In the austral summer of 2016/2017, the British Antarctic Survey Polarimetric-radar Airborne Science Instrument (PASIN2) ice sounding radar system was used to acquire ∼ 3000 line km of radio-echo sounding (RES) data along the English Coast of western Palmer Land, at ∼ 3–5 km line spacing (Corr and Robinson, 2020). PASIN2 operates at a frequency of 150 MHz, using a pulse-coded waveform at an effective acquisition rate of 312.5 Hz and a bandwidth of 13 MHz. Technical details of the RES system are available in Corr et al. (2007). Differential GPS was used to record aircraft position (with an accuracy better than ±1 m) and RES data were collected at an average flying velocity of 55 m s-1. Along-track processing of the data results in an output data rate of 5 Hz, which produces an average spacing between radar traces of 11 m. Section 7 details the information we extract from the online data repository for use in this paper.

For the processing of the data, a coherent moving-average filter, commonly referred to as an unfocused SAR, was used on the range compressed data. The onset of the bed reflector was first automatically picked using first-break picker of the ProMAX (version 5000.10.0.0; Landmark Software and Services) seismic processing software with all picks then checked afterwards and corrected by hand if necessary. The delay time of the bed reflector picks were covered to range using a standard electromagnetic wave propagation speed in ice of 0.168 m ns-1 and a correction of 10 m to account for the near-surface high-velocity firn layer (Dowdeswell and Evans 2004; Vaughan et al., 2006). Ice thickness was calculated by subtracting surface elevation measurements (derived from radar/laser altimeters for aircraft terrain clearance) from bed reflector depth picks. Internal crossover analysis (measurements of ice thickness at the same position) yield a standard deviation of 13 m at line intersections, with no systematic line-to line biases. Independent crossover analysis, with NASA's airborne Operation IceBridge (OIB) radar data (Paden et al., 2010) (collected from November 2010–November 2016), yields a higher standard deviation of 48 m (when high-elevation OIB flights are removed from analysis). As this standard deviation is skewed by a relatively small number of high crossover misfits over steep subglacial topography (where the outlet ice flows are located), we use the internal crossover analysis value of 13 m for our RES errors.

RES transects were visualized in 2D in Reflexw radar processing software (version 7.2.2; Sandmeier Scientific Software) where an energy decay gain was applied to compensate for geometric spreading losses in the radargram (Daniels et al., 2004). OpendTect seismic interpretation software (version 6.4.0; dGB Earth Sciences) was employed to plot radargrams in real space using DGPS co-ordinates, to enable three-dimensional analysis of RES data.

Airborne RES data presented in this paper have been incorporated in the new BedMachine dataset; a self-consistent dataset of the Antarctic Ice Sheet based on conservation of mass, which has a resolution of 500 m (Morlighem, 2019; Morlighem et al., 2019). As a result, data presented in this paper have already been combined with numerous other RES survey data (including OIB data) to produce continent-wide ice thickness and subglacial topography maps (Fig. 2b). Whilst Morlighem et al. (2019) report potential vertical errors of ∼ 100 m in central Palmer Land, these values decrease towards the coast, where RES measurements are more frequent (Morlighem, 2019).

Surface flow speeds are extracted from MEaSUREs phase-based Antarctica ice velocity map which has a resolution of 450 m (Mouginot et al., 2019a) (Fig. 1a). This dataset combines interferometric phases from multiple satellite interferometric synthetic-aperture radar systems, with additional data, including tracking-derived velocity to maximize coverage from 1996 to 2018 (Mouginot et al., 2019b). Across western Palmer Land the average flow speed error is estimated to be less than 4 m a-1.

