Text S4: Field collection and data processing of GPS ice surface motion measurements
Records of positional location were collected with a Trimble R7 dual-frequency global positioning system (GPS) receiver and Trimble Zephyr Geodetic antenna (Figure S6 ). The system was installed ~750 m SSE of the moulin near the ADCP gauging site at location 67.048°N, -49.018 °W, elevation 1211.43 m) The antenna was affixed to a 3.3 m schedule-40 aluminum rod drilled vertically 3 m into the ice. The aluminum rod-antenna setup was allowed to freeze overnight. The system was powered by a 40 W solar panel attached to a weatherproof Pelican hard case that enclosed the GPS receiver, batteries, and cables adjacent to the antenna. The ice sheet thickness at this location is ~934 m based on Bedmachine v3 (Morlighem et al., 2017). The entire system was provided by UNAVCO (formerly University NAVSTAR Consortium), with protocols for field installation and GPS receiver settings provided by UNAVCO geodetic support engineers. The GPS station recorded positions at 5-s intervals between 5 and 13 July 2016. A base station was also established on bedrock near the ice sheet terminus (67.150°N, 50.058°W, elevation 581.19 m) and recorded positions at 5-s intervals between 4 and 15 July 2016.
Trimble binary receiver files were converted to RINEX observation files using runpkr00 v5.40 and TEQC utilities (Estey and Meertens , 1999). On-ice kinematic GPS positions were estimated using carrier-phase differential processing relative to the bedrock mounted reference station (baseline of ~47 km) using TRACK v1.28 (Chen , 1998) and final International GNSS Service satellite orbits following Andrews et al., 2018; Hoffman et al., 2011 . During processing, kinematic station motion was constrained on an epoch-by-epoch basis to 2,000 m yr-1 to permit rapid, short-term velocity changes. The 5-s time series was then smoothed with a 6-hr phase-preserving filter to eliminate spurious signals associated with GPS uncertainties and decimated to a 15-min time series. The smoothed x and y positions were used to calculate 6-hr velocities using a centered time window to limit aliasing that may result from using discrete time intervals. Uncertainties presented here are +/- one standard deviation of the 15-min binned 5-s position data.