EGU21-16349
https://doi.org/10.5194/egusphere-egu21-16349
EGU General Assembly 2021
© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

A record of slush and water extent on Antarctic ice shelves from 2013 to present day

Rebecca Dell1, Alison Banwell2, Neil Arnold1, Ian Willis1, Anna Ruth W. Halberstadt4, Tom Chudley1, and Hamish Pritchard3
Rebecca Dell et al.
  • 1Scott Polar Research Institute, Lensfield Road, Cambridge, United Kingdom, CB2 1ER
  • 2Cooperative Institute for Research in Environmental Sciences, University of Colorado Boulder, Boulder, 80309,United States,
  • 3British Antarctic Survey, High Cross, Madingley Road, Cambridge, CB3 0ET
  • 4Department of Geosciences, University of Massachusetts, Amherst, MA 01003, United States

Supraglacial melt is observed across the majority of Antarctic ice shelves and is expected to increase in line with rising air temperatures. Surface meltwater may run off the ice shelf edge and into the ocean, or be stored within firn pore spaces (slush) and supraglacial water bodies (ponds, lakes or streams). When stored either as slush or supraglacial water bodies, the water can indirectly impact ice shelf dynamics, and potentially facilitate ice shelf collapse. Numerous studies have quantified ice shelf meltwater in supraglacial water bodies, however, despite its importance, no studies exist that quantify the extent of slush on a pan-Antarctic scale.

Here, we develop a supervised classifier in Google Earth Engine capable of identifying both slush and ponded water on a pan-Antarctic scale using Landsat 8 imagery. We train and test our classifier on six ice shelves: (1) Nivlisen, (2) Roi Baudouin, (3) Amery, (4) Shackleton, (5) Nansen, (6) George VI. A k-means clustering algorithm is applied to selected Landsat 8 training scenes, and the output clusters are manually interpreted to form training classes (i.e. slush, water, and other surface types (e.g. blue ice, dirty ice)). These training classes are then used to train a Random Forest Classifier, and the accuracy of the outputs are assessed using expert elicitation. Overall, the classifier accuracy for water and slush is 78 % and 70 % respectively. The validated classifier is then applied to numerous ice shelves across Antarctica, in order to produce estimates of slush and water extent from 2013 to the present day.

How to cite: Dell, R., Banwell, A., Arnold, N., Willis, I., Halberstadt, A. R. W., Chudley, T., and Pritchard, H.: A record of slush and water extent on Antarctic ice shelves from 2013 to present day, EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-16349, https://doi.org/10.5194/egusphere-egu21-16349, 2021.

Displays

Display file