁺ are not shown in the figure because these are similar to Na⁺ (r > 0.99), but the ratio of Cl^–/Na⁺ is shown. Non-sea-salt portions are indicated for Ca²⁺. The raw values (wide line) of δ¹⁸O, δD and MSA shown with standardized profile in thin line. Enrichment factor (EF) of nssSO₄^2– was shown in black line and horizontal red line indicates the average nssSO₄^2–. Brown shading represents depth range to increased nssSO₄^2– which indicate the period of Pinatubo and Cerro Hudson volcanic signals.

[Figure omitted. See PDF]

The age determination was based on the seasonal peaks of δ¹⁸O (δ¹⁸O preferred and the correlation to δD is 0.99) with MSA and nssSO₄^2– (e.g. Udisti and others, 1996; Kwak and others, 2015; Caiazzo and others, 2016), starting from the year of the drilling. The elevated values of δ¹⁸O identified as midsummer (1 January) (e.g. Markle and others, 2012; Rhodes and others, 2012; Tuohy and others, 2015). Also, the age scale was adjusted with nssSO₄² signal of Pinatubo and Cerro Hudson volcanic eruption. The age for each sample was determined by the linear interpolation between subsequent two peaks. The number of samples per year ranged between 4 and 14 (averagely nine samples) and the annual accumulation rate was estimated in adjusted water-equivalent by multiplying its density.

Three stages of densification have been identified which reach 550, 820–840 and 917 kg m⁻³ in the first, second and third stages, respectively (Benson, 1962; Anderson and Benson, 1963; Herron and Langway, 1980). The density of the firn core (<500 kg m⁻³) is lower than the critical density of 550 kg m^–3, thus only the first stage of the densification was considered. Within this stage, the firn becomes denser rapidly by the processes of grain-boundary sliding and grain growth due to the overlying weight of snow layers and the changes in temperature affects the densification rate (Herron and Langway, 1980). In the firn densification model 1 (M1) (Herron and Langway, 1980) and model 2 (M2) (Morris, 2018), the annual mean temperature of −32.5°C at 15 m borehole measurement was used (Yang and others, 2018). Age–depth calculation was performed with the mean accumulation rate of 150 kg m^–2 a^–1 which can be averaged from our annual layer counting (excluded a low value in 2014). The M2 used the different transition density and scaling factor from the M1, which allowed modeling a smooth transition between the densification stages 1 and 2. The age scale by the annual layer counting was compared to the model results.

The forecast data (hourly average) of the precipitation (precipitation = total precipitation − evaporation) by the ERA-Interim reanalysis dataset produced by the European Centre for Medium-Range Weather Forecasts (ECMWF) in the nearest (~4 km) gridpoint (73°52.30′ S, 163°37.30′ E) to the study area were selected and compared to our estimated snow accumulation rates (Dee and others, 2011). We also compared the annual mean isotopic composition (δ¹⁸O) with the instrumental temperature record from automatic weather station (Lola AWS), which locates ~40 km south of the site (http://www.climantartide.it) and 2–m temperature data from ERA-Interim reanalysis (Dee and others, 2011), and SIE (Parkinson and Cavalieri, 2012; Data archive at National Snow and Ice Data Center).

3. Results and discussion

3.1. Temporal variations of the isotopes and ionic species

The fluctuations in the concentrations of the ionic species with depth (depth is in snow depth) (Fig. 2) and the descriptive statistical parameters are displayed (Table 1). Strong positive correlations (r) were observed between Na⁺ and other ionic species (r > 0.9 for K⁺, Mg²⁺, Cl^–, SO₄^2–, and r = 0.5 for Ca²⁺, n = 197, p < 0.001). These species are the main components of sea-salt aerosols, which originated from the surrounding oceanic source (Piccardi and others, 1994; Traversi and others, 2004). The mean values of nss fractions of ionic species, including K⁺ (19.2%), Mg²⁺ (19.1%) and Cl^– (25.1%) were lower than the Ca²⁺ (74.0%) and SO₄^2– (65.5%) (Table 1). The ion distributions were roughly right-skewed, related to the inconsistent increased sea-salt aerosols. An abrupt increase was noticed in the ionic concentrations, including Na⁺, K⁺, Mg²⁺, Ca²⁺, Cl^– and SO₄^2–, at a depth of 6.47 m. The highest fraction of sea-sourced Ca²⁺ (~99.9%) and the ratio of Cl^– to Na⁺ (1.86), which is similar to the theoretical ratio of sea water (Pilson), observed at this depth, indicating the abnormal sea-salt aerosol input.

