Supporting data for: Rapid ice-marginal lake growth in Alaska driven by glacier retreat through bed overdeepenings

In this study, we mapped ice-marginal glacial lakes in Alaska in 2018 and 2024 using Sentinel-2 optical imagery. We mapped glacier-bed overdeepenings using existing datasets and evaluated how much lake growth has occurred within these mapped overdeepenings as well as how much additional growth is possible. Here we provide seven supporting data products: i) a raster of mapped glacier-bed overdeepenings using best estimate of ice thickness, ii) a raster of glacier-bed overdeepenings using ice thickness plus uncertainty, iii) a raster of glacier-bed overdeepenings using ice thickness minus uncertainty, iv) a 2018 ice-marginal lake shapefile, v) a 2024 ice-marginal lake shapefile, vi) a 2024 ice-marginal lake shapefile joined to glacier-bed overdeepening area, and vii) RGI 6 glacier inventory joined to glacier-bed overdeepening area. 

Methods
Using an existing glacial lake inventory (Rick et al., 2022), we identified glacial lakes in Alaska greater than 0.5 km2 (n=211) for the period 2016 to 2019 (median year 2018). Using Sentinel-2 optical imagery from 1 May to 30 September 2018 and 2024, we manually mapped lake extents at 1:10,000 scale for these two years in the open-source QGIS software program. We estimate error in lake area by multiplying the lake perimeter (in km) by half the Sentinel-2 pixel resolution (5 m or 0.005 km). We also reviewed and updated the lake connection category, which describes the lake's spatial relationship to its source glacier. These categories include: proglacial (lakes at terminus of glacier, in contact with the ice), supraglacial (lakes on surface of glacier ice), ice (ice-dammed lakes found at ice margins or tributary valleys), detached (lakes fed by glaciers but not in contact with the ice), or unconnected (detached lakes that are not fed by glaciers). This study solely assesses ice-marginal lakes (n=147 lakes), defined as having contact with a glacier and thus excludes detached and unconnected lakes.

To map glacier-bed overdeepenings in the region, we combined the Copernicus GLO-30 Digital Elevation Model (DEM) and previously modelled ice thickness estimates derived from observed ice velocities and the shallow-ice approximation (Millan et al., 2022). Imagery for GLO-30 was acquired between 2010 and 2015, while ice velocities used in Millan et al. (2022) are from 2017–18. The GLO-30 DEM was mosaicked, regridded to 50 m, and reprojected to NAD83/Alaska Albers (EPSG: 3338). The modeled ice thicknesses were mosaicked and reprojected and then subtracted from the surface DEM to generate a bed DEM. We created a filled bed surface using the “fillsinks” function in TopoToolbox in Matlab. In QGIS, we subtracted the filled bed DEM from the original bed DEM to map glacier-bed overdeepening spatial extents and magnitudes. To account for uncertainty in ice thickness estimates, we repeated the above workflow with revised ice thickness rasters that accounted for stated uncertainties.