Basal channel characteristics and formation

Basal channels, observed in satellite imagery over ice shelves throughout Antarctica and on some floating ice tongues in Greenland, are typically ~1–3 km in width, ~50–400 m in height from the base of the ice shelf to the channel apex, and a few to many tens of kilometers in length in the approximate along-ice-flow direction, based on observations from MODIS, Landsat, ice-penetrating radar (e.g. Rignot and Steffen, Reference Rignot and Steffen2008; Stanton and others, Reference Stanton2013; Alley and others, Reference Alley, Scambos, Siegfried and Fricker2016) and the Reference Elevation Model of Antarctica (Alley and others, Reference Alley, Scambos, Alley and Holschuh2019; Howat and others, Reference Howat, Porter, Smith, Noh and Morin2019; Chartrand and Howat, Reference Chartrand and Howat2020). These relatively large basal channels can be observed in satellite imagery because of the hydrostatic relaxation of the overlying ice, creating surface depressions visible on the ice-shelf surface (Fig. 1). Many ice shelves may also have smaller basal channels, revealed by ice-penetrating radar, that are invisible in satellite imagery because they are narrow enough for bridging stresses to prevent easily detectable hydrostatic adjustment (e.g., Langley and others, Reference Langley2014). The relatively large basal channels observed in satellite imagery are distributed widely around the Antarctic continent, but are most common on ice shelves with high basal melt rates and especially those that are directly influenced by modified Circumpolar Deep Water (Alley and others, Reference Alley, Scambos, Siegfried and Fricker2016), the water mass responsible for much of the observed recent retreat of the West Antarctic Ice Sheet (e.g., Hillenbrand and others, Reference Hillenbrand2017).

Basal channel formation is driven by the activity of buoyant plumes of water that persist in approximately the same location over time. Many of these plumes are initiated by the outlet of subglacial meltwater into the ice-shelf cavity (Fig. 1a). Though relatively cold, the fresh meltwater is buoyant enough to rise along the ice-shelf base, entraining ocean water that generally is well above the in situ melting temperature and melting a trough in the ice through turbulent heat transfer; many plumes eventually super-cool and lead to freeze-on as the dropping pressure along-flow raises the melting temperature and the ice-shelf base flattens, reducing entrainment (Le Brocq and others, Reference Le Brocq2013; Drews and others, Reference Drews2017). The channels formed by this process, which begin at the grounding line and tend to dissipate downflow, can be referred to as subglacially sourced channels (Alley and others, Reference Alley, Scambos, Siegfried and Fricker2016). At least some of these subglacially sourced channels may be influenced by subglacial lake drainage events (e.g. Whiteford and others, Reference Whiteford, Horgan, Leong and Forbes2022), suggesting that they experience basal melt rates that are highly temporally variable.

In contrast, the heads of many channels are detached from the grounding line, and their formation is inferred to be driven by plumes that are buoyant due to the presence of warm ocean water combined with meltwater entrained as the plume rises along the ice-shelf base (Alley and others, Reference Alley, Scambos, Siegfried and Fricker2016). These ocean-sourced channels (Fig. 1b) are likely to exploit local high spots in the basal topography, which may be imparted by grounding-line irregularities (Gladish and others, Reference Gladish, Holland, Holland and Price2012), suture zones (Dow and others, Reference Dow2018), or by larger-scale patterns related to stretching and thinning of the ice shelf (Alley and others, Reference Alley, Scambos, Alley and Holschuh2019). They may also form spontaneously, as ocean water accelerates up an ice-thickness gradient and encounters an irregularity, which deepens as more water is channelized in a positive feedback (Sergienko, Reference Sergienko2013). Ocean-sourced channels frequently terminate in persistent open-water polynyas at the ice edge, where the outflowing plume has enough buoyancy and heat energy to prevent the formation of fast ice (Bindschadler and others, Reference Bindschadler, Vaughan and Vornberger2011; Mankoff and others, Reference Mankoff, Jacobs, Tulaczyk and Stammerjohn2012).

Grounding-line-sourced channels (Fig. 1c) fall somewhere in-between subglacially sourced and ocean-sourced channels, as they show an apparent intersection with a grounding line but without evidence of coincident subglacial outflow, which may reflect either deficiencies in subglacial hydrologic modeling or warm-ocean influence very close to relatively shallow grounding lines (initiating an ocean-sourced channel at or close to the grounding line, in thin enough ice for hydrostatic adjustment to be visible at the surface). More recent work suggests that there are no sharp boundaries between these channel categories, with many channels of hybrid origin. For example, ocean-sourced channels on Getz Ice Shelf, with a clear visible separation between the grounding line and the channel head and which terminate in persistent polynyas, are in fact likely driven by subglacial meltwater drainage and then enhanced by warm ocean water (Wei and others, Reference Wei2020). Subglacially sourced channels could also appear as ocean-sourced channels if grounding-line retreat outpaces headward channel growth, particularly in areas with very thick ice where hydrostatic relaxation is limited by bridging stresses. Similarly, there is likely a continuum between basal channels and basal fractures. The hydrostatic relaxation that makes channels visible on the surface also imparts stresses that can create fractures at channel apices (Vaughan and others, Reference Vaughan2012), which we discuss in more detail in a later section. In addition, basal fractures form high spots that buoyant plumes may exploit (Fig. 2).

