Abstract
Supraglacial lakes (SGLs) on the Greenland Ice Sheet (GrIS) influence ice dynamics if they drain rapidly by hydrofracture. MODIS data are often used to investigate SGLs, including calculating SGL area changes through time, but no existing work presents a method that tracks changes to individual (and total) SGL volume in MODIS imagery over a melt season. Here, we develop such a method by first testing three automated approaches to derive SGL areas from MODIS images from the MOD09 level-2 surface-reflectance product, by comparing calculated areas for the Paakitsoq and Store Glacier regions in West Greenland with areas derived from Landsat-8 (LS8) images. Second, we apply a physically-based depth-calculation algorithm to the pixels within the SGL boundaries from the best performing area-derivation method, and compare the resultant depths with those calculated using the same method applied to LS8 imagery. Our results indicate that SGL areas are most accurately generated using dynamic thresholding of MODIS band 1 (red) MOD09 data with a 0.640 threshold value; calculated values from MODIS are closely comparable to those derived from LS8. Third, we incorporate the best performing area- and depth-detection methods into a Fully Automated SGL Tracking (“FAST”) algorithm that tracks individual SGLs between successive MODIS images. Finally, we apply the FAST algorithm to the two study regions, where it identifies 43 (Paakitsoq) and 19 (Store Glacier) rapidly draining SGLs during 2014, representing 21% and 15% of the respective total SGL populations, including some clusters of rapidly draining SGLs. The FAST algorithm improves upon existing automatic SGL tracking methods through its calculation of both SGL areas and volumes over large regions of the GrIS on a fully automatic basis. It therefore has the potential to be used for investigating statistical relationships between SGL areas, volumes and drainage events over the whole of the GrIS, and over multiple seasons, which might provide further insights into the factors that trigger rapid SGL drainage.
Highlights
The Greenland Ice Sheet (GrIS) is losing mass at an accelerating rate, which is predicted to continue for at least the coming century; developing a robust understanding of the processes contributing to this mass loss forms a fundamental research agenda (Rignot et al, 2011; Vaughan et al, 2013; van den Broeke et al, 2016; Noël et al, 2016)
Dynamic thresholding of the red band generally outperforms the other approaches for both regions in terms of root mean square error (RMSE) and the percentage of Supraglacial lakes (SGLs) reported by MODerate-resolution Imaging Spectroradiometer (MODIS) relative to the total number of SGLs reported in the LS8 dataset (Table 1)
The best performing values for dynamic thresholding are nearly identical between Paakitsoq and Store Glacier, with the threshold value of 0.640 at Store Glacier only marginally outperforming the 0.645 value for Paakitsoq in terms of number of SGLs reported (Table 1)
Summary
The Greenland Ice Sheet (GrIS) is losing mass at an accelerating rate, which is predicted to continue for at least the coming century; developing a robust understanding of the processes contributing to this mass loss forms a fundamental research agenda (Rignot et al, 2011; Vaughan et al, 2013; van den Broeke et al, 2016; Noël et al, 2016). They influence surface melt rates, through their effect on lowering albedo (Lüthje et al, 2006; Tedesco and Steiner, 2011; Tedesco et al, 2012) They affect ice-dynamic processes, since their rapid drainage by hydrofracture allows large pulses of surface meltwater to reach the GrIS's bed, which may impact subglacial effective pressures, raising basal water pressures and surface ice velocities (Zwally et al, 2002; Alley et al, 2005; Shepherd et al, 2009; Bartholomew et al, 2010, 2011a, 2011b, 2012; Schoof, 2010; Sundal et al, 2011; Hoffman et al, 2011; Colgan et al, 2011a; Cowton et al, 2013; Joughin et al, 2013; Sole et al, 2013; Tedstone et al, 2013; Andrews et al, 2014; Bougamont et al, 2014; Dow et al, 2015; Stevens et al, 2015). Hydrofracture opens up moulins that can continue delivering meltwater to the bed throughout the remainder of the melt season, influencing basal water pressures and sliding over longer timescales (Palmer et al, 2011; Colgan et al, 2011b; Sole et al, 2013; Banwell et al, 2013, 2016; Tedstone et al, 2014)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