Table 1. Summary of statistics of water stable isotopes in ‰ (n = 227) and chemical species in μg L^–1 (n = 197) in the firn core

	Min	Max	Range	Mean	Median	Std. error	Std. deviation	Mean nss %
δ18O	−43.13	–26.70	16.43	−34.92	−34.85	0.21	3.13	–
δD	−340.63	−205.17	135.46	−273.11	−272.99	1.68	25.32	–
MSA	1.34	53.71	52.37	8.12	5.25	0.51	7.17	–
Cl–	6.58	16 394.61	16 388.03	339.42	151.98	88.34	1239.98	25.1
SO42–	13.38	2506.58	2493.20	91.63	65.80	13.43	188.53	65.5
Na+	1.20	8802.32	8801.12	171.07	64.45	47.81	671.04	–a
K+	0.43	373.13	372.70	7.38	2.61	2.04	28.60	19.2
Mg2+	1.07	1042.42	1041.35	20.86	10.01	5.53	77.67	19.1
Ca2+	2.73	362.57	359.83	26.97	9.91	3.64	51.08	74.0

^a Assumed to be solely originated from sea spray.

PCA was performed on the isotopic (δ¹⁸O, δD and d-excess) and ionic (Na⁺, K⁺, Mg²⁺, Ca²⁺, MSA, Cl^–, ssSO₄^2– and nssSO₄^2–) values for same depth resolution of the firn core (n = 197). Principal component (PC) 1 explained 49.52% of the total variance with the strong loadings of the positively correlated sea-salt ions including Na⁺, K⁺, Mg²⁺, Ca²⁺, Cl^– and ssSO₄^2– (Table 2). This result highlights the dominant contribution of the oceanic source in line with other studies (Kwak and others, 2015). The second principal component (PC 2) has high loadings of δ¹⁸O, δD, nssSO₄^2– and MSA, accounting for 23.14% of the total variance and indicates the identical seasonal nature of these variables. PC 1, PC 2 and PC 3 captured 81.52% of the total variance (Table 2), including 8.86% from PC 3 with high loading of d-excess. PCA classified the variables into species representing an input of sea-salt aerosol (PC 1) and species signing the seasonal cycle, related to the high temperature (PC 2). PC 2 score suggests the potential seasonal markers including δ¹⁸O, δD, nssSO₄^2– and MSA (e.g. Stenni and others, 2000; Kwak and others, 2015).

Table 2. Loadings of 11 variables for first three principal components (PC) in the firn core with larger values highlighted in bold

	PC 1	PC 2	PC 3
δ18O	−0.12	0.92	−0.24
δD	−0.12	0.93	−0.17
d-excess	−0.03	0.33	0.92
MSA	0.15	0.60	0.09
Cl–	0.99	−0.01	0.02
ssSO42–	0.99	−0.01	0.02
nssSO42–	0.33	0.60	0.01
Na+	0.99	−0.01	0.02
K+	0.99	−0.02	−0.01
Mg2+	0.99	0.00	0.02
Ca2+	0.62	−0.03	−0.19
Variance explained (%)	49.52	23.14	8.86
Cumulative percent (%)	49.52	72.66	81.52
Number of variables	197	197	197

