Radio-echo sounding (RES) is commonly used to measure ice thickness and to understand subglacial topography and basal ice sheet conditions (Dowdeswell and Evans, 2004; Bingham and Siegert, 2007). A series of airborne geophysical explorations were conducted across East Antarctica in the 1970s (Robin et al., 1977; Dean et al., 2008; Turchetti et al., 2008; Naylor et al., 2008), which led to the first compilation folio of maps of subglacial bed topography, ice sheet surface elevation, and ice thickness in Antarctica (Drewry and Meldrum, 1978; Drewry et al., 1980; Jankowski and Drewry, 1981; Drewry, 1983). Since then, multiple efforts have been made to collect and compile RES data in order to expand the subglacial topographic database across the continent (Lythe et al., 2001; Fretwell et al., 2013).

Geophysical surveys of the coast of Princess Elizabeth Land (PEL) began in 1971, providing basic ice thickness, bed topography, and magnetic field data (Popov and Kiselev, 2018; Popov, 2020). Prior to our work, virtually no RES data had been acquired upstream of ∼300 km from the grounding line of PEL, however. Hence, this region has been described as one of the so-called “poles of ignorance” (Fretwell et al., 2013), and its representation in recent bed DEMs (e.g., Bedmap2 and BedMachine Antarctica) is as a zone of flat topography, reflecting the absence of data (Morlighem et al., 2020). Other data gaps (Recovery Glacier system; Diez et al., 2019; South Pole; Jordan et al., 2018) have been filled recently, leaving PEL as the last remaining significant region in Antarctica to be surveyed systematically.

In the absence of bed data, glaciologists have had to rely on satellite imagery, inversion from poor-resolution satellite gravity observations, and ice flow modelling to infer the subglacial landscape and its interaction with the ice above (Fretwell et al., 2013; Jamieson et al., 2016). For example, a combination of three satellite-derived mosaics and some initial exploratory RES data (Blankenship et al., 2017) have been used to hypothesize the subglacial features of PEL, revealing the presence of a potentially large (>100 km long) subglacial lake (white box; Fig. 1a and b) and an expected canyon morphology across the PEL sector. A study by Dongchen et al. (2004) adopted interferometric synthetic aperture radar (InSAR) satellite technology to generate an experimental subglacial bed elevation model across the ice sheet margin. While the result contains a certain level of detail, it has an obvious limitation in that the bed elevation was based solely on the satellite data and without direct measurements of the subglacial landscape. Hence, the bed topography in PEL is the poorest defined of any region in Antarctica – and indeed of any land surface on Earth.

Here, we present the first detailed ice thickness DEM for PEL, based on new RES measurements collected since 2015, which we refer to as the ICECAP2 DEM. We briefly discuss the differences between the ICECAP2 DEM and its representation in both Bedmap2 and BedMachine Antarctica. The ICECAP2 bed DEM is relative to ice surface elevations from the Reference Elevation Model of Antarctica (REMA; Howat et al., 2019). The ice thickness DEM can be easily integrated with updated surface DEMs (i.e. Helm et al., 2014) and, in particular, the upcoming Bedmap3 product.

The PEL sector of East Antarctica is bounded on the west by the Amery Ice Shelf and on the east by Wilhelm II Land (Fig. 1). The region covered by the ICECAP2 DEM extends ∼1300 km from east to west and ∼ 800 km from north to south. We use the differential interferometry synthetic aperture radar (DInSAR) grounding line (Rignot et al., 2011) to delimit the ice-shelf-facing margin of the ice sheet. The DEM was built from recently acquired airborne geophysical data collected across PEL by the ICECAP2 programme over four austral summer seasons from 2015 to 2019 (Fig. 1c).

Field data acquisition was achieved using the Snow Eagle 601 aerogeophysical platform and a BT-67 aeroplane operated by the Polar Research Institute of China for the Chinese National Antarctic Research Expedition (CHINARE) programme (Fig. 2a and b). The suite of instruments configured on the aeroplane includes a phase-coherent RES system, functionally similar to the high-capability airborne radar sounder (HiCARS) developed by the University of Texas Institute for Geophysics (UTIG; i.e. Young et al., 2011; Greenbaum et al., 2015). HiCARS operates at a central frequency of 60 MHz and a peak power of 8 kW, making it capable of penetrating deep (>3 km) ice in Antarctica. The system has an along-track spatial sampling rate and a vertical resolution of ∼20 and ∼5.6 m, respectively. Further details on the parameters and introduction of the RES system can be found in Cui et al. (2018). A Javad GNSS Inc. GPS receiver and its four antennas were mounted at the aircraft's centre of gravity (CG), tail, and both wings. GPS data from the antenna at the aircraft's CG were used for RES data interpretation.

During the first field season (2015–2016), a survey acquiring exploratory, fan-shaped radial profiles, to maximize range and data return on each flight, was completed across the broadly unknown region of PEL. These flight lines extend from Progress Station at the coast to the interior ice sheet divide at Ridge B (Fig. 1). In the second and third seasons (2016–2017 and 2017–2018), a survey grid was completed, targeting enhanced resolution over a proposed subglacial lake and a series of basal canyons (see Jamieson et al., 2016). In the fourth season (2018–2019), a few additional transects were completed to fill the largest data gaps within aircraft range.

