Field site

The (unofficially named) RBIS is located on the Princess Ragnhild Coast, East Antarctica, at ~70˚ S, 24˚ E (Fig. 2a). The ice shelf extends southwest–northeast for ~350km along the coast and has a flow-parallel width of ~50– 100 km. The rift investigated herein is located within ~5km of the seaward edge of the ice shelf and both are aligned approximately parallel to each other (Fig. 2b). The rift forms the easternmost limb of a suite of fractures related to the presence of an ice rise located ~10 km upflow. Measured ice velocity in the region of the rift is ~350–380ma^–1 (Reference Pattyn, de Brabander and HuyghePattyn and others, 2005), markedly lower than that predicted theoretically in the absence of the ice rise. The ice shelf overlies a relatively shallow ocean cavity with a sub-ice-shelf water column thickness of ~200–300m (Pattyn and others, unpublished information). We drilled 13 boreholes, each 5–66m long, in either the rift or the adjacent ice shelf. Of these, six penetrated the base of the rift, extending into the underlying cavity, and six were logged by OPTV (Fig. 2b; Table 1).

Table 1. Summary data for boreholes cored on the RBIS in 2008 and 2010

Fig. 2. Location map of (a) the general study site on the Roi Baudouin Ice Shelf (PES is Princess Elisabeth Station) and (b) the individual boreholes reported herein. Boreholes logged by OPTV are represented as open circles.

Methods

Boreholes were drilled (during December 2008 or December 2010) either by a mechanical SIPRE-type corer or by an adapted Eclipse ice drill. The latter was modified to enable coring through permeable ice saturated with sea water, requiring contacts within the drill assembly (including those within the slip ring) to be electrically isolated. Importantly, these adaptations allow the Eclipse drill to operate in boreholes filled with sea water, permitting coring below the local water level to the base of the ice shelf. Once drilled, OPTV analysis was carried out as soon as possible and usually within some tens of minutes in order to avoid borehole closure by freezing and, where boreholes penetrated the cavity, clogging by buoyant platelets rising up the sea-water column in the boreholes. The latter effect was unfortunately common, generally occurring within some minutes of the corer being removed from the hole and effectively preventing OPTV logging of several rift boreholes. Once logged, OPTV images were collated, analysed and prepared for presentation (including rolling to create virtual core images) using WellCAD software. This analysis included calculating the luminosity (expressed in non-dimensioned units of RGB pixel brightness) of each 1mm depth step, represented by the mean value of each ring of 720 pixels. Within some metres (depending on the optical transmissivity of the material being cored) of the borehole surface, recorded light is dominated by that transmitted from the surface, but below this zone it is exclusively composed of that reflected back from the borehole walls to the OPTV sensor. Since borehole illumination, achieved by a circular array of white LEDs, is uniform in time and therefore also in space as the probe moves along a borehole, the net luminosity of the recorded signal varies with the reflectivity of the material forming the borehole wall.

Ice-shelf boreholes

The OPTV logs of the two boreholes cored into the ice shelf proper, 08-S1 and 10-S1 (Fig. 2b; Table 1), reveal similar material properties, with 66m long 10-S1 providing the longer record. The log of 10-S1 (Fig. 3 and 4), has three notable features. First, regularly repeated dark layers, which have a typical luminosity of <100 units lower than the local image background, can be observed along most of the length of the core. The spacing of these layers gradually decreases from ~1.0m near the borehole’s upper surface (e.g. five regularly spaced darker layers between 4 and 9 m depth in Fig. 3) to ~0.15m near its base (e.g. seven regularly spaced darker layers between 51.5 and 52.5 m depth in Fig. 3 and 4b). Although still visible, these layers are more difficult to distinguish from the (now darker) background towards the base of the borehole. Second, sharply defined very dark layers, typically with a luminosity that is >100 units lower than the local image background, are observed intermittently along the full length of the borehole. For example, four distinctive dark bands are located between 30 and 31 m depth, each of which is 5–20 cm thick (Fig. 3 and 4a). Thus, the luminosity of these layers is typically similar to, or darker than, that of the thinner and more regular layers noted above. Finally, if both the above sets of dark layers are removed from the analysis, the background reflected light intensity of the OPTV log of 10-S1 decreases consistently down the borehole, from typical values of ~400 near the ice-shelf surface to ~150 at the base of the borehole, visible in both Figures 3 and 4.

