Determining Crevasse Sequences in Surging Glaciers using Neural Network Classification from Remotely Sensed Images of Bering Glacier, AK

Jessica Bobeck, University of Colorado at Boulder, Department of Geography, Boulder, CO, United States, Ute C Herzfeld, University Colorado Boulder, Geomathematics, Remote Sensing and Cryospheric Sciences Laboratory; Electrical, Computer and Energy Engineering, Boulder, United States; Cooperative Institute for Research in Environmental Sciences, University of Colorado Boulder, Boulder, United States, Lukas Goetz-Weiss, Univ Colorado Boulder, Electrical, Computer and Energy Engineering, Boulder, CO, United States and Griffin Hale, Sierra Nevada Corporation, Louisville, CO, United States
Abstract:
Bering Glacier in Alaska is a surging glacier; one of the most understudied glacier classes in the cryospheric sciences. By using a neural network created with a new method for extracting crevasse patterns, and remotely sensed images acquired from World View 1, it is possible to not only to classify crevasses formed during Bering's surges, but also to determine whether glacier crevasses form in pattern sequences. In order to understand the relationship between the geographic location of the crevasse types and the geophysical formation of those crevasses, an analysis of glacier flow, velocity, and the dependency on the type of force acting upon the glacier, will be used over a time sequence of Wold View 1 images. The importance of this study will allow for better understanding of the geophysical processes that occurs on surging glaciers, along with allowing for future prediction of crevasse formation which will be useful in determining hazardous regions of Bering glacier, ultimately allowing for higher safety for researchers.
Back to: Remote Sensing of the Cryosphere III Posters