A Fully Automated Supraglacial lake area and volume Tracking (“FAST”) algorithm: Development and application using MODIS imagery of West Greenland

Toward Monitoring Surface and Subsurface Lakes on the Greenland Ice Sheet Using Sentinel-1 SAR and Landsat-8 OLI Imagery

“…The FAST algorithm creates a maximum lake extent array from all available summer lake extents from the relevant imagery, and then tracks changes to the individual lake areas within these maximum extents over the sequence of imagery. For each lake, a rapid drainage event was identified if a lake lost >80% of its area within ≤96 h. Although some rapid lake drainage events are known to occur in under 24 h , this four-day threshold is in line with definitions for rapid drainage events used in previous remote sensing studies: two days (Selmes et al, 2011(Selmes et al, , 2013; four days (Doyle et al, 2014;Fitzpatrick et al, 2014;Williamson et al, 2017); five days (Liang et al, 2012); and six days (Morriss et al, 2013). Slow drainage events were defined as when a lake lost >20% of its area over any time period, after any previously identified rapidly draining lakes were removed from the dataset.…”

“…Previous remote sensing studies in the area have quantified supraglacial lake area and volume (e.g., Box and Ski, 2007;Georgiou et al, 2009;Williamson et al, 2017) and found that lakes have been forming and draining at higher elevations as melt rates have increased in recent years (Selmes et al, 2011;Howat et al, 2013;Fitzpatrick et al, 2014;Pope et al, 2016), particularly in very warm years, when lakes also drain earlier (Liang et al, 2012). The study region was chosen to match the total extent of the selected Landsat-8 scenes (paths 006, 007, 008, 009, and 010, and rows 011 and 012; Figure 1).…”

“…These three separate analyses allowed us to assess the extra information about lake filling and draining that can be obtained using Sentinel-1 data instead of, or in addition to, Landsat data. Our lake drainage detections were derived using Williamson et al's (2017) Fully Automated Supraglacial lake Tracking (FAST) algorithm, which was originally developed for MODIS imagery. The FAST algorithm creates a maximum lake extent array from all available summer lake extents from the relevant imagery, and then tracks changes to the individual lake areas within these maximum extents over the sequence of imagery.…”

“…Previous studies investigating surface lakes have primarily used the optical and infrared wavelengths of MODIS (Box and Ski, 2007;Selmes et al, 2011;Morriss et al, 2013;Fitzpatrick et al, 2014;Williamson et al, 2017), or Landsat (McMillan et al, 2007Arnold et al, 2014;Banwell et al, 2014;Moussavi et al, 2016;Pope et al, 2016). MODIS has a high temporal resolution (daily) whereas the Landsat satellites have a high spatial resolution [30 m for Landsat-8 Operational Land Imagery (OLI)].…”

“…However, both the low spatial resolution of MODIS (250 or 500 m) and low temporal resolution of Landsat-8 OLI (at best a four-day repeat, although frequently up to 16 days) makes them poorly suited to the investigation of small lakes that may change rapidly. Furthermore, such sensors are unable to image through cloud or in darkness, restricting the number of useful scenes that can be used (Selmes et al, 2013;Williamson et al, 2017).…”