Marine-based ice sheet stability is strongly influenced by perturbations near the grounding line; the most downstream location ice is in contact with the underlying bed (e.g. Schoof, 2011; Robel et al., 2014). The grounding line position is fundamentally determined by ice thickness relative to water depth, where ice is sufficiently thick to overcome buoyancy (Fig. 1a), and where ice thickness in turn is determined by mass balance at the grounding line. A broad suite of processes and conditions that locally dictate both buoyancy and mass balance make it difficult to reliably distinguish and define grounding line positions as “stable” versus “unstable”. Yet predicting how ice sheet sectors will respond to their grounding lines being dislodged by enhanced melt or rising sea level under future warming scenarios or, conversely, how grounding lines will respond to changes in interior ice flow behaviour is an urgent endeavour.

The flux of ice to grounding lines is highly spatially variable, determined by the overall flow structure of the ice sheet (Bamber et al., 2000; Rignot et al., 2011), its basal thermal regime (Kleman and Glasser, 2007), basal slipperiness due to the distribution and style of meltwater drainage (Stearns et al., 2008), cyclic responses of subglacial till rheology to tides (Doake et al., 2002; Anandakrishnan et al., 2003; Gudmundsson, 2007), and the effects of ice shelf buttressing (Rignot et al., 2008; Hulbe et al., 2008). Mass loss occurs by calving and by submarine melting of the ice front and ice shelf, the balance between which can vary enormously, with orders-of-magnitude variability in melt rates (Depoorter et al., 2013; Rignot et al., 2013). Ocean-driven basal melting of ice shelves is thought to be concentrated near grounding lines (e.g. Jenkins and Doake, 1991; Rignot and Jacobs, 2002), and channelised subglacial freshwater emanating at grounding lines can lead to locally enhanced ice shelf melting (Le Brocq et al., 2013; Marsh et al., 2016). While the magnitude of these processes and changes therein may predispose an ice sheet grounding line to advance or retreat, the position of the grounding line – and, one might expect, the duration with which it is held – is dictated by the buoyancy of ice. Since water depth is a primary control, grounding line position may be sensitive to sea level change (Thomas and Bentley, 1978; Schoof, 2007; Katz and Worster, 2010), the modulating effects of glacial isostatic adjustment (Gomez et al., 2010), and bed topography – in the form of either antecedent topography (e.g. Jenkins et al., 2010; Matsuoka et al., 2015; Halberstadt et al., 2016) or the sedimentary construction of relief at the grounding line (Anderson, 1999; Alley et al., 2007).

In the last decade, observations and measurements from direct access as well as from remote sensing and geophysical data have helped characterise contemporary grounding line environments and the processes acting at the time of observation. At the Whillans Ice Stream grounding line, one of the best studied contemporary grounding lines, a grounding zone wedge is actively forming (Anandakrishnan et al., 2007) and processes including channelised meltwater delivery (Horgan et al., 2013), tidally induced compaction of till (Christianson et al., 2013), and basal melt-out of englacial debris (Christianson et al., 2016) are thought to contribute to grounding line dynamics. Observations and modelling results demonstrate coupling between ice shelf change and grounding line movement, indicating that grounding lines are sensitive to ice shelf buttressing (e.g. Shepherd et al., 2004; Goldberg et al., 2009). Longer-term and larger-scale modelling has shown that grounding lines are sensitive to bed geometry and the presence or absence of topographic pinning points (Jamieson et al., 2012). Despite these advances, most observations of grounding line processes and, importantly, the response of the grounding line to those processes, are limited in spatial coverage and relate to timescales of years to decades at best. A comprehensive understanding of grounding line stability and the rates, magnitudes, and timescales of change is therefore precluded.

Grounding line landforms (grounding zone wedges and moraines, Fig. 1b–d) directly mark present and former grounding line positions and represent the history of sedimentation during periods of grounding line position stability. Sediment is transported by glacial and glaciofluvial processes to the grounding line, where it is either deposited and a landform builds or is further transported into the marine environment by sediment plumes. Terminal moraines, here referring to any moraines that form at a grounding line position, are thought to form by a variety of sedimentation processes, including lodgement and deformation of subglacial till; pushing and squeezing of ice-marginal sediments; rockfall, dumping, and melt-out of englacial debris; and glaciofluvial sediment delivery and suspension settling (Powell and Alley, 1997; Batchelor and Dowdeswell, 2015). Grounding zone wedges are rather distinct landforms with an asymmetric morphology (e.g. Anderson, 1999; Anderson and Jakobsson, 2016; Batchelor and Dowdeswell, 2015). Previously described as till tongues (King et al., 1991), till deltas (Alley et al., 1987, 1989), and diamict aprons (Hambrey et al., 1991; Eittreim et al., 1995), grounding zone wedges are composed of prograding strata of dilatant deforming till (King, 1993; Powell and Alley, 1997; Anderson, 1999; Dowdeswell and Fugelli, 2012; Batchelor and Dowdeswell, 2015; Simkins et al., 2017a). Both types of grounding line landforms have been observed to contain features described as grounding line fans: lobate or bulbous deposits building from a point source and linked to both glaciofluvial deposition at the mouth of a subglacial channel and re-mobilisation of grounding line sediments by gravity flows (Powell and Alley, 1997; Bjarnadóttir et al., 2013).