Fig. 2.

An example of basal channel change on Pine Island Glacier. The arrow in the first panel indicates a basal channel, which grows first towards another channel and the ice edge, which it reaches by 1996, and then headward, which is clearly visible by 2017. Retreat occurs along the channel throughout the time series, and transverse rifts can be seen opening from the channel from 1996 onwards. These transverse rifts that initiate from the channel frequently result in calving events and ice front retreat. The grounding line shown in red is from the 2014 MODIS Mosaic of Antarctica (Haran and others, Reference Haran2018). Images come from Landsat 5, 7, and 8.

Basal channel importance in ice-shelf stability

The formation of basal channels influences ice-shelf basal melt rates and patterns, an important variable controlling ice-shelf stability. Fundamentally, basal channels form due to positive feedbacks that concentrate melt in high spots in the ice-shelf base. On ice shelves with a high concentration of basal channels, shelf-average melt rates may decrease with increasing channelization, likely because the concentration of flow in buoyant plumes prevents much of the entrained warm water from reaching the ice-shelf base (Gladish and others, Reference Gladish, Holland, Holland and Price2012; Millgate and others, Reference Millgate, Holland, Jenkins and Johnson2013). The concentrated melting that forms basal channels is opposed by ice flow that seeks to fill the trough through ice creep (Drews, Reference Drews2015; Wearing and others, Reference Wearing, Stevens, Dutrieux and Kingslake2021; Humbert and others, Reference Humbert2022). The fact that channels persist over long distances and many decades, and often even deepen towards the ice edge (Alley and others, Reference Alley, Scambos, Siegfried and Fricker2016), suggests that basal melt in most large channels is sustained, or occurs frequently, and over the long-term equals or exceeds creep closure.

Buoyant plume flow within basal channels is directed primarily towards the ice-shelf edge, which is consistent with large-scale patterns of overturning circulation within ice-shelf cavities that directs buoyant water up the ice-shelf basal slope. This outward flow is also evidenced by the commonly observed formation of persistent polynyas where basal channels meet the ice-shelf edge (Bindschadler and others, Reference Bindschadler, Vaughan and Vornberger2011; Mankoff and others, Reference Mankoff, Jacobs, Tulaczyk and Stammerjohn2012; Alley and others, Reference Alley, Scambos, Siegfried and Fricker2016, Reference Alley, Scambos, Alley and Holschuh2019) and by the results of numerical modeling studies (Gladish and others, Reference Gladish, Holland, Holland and Price2012; Millgate and others, Reference Millgate, Holland, Jenkins and Johnson2013; Sergienko, Reference Sergienko2013). Within basal channels, outward plume flow, concentrated on the Coriolis-favored side of the channel, may be accompanied by an adjacent opposing return flow on the non-Coriolis-favored side (Millgate and others, Reference Millgate, Holland, Jenkins and Johnson2013). Flow velocity is likely to be strongly heterogeneous, with higher plume velocities near the head of the channel where the ice-shelf basal topography tends to be steeper, and on the Coriolis-favored side of the channel (Gladish and others, Reference Gladish, Holland, Holland and Price2012; Millgate and others, Reference Millgate, Holland, Jenkins and Johnson2013; Sergienko, Reference Sergienko2013). In general, melt rates are expected to follow roughly the same pattern as velocities, with the highest melt rates near the grounding line (Rignot and Steffen, Reference Rignot and Steffen2008; Gladish and others, Reference Gladish, Holland, Holland and Price2012; Dutrieux and others, Reference Dutrieux2013; Humbert and others, Reference Humbert2022) where ocean forcing is greatest and the subglacial topography steepest, and on the Coriolis-favored side of the channel (Sergienko, Reference Sergienko2013).