3.1.1. Temporal variations of stable isotopic composition

The δ¹⁸O, δD and d-excess with the increasing depth were presented in Figure 2. The δ¹⁸O (δD) values (n = 227) ranged from −43.13‰ (−340.6‰) to  −26.70‰ (−205.2‰), with a mean value of −34.92‰ (−273.1‰), which were analogous to other records on the Styx Glacier (Stenni and others, 1999; Kwak and others, 2015). Moreover, mean δ¹⁸O averaged from database, compiled in Masson-Delmotte and others (2008), showed similar values of −30.19‰ in the coastal Antarctic sites (~40–80 km from coast) and −30.91‰ in the sites, located in the elevation range between 1400 and 1800 m a.s.l. The isotopic compositions revealed maximum (summer) and minimum (winter) values with a distinct seasonal bias. Summer peak values (assumed as January) of δ¹⁸O (δD) ranged between −36.50‰ (−287.2‰) and −27.22‰ (−210.9‰), respectively. There was not a clear increasing or decreasing trend in δ¹⁸O (δD) in the whole record down to 8.84 m (Fig. 2). The seasonal pattern of δ¹⁸O (δD) was used for stratigraphic dating with the assumption of less post-depositional effects due to the relatively high accumulation rate (e.g. Stenni and others, 2000; Kwak and others, 2015; Furukawa and others, 2017).

The linear relationship between δ¹⁸O and δD appeared with a slope of 8.06 (δD = 8.06 × δ¹⁸O + 8.56, R² = 0.99, n = 227, p < 0.001), which is fairly consistent with the GMWL (δD = 8 × δ¹⁸O + 10) (Dansgaard, 1964). Masson-Delmotte and others (2008) reported a meteoric waterline (δD = 7.75 × δ¹⁸O − 4.93, R² = 0.98, n = 789) at the continental scale, which is slightly different to our estimate. This negligible alteration of the slope leads us to assume that the accumulated snow layers are fairly well preserving the precipitation isotopic content with less influences by post-depositional processes, including temperature gradient metamorphism (Lee and others, 2010), diffusion process (Johnsen and others, 2000) and sublimation (e.g. Steen-Larsen and others, 2014).

The slope in summer (8.91) was larger with higher d-excess values than those in a winter period (7.99). Moreover, the less variation in the seasonal cycle of d-excess (mean value of 6.28 ± 1.77‰) values signify the dominant moisture transport from the neighboring stationary ocean source (e.g. Jouzel and Merlivat, 1984; Jouzel and others, 2007; Fujita and Abe, 2006). Although the distribution of d-excess differed from the oscillations of δ¹⁸O and δD, the relative consistency in the variation is likely to be linked to the dominant contribution of the steady marine source and local transport of the precipitation (Delmotte and others, 2000). It is because the variation of the sea surface temperature, relative humidity and wind speed at the moisture source region strongly control the variations of d-excess (Merlivat and Jouzel, 1979). Stenni and others (2000) suggested the main influences are originated from the Pacific Ocean sector and the Ross Sea.

3.1.2. Temporal variations of ionic compositions

The sea-salt (process of bubble bursting and wind blowing on the wave crests on open-water and ice-covered area) and oceanic biogenic activities are considered as two major sources of ionic species in the coastal Antarctic firn cores (e.g. Delmas and others, 1992; Wagenbach and others, 1998; Kreutz and Mayewski, 1999; Benassai and others, 2005). Higher concentrations of sea-salt aerosols (e.g. Na⁺, Mg²⁺, K⁺, Ca²⁺, Cl^– and ssSO₄^2–) are observed in the winter layer, related to the thin layer of sea-salt-enriched crystal formed above sea ice and its movement with blowing snow (Rankin and other, 2000; Abram and others, 2007; Jiayue and others, 2017; Vega and others, 2018), which is larger than those transported from open sea during warm periods. Coupled with high cyclonic activities in the Ross Sea, mineral dusts (partial of K⁺, Mg²⁺ and Ca²⁺) showed an enhancement in spring and summer in northern Victoria Land (e.g. Caiazzo and others, 2017). This relates to the input from local ice-free areas as well as long-range transport from South America (Kreutz and others, 1999). Na⁺ is a more reliable proxy for sea-spray aerosol due to the absence of extra sources, particularly in coastal regions, and the steady concentration during transport and preservation without fractionation (Traversi and others, 2004).