Using surface flow speeds (Mouginot et al., 2019a) and ice thickness measurements from the 1us radargrams in the online data repository (see Sect. 7), we calculate ice flux across fixed gates delineated for each of the named ice streams and glaciers along the English Coast (Fig. 2). These flux gates are delineated along RES transects immediately upstream of the grounding line and they span the width of each outlet. Ice flux (q) for each ice stream or glacier (j) is calculated following Eq. (1): 1qj=∑i=1nhijwijvijρ,

where i is an equally spaced bin along the length of the flux gate, w is the bin width (which is fixed to 1 m for all outlets and is sufficiently small that the solution is not sensitive to a bin width smaller than this), and v is the velocity normal to the flux gate. We use an ice density value of ρ=917 kg m-3, which is consistent with densities used in both Bedmap2 and BedMachine datasets (Fretwell et al., 2013; Morlighem, 2019). For simplicity, we are assuming that surface velocities and ice density are constant with depth. To examine the impact of incorporating high-resolution RES data into gridded bed topography datasets, we directly compare ice flux from Bedmap2 (Fretwell et al., 2013) (which has a resolution of 1 km) with the radar picks described in Sect. 3.1, which are included in BedMachine (Morlighem, 2019) (Fig. 2c). For these calculations we use the same flux gates, phase-based ice velocities, and ice density, simply replacing RES ice thickness for Bedmap2 ice thickness. Using available errors in velocity and ice thickness datasets, we calculate errors in our calculated ice flux (σq) for each glacier following Eq. (2): 2σq=σv2+σh2, where σv=∑i=1nhijwijdvijρ and σh=∑i=1ndhijwijvijρ are the contribution of errors in velocity (dv) and ice thickness (dh) to the errors in ice flux respectively. Ice flux and associated error bars for each outlet are shown in Fig. 2c.

Close to the grounding line, Ers Ice Stream reaches a maximum flow speed of just over 940 m a-1 (averaging out at ∼ 2.5 m d-1) (Mouginot et al., 2019a). This ice originates from central Palmer Land (Fig. 1), where ice flows across the west of the Antarctic Peninsula, towards Ers Ice Stream. In the upper catchment, flow speeds of ∼ 400 m a-1 (Mouginot et al., 2019a) are recorded along RES transect Ers 6 (Fig. 3a). A succession of airborne RES transects in Fig. 3c show how this fast-flowing ice is channelized towards the coast, through a subglacial depression ∼ 8–14 km wide. As ice flows through this channel, towards the local grounding line (marked in white in Fig. 3a), ice thickness reduces from a maximum of ∼ 1400 m (along transect Ers 6) to between 580 and 610 m (along transect Ers 1), where the ice flow is grounded ∼ 400 m below sea level. Ice flux calculated along this radar transect suggests that Ers Ice Stream contributes over 7.24 ± 0.15 Gt a-1 to George VI Ice Shelf (Fig. 2c). Although this flux gate represents the main trunk of Ers Ice Stream (Fig. 2a, b), neighbouring ice flow from the lateral margins of the ice stream (where ice flows at ∼ 210–390 m a-1) will, of course, add to this value.

A central flow unit, more than 14 km wide, distinguishes Cryosat Ice Stream from neighbouring regions of slower-flowing ice along the English Coast (Fig. 1). Whilst surrounding ice flows at ∼ 100 m a-1, flow speeds in the ice stream range from 400–500 m a-1 inland (along RES transect Cryosat 3) to 950 m a-1 (∼ 2.6 m d-1) (Mouginot et al., 2019a) along RES transect Cryosat 1 – which was traversed close to the Antarctic Surface Accumulation and Ice Discharge (ASAID) grounding line (Bindschadler et al., 2011) (Fig. 4a). Figure 4a shows how the main flow of Cryosat Ice Stream is joined by a smaller tributary to the south, where ice flow speeds increase from 180 m a-1 along transect Cryosat 4 to over 400 m a-1 (Mouginot et al., 2019a) along transect Cryosat 1 (traversed ∼ 5 km from the ASAID grounding line). In both flow units, the subglacial bed remains well below sea level along the length of each transect. Close to the grounding line, along transect Cryosat 1, the glacial bed is between 450 and 800 m below sea level, where overlying ice is 500–900 m thick. Although subglacial topographic depressions are visible in-land (where subglacial peaks which reach over 500 m from the bed help to define the low-elevation topography), subglacial troughs become more defined towards the coast, where ice is guided through several almost U-shaped troughs (Fig. 4b). This is most obvious in RES transect Cryosat 1, where the main flow of ice is channelled through a ∼ 14 km wide, 300 m deep subglacial trough close to the local grounding line, whilst the smaller (southern) tributary flow is directed through a ∼ 400 m deep trough, which is ∼ 3 km wide at its base (Fig. 4b). The flow units of Cryosat Ice Stream collectively discharge 5.99 ± 0.14 Gt a-1 of ice across the grounding line. The ice flux profile in Fig. 4b shows how much of this flux is discharged through the deep and fast-flowing central sector of the ice stream, rather than the deeper southern tributary.