Ice thickness measurements were derived from two RES data products from which the ice bed interface was traced and digitized, namely (a) 2D-focused synthetic aperture radar (SAR) processed data, applied to RES data from the first two seasons, and (b) unfocused field RES data from the third and fourth seasons. Raw RES data were first separated to differentiate the project/set/transect (PST) during the field data processing. Pulse compression, filtering, 10 traces coherent stacking, and five traces incoherent stacking were then applied to generate a field RES data product. The field RES data can be used for quality control and are also good enough for initial ice bed interface measurements, leading to the calculation of first-order ice thicknesses and an initial bed DEM. To achieve better quality RES images, 2D-focused SAR processing was applied to data from the first two seasons (Peters et al., 2007). The ice bed interface was picked in a semi-automatic manner, using a picking programme used previously by the ICECAP programme on data from the Aurora and Wilkes subglacial basins (Blankenship et al., 2016, 2017). Ice thicknesses were calculated by multiplying two-way travel time by the velocity of electromagnetic waves in the ice (i.e. 0.168 mns-1; Cui et al., 2018). Firn corrections were not applied and, thus, may be subject to a small systematic error. The precise point positioning method was used in the GPS processing to improve location accuracy since the flight distance is too far from the GPS base station for post-airborne GPS data processing. Processed GPS data were interpolated and fitted to the RES traces according to time stamps generated by the integrated airborne system. The aircraft to ice surface range was calculated by multiplying the two-way travel time of the RES reflections off the ice surface by the velocity in air (0.3 mns-1). Figure 2c shows examples of the RES data collected in 2017–2018.

To derive the ice thickness map (Fig. 4a), we employed a variety of techniques, depending on the ice speed, following the approach described in Morlighem et al. (2020). In fast-flowing regions (i.e. velocity >30 myr-1), we relied on mass conservation (MC; Fig. 3) constrained by the ICECAP2 RES data and additional RES data that were available as part of BedMachine Antarctica (Morlighem et al., 2020). In the slower moving regions inland, we relied on a streamline diffusion interpolation to fill the area between data points (Fig. 3).

For the purpose of comparing the ICECAP2 DEM (Fig. 4b) with Bedmap2 (Fig. 4c) and BedMachine Antarctica (Fig. 4d), the 500 m ice surface elevation DEM from the REMA (Howat et al., 2019) was used. Prior to the subtraction process, the Bedmap2 and BedMachine ice thickness DEMs were transformed from the g104c geoid vertical reference to the World Geodetic System (WGS) 1984 vertical reference frame. The ice thickness for both Bedmap2 and BedMachine are in ice equivalent rather than an estimate of the ice thickness from firn corrections. The Bedmap2 and BedMachine ice thickness DEMs were resampled, using the bilinear function in ArcGIS, to a 500 m spacing and referenced to the polar stereographic projection (Snyder, 1987). The ice thickness in all three models was then subtracted from the ice surface elevation DEM (Howat et al., 2019) to produce bed DEMs at a 500 m resolution. Difference maps were then computed by subtracting the Bedmap2 (Fig. 4e) and BedMachine (Fig. 4f) bed DEMs from the ICECAP2 bed DEM. Crossover analyses revealed root mean square (RMS) errors within the ICECAP2 RES data of 24.2 m (2015–2016), 39.2 m (2016–2017), 10.4 m (2017–2018), 7.5 m (2018–2019), and 35.4 m (for the full data set).

The ICECAP2 ice thickness and bed elevation models of the PEL sector are available in 500 m horizontal resolutions at 10.5281/zenodo.4023343 (Cui et al., 2020a). The airborne RES ice thickness measurements used to generate the products, recorded here in comma-separated values (CSVs) format, are accessible from 10.5281/zenodo.4023393 (Cui et al., 2020b). The 500 m ice sheet surface elevation DEM derived from the REMA (Howat et al., 2019) can be obtained from https://www.pgc.umn.edu/data/rema/ (last access: 12 September 2020). If the users wish to modify the bed DEM, our model can be easily integrated with the updated surface elevation models (Bamber et al., 2009; Helm et al., 2014). Auxiliary details for the MEaSUREs InSAR ice velocity map of Antarctica can be found at 10.5067/D7GK8F5J8M8R (Rignot et al., 2017b). The satellite images from the MODIS Mosaic of Antarctica (2008–2009) are obtainable from 10.7265/N5KP8037 (Haran et al., 2014, updated in 2019). A summary of the data used in this paper and their availability is provided in Table 1.

We have presented RES data from the first dedicated airborne geophysical survey of PEL. From the data (and using a combination of interpolation and modelling techniques), we have generated a bed DEM at a resolution of 500 m over an area of ∼900000 km2 – the ICECAP2 DEM. Considerable differences between this DEM and both Bedmap2 and BedMachine Antarctica are observed, particularly at the centre of the DEM where a broad subglacial basin has been identified and measured. The ICECAP2 DEM completes the first-order data coverage of subglacial Antarctica – a feat spanning around 70 years of international collaboration (Turchetti et al., 2008).