Fig. 3. OPTV log of the full length of ice-shelf core 10-S1. The raw OPTV image is plotted on the left-hand side of each panel and its rolled equivalent is plotted on the right. The luminosity trace overlaid on the raw OPTV image is sampled each millimetre in the vertical and is scaled to decrease, over the range 450–100 (non-dimensioned) units, to the right.

Fig. 4. Expanded rolled OPTV images of two 1 m long virtual core segments from ice-shelf core 10-S1 (Fig. 3): (a) 30–31m depth and (b) 51.5–52.5m depth.

Rift boreholes

We drilled ten boreholes into the base of the rift proper and one was drilled into the ramp formed in the rift tip (Fig. 2b; Table 1). Several of these boreholes intersected apparently fundamentally different material facies. The OPTV log of 08-T1, cored for 38m into the rift tip, was of uniformly high luminosity, similar to the material forming the matrix between the dark layers in 10-S1 (above). Indeed, the OPTV log of 08-T1 contrasts with that of 10-S1 in that the former is characterized by (1) fewer dark layers and (2) no apparent systematic decrease in the intensity of the background reflected light with depth.

All of the remaining ten boreholes were drilled directly into the base of the rift (Fig. 2b), intersecting a series of material facies which always appeared in the same order but which were not all present at every borehole. These facies are as described below.

Marine ice

Progressing down-borehole, granular ice gradually gave way to a less massive and more strongly layered ice facies that was very similar in character to the marine ice imaged on the Amery Ice Shelf (Reference CravenCraven and others, 2005, Reference Craven, Allison, Fricker and Warner2009). This facies was present in all rift boreholes, either on its own or beneath the granular facies (and never above it), and showed an increased prevalence further west, away from the rift tip, as the thickness of the granular ice diminished. Indeed, in some of the most westerly boreholes (e.g. 08-R3), the marine ice facies extended the full thickness of the rift, cropping out at its upper surface (Fig. 5). Although this facies appeared to be homogeneous in OPTV images, containing no notable bubble-defined layering, it was characterized by a green hue, particularly under transmitted light near the surface. We interpreted this as due to the presence of a low concentration of chlorophyll-bearing marine algae, supported by occasional observations of dense patches of algae in core sections recovered from this facies. Such algae were not observed within the overlying granular ice. Although not clearly imaged by OPTV, core sections of marine ice did reveal the development of a strong sub-horizontal crystal alignment with depth, giving the facies a fissile texture.

Fig. 5. OPTV log of the full length of rift core 08-R3 with (progressing left to right) the raw OPTV image, the rolled OPTV image and annotations. Note that no solid core was retrieved from below 13.26m depth where unconsolidated platelet ice was encountered. Although not visible in the OPTV log, the ‘core base’ at 13.26m marks the point at which no more core could be recovered by mechanical coring. An annotated expansion (represented in three dimensions as a virtual core) of the unconsolidated platelet ice forming the lowermost part of this log is presented in Figure 8.

One further notable aspect of this facies is that it contained sinuous tubular channels, similar in appearance to large worm-holes (Fig. 6). These tubes were intersected by our cores on several occasions (e.g. 10-R3) and are typically 1–3cm in diameter and appear to have no preferential orientation. Unfortunately, no OPTV log was recovered from a borehole segment intersecting such a tube, principally because most of the rift holes became blocked with buoyant ice platelets before OPTV logging was possible. However, the tubes were observed directly in the recovered core sections (Fig. 6).

Fig. 6. Photographic image of a typical worm-hole-like tubular conduit located within core 10-R3. The core is retained within a wooden core holder and the tube extends approximately vertically along the top of the core. The tube is ~25mm in diameter.

Ice platelets

Towards the base of the rift cores, the marine ice became so fragile that it formed only a weakly consolidated mass of thin platy crystals (Fig. 7). Below this point, the crystals were effectively unconsolidated and samples could no longer be retrieved by traditional coring (since the retaining core dogs could no longer hold the unconsolidated slurry within the barrel), effectively forming a sea-water-saturated ‘mushy layer’ (Reference Feltham, Untersteiner, Wettlaufer and WorsterFeltham and others, 2006). This transition marked the indistinct interface between the base of the solid ice shelf and the platelet-rich uppermost layers of the underlying sea water. One OPTV log, recovered from 08-R3 (Fig. 5 and 8), did extend for ~4m below the point at which solid core was no longer retrievable, thereby presumably penetrating the uppermost layers of the platelet-rich cavity. This lowermost section of the OPTV log of 08-R3 revealed the presence of unconsolidated material that was characterized by small-scale sub-horizontal wavy layering (Fig. 8), strikingly similar in appearance to that imaged by directional video at the base of the Amery Ice Shelf (fig. 13 in Reference CravenCraven and others, 2005). These layers were repeated every few centimetres throughout the facies. Unfortunately, OPTV images could not be obtained from deeper into this facies because the buoyancy of the unconsolidated mass prevented further OPTV probe penetration.