It should follow that if different sets of processes build contrasting landforms, then the landforms themselves can be used to infer conditions and processes operating at palaeo-grounding lines. However, a consistent view of what fundamentally controls why one landform type is produced rather than another is lacking. Whereas terminal moraines are observed in both marine and terrestrial settings, grounding zone wedges are only documented in marine settings largely associated with fast-flowing ice in cross-shelf troughs and fjords (e.g. Batchelor and Dowdeswell, 2015). The presence of an ice shelf is argued by some to be critical to the production of grounding zone wedges (e.g. King, 1993; Dowdeswell and Fugelli, 2012; Batchelor and Dowdeswell, 2015). The restricted vertical accommodation space underneath an ice shelf accounts for the asymmetric wedge morphology and promotes growth by progradation; conversely, a moraine ridge can build at an ice cliff terminus where its vertical growth is unrestricted. Powell and Alley (1997) argue that an ice shelf is not critical, but rather the subglacial thermal and hydrological regimes and their effects on the mode of sediment delivery control terminal landform development are. Dilatant deforming sediment, the production of which is encouraged by meltwater drainage through sediment pore space (Darcian processes), has a low angle of repose and will build a low-relief wedge irrespective of accommodation space. Where meltwater is instead in high abundance and drains through a channelised system, subglacial sediments are less easily deformed and sediment delivery to the grounding line via till deformation may decrease (Simkins et al., 2017b). Where grounding line sedimentation is not dominated by the transport of dilatant till, terminal moraines and fans may build with a higher angle of repose. Bjarnadóttir et al. (2013) challenge this meltwater–sediment delivery model for grounding line landforms, reporting observations of meltwater fans (channelised meltwater) within grounding zone wedges (distributed meltwater). However, in all these cases net addition of sediment to the grounding line implies that the size of an eventual landform will reflect a combination of sediment flux/accumulation and time and should therefore provide some measure of grounding line “stability”.

Enhanced coverage and resolution of bathymetric data (e.g. multibeam sonar) acquired over the last 10–15 years from numerous continental shelf and ice sheet settings reveal vast swathes of grounding line landforms. These provide a wealth of data on grounding line retreat following the last glacial maximum and offer an opportunity to extract information about grounding line processes and sensitivity across a range of glaciological, topographic, and oceanographic settings. Here we characterise morphological traits and the spatial distribution of 6275 grounding line landforms from the western Ross Sea continental shelf, formerly occupied by a marine-based sector of the East Antarctic Ice Sheet, to characterise landform morphology, examine those factors that control landform morphology and distribution, and explore drivers of grounding line stability and instability.

Multibeam bathymetry was collected on cruise NBP1502A aboard the RV Nathaniel B. Palmer using a Kongsberg EM 122 system in dual-swath mode with a 1∘×1∘ array and 12 kHz frequency, surveying with approximately 30 %–60 % swath overlap and with regular sound velocity control. At this frequency, the system vertical resolution is on the order of centimetres and, in 600 m water depths typical of our study area, horizontal resolution is ∼6 m (following Jakobsson et al., 2016). The NBP1502A data were cleaned and gridded at 20 m cell size and combined with re-processed legacy multibeam data from the Lamont-Doherty Earth Observatory, Columbia University Marine Geoscience Data System archive at www.marine-geo.org (last access: 9 August 2018) at a grid cell size of 20–40 m, depending on the resolution of the original component datasets. Sub-bottom acoustic data were collected with a Knudsen chirp 3260 system during cruise NBP1502A using a frequency of 3.5 kHz and a 0.25 ms pulse length. Two-way-travel time was converted to depth using a sound velocity of 1500 m s-1 and to sediment thickness using velocities of 1500–1750 m s-1.

Grounding line landforms were mapped based on visual identification and interpretation (Fig. 2). Morphometric parameters of individual landforms were calculated using standard line geometry tools in ArcGIS and a peak picking function in MATLAB from transects across grounding line landforms (Fig. 2). We explore correspondence among morphometry, landform distribution, topography, and sediment distribution. Analyses are detailed in the Supplement (Sect. S1).

