Estimating Supraglacial Lake Depth with Landsat 8: Algorithm development, Greenland case study, and sharing data & code

Allen Pope1,2, Ted Scambos1,2, Mahsa Moussavi2,3, Marco Tedesco4, Mike Willis5,6, and Shane Grigsby2,3

Abstract
Supraglacial lakes play a significant role in glacial hydrological systems. As a storage mechanism for surface melt, supraglacial lakes play an important role in ice hydrofracture. Thus they are a key step in transporting water to the glacier bed in Greenland or ice shelf disintegration in Antarctica. Multispectral remote sensing provides multiple methods for estimating supraglacial lake depth – either through single-band or band-ratio methods, both empirical and physically-based. Landsat 8 is the newest satellite in the Landsat series. With new bands, increased acquisition rates, higher dynamic range, and higher radiometric resolution, the Operational Land Imager (OLI) aboard Landsat 8 has high potential to facilitate a significant improvement in mapping of stored melt volume and tracking seasonal volume variations. This study uses in situ reflectance spectra and lake depth measurements to investigate the ability of Landsat 8 to estimate depths using multiple methods, as well as quantify the improvements over Landsat 7’s ETM+. Promising methods are applied to Landsat 8 OLI imagery of case study areas in Greenland allowing calculation of regional volume estimates using 2013 and 2014 summer-season imagery. Altimetry from WorldView DEMs are used to validate lake depth estimates. The optimal method for supraglacial lake depth estimation with Landsat 8 is shown to be an average of single band depths using the red and panchromatic bands. In addition, this poster will show work that has been done to document data provenance and the scientific workflow which led to the results of this study. As part of an EarthCube project (geosoft-earthcube.org), all data and code will be shared and documented as fully as possible, from ingestion through to results. Data and metadata have DOIs (Digital Object Identifier) and are shared via ACADIS (aoncadis.org) and FigShare (figshare.com). Code also has a DOI and is licensed and shared via GitHub (github.com). This case study serves as a learning example for more open and reproducible Earth science research.