Fig. 7. Photographic image of a sample of highly flaky marine ice just above the point at which it becomes completely unconsolidated and can no longer be retrieved by coring.

Fig. 8. Rolled OPTV virtual core image of a 1 m interval (15–16 m) near the base of core 08-R3 (Fig. 5): (a) actual OPTV image, (b) interpretative sketch. Brighter light reflected off repeated irregular sub-horizontal layering is interpreted as marking the boundary of layers of aggregated sub-ice-shelf platelets.

Ice-shelf boreholes

The ice-shelf cores (08-S1 and 10-S1) both began in surface snow. However, while the former terminated in firn, the latter (being >66m long) penetrated through the firn and into the underlying ice. Three specific characteristics of the OPTV images retrieved from 10-S1 were reported: (1) regular layering by dark bands, (2) intermittent layering by thicker dark bands and (3) a general down-core decrease in luminosity. We interpret the first of these, the ‘primary’ layering, as annual layers similar to those previously identified on the basis of directional video by for example Reference Hawley, Waddington, Alley and TaylorHawley and others (2003) and Reference Hawley and MorrisHawley and Morris (2006). In such logs, the darker zones mark the more melt-influenced icy layers formed during the summer, and the lighter zones the colder winter accumulation. The summer layer spacing of ~1.0m near the surface of 10-S1 accords with regional mass-balance approximations which indicate an accumulation of ~0.3mw.e. a^–1 (Reference Van de Berg, Van den Broeke, Reijmer and Van MeijgaardVan de Berg and others, 2006). However, this primary layering was frequently disrupted by the second layer type, which was darker (indicating less reflected light, consistent with bubble-poor ice) and had sharper boundaries with the matrix material. These characteristics, along with the intermittent occurrence of this ‘secondary’ layering along the borehole, are consistent with an interpretation as refrozen surface melt layers. Although comparison with the spacing of annual layers (below) suggests that these melt events do not occur each summer, they do appear throughout the full borehole length. The presence of these secondary melt layers also makes it difficult to derive an age–depth relationship for 10-S1 by primary-layer counting. However, the age range of this borehole may be approximated by interpolating primary-layer spacing from zones that are devoid of melt layers (e.g. 5–9, 17–19, 22–25, 36–40, 50–51 and 56–60m in Fig. 3), suggesting that the record extends back for ~150 years from present. Although the frequent melt layers disrupt the core’s potential to provide an undisturbed palaeoenvironmental record, they do provide independent information on the scale and timing of major surface melting events and directly affect only a minor proportion of the core’s total length. Palaeoenvironmental reconstructions should therefore still be possible from cores such as this, as long as the physical influence of melting events is isolated and removed from the analysis. Finally, we interpret the general down-hole decrease in the luminosity of the OPTV image of 10-S1 (at a rate of 3.5 unitsm^–1 depth, averaged linearly over the full 66m borehole length; Fig. 3) in terms of progressive firnification, specifically the gradual isolation and coalescence of bubbles resulting in a net increase in the optical transmissivity (and concomitant decrease in reflectivity) of the borehole wall.

Rift boreholes

The material imaged in the rift tip (08-T1) was similar to that forming the matrix of 10-S1. However, in contrast to the OPTV log of 10-S1, that of 08-T1 is uniform and includes only three slightly darker layers that are each only a few centimetres thick. Further, luminosity does not decrease measurably with depth along 08-T1. We interpret these properties, in association with the location of the borehole in the rift tip, as indicative of infilling by blown snow (as opposed to surface firnification on the ice shelf; see above). This process, described by Reference Leonard, Tremblay, MacAyeal and JacobsLeonard and others (2008) for a rift on the Ross Ice Shelf, would result in little or no seasonal signal, while the possible rapidity of aeolian infilling may also explain the absence of firnification-related bubble nucleation with depth.