However, basal melt observations suggest a more complex and heterogeneous story. Measurements from autonomous underwater vehicles and radar observations have revealed terraces - relatively flat and nearly horizontal regions extending for long distances along flow, separated by steeper slopes - on the sides of some basal channels (Dutrieux and others, Reference Dutrieux2014), with calculated rates of basal melt highest on basal channel flanks, and lowest at the channel apices. Dutrieux and others (Reference Dutrieux2014) found this pattern in basal channels on both Pine Island Glacier in West Antarctic and Petermann Glacier in northwest Greenland. In contrast, Stanton and others (Reference Stanton2013), working on a different area of Pine Island Glacier, and Rignot and Steffen (Reference Rignot and Steffen2008) on a different spot on Petermann found exactly the opposite pattern: smooth channels, with the highest melt rates found at channel apices. All channels must at some point have high melt at channel apices in order to create the observed troughs. However, as the buoyant plume moving along the channel loses energy with the decrease in ice-shelf basal slope, models suggest the plume may rapidly detrain (Millgate and others, Reference Millgate, Holland, Jenkins and Johnson2013), and a stratified layer near the channel apex may develop (Dutrieux and others, Reference Dutrieux2014). It is, perhaps, at this point that a smooth channel dominated by apex-melt may transition to a terraced channel dominated by flank-melt, but this hypothesis requires much more investigation.

Along with changing patterns of ice-shelf basal melt, basal channels directly impact ice-shelf stability through complex interactions with ice-shelf fracture (see Fig. 3 for a summary diagram). Vaughan and others (Reference Vaughan2012) showed that stresses associated with hydrostatic adjustment can be high enough to induce along-channel (longitudinal) basal fractures at channel apices, and this pattern has also been observed in radar transects (Dutrieux and others, Reference Dutrieux2014). In cases of many basal channels parallel to each other, swarms of surface fractures can form above channel keels as a result of the same stresses. In a pattern likely to be related to the weakening due to channel apex factures, ice shelves are often observed to develop full-thickness, along-channel rifting, and then to retreat along basal channels (e.g. Rignot and Steffen, Reference Rignot and Steffen2008; Alley and others, Reference Alley, Scambos, Siegfried and Fricker2016). An example of this behavior is demonstrated in Figure 2, which shows a time series of basal channel change on Pine Island Glacier. Rifting and retreat along the channel occurs repeatedly, but is particularly clear in the 2020 image.

Fig. 3.

Summary of suggested basal channel research priorities.

Basal channels can also influence the transverse fractures that control ice-shelf calving. Dow and others (Reference Dow2018) observed repeated calving events on Nansen Ice Shelf that occurred along rifts that initiated and extended from the location of a basal channel. This pattern of transverse fractures (either full-thickness rifts or the surface expression of basal fractures) originating from basal channels is commonly observed around Antarctica. For example, the channel on Pine Island Glacier (Fig. 2) is both fracturing along-channel, as discussed above, and generating transverse fractures that open outwards from the channel, which are visible in every panel starting in 1996. It is not yet clear whether the association between basal channels and transverse fractures is related purely to ice dynamics or if pre-existing fractures are being widened by the plumes flowing through the channels. Basal channels form disproportionately often beneath shear margins because of advection of thinned ice from shear margins of grounded ice streams, and these ice-shelf shear-margin channels further localize thinning and weakening (Alley and others, Reference Alley, Scambos, Alley and Holschuh2019). Many large calving and retreat events initiate from these shear margins, and basal channels may play a role in weakening them and initiating these events.

Curiously, the opposite effect also seems to occur in some locations, where a propagating fracture is stopped when it reaches a basal channel. For example, the channel on the Thwaites Eastern Ice Shelf in Figure 1b appears to be arresting the propagation of rifts growing from the top of the image towards the bottom. This effect might help to stabilize an ice shelf, at least temporarily. The process may be similar to a pattern noted on the Larsen C Ice Shelf, where propagating rifts are arrested at suture zones, which contain large thicknesses of relatively soft and warm marine ice (McGrath and others, Reference McGrath2012). Marine ice formation in basal channels has been inferred by thermodynamic conditions (Le Brocq and others, Reference Le Brocq2013) and observed in radar transects (Whiteford and others, Reference Whiteford, Horgan, Leong and Forbes2022), although it has not been observed on ice shelves with relatively warm cavities (e.g., Rignot and Steffen, Reference Rignot and Steffen2008; Stanton and others, Reference Stanton2013). While the characteristics of the ice within and surrounding the channel might be responsible for arresting fracture propagation, it is also possible that the behavior could be explained by a change in ice-shelf stress state due to the presence of a basal channel focusing shear.

Along with these significant influences on ice-shelf stability, basal channels have been observed to change over time, sometimes very rapidly. A basal channel on Getz Ice Shelf grew ~20 km upstream since Landsat observations began in the 1970s (Alley and others, Reference Alley, Scambos, Siegfried and Fricker2016), with incision rates near the head of the channel of up to 22 m/year (Chartrand and Howat, Reference Chartrand and Howat2020). The Pine Island channel in Figure 2 grew downstream to reach the ice edge from its 1986 position, and then grew headward, eventually intersecting with a basal fracture that formed from extensional stresses just downstream of the grounding line. As basal channels are most common in areas where ice-shelf basal melt rates are high and are closely associated with modified Circumpolar Deep Water intrusion in the sub-shelf cavity, it is reasonable to assume that the increase in basal channel length may be tied to increasing access of modified Circumpolar Deep Water to West Antarctic ice shelves in recent decades (e.g., Jacobs and others, Reference Jacobs, Jenkins, Giulivi and Dutrieux2011). Basal channels have also been observed to migrate laterally across ice shelves, independent of ice flow (Chartrand and Howat, Reference Chartrand and Howat2020; Drews and others, Reference Drews2020).