In the MEaSUREs velocity map, Sentinel Ice Stream appears to have the widest outflow of the English Coast, reaching a width of over 20 km. The main trunk of the ice stream curves round from an almost southerly flow direction, to a more westerly direction along its length (Fig. 4c) as ice flow speeds increase from ∼ 350 m a-1 (along RES transect Sentinel 5) to ∼ 800 m a-1 (closer to the grounding zone, along transect Sentinel 1) (Mouginot et al., 2019a). Whilst the subglacial bed remains well below sea level in all transects (at elevations in the region of -500 to -680 m), fluctuations in subglacial topography and ice thickness are recorded along- and down-flow in successive RES transects (Fig. 4d). The largely unconfined ice flow in transect Sentinel 5 becomes more confined down-flow due to the emergence of higher-elevation subglacial topography along the lateral margins of Sentinel Ice Stream. These subsurface conditions are concurrent with ice thickness measurements (where maximum ice thickness decreases down-flow, from ∼ 1200 m in transect Sentinel 5 to ∼ 550 m in Sentinel 1), as well as surface velocity measurements, which reveal increasing flow speeds in the central trunk of Sentinel Ice Stream with distance down-flow. The total flux of Sentinel Ice Stream is 6.01 ± 0.14 Gt a-1. Whilst this flux will be added to by flow from the south (where enhanced flow speeds are recorded, but they are about 3 times slower than the central trunk of Sentinel Ice Stream), there will be much less flux in the north, where ice flows at a few tens of metres per year (Mouginot et al., 2019a) (Fig. 4c), over higher-elevation subglacial topography (∼ 400 m higher than the base of the subglacial trough).

Hall Glacier is the most northern tributary flow of the Stange Ice Shelf (Fig. 1). Surface flow speeds increase from RES transect HNL 6 (close to the onset of streaming flow) – where ice flows just over 100 m a-1, to RES transect HNL 1 (∼ 1.5–9.5 km from the ASAID grounding line and 14 km from HNL 6), where ice flow speeds reach 380 m a-1 (Mouginot et al., 2019a) (Fig. 5a). These enhanced flow speeds clearly differentiate Hall Glacier from the almost stagnant neighbouring ice flow (< 10 m a-1) along its lateral margins in Fig. 5a. This figure shows how the fast-flowing portion of the outlet glacier decreases in width from ∼ 15 km inland to ∼ 8 km along RES transect HNL 2. This reduction in width coincides with a change in subsurface topography and ice thickness (Fig. 5c). Whilst a shallow subglacial depression is apparent upstream, in RES transect HNL 5 (where the subglacial bed is ∼ 500 m below sea level and ice thickness reaches a maximum of 750 m), a much deeper channel is recorded down-flow, where ice up to 930 m thick is channelized through high-elevation subglacial topography. The profile in Fig. 5b (derived from the flux gate marked in Fig. 5a) shows the impact this subglacial topography and ice thickness have on ice flux. Flux is greatest along the central trunk of Hall Glacier where a ∼ 7 km wide subglacial channel supports ice flow speeds of more than 350 m a-1 (Mouginot et al., 2019a). Over the whole flux gate, Hall Glacier contributes ∼ 1.87 ± 0.04 Gt a-1 of ice to the Stange Ice Shelf, which drains into the Bellingshausen Sea sector of the Southern Ocean.

