Toward Automated Mapping of the Greenland Ice Sheet’s Terrestrial Margins from Satellite Imagery

The Greenland Ice Sheet (GrIS) is rapidly losing mass and contributing to rising global sea levels. Accurate mapping of the position of the ice-sheet margin is critical for correctly apportioning this mass loss. Most studies have focused on its marine-terminating margins, which change rapidly and calve regularly, leading to readily observed terminus retreat. However, the vast majority (>93%) of the ice sheet’s perimeter is land-terminating. This portion of the terminus is presently delineated manually, limiting the temporal resolution and/or spatial coverage of terminus position datasets, so it is unknown if the terrestrial margin has moved appreciably in recent decades at large scales. Such change in the terrestrial terminus position could substantially contribute to mass loss as the majority of the ice sheet’s perimeter terminates on land. Given the length of the land-terminating ice-sheet margin (>50,000 km), automated methods are required to produce frequent assessment of any change therein, but have greater risk for misclassification (e.g., shadowed terrain, debris-covered ice, medial moraines, melt ponds and proglacial lakes). Here we present a method for automatic mapping of the land-terminating portion of the GrIS, with the goal of creating an annual, 10-m-resolution delineation of this terminus from ~2018 onward. Our method uses Google Earth Engine to combine random forest surface classification of late-summer Sentinel-2 imagery, Simple Non-Iterative Clustering (SNIC) segmentation, and post-classification binary image processing. The automatic terrestrial terminus positions for 2022 are validated against a recent high-resolution manual delineation for the same year from the Geological Survey of Denmark and Greenland (GEUS) (PROMICE-2022 Ice Mask, https://doi.org/10.22008/FK2/O8CLRE). We present initial maps of the land-terminating margin across the entire GrIS for both 2019 and 2022 and demonstrate their potential for improved understanding of ice-sheet-wide terminus change.