Using high-resolution multibeam bathymetry data, we mapped 6275 grounding line landforms that visually present two distinct populations (Fig. 2): quasilinear, closely spaced, symmetric ridges interpreted as moraines (Fig. 3a–d; n= 4586) and asymmetric ridges with a smeared appearance, interpreted as grounding zone wedges (Fig. 3e–g; n= 1689). Whereas grounding zone wedges are occasionally overprinted by glacial lineations (Fig. 3e), glacial lineations never overprint the mapped moraines nor terminate at the proximal (i.e. subglacial) side of the moraine ridges (Fig. 3a–b). In all cases, moraines occur in fields recording a retreat sequence; therefore we refer to them as recessional features. Amid one field of recessional moraines, we observe a group of irregular ridges with variable amplitudes and orientations that both cut across and partially align with the moraines (Fig. 3h–i; n= 189). We interpret these as basal crevasse-squeeze ridges (e.g. Ottesen and Dowdeswell, 2006; Evans et al., 2016) likely formed in a subglacial, yet near-grounding line, setting. Although crevasse-squeeze ridges are commonly associated with land-based surging glaciers (e.g. Evans and Rea, 1999), they are also observed in palaeo-ice stream troughs and inferred to indicate phases of ice stagnation or break-up (e.g. Rüther et al., 2013; Klages et al., 2015; Greenwood et al., 2017).

Morphological analyses show that as a population, landforms interpreted as recessional moraines are low amplitude (μ=2.0 m, SD = 1.2), are narrow in the cross profile (i.e. along ice flow) direction (μ=83 m, SD = 39.1), are spaced typically less than 1 km apart (μ=419 m, SD = 328), tend towards a symmetric cross profile, and have a straight form (Fig. 4a). Landforms classified as grounding zone wedges are typically higher amplitude (μ=6.2 m, SD = 8.0), wider in the cross-profile direction (μ=522 m, SD = 724), more widely spaced (μ=2100 m, SD = 3430), asymmetric, and relatively sinuous (Fig. 4b). Among the whole population of landforms, width is found to scale with amplitude (Fig. 5a), a trait that is consistent with grounding line landforms reported from other marine-terminating ice sheet settings worldwide (Fig. 5b–c). A notable trend, however, is that western Ross Sea grounding line landforms have smaller widths and amplitudes than most landforms observed elsewhere, with western Ross Sea grounding zone wedges being the smallest documented grounding zone wedges (Fig. 5b) and western Ross Sea recessional moraines overlapping with landforms elsewhere identified as de Geer moraines (Fig. 5a; Ojala et al., 2015; Todd, 2016).

We additionally find that the larger the two-dimensional form of the landform, the greater asymmetry it has developed (Fig. 5d), while Fig. 5e, f illustrate that these properties are also correlated with landform sinuosity. Grounding zone wedges in general are found to be more variable in size, sinuosity, and asymmetry compared to the tight distributions and consistent form of recessional moraines. Individual morphometric parameters show overlapping distributions and imply a continuity of form between recessional moraines and grounding zone wedges (Fig. 4). However, a more holistic description that considers two or more landform properties (Fig. 5) tends to separate the landform population into two groups, consistent with visual interpretation of two distinct landform types. Grounding zone wedges and recessional moraines occur within clusters of numerous landforms of the same morphotype, which abruptly transition from one morphological endmember to another both laterally across a single time-synchronous grounding line (Fig. 6a–c) and within a grounding line retreat sequence (Fig. 6d).

Our observations point to variability in grounding line processes and environments that can lead to a spatial (lateral) and/or temporal switch between two distinctly different landform products. We now ask what grounding line settings or processes may control the production of contrasting landforms and what, consequently, can we learn from the style and distribution of grounding line landforms about the (in)stability of a retreating ice sheet?

A state of grounding in marine settings is fundamentally a function of ice thickness and water depth. A range of glaciological and oceanic processes and topographic settings can affect this relationship (Fig. 1a) and, one may hypothesise, also affect the landform product of grounding. Bed topography has a direct control on the relationship between ice thickness and water depth, and thus the grounding line position. Topography also exerts an indirect control on grounding line processes by creating variability in ice flow velocity and basal sediment flux, influencing tides and near-grounding line ocean circulation, and thereby affecting both the tendency towards buoyancy and grounding line mass balance. Grounding line sedimentation, importantly, serves to build relief at the grounding line, which has been identified as a potential feedback on grounding line position stability (e.g. Alley et al., 2007; Christianson et al., 2016). Processes at the ice–bed interface determine subglacial sediment transport mechanisms and fluxes, and basal and near-grounding line sedimentary processes have therefore been considered fundamental to the production of different grounding line landforms (e.g. Powell and Alley, 1997; Bjarnadóttir et al., 2013). Finally, the presence or absence of an ice shelf exerts a major control on ice flow by providing back stress to grounded ice (Fig. 1a; e.g. Scambos et al., 2004; Fürst et al., 2016), affects mass balance at the grounding line via effects on submarine melting and on calving rate, and places a limit on sediment accommodation at the grounding line.