In the firn core, Na⁺ show increase in some winter period which may be associated with the high concentrations of sea-salt aerosols induced by the changes in sea ice and the intensive winter storm events, moving sea-salt aerosols (Brimblecombe, 1996; Udisti and others, 1998; Rankin and others, 2000; Stenni and others, 2000). According to the wind data from the Lola AWS (1990–2014), the higher velocities >30 m s⁻¹ were observed mostly in late autumn to early spring (69% of total counts) and highest frequencies occurred in winter (between June and September) with 50.4% of the total counts. Moreover, the western and southwestern winds were dominant in these data, suggesting the air masses originated from ocean. However, the Na⁺ concentration peaks as well as K⁺ and Mg²⁺ were not consistent with all case in winter values (Fig. 2). It may indicate the event-based distribution relating to variability of atmospheric pressure pattern (e.g. Tuohy and others, 2015). The slope between the Na⁺ and Cl^– ([Cl^–] = 1.846 × [Na⁺] + 23.47, R² = 0.99, n = 197, p < 0.001) was similar to the ratio in sea water (1.81), which represents the dominant contribution of the oceanic source (Traversi and others, 2004). However, with this intercept of 23.47 and the Cl^–/Na⁺ ratio for each sample for each depth was larger than the sea-water ratio of 1.81 indicate the additional source for Cl^– (Benassai and others, 2005). Due to this inconsistent distribution trend of the sea-salt aerosols, we were not able to use them as time markers for age dating.

Shifts in the distribution of Ca²⁺ ions suggest contribution from various sources in this area. The average value of the nss fraction of Ca²⁺ is 74.0%, indicating the less dominantly sourced from sea spray. Other sources are considered as a crustal dust input to the total budget of Ca²⁺ (Fig. 2). Although, the high concentration of nssCa²⁺ could indicate the spring and summer season due to the mesoscale cyclonic activities (Caiazzo and others, 2017) and generally lithophile elements are influenced by the wind events in the Ross Sea region (Rhodes and others, 2012). Our result shows no distinguishable seasonal pattern for Ca²⁺, which thus cannot be used for age determination of the firn core.

Because the site is near to coast, nssSO₄^2– is mainly derived from the marine biogenic activity (oxidation of dimethylsulfide [DMS] emitted from marine algae and phytoplankton) and partially from crustal erosion, and volcanic emissions (Handler, 1989; Delmas and others, 1992). Moreover, the spikes in nssSO₄^2– indicate the timing of volcanic eruption in polar ice cores (e.g. Legrand and Mayewski, 1997; Cole-Dai and others, 1999). Together with the increased MSA, which was solely from marine biogenic activity, the values of nssSO₄^2– and MSA illustrate noticeable signal of spring and summer snow layers in the northern Victoria Land (e.g. Legrand and others, 1992; Udisti and others, 1998; Saltzman and others, 2006; Rhodes and others, 2012; Becagli and others, 2016). The MSA and nssSO₄^2– are observed with a similar concentration range to other studies (Stenni and others, 2000) and the maximum values are slightly differed from each other (Fig. 2). This discrepancy is likely to be linked to the distinct characteristic of the photochemical oxidation rate (Preunkert and others, 2008) and difference in the size and movement trajectory of them even both enhanced during the spring/summer period (Becagli and others, 2012). The MSA record reveals the weak correlation with nssSO₄^2– (r = 0.34, p < 0.001, n = 197). The culmination values of nssSO₄^2– and MSA were detected mostly in the late spring/summer before the summer maximum of δ¹⁸O associated with the enrichment of marine biogenic activity with a presence of open-water area (Fig. 2) (e.g. Udisti and others, 1998; 1999; Stenni and others, 2000; Kwak and others, 2015). The spring/summer elevated values matched with the earlier studies (e.g. Whitlow and others, 1992; Stenni and others, 2000; Caiazzo and others, 2016).

The negative values of nssSO₄^2–, which are principally observed in coastal/low-latitude areas and detected at depths of 6.29 and 8.52 m in the firn core, is likely to be the winter high sea-spray content (Piccardi and others, 1996; Udisti and others, 1998). The nssSO₄^2– concentration reached a maximum value of 523.0 μg L^–1 at a depth of 8.04 m with relatively higher concentrations at the subsequent depth intervals (8.07–8.15 m), compared to the mean concentration of 48.5 μg L^–1. Moreover, this increasing trend with double peaks continues to the depth of 7.17 m with a relatively higher concentration of nssSO₄^2– and the higher enrichment factor of SO₄^2– (Fig. 2).

