3.1. Vertical profiles of sea-ice salinity, δ18O values, and sea-ice structure
Over the study period, first-year ice and snow on ice were ranged between 1.32 and 1.55 m thick and 0.05 and 0.20 m deep, respectively (Table 1). Multi-year ice and snow on ice were ranged between >2.47 and ca. 6 m thick and 0.34 and 1.68 m deep, respectively (Table 1). Figure 2 shows vertical salinity and δ18O profiles for both first-year and multi-year ice cores. The salinity of the first-year ice was close to 0 near the ice–snow interface and increased with increasing depth until reaching a maximum, after which it decreased with increasing depth (Figure 2a). Salinity again increased at the bottom of some cores, giving the salinity profiles of first-year ice a characteristic “C” shape (excluding the topmost low-salinity ice). δ18O values in first-year ice were as low as −20‰ near the ice–snow interface (Figure 2b) and increased with increasing depth, reaching values above that of the under-ice water (−0.7‰) by the middle of the core. The profiles record low salinity and δ18O values near the ice–snow interface because the cores were collected in late summer (Table 1), by which time the accumulated snow had become snow-origin ice due to melting of the sea-ice surface in above-freezing air temperatures and strong sunlight (Nomura et al., 2011a). Furthermore, in late summer, the nighttime air temperature becomes negative in this area (Nomura et al., 2011a), freezing any water that had melted during the day. Because the snow depths were quite small relative to the ice thickness (Table 1) and the salinity was extremely low, especially near the ice surface, we considered that the ice formed was superimposed ice rather than snow ice because snow ice should only form when the snow depth is large with respect to the ice thickness (Maksym & Jeffries, 2000; Sturm & Massom, 2017). Indeed, in thin section, we observed large, polygonal, granular crystal textures (photograph in Figure 2a) characteristic of superimposed ice (Kawamura et al., 2004) near the ice surface.
The salinity of multi-year ice was 0 from the ice–snow interface to depths of around –100 cm (Figure 2c), then generally increased with increasing depth. The maximum observed salinity was 3.8 in core JARE58_KU3. At the bottom of the ice column, salinities were about 1–3. The δ18O values of sea ice were similar to those of the overlying snow from the ice–snow interface to depths of –200 cm (Figure 2d), then increased with increasing depth. The ice δ18O values reached that of the under-ice water at depths of –300 cm in core JARE56_NMS and –380 cm in core JARE57_P31.5. These results indicate that snow cover affects to a maximum depth of –380 cm.
Figure 3 compares the salinity and δ18O profiles of multi-year ice core JARE57_P31.5 to the interpreted ice structure and photographs of representative thin sections. Up to a depth of –117 cm, salinity and δ18O were maintained at 0 and −20‰, respectively (Figure 3a). Because most of this upper section of the core comprised coarse-grained (large polygonal) granular ice, we concluded that it was superimposed ice (Figure 3b, c; Kawamura et al., 2004). In addition, fine-grained granular ice was present at depths from –100 to –117 cm (Figure 3b, d). Based on the salinity and textures down to a depth of –117 cm, it is unlikely that snow ice was formed by the infiltration of seawater. In addition, the small crystal size at depths of –100 to –117 cm (Figure 3d) indicated that this ice corresponded to superimposed ice formed from wet, unmelted snow that had re-frozen at negative temperatures, which we refer to as fine-grained superimposed ice.
At depths of –117 to –170 cm, sea ice salinity and δ18O values increased (Figure 3a) and the ice structure became columnar (Figure 3b, e), with larger grains than in the overlying fine-grained granular ice. However, the columns were neither as wide nor as vertically extended as those of columnar ice formed from seawater (Figure 3f, see next paragraph). Based on the salinity and δ18O values of the ice, this structural transformation occurred because the ice formed snowmelt with a slight seawater influence. Therefore, we suspect that this ice formed from a refrozen mixture of seawater and snowmelt. Indeed, in this area, it has been reported that a layer of superimposed ice forms under the snow, and that a slush layer (puddles in the sea ice that are likely snowmelt ponds) forms under the superimposed ice (Nomura et al., 2018). In that study, the salinity and δ18O values of the slush layer were ≤1 and −20‰, respectively, because snowmelt was supplied mainly from above (as snow or superimposed ice), and a small amount of seawater was transported upwards through a crack and then horizontally transported into the slush layer. The salinity and δ18O values of their slush layer were approximately the same as those of the columnar ice measured in our study (Figure 3a). The slush layer is about 10–20 cm thick (Nomura et al., 2012) and, in winter, is thought to freeze from the top faster than columnar ice develops at greater depths in the sea ice. Therefore, we consider that the columns in this zone are thinner and shorter than those in columnar ice formed from seawater (Figure 3e, f) because the freezing speed affects the crystal structure with slower growth on the c-axis direction than the basal plane (e.g., Hilling, 1959). Both fine- and coarse-grained granular ice have random crystal texture (Figure 3c-d). As the freezing develops, geometric selection sweep away crystals with c-axis perpendicular to the ice–ocean interface (Kolmogorov, 1949). The crystals in the slush layer that have the c-axes pointing radially (perpendicular to the ice–ocean interface) grow faster than those with c-axes pointing parallel to the ice–ocean interface. Therefore, it seems likely that the columns in slush layer was thinner due to the initial stage of geometric selection and shorter than those in columnar ice formed from seawater constrained by the slush layer thickness. In this study, we refer to this particular columnar ice texture as ‘slush-origin columnar ice’, in contrast to columnar ice formed from seawater (‘seawater-origin columnar ice’).
A distinct unit comprising a fine-grained granular ice layer overlying a coarse-grained granular ice layer was sandwiched between slush-origin columnar ices at depths of –170 to –190 cm. This unit formed because superimposed ice formed under the snow (Nomura et al., 2018) and a slush layer below it. This is thought to reflect greater amounts of snowfall in following years, with the slush layer being repeatedly formed and refrozen. The underlying slush-origin columnar ice layer extended to a depth of –300 cm, where the ice structure became columnar, but of seawater origin.
In contrast, the first-year ice was predominantly columnar (Table 2). From the above discussion, snow accumulates on the ice year-by-year, causing flooding and slush formation, then refreezing in winter. Furthermore, when the ice thickness increases, flooding ceases, and the snowmelt is refrozen. This upward growth is characteristic of Antarctic multi-year ice (Kawamura et al., 1997; Nomura et al., 2018), although the collection of a nearly 5-m sea-ice core has never been reported. This is thus the first time that the characteristics of the upward growth of multi-year fast ice have been shown based on multiple years of repeated ice coring.