We use our dataset to evaluate three groups of potential controls on grounding line landform morphology: (i) topographic setting, (ii) grounding line sedimentation, and (iii) presence or absence of an ice shelf.

Grounding line stability can be conceptualised in numerous ways:

sensitivity to change in position (e.g. likelihood of a grounding line response to certain forcings);

Given these different facets of stability, should we define, for example, retreat events of small (large) magnitude punctuated by short (long) periods of grounding line position occupation as stable or unstable retreat? We find from our data that we must consider stability as a multifaceted concept.

Grounding line landforms have the potential to inform us of the processes governing the stability and retreat of palaeo-ice sheet grounding lines. From a large dataset of mapped grounding line landforms, individual morphometric properties indicate a continuum of form. However, multi-parameter analyses support a visual classification of a binary landform product that expresses lateral (i.e. along a single grounding line) and temporal (i.e. within a retreat sequence) transitions between clusters of two endmember morphotypes: moraines and grounding zone wedges. It is an appealing idea that a different set of controls and/or processes should dictate the formation of two different landform types. Yet, of the potential controls on landform morphology that we have explored here, we find inconclusive evidence that a distinct set of controls/processes can wholly explain the formation of either morphotype.

Landform morphotype is not fundamentally controlled by water depth or bed slope, although grounding zone wedges are observed on isolated pinning points likely associated with locations of enhanced grounding line position stability. Neither can the presence or absence of an ice shelf be convincingly demonstrated to control the type of landform that results. Inconsistent spatial arrangements of moraines and grounding zone wedges with respect to topography are difficult to reconcile with plausible ice shelf–ice cliff configurations, and the greater amplitude of grounding zone wedges than moraines suggests vertical accommodation space does not dictate landform morphology. This argument does not reject an ice shelf–cliff control, but additional factors are required to limit moraine growth in this setting.

We find that both sediment supply to the grounding line and the duration of grounding line position occupation are important, most notably expressed in cases in which grounding zone wedges laterally transition to recessional moraines: grounding zone wedges represent both a higher basal sediment flux and a longer duration of grounding than do recessional moraines. This is consistent with the development of landform shape (asymmetry and sinuosity) with the size of the landform. A tempting conclusion is that given sufficient time and supply, a moraine would seed and develop into a wedge. This remains, however, difficult to reconcile with larger terminal moraines in other glaciated settings.

With this large dataset of morphological features associated with palaeo-grounding lines that progress tens to hundreds of kilometres in the retreat direction, we are able to explore what landforms reveal about grounding line stability. Recessional moraines are associated with short-lived grounding line positions yet record steady, small-magnitude retreat events. This suggests that a regular process drives grounding line retreat, linked to steady and cyclic net loss of mass. While grounding zone wedges represent longer periods of position stability, the magnitude of retreat events is larger and more variable. Reduced ablation or a grounding line buffered against cyclic ablation processes may prolong grounding line occupation, while sediment aggradation and progradation in wedge growth may independently enhance grounding line stability. This stable phase is reflected as asymmetry in landform morphology and in lineations on the wedge topset. However, some threshold of stability is reached to result in large unstable retreat events. The development of landform sinuosity due to spatial variability in sediment transport to and deposition rates at grounding lines could potentially destabilise otherwise stable grounding lines. In this regard, channelised meltwater delivery to the grounding line, ice sheet–shelf configuration, and the access of ocean water to the grounding line are likely of fundamental importance in governing grounding line shape and therefore ultimate stability.

Grounding line retreat in the western Ross Sea is characterised by either (i) short-lived grounding line positions that backstep with small-magnitude retreat events or (ii) longer-duration grounding line positions followed by major destabilisation in the form of larger-magnitude retreat events (Fig. 13a, b). These contrasting behaviours vary abruptly in space and time, yet neither can be explicitly characterised as slow or fast retreat, nor can a single descriptor as stable or unstable be applied without further qualification. Stability may be conceptualised in numerous and sometimes contradictory ways. Bart et al. (2017) describe prolonged grounding line occupation and large-magnitude retreat as a paradox; here we find this is a common trait of grounding line behaviour. Given non-uniform ablation and non-uniform sediment supply to a prograding landform at the grounding line, an ice margin may, over time, become increasingly prone to unstable (large-magnitude) retreat. This study highlights the importance of understanding thresholds – potentially in the grounding line sedimentation system itself – which may destabilise a system from an apparent state of stability, and of controls on grounding line dynamics on short (annual) to long (centennial to millennial) scales in order to project future changes in ice sheet mass balance.