Further by downloading this data the user acknowledges that they agree with the NERC data policy (), and the following conditions:

1. To cite the data in any publication as follows:

DATA REFERENCE

Jordan, T., & Robinson, C. (2021). Processed airborne radio-echo sounding data for the Thwaites Glacier 2019 survey, West Antarctica (2019/2020) (Version 1.0) [Data set]. NERC EDS UK Polar Data Centre. https://doi.org/10.5285/E7ABA676-1FDC-4C9A-B125-1EBE6124E5DC

2. The user recognizes the limitations of data. Use of the data is at the users' own risk, and there is no warranty as to the quality or accuracy of any data, or the fitness of the data for your intended use. The data are not necessarily fully quality assured and cannot be expected to be free from measurement uncertainty, systematic biases, or errors of interpretation or analysis, and may include inaccuracies in error margins quoted with the data.

Methodology:

** Instrumentation and Processing:

Radar data were collected using the new bistatic PASIN-2 radar echo sounding system mounted on the BAS Twin Otter aircraft "VP-FBL" and operating with a centre frequency of 150 MHz and using a 4-microseconds, 13 MHz bandwidth linear chirp. The PASIN system transmitted 5 separate pulses from the wing arrays as follows; Port 4 microseconds chirp; Starboard 4 microseconds chirp; Port 4 microseconds chirp 180deg phase shift; Starboard 4 microseconds chirp 180deg phase shift; Port 1 microseconds chirp. Data for every antenna and pulse is recorded separately, in 20 second segments.

Chirp compression was applied using a Blackman window to minimise sidelobe levels, resulting in a processing gain of 10 dB. The chirp data was processed using a coherent averaging filter (commonly referred to as unfocused Synthetic Aperture Radar (SAR) processing) with Doppler beam sharpening to enhance the signal to clutter ratio of the bed echo and improve visualisation. The chirp data is best suited to assess the bed and internals in deep ice conditions. The coherent pulse-data (0.1-microseconds) was processed using a coherent averaging filter. This data is best used to assess the internal structure and bed in shallow ice conditions.

For picking the along-track bed elevation, only data from the port antenna array was used. Data from all the port antennas was combined, and the 4 microseconds and microseconds 180° phase shift pulses were combined to enhance the array gain and minimise coherent noise. The use of the same transmit/receive array theoretically minimises the power loss due to cross polarisation of the opposite wing. Along-track SAR focusing was applied to the combined port wing data. The segy files for each flight were then imported into the Promax seismic processing package. Down trace 'time' is simply the sample number across the 64 microseconds window, digitised at 120 MHz (i.e. max sample number = 7680 = 64 microseconds). Weighted trace mixing across 5 traces was applied to improve the signal to noise ratio of the data and down-trace automatic gain control was applied to reveal low amplitude returns from deeper reflectors. An initial window ~100 samples above the bed reflection was manually defined (top mute). An automated first break pick algorithm was then run to locate the precise bed return below the top mute. Subsequent manual picking removed un-realistic spikes, and selected the most physically appropriate bed surface in cases where multiple reflections were seen close to the bed. Generally the shallowest reflector was assumed to be the bed, as off-axis reflectors would likely appear later (deeper) in the section. In some cases strong reflectors which appeared deeper were chosen, with shallower week reflectors assumed to reflect entrained debris, accreted ice, or un-compensated refraction hyperbole close to the bed.

The PASIN radar system does not resolve the ice surface well. Range to surface from coincident LIDAR data, or calculated from an accurate DEM is therefore preferred. However, to estimate ice thickness and hence correct bed elevation, the location of the surface reflector in the radargram must be known. To calculate the theoretical surface pick location from the LIDAR or DEM range to ground these measurements must be calibrated. To do this the ice surface location for a single flight (T05) was picked from the Port to Starboard (P2S) radar dataset. Use of this dataset avoided the problem of the transmit pulse and switching period overlapping with the surface reflection. The location of the surface reflector was picked in Promax following a similar approach to the bed, with an additional bottom mute defined ~100 samples below the surface reflection. The Promax surface pick was then plotted against the LIDAR range to ground and a linear trend fit to the data. The resulting slope a...(104)

Data collection:

** Instrument:

Radar data were collected using the new bistatic PASIN-2 (Polarimetric radar Airborne Science Instrument) radar echo sounding system operating in full polarimetric mode and mounted on the BAS Twin Otter aircraft "VP-FBL" and operating with a centre frequency of 150 MHz and using a 4-microseconds, 13 MHz bandwidth linear chirp (deep sounding). The Pulse Repetition Frequency was 15,625 Hz (pulse repetition interval: 64

microseconds).

** Antenna configuration:

8 transmitters (4x port, 4x starboard)

12 receivers (8 wings, 4x belly)

Antenna gain: 11 dBi (with 4 elements)

Transmit power: 1 kW into each 4 antennae

Maximum transmit duty cycle: 10% at full power (4 x 1 kW)

** Waveform details:

Four waveforms, 4uS Tukey port, 4uS Tukey starboard, 1uS Tukey port, 1uS Tukey starboard.

** Radar receiver configuration:

Receiver vertical sampling frequency: Receiver vertical sampling frequency: 120 MHz (resulting in sampling interval of 8.33 ns)

Receiver coherent stacking: 25

Receiver digital filtering: -50 dBc at Nyquist (11 MHz)

Effective PRF: 156.25 Hz (post-hardware stacking)

Sustained data rate: 10.56 Mbytes/second

Data quality:

Analysis of 52 crossover points within the survey area indicates a standard deviation for the bed elevation of ~22m, which is in-line with the values suggested for previous radar surveys. Lidar derived surface elevation has a crossover error of ~10 m. The relatively high value is attributed to use of Lidar nadir range to ground un-corrected for aircraft roll, pitch and yaw.

Line spacing is on the order of 15 km (designed to inter-leave with previous surveys). Along line data spacing is ~24 m.