Within the rift proper, we interpret the uppermost granular ice facies as snow and firn that have become saturated by percolating saline water, causing the observed grain rounding. This interpretation is also consistent with the general thinning of this facies westwards, away from the apex of the rift, where the rate of surface accumulation by snow trapping is expected to be lower. Progressing westwards and with depth into the base of the rift, this facies gives way to the marine ice facies. All of the properties of this facies, and in particular its occasional high algal content and gradual disintegration into unconsolidated platelets at the base of the rift boreholes, suggest formation by the progressive accumulation and compaction of buoyant platelets formed within the sub-ice-shelf cavity, consistent with previous interpretations. The fact that these platelets rose up the water column rapidly in completed boreholes, commonly preventing OPTV access within some minutes of drilling penetrating the cavity (08-R3 being the sole exception in this study), indicates that the platelets were both buoyant and mobile. The ubiquitous presence of this facies within our rift boreholes, typically to a thickness of some metres to tens of metres, also indicates that platelet marine ice formed throughout the rift.

Finally, the unique observations of worm-hole-like tubes within the granular ice and marine ice facies provide clear evidence of conduit-based water flow through the lowermost layers of the rift melange. We assume that these conduits are formed during the early stages of ice formation when the material is still erodible and then become frozen-in as the material consolidates and (possibly) de-waters. It is therefore apparent that this marine ice is highly porous and permeable, consistent with previous indications (e.g. Reference CravenCraven and others, 2005). One characteristic of marine ice that remains poorly understood is its very low measured bulk salinity (typically 0.05–0.5%; e.g. Reference Tison, Khazendar and RoulinTison and others, 2001) relative to that measured in sea ice (typically 1–20%; Reference WeeksWeeks, 2010), despite both ice types being formed by the aggregation and consolidation of frazil ice frozen from sea water. Reference Eicken, Oerter, Miller, Graf and KipfstuhlEicken and others (1994) argued that the standard mechanisms of post-formational desalination proposed for sea ice are in fact insufficiently effective to reduce the salinity of marine ice to measured values. Instead, these authors tentatively invoked a mechanism of salt-water expulsion from already low-salinity platelets during buoyancy-driven aggregation and densification. Reference TabrahamTabraham (1998) introduced the ‘mushy layer’ concept into a solidification model for marine ice. In this approach, relatively recently proposed to describe the desalination process in sea ice, convective movements in the interstitial liquid are driven by density instabilities due to salinity gradients in a temperature field increasing downwards. This leads to the development of convective ‘chimneys’ known as ‘brine channels’ in sea ice, exporting salts from the mushy layer to the ocean below. However, Reference TabrahamTabraham (1998) also recognized that, although the model of drainage through channels produced some desalination in marine ice, the amount of desalination was found to be less than the levels observed within actual marine ice, the main problem being shutdown of the flow with continuing solidification. The author suggested that combining mushy-layer desalination with compaction might be sufficiently efficient to reach the observed low salinity values. Alternatively, Reference Tison, Khazendar and RoulinTison and others (2001) showed that by treating marine ice as a two-phase compound (pure frazil ice crystals in a consolidating interstitial fluid) and applying a boundary-layer model for consolidation of the interstitial liquid, marine ice salinities could be reproduced if fractionation coefficients derived from the solidification of freshwater ice were used rather than those for columnar (dendritic skeletal layer) sea-water ice.

The 0.05–0.5% salinity range discussed above has generally been observed in thick (102–10³ m) marine ice layers, and to the best of our knowledge no internal desalination chimneys (either active or relict) have been described in such layers. The marine ice salinity range in our RBIS dataset covers a range more typical of sea ice (<9% in the lower layers), with very low salinities (0.1%) only being measured in the upper few metres over a total thickness of a maximum of 10–20m (Pattyn and others, unpublished information). The occurrence of the wormhole-like tubes in the less consolidated lower layers could therefore represent the signature of mushy-layer-like convection processes in the early stages of consolidation of the marine ice layer. The fact that these tubes lack the typical vertical tree-like structure of sea-ice brine channels might reflect the geometrical control of the sub-horizontal accumulation pattern of the loose large frazil ice platelets. An alternative, or supplementary, hypothesis is that once the marine ice layer is formed, salt would continue to be expelled from the layer by freezing-front rejection accompanying ongoing recrystallization within the layer. The removal of that salt-rich water through an effective internal drainage system would then decrease the bulk salinity of the remaining marine ice. Such a process would also involve the delivery of relatively high-salinity water to the underlying water column.