We have assumed that the peak values at this depth range to be sourced from a volcanic eruption that occurred in 1991 (Stenni and others, 2002; Karlöf and others, 2005; Nardin and others, 2020). The Pinatubo eruption (June 1991, Philippines, 15°08′ N, 73°00′ E) released 18 ± 2 Mt SO₂ (Cole-Dai and others, 1999) and Cerro Hudson eruption (August 1991, Chile, 45°55′ S, 73°00′ W) (Legrand and Wagenbach, 1999) injected 2 Mt of SO₂ into the atmosphere (Doiron and others, 1991). In the previous studies, the Cerro Hudson plume was centered over the South pole in September 1991 and Pinatubo plume was indicated in November 1991 (Cacciani and others, 1993; Saxena and others, 1995). The annual layer counting revealed that the depth of the higher concentrations of nssSO₄^2– (particularly at 7.61–7.84 m) corresponds to late-1992 to early-1994. This is sufficiently consistent with other studies, showing that the aerosol mass of Pinatubo spread in mid-1992 and late-August or early-September of 1992 (Trepte and others, 1993; Hitchman and others, 1994; Cole-Dai and others, 1999; Severi and others, 2009). Meanwhile, the depth range of 8.04–8.15 m with the sharp sulfate peak (late-1991 to early-1992) is likely to represent the volcanic aerosol input from the Cerro Hudson with the early detection associated with the relatively near distance to the site (e.g. Cole-Dai and others, 1999; Karlöf and others, 2005). This separate detection of two volcanic events was also observed in other sites (Cole-Dai and others, 1999; Stenni and others, 2002; Karlöf and others, 2005; Nardin and others, 2020) however there are some cases of single detection of only Pinatubo eruption (smooth peak) even in the site from northern Victoria Land (e.g. Severi and others, 2009). Thus, this result suggests the potential of the firn cores in this site to recover high-resolution records. Despite the potential local inputs from Mount Erebus (>300 km far from the Styx Glacier) and Mount Melbourne volcano (~40 km distance from the Styx Glacier), from this point of view, our findings confirm the notable nssSO₄^2– value indicate volcanic signals as well as a seasonal indicator (spring/summer) with MSA for dating purposes.

3.4. Dating and accumulation rate

The annual signals of peak values (δ¹⁸O) with MSA and nssSO₄^2– were observed in every 0.34 m (averagely eight samples per year for isotopic profile) in snow depth and suggest a 25-year record (1990–2014). The nssSO₄^2– of the volcanic events (Pinatubo and Cerro Hudson) in 1991 was recognized in the time period between 1991 and 1994 in the dating scale. The dating difference is due to the transportation time of the volcanic gas to the high latitude sites in Antarctica (Legrand and Mayewski, 1997).

The density of the firn core ranged between 330 and 500 kg m^–3 with increasing trend as depth increases. The annual mean accumulation rate was calculated by multiplying the snow density and was estimated to be 146 ± 60 kg m^–2 a^–1 (mean value ± std dev.) during the period of 1990 to 2014. The depth–age relationship by the layer counting method was plotted together with the densification model results in Figure 3. To compare the age by annual layering method with the densification model, the depth values of the annual layering was converted to the same values of the models. Our estimate of annual mean accumulation rate is reasonable because there was an average value of 1.6- and 0.8-years difference with M1 and M2 (n = 22), respectively. Our estimate on annual mean snow accumulation is closer to the model-based estimation of 130 kg m^–2 a^–1 (Han and others, 2015) and comparable to other quantities of accumulation rate on Styx Glacier having values of 160, 203 and 226 kg m^–2 a^–1 (Udisti, 1996; Stenni and others, 2000; Kwak and others, 2015) (Table 3). Although the mean precipitation rate from ERA-Interim reanalysis (154 ± 37 kg m^–2 a^–1) is similar to our estimated value (local estimate), the temporal variation of snow accumulation during the corresponding period was different (Fig. 4). This inconsistency in the annual accumulation rate can be resulted by the post-depositional effects, including mainly by the wind-driven snow redistribution and partially by the no existence of precipitation, surface and snowdrift sublimation in snow layers (Fig. 4) (e.g. Frezzotti and others, 2004; Karlöf and others, 2005; Sinclair and others, 2010; Rhodes and others, 2012), and the influence by the temperature change, regime of precipitation and changes in cyclonic systems (e.g. frequency, path, timing and strength) (Kreutz and others, 2000; Goodwin and others, 2003; Kaspari and others, 2004). Jang and others (2019) have mentioned that the strong wind effect (as blizzards) could have induced the large variation in density of snow in the Styx Glacier.