Situated between Hall Glacier and Lidke Ice Stream, Nikitin Glacier maintains flow speeds in the region of 200–450 m a-1 (Mouginot et al., 2019a), as ice flow from central Palmer Land begins to stream towards the Stange Ice Shelf (Fig. 5). For much of its length, Nikitin Glacier flows through a 15 km wide subglacial channel, where ice thicknesses up to 1000 m flow over a glacial bed situated well below sea level (with elevations of -400 to -700 m). This low-elevation subglacial topography combined with thick ice flows and enhanced ice flow speeds enable Nikitin Ice Stream to contribute over 2.13 ± 0.05 Gt a-1 of ice to the Stange Ice Shelf. Whilst it is difficult to precisely define the point at which this ice begins to float in our radargrams, it is worth noting that complex and highly reflective RES returns beneath Nikitin Ice Stream in transect HNL 1 suggest that the ice stream could be afloat here. This finding is coincident with the positioning of the ASAID grounding line (Bindschadler et al., 2011) (marked as a white line in Fig. 5a), which is derived from satellite data.

The MEaSUREs dataset (Mouginot et al., 2019a) presented in Fig. 5a, shows how Lidke Ice Stream is fed by two tributary flows which coalesce close to RES transect HNL 4, where ice begins to flow along a central trunk at flow speeds in the region of 350–420 m a-1 (Fig. 5a). Although Lidke Ice Stream is linked to neighbouring Nikitin Ice Stream in its upper catchment, a clear separation between the two ice streams is recorded down-flow, where the enhanced flow units become separated by a region of almost stagnant ice (< 10 m a-1). RES transects in Fig. 5c show how this slow-moving ice sits on top of relatively high-elevation subglacial topography (with elevations of -380 to -500 m). This raised topography helps to define the northern margin of Lidke Ice Stream, which flows through much lower-elevation subglacial topography, situated ∼ 600–800 m below sea level.

In RES transects HNL 4 and HNL 5 (traversed close to the onset of streaming flow) numerous peaks and troughs dominate the subglacial topography returns, resulting in spatially variable ice thickness and ice flux. However, further down-flow, and closer to the grounding line, subglacial topography is more subdued, with the emergence of a depressed subglacial channel (reaching a maximum depth of 810 m below sea level), where ice up to 1250 m thick achieves surface flow speeds in the region of 400 m a-1 (Mouginot et al., 2019a) at the grounding zone. In a flux gate along HNL3 (marked in Fig. 5a), Lidke Ice Stream is calculated to contribute > 2.71 ± 0.01 Gt a-1 to the Stange Ice Shelf. The flux profile in Fig. 5b shows how this value is distributed across the glacier – with high flux values recorded in areas which have low-elevation subglacial topography, thicker ice, and fast ice flow.

Landsat Ice Stream (situated close to the catchment-defined boundary between the Antarctic Peninsula and West Antarctica) is formed of a northern and southern tributary, with ice flow converging at or close to the ASAID grounding line (Fig. 6). Both tributaries have similar characteristics: they each reach flow speeds in excess of 500 m a-1 in the centre of the ice flow (along RES transect Landsat 3) before flow begins to accelerate downstream (to over 700 m a-1 near transect Landsat 1) (Mouginot et al., 2019a). Between the two tributaries, flow speeds are much lower, ranging from 40 m a-1 (25 km inland of the grounding line) to ∼ 100 m a-1 (along RES transect Landsat 1, traversed close to the local grounding line) (Mouginot et al., 2019a). A sequence of airborne RES transects in Fig. 6c show that these flow speeds reflect subglacial topography. Both tributaries flow through deep subglacial basins (situated ∼ 700 m below sea level), where ice flows up to 900 m thick are increasingly channelized towards the coast by higher subglacial topography along the ice stream's lateral margins. Along RES transect Landsat 2, ice flux gates across the north and south tributary flows combine to produce a total ice flux of 7.23 ± 0.13 Gt a-1. Between these two flow units ice flux is substantially lower, because of lower surface flow speeds, elevated subglacial topography, and reduced ice thickness.

