Supervised classification of slush and ponded water on Antarctic ice shelves using Landsat 8 imagery

<jats:title>Abstract</jats:title><jats:p>Surface meltwater is becoming increasingly widespread on Antarctic ice shelves. It is stored within surface ponds and streams, or within firn pore spaces, which may saturate to form slush. Slush can reduce firn air content, increasing an ice...

Full description

Bibliographic Details
Main Authors: Dell, RL, Banwell, AF, Willis, IC, Arnold, NS, Halberstadt, ARW, Chudley, TR, Pritchard, HD
Format: Article in Journal/Newspaper
Language:English
Published: Cambridge University Press (CUP) 2022
Subjects:
Online Access:https://www.repository.cam.ac.uk/handle/1810/329873
https://doi.org/10.17863/CAM.77318
Description
Summary:<jats:title>Abstract</jats:title><jats:p>Surface meltwater is becoming increasingly widespread on Antarctic ice shelves. It is stored within surface ponds and streams, or within firn pore spaces, which may saturate to form slush. Slush can reduce firn air content, increasing an ice-shelf's vulnerability to break-up. To date, no study has mapped the changing extent of slush across ice shelves. Here, we use Google Earth Engine and Landsat 8 images from six ice shelves to generate training classes using a <jats:italic>k</jats:italic>-means clustering algorithm, which are used to train a random forest classifier to identify both slush and ponded water. Validation using expert elicitation gives accuracies of 84% and 82% for the ponded water and slush classes, respectively. Errors result from subjectivity in identifying the ponded water/slush boundary, and from inclusion of cloud and shadows. We apply our classifier to the Roi Baudouin Ice Shelf for the entire 2013–20 Landsat 8 record. On average, 64% of all surface meltwater is classified as slush and 36% as ponded water. Total meltwater areal extent is greatest between late January and mid-February. This highlights the importance of mapping slush when studying surface meltwater on ice shelves. Future research will apply the classifier across all Antarctic ice shelves.</jats:p> Rebecca L. Dell was funded by a Natural Environment Research Council (NERC) Doctoral Training Partnership Studentship (CASE with the British Antarctic Survey, grant no. NE/L002507/1). Alison F. Banwell received support from the U.S. National Science Foundation (NSF) under award no. 1841607 to the University of Colorado Boulder. Ian Willis was supported by the UK Natural Environment Research Council under NE/T006234/1 awarded to the University of Cambridge