Text S2: Field collection and data processing of Acoustic Doppler Current Profiler (ADCP) supraglacial river discharge measurements
Over the period 5-13 July 2016, a total 847 ADCP transects were acquired at a fixed cross-section (location 67.0499°N, -49. 0180°W) in the main-stem Rio Behar supraglacial river, using field methods based onSmith et al. (2017) (Figures S1-S4 ). Of these 847 transects, 677 later passed rigorous quality-assurance screening and were used to compute 174 in situ supraglacial river discharge estimates (Tables S1-S2; Figure S5 ). The 174 measurements were acquired between 13:00:09 UTC on 5 July 2016 and 10:37:57 UTC on 13 July 2016. Of these measurements, 168 were collected consecutively every hour starting 11:34:50 UTC on 6 July 2016 and ending 10:37:57 UTC on 13 July 2016. These 168 consecutive hourly measurements (1 full week) are the moulin input dataset analyzed in this study. The additional 6 discharge measurements collected intermittently on 5/6 July 2013 are excluded from our analysis because they do not fully capture the diurnal cycle, but are included in the archival dataset.
All surveys were conducted using a SonTek River Surveyor® M9 ADCP mounted on a SonTek HydroBoard II and a moving-boat survey type. To complete each survey, the M9 system was towed, in-transect, back and forth across the Rio Behar Channel, using a custom bank-operated cableway that enabled single-side tensioning and operation (Figures S1-S4 ). Between 3-9 individual hydrographic profiles or transects of channel cross-section, wetted perimeter, and flow velocity were collected during each measurement hour, yielding a total of 847 transects acquired over the field experiment study period (Tables S1-S2, Additional Supporting Information Datasets S1-S3 )
ADCP data were later processed into high-quality discharge retrievals using the following quality assurance (QA) and quality control (QC) workflow. This QA/QC workflow is similar to that described inSmith et al. (2017) and consists of the following:
  1. Open all ADCP output files for a given hour in River Surveyor Live (RSL) software and manually check/edit system settings. For all files, the Transducer depth was set to 0.1 m, the magnetic declination was set at -29, GPS reference was set to GGA, and the depth reference was set to Vertical Beam (VB) (rather than Bottom Track, BT).
  2. Instrument performance was also validated in RSL. Our system quality checks include: ensuring system power or voltage >9.5, GPS quality >= 3, Horizontal Dilution of Precision (HDOP) <= 2, the track reference was >0. Quality checks were initially conducted manually, and were later automated using Matlab.
  3. The edge or bank data for each measurement were manually inspected to confirm that the ADCP was receiving velocity and depth data near the profile edges. Profiles with no edge data (for either or both edges) were discarded from the final hourly discharge estimate.
  4. The depth data for each profile were inspected by comparing both the VB and BT data series and determining which depth reference was higher quality (i.e. had fewer outliers and less dropout). If both VB and BT were of equal quality, VB was selected as the depth reference. If VB had substantial dropout or anomalies, BT was selected. If either VB or BT had data dropout whereas the other depth reference contained data, composite tracks were selected such that RSL fills gaps in depth data series. Each profile was manually ranked on a scale from 0 to 3, where 0 or 1 indicates a poor or unusable transect due to insufficient depth data, 2 indicates a profile with minimal outliers and dropout, and 3 indicates a profile with no outliers or dropout. Profiles ranked as 0 or 1 were discarded from the final hourly discharge estimates, unless all transects in a given measurement hour were ranked as 0 or 1. In this instance, all transects were kept unless certain transects had notable more outliers or data dropout than other transects, in which case lower quality transects were removed from the final hourly discharge estimate.
  5. Velocity vectors and the signal-to-noise ratio were also inspected manually. Velocity vectors were ranked on a scale of 1 to 3, where 1 indicates minimal perpendicular vectors, substantial drift, or no data, 2 indicates vectors with moderate drift and some vector crossover, and 3 indicates minimal to no drift or crossover. Profiles with a ranking of 1 were discarded from the final hourly discharge estimate.
  6. All QA/QC’d data files were exported from River Surveyor Live as Matlab files. Both original ADCP data files (.riv or .rivr) readable in River Surveyor Live (which can be freely downloaded from the SonTek/Xylem website after registering with an email address) and output to Matlab files are available via Additional Supporting Information Dataset S3 .
  7. Following manual/automated QA/QC checks, resultant ADCP data and associated variable descriptions were summarized for each measurement hour. These summary data are presented in Tables S1-S2 ; and in as Additional Supporting Information in Excel spreadsheet (Dataset S1 ) and .txt (Dataset S2 ) formats.