Ers Ice Stream, at the northern extremity of our study site, produces the largest ice flux of all English Coast ice streams (Fig. 2c). This is the result of elevated surface flow speeds (Mouginot et al., 2019a), substantial ice thicknesses, and pronounced subglacial topography, which, for the most part, channelizes ice through a wide subglacial depression (Fig. 3c). Enhanced ice flow is also recorded on either side of the subglacial channel, where surface flow speeds greater than 100 m a-1 (Mouginot et al., 2019a) contribute over 1 × 10-4 Gt of ice to George VI Ice Shelf per year. This enhanced ice flow makes it difficult to precisely map the lateral margins of the ice stream and fully assess the individual contribution of Ers Ice Stream to English Coast ice flux. However, it is clear that this area of the English Coast contributes substantial and continued ice flux to George VI Ice Shelf, as a result of high surface flow speeds, thick ice, and deep subglacial topography.

Although ice flux from Cryosat Ice Stream is more than 50 % lower than that of neighbouring Ers Ice Stream, it boasts the greatest surface flow speeds of the English Coast: flowing at a maximum of 950 m a-1 (Mouginot et al., 2019a) (averaging out at ∼ 2.6 m d-1). These enhanced ice flow speeds are recorded along the width of the ice stream, where thick ice flows through and over multiple deep incisions in the basal topography (Fig. 4b). Figure 4a shows how these ice flow speeds are maintained across the grounding zone, as ice flows into George VI Ice Shelf. As the ice shelf buttresses the inland ice flow of Cryosat Ice Stream, further thinning of the ice shelf could reduce resistive stress (buttressing) at the grounding line, subsequently increasing ice discharge in this region (Tsai et al., 2015; Minchew et al., 2018).

Pronounced topographic depressions in most of the cross-flow radar lines that transect Sentinel Ice Stream (Fig. 4d) suggest a degree of topographic confinement for Sentinel Ice Stream, which is grounded more than 500 m below sea level. Whilst this confinement helps to channelize 6.00 Gt a-1 of ice towards the local grounding line currently, along-flow radargrams in Fig. 7a show how the ice stream might respond to future ingress of the grounding line position (e.g. Christie et al., 2016). Ice stream thickness fluctuates in conjunction with subglacial topography down the main trunk of the ice stream – from the upper catchment of the ice stream to the floating ice tongue, which is recorded by bright, white RES reflectors in Fig. 7a. These bright reflectors help to highlight the grounding zone (MacGregor et al., 2011), where ice flexes in response to tidal modulation (e.g. Rosier and Gudmundsson, 2018). Annotations in Fig. 7a point out a range of previously unknown subglacial features beneath Sentinel Ice Stream, like reverse subglacial slopes close to the grounding zone (which decline inland at ∼ 5.5–4.5∘ km-1), as well as more raised topographic features further inland. These measurements are critical for simulations of groundling line retreat. They show that a retreat of the grounding line into deeper water could allow thicker ice to reach floatation, which would increase glacier driving stress and ice flux across the grounding line (Tsai et al., 2015), with immediate implications for ice flow speed, ice discharge, and meltwater contribution to the Southern Ocean (Minchew et al., 2018). RES measurements inland of the present-day grounding line reveal a steep reverse bed slope, which after an initial retreat of the grounding line (due to some forcing) could promote unstable (runaway) grounding retreat (e.g. Schoof 2007; Jamieson et al., 2012; Kleman and Applegate, 2014). However, elevated subglacial topography ∼ 10 km inland of the current grounding line could potentially act as a pinning point for future ice stream re-grounding (Favier et al., 2016) (Fig. 7a). Our RES measurements will allow these potential instabilities to be explored in new, high-resolution numerical modelling simulations.