Future research priorities

Based on these observations of basal channels, their impacts, and their evolution, we suggest three priorities for future research, with components summarized in Figure 3. The first research priority we suggest is to quantify patterns and rates of basal channel change, with further consideration of how channels form and migrate. Basal channel change has been observed in all dimensions (represented in Fig. 3) and many mapped basal channels (Alley and others, Reference Alley, Scambos, Siegfried and Fricker2016) have not been analyzed in any dimension over time. It is very difficult to draw conclusions about large-scale basal channel change from the few isolated studies that have been carried out. Important remaining questions include: Are most basal channels in steady-state, or is change common? Is basal channel change widespread throughout Antarctica, or concentrated in certain areas? Are changes brief or sustained in time? Many available remote sensing datasets provide valuable information to answer these questions. Long-term optical time series, such as those available from Landsat, can be used to track lateral migration and changes in channel length. High-resolution DEMs and altimetry missions constrain channel deepening and changes in cross-sectional profiles. Single profiles from ground-penetrating radar record past channel migration in the imaged layers, and repeat profiles reveal details of more recent change.

Perhaps a harder challenge is to constrain the styles and patterns of basal melt associated with basal channels. As described above, some basal channels have been observed to develop terraces (Dutrieux and others, Reference Dutrieux2014), while others are smooth (Rignot and Steffen, Reference Rignot and Steffen2008; Stanton and others, Reference Stanton2013). Along-flow melt variability has been noted in calculated basal melt rates (Rignot and Steffen, Reference Rignot and Steffen2008; Dutrieux and others, Reference Dutrieux2013) and numerical models (Gladish and others, Reference Gladish, Holland, Holland and Price2012; Millgate and others, Reference Millgate, Holland, Jenkins and Johnson2013; Sergienko, Reference Sergienko2013). Across-channel melt variability has shown contradictory signals, with some sections of channels melting fastest at the apex (Stanton and others, Reference Stanton2013) and others on their flanks (Dutrieux and others, Reference Dutrieux2014). Persistent polynyas at the ends of basal channels give clues to the thermal forcing and temporal variability of the plumes flowing through basal channels (Bindschadler and others, Reference Bindschadler, Vaughan and Vornberger2011). Big-picture questions remain about what controls the strength, temperature, and variability of basal channel plumes, and how to reconcile the varying observations, which suggest significant spatial and/or temporal heterogeneity in basal melt patterns. While some of these questions might be answered through detailed analysis of polynya surface temperatures and sizes, most will require in situ measurements and numerical modeling. We need better constraints on basal channel geometry using ice-penetrating radar and sub-ice-shelf multibeam data, and point measurements of basal melt from phase-sensitive radar. It is also important to collect more direct measurements of plume flow in basal channels, using autonomous underwater vehicles as well as oceanographic data from instruments lowered through ice-shelf boreholes.

Finally, perhaps the most consequential basal channel research priority for understanding future ice-shelf stability and impacts on sea-level rise is to constrain the relationship between basal channels and ice-shelf fracture and calving. Fracturing has been observed to propagate both longitudinally, along basal channels, and transversely, away from basal channels, both of which can result in accelerated calving and ice-shelf retreat (Dow and others, Reference Dow2018; Alley and others, Reference Alley, Scambos, Alley and Holschuh2019; Figure 2). Channels preferentially form in already-weak shear margins, where many calving events initiate (Alley and others, Reference Alley, Scambos, Alley and Holschuh2019). Conversely, channels may arrest the propagation of fractures, in a manner similar to suture zones filled with marine ice (McGrath and others, Reference McGrath2012). We need to investigate the magnitude of change to ice-shelf strength and dynamically induced stresses that is imparted by basal channel formation, and combine these results with observations of channel change to predict future ice-shelf behavior. Much of this work will be ultimately underpinned by numerical modeling, a daunting task as ice-shelf fracture is not well-represented in most numerical models. High-temporal-resolution records from optical and radar imagery will also be important to track the details of fracture propagation and calving patterns.

Basal channels are complex systems influenced by ice creep, advection, and ocean control. We have enough information already to understand that basal channels have significant impacts on ice-shelf basal melt and structural stability, and we know that they can change significantly and rapidly. Much work remains to constrain in detail the influence that basal channels will have as the world warms and ice shelves evolve in the future.