Fig. 3.

The comparison of the depth–age relationship of the annual layer counting and the firn densification models.

[Figure omitted. See PDF]

Fig. 4.

Comparison of annual accumulation rate (Styx-B firn core, ERA-Interim, Stenni and others, 2000), annual mean δ¹⁸O, standardized annual flux of MSA and nssSO₄²⁻, SIE during cold period (JJASON) and temperature record from Lola AWS and ERA-Interim reanalysis data. The linear trendlines shown for annual accumulation rate (Styx-B firn core for the period of 1990–2014 together with Stenni and others (2000) up to 1971 from 1990) and trend of SIE between 1979 and 2014.

[Figure omitted. See PDF]

Table 3. Mean accumulation rates estimated near to the Styx Glacier for comparison to the firn core

Locations	Accumulation rate, kg m–2 a–1	Time period	Reference
Styx Glacier	146 ± 60	1990–2014	This study
ERA-Interim	154 ± 37	1990–2014	Dee and others, 2011
ERA-Interim	150 ± 36	1979–2014
Hercules Neve	160	1971–1992	Udisti (1996)
Styx Glacier	203 (111–335)	1971–1990	Stenni and others (2000)
Talos Dome	86.6	1965–2007	Stenni and others (2002)
Styx Glacier	226	2009–2012	Kwak and others (2015)
Styx-M (densification model for Styx-M-core)	130	~1360–2014	Han and others (2015)
GV7	242 ± 71	2008–2013	Caiazzo and others (2017)

Stenni and others (2000) reported the annual accumulation range of 111–335 kg m^–2 a^–1 for the period between 1971 and 1990 and this estimation is plotted together with the firn core (Fig. 4). There is an overall decreasing trend in accumulation rate in our estimate (r = −0.53, p < 0.001, n = 25). The accumulation profile was expanded with Stenni and others (2000) (r = −0.56, p < 0.001, n = 45). The highest (>200 kg m^–2 a^–1) annual accumulation rate was observed in years 1992, 1995, 1996, 1999, 2002 to 2004, while the three lowest annual accumulation rates occurred during 2012–2014. Stenni and others (2000) highlighted that there was an important role of the Ross Sea SIE on the snow precipitation in northern Victoria Land. The mean SIE in the Ross Sea region during cold period (months between June and November [JJASON]) shows a weak negative correlation with both the snow accumulation rate (r = –0.41, p < 0.01, n = 25) and an annual mean δ¹⁸O (r = −0.32, p < 0.05, n = 25) of the firn core.

A negative correlation between d-excess and SIE was observed with a correlation coefficient of −0.53 between d-excess record of the WhiteHall Glacier site and SIE of the Ross Sea (Sinclair and others, 2014). However, we find a weaker correlation between our d-excess and the SIE (r = −0.21, p < 0.01, n = 25) as well as a weak negative correlation of SIE with the annual flux of MSA (r = −0.51, p < 0.01, n = 25) and nssSO₄^2– (r = −0.39, p < 0.01, n = 25), while the annual flux of Na⁺ does not show significant correlation (Fig. 4). It can be explained by single strong source of open-water area for the emission of DMS to the concentrations of MSA and nssSO₄^2– (e.g. Rhodes and others, 2009; Sinclair and others, 2014). While several enhancing factors (may likely to individual storm events) rather than only SIE was assumed for the Na⁺ concentration in the firn core. The increasing (trend) of SIE during the period of 1990 to 2014 in the Ross sea region and evidence in ice core from WhiteHall Glacier (Sinclair and others, 2014) and decreasing precipitation from the firn core is likely to be coupled with the positive trend of Southern Annular Mode (Marshall and others, 2018) and deepening of Amundsen Sea Low pressure and its location (Hosking and others, 2013; Sinclair and others, 2014), which enhances westerly winds and the cooling of air temperature (−0.55 ± 0.97°C/100 a^–1) in the northern Victoria Land coast (Stenni and others, 2017; Yang and others, 2018).