Further down the English Coast, Hall Glacier, Nikitin Glacier, and Lidke Ice Stream are clearly discernible in maps of surface ice flow speeds (Mouginot et al., 2019a) (Fig. 1) and subsurface topography maps, like Bedmap2 (Fretwell et al., 2013) and the newer, higher-resolution BedMachine (Morlighem, 2019) (Fig. 2). These maps show how discrete ice flow units develop in accordance with subglacial depressions, where elevated subglacial topography between tributaries helps to promote independent, channelized ice flow towards the coast (Fig. 5). All three ice flows converge in the floating Stange Ice Shelf, where they release a combined ice flux of ∼ 6.72 Gt a-1. The zone between grounded and floating ice is discernible in satellite data (Bindschadler et al., 2011) (noted by the ASAID grounding line in Fig. 5a) and in our RES dataset, where bright subglacial reflections suggest water ingress (MacGregor et al., 2011) along line HNL 1 (Fig. 5). These independent datasets mark the same grounding zone position along the English Coast. Whilst our radargrams do not extend seaward of transect HNL 1, we hypothesize that the 8 km digression of the ASAID grounding line in Fig. 5a could reflect the subglacial extension of the deep subglacial trough beneath Hall Glacier. This relative extension of the grounding line shows the impact subglacial troughs can have on grounding line location and potentially grounding line stability (as noted in other regions of Antarctica by Jamieson et al., 2012). Should the grounding line migrate in the future, relatively small-scale subsurface features like these could result in substantially different reactions from neighbouring ice flows, like Hall Glacier, Nikitin Glacier, and Lidke Ice Stream.

The final radar transects in our survey were flown across Landsat Ice Stream (Fig. 6). These radargrams reveal topographically confined ice flow along two discrete tributaries (north and south) for more than 15 km. These ice streams, which flow at speeds greater than 500 m a-1 (Mouginot et al., 2019a) contribute over 7.23 Gt a-1 of ice to the Bellingshausen Sea. Along-flow lines presented in Fig. 7b show the differences in ice thickness and subglacial topography between the north and south tributaries of Landsat Ice Stream, which are each grounded more than 700 m below sea level. The north tributary flows across a remarkably flat bed for most of its length, but this is punctuated by a region of elevated subglacial topography ∼ 5 km inland of the current grounding line, which is ∼ 100 m higher than surrounding bed returns (Landsat 6 transect, Fig. 7b). Whilst this generally flat, low-elevation subglacial bed could enable rapid grounding line retreat in response to mass balance changes and/or applied oceanic forcings (Weertman, 1974; Jamieson et al., 2012), this region of elevated subglacial topography could act as a temporary pinning point for re-grounding in a retreating-ice-sheet scenario. A similar potential pinning point is located much further inland of the grounding zone on the south tributary of Landsat Glacier (RES transect Landsat 7, Fig. 7b). Here, flat subglacial topography (situated ∼ 600 m below sea level) extends ∼ 12 km inland of the current grounding line, until bed topography lowers slightly and then inclines by 120 m over 2 km. Beyond this point, there is a reverse slope, dipping inland at 3.5∘ km-1. This subglacial topography correlates with satellite-derived surface ice flow speeds (recorded by Mouginot et al., 2019a): enhanced flow is recorded along RES transect Landsat 1, where bright subglacial reflectors suggest the presence of subglacial water (MacGregor et al., 2011). These reflections, which extend inland of the ASAID grounding line, could provide the subglacial evidence to corroborate recent satellite-derived measurements of inland grounding-line migration in this region of Antarctica (Christie et al., 2016; Konrad et al., 2018). As warm circumpolar deep water resides at ∼ 300 m depth in the neighbouring ocean (Kimura et al., 2015) any relatively warm water ingress inland could promote ice dynamical imbalance in this region of Antarctica and lead to further drawdown of ice from the interior (as reported by Hogg et al., 2017).