The melting of ice by ocean waters along the periphery of ice sheets is a major physical process driving their evolution in a warming climate. Using the fiber-optic-tethered Grounding line Remote Operated Vehicle (GROV), we explored the ice shelf cavity of Petermann Glacier, in Northwestern Greenland, in May 2023, using a novel interferometric multibeam sonar operating at 117 KHz with 360° viewing capability. The seafloor depth is uniform at 820 m and 200 m deeper than anticipated. At the ice shelf base, we find a succession of terraces interrupted by 20-40 m ice cliffs that have no signature at the surface, but are consistent with double-diffusive convection. The central melt channel deviates by ± 80 m from flotation, is smoother than indicated by the surface, and reveals asymmetric melt. The results demonstrate the fundamental importance of surveying the geometry of ice shelf cavities to document ice-ocean interaction.

Petermann Glacier is a major outlet glacier in North Greenland (60°W, 81°N) that forms a 20-km wide and 45-km long floating ice shelf (Fig. 1). Petermann drains 4% of the Greenland Ice Sheet in area. The ice shelf undergoes massive ice melt in contact with the ocean waters that removes more than half of its ice thickness in the first 10 km. In 2010 and 2012, the ice shelf underwent two major calving events, which truncated its length by 25 km. In 2018-2022, the glacier grounding line retreated for the first time in this century, by 7 kilometers at the center, 4 km on the sides, and the glacier sped up by 100 - 150 m/yr.

GROV: The``Grounding line Remote Operate Vehicle” (GROV) uses the Graaltech, torpedo shaped, X-300 Autonomous Underwater Vehicle (AUV), which is a portable, modular AUV, with an open control system, fully holometric five degrees of freedom (no fin), 15.5-cm diameter, 29 kg in weight, and a 14-hour endurance (untethered). We built a sonar module based on the Bathyswath-3 multibeam, phase-differencing, interferometric sonar from ITER-Systems Ltd. operating at a frequency of 117 KHz (12 mm in water). The interferometric capability achieves a wider swath in shallow water (< 200-400 m depth) compared to standard MBES systems, typically 10 times the water depth instead of 3-4 times, which is ideal for cavity exploration. We molded 12 transducer antennas on a cylindric module, with 4 transmitting/receiving and 8 receiving-only elements to enable 360° imaging with a 400-m nominal range, 600-m maximum. The sonar module is 22 cm in diameter. The Inertial Navigation Unit is a Spatial-FOG dual INS from Advanced Navigation which measures roll and pitch within 0.01°, heading within 0.01°, and with a bias instability of 0.05° per hour without external aiding (e.g. GNSS or magnetometers) . Navigation is assisted by Doppler Velocity Logger (DVL) DVL-A125 from Waterlink, with 5 cm-125 m range, 600 m depth rating, and maximum velocity of 9 cm/s, which locks on the ice bottom. Initial position and heading is determined using a the dual GNSS receiver build-in the INS. During the dive, DVL-aided navigation, East and North uncertainty is expected to be within 0.1% of the distance travelled.

CTD: We use a Conductivity Temperature Depth (CTD) sensor Star-Oddi DST online, with a temperature accuracy of 0.1°C and a precision of 0.032°C; a salinity accuracy of 0.1 psu and a precision of 0.02 psu. The fiber optic tether is lightweight (4.2 mm diameter, 95 kg with 5 km cable and motored winch) and high resistance (200 kg breaking strength). The GROV is depth-rated to 600 m, capable of a 1 km swath, weighs 56 kg in air, and is 2.2 m in length. 

Access Hole: We drilled a 25-cm hole in newly-formed ice with an ice Auger into 10 cm thick ice. The three-person science team plus a two-person film crew established a camp above the bore hole. We lowered a Minos XL CTD down the hole to record temperature, salinity, and depth up to 700 m depth, which was above the sea floor. The GPS location was noted every day before each dive. We lowered the GROV through the access hole and performed a series of eleven dives, with the longest dive corresponding to a tethered-traveling distance of 4 km under the ice shelf over a duration of 4 hours. The GROV was brought back to the surface regularly to re-charge its batteries in about 5 hours.  

Data Processing: The MBES data were processed using ITER System freeware software and post-processed with self-developed python-based interferogramm method and filtered using Cloud Compare freeware. The seafloor was not detected in the first couple of dives performed 50 m below the ice bottom. In subsequent dives, we unlocked the DVL from the ice base and surveyed at 400-500 m depth, at which point we detected both top and bottom of the cavity. Navigation was controlled by maintaining a constant heading and keeping track of distance with the fiber optic markings. Prior testing had indicated an exceptional stability for the IMU when the DVL had a good lock on the ground, i.e. a deviation of less than 2 meter in 1.5 km of travelled distance. As we collected data both on the way in and on the way out, we also overlapped tracks for consistency. The data are processed in x, y, z points available as a point cloud data set and gridded on a 25 m x 25 m horizontal using data averaging. The point cloud data include recordings from the CTD probe at each x,y,z location. 

Ancillary data: We employed a World View Digital Elevation Model (DEM) referenced to mean sea level and ortho-rectified imagery at 2 m spacing dated May 10, 2023 to conduct the data analysis. We used WV imagery from April 2023 to forecast the field operation. The data spacing is 8 m with a vertical precision of 1 m.