With respect to the conventional correlation of annual mean values of isotopic ratio and air temperature (T − δ¹⁸O) (Dansgaard, 1964; Petit and others, 1999), the isotopic composition of the firn core does not quantitatively represent this relationship with temperature record of both Lola AWS and ERA-Interim due to the alteration of seasonality of δ¹⁸O with increasing depth and coupled influencing factors (Fig. 4). Annual mean δ¹⁸O (δD) fluctuated between −40.65‰ (−320.9‰) and −30.24‰ (−236.1‰) with a mean of −34.96‰ (−273.6‰) and std dev. of 2.17‰ (17.68‰), while the atmospheric temperature measured at the nearest Lola AWS was less variated (ranged between −23.41 and −18.83°C with the mean of −22.44°C and std dev. of 1.02°C). Stenni and others (2000) compiled the annual mean temperature (both AWS and borehole temperature measurement) and mean δ¹⁸O values from other firn cores in northern Victoria Land and suggests a slope of 0.81‰ °C^–1 (r = 0.9). Slightly different slopes of 0.44‰ °C^–1 (van Ommen and Morgan, 1997), 0.60‰ °C^–1 (Stenni and others, 2002) and 0.62‰ °C^–1 (Sinclair and others, 2012) were reported in East Antarctic sites. Moreover, Stenni and others (2017) reported a slope of 1.05‰ °C^–1 in East Antarctic Plateau and 1.21‰ °C^–1 in Victoria Land coast by model simulation.

In our study, the gradients of 1.07 and 1.20‰ °C^–1 were found by comparing mean δ¹⁸O value (−34.96‰) with borehole temperature (−32.5°C) and mean annual temperature from ERA-Interim (−29.07°C), respectively. Large difference in mean temperature recorded in the Lola AWS (−22.44°C) compared to the estimates in borehole and ERA-Interim suggests the importance of a reliable instrumental records. The various factors including changes in evaporation conditions, transport pathway, and changes in precipitation seasonality (Werner and others, 2018) and post-depositional processes (Sinclair and others, 2010; Rhodes and others, 2012; Goursaud and others, 2017; Casado and others, 2018) need to be considered to this high-resolution firn core record. Moreover, there are studies that reported a weak or no correlation between annual mean temperature and δ¹⁸O in the coastal Antarctic sites (Bertler and others, 2011, 2018; Goursaud and others, 2019). Goursaud and others (2018) suggested challenges to obtain a correlation between annual mean temperature and δ¹⁸O by the ECHAM5-wiso simulation during the period between 1979 and 2013. With the consideration of short-temporal data for comparison, at this stage the isotopic and ionic species (MSA and nssSO₄^2–) of the firn core are likely to preserve the sign of oceanic environment. This study can be applied for the interpretation of the longer ice core drilled at the same site which is being in analysis.

4. Summary

The chronology of the firn core from Styx Glacier in northern Victoria Land, Antarctica was studied through the temporal variations of its isotopic and ionic compositions to estimate the snow accumulation rate. The seasonal variations of δ¹⁸O, MSA and nssSO₄^2– which represent the ocean-sourced precipitation and reference horizon of volcanic sulfate signal were used to identify the age of the firn core. The 25 annual signals spanning back from 1990 to 2014 were measured with a mean accumulation rate of 146 ± 60 kg m⁻² a⁻¹. A decreasing trend of the accumulation was observed through the profile. The annual mean accumulation rate and δ¹⁸O, flux of MSA and nssSO₄^2– were shown to be correlated with sea-ice extent (JJASON) in the Ross Sea region. Although the records are short, at this stage this site is likely to preserve the changes in oceanic environment which also can be confirmed in the long ice core. This study will be useful for the interpretation of long ice core drilled at the same site.

Acknowledgements

This study was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (2017R1D1A1A09000732) and KOPRI research grant (PE20190). We thank Dr. Elizabeth Morris for sharing the results of firn densification models and comments on the manuscript. We are also grateful to the Italian PNRA (Programma Nazionale di Ricerche in Antartide) project for providing meteorological data from Lola AWS (http://www.climantartide.it) and the ERA-Interim dataset from the ECMWF. We appreciate the scientific editor and two anonymous reviewers, whose comments led to significant improvements.