Modeling Ocean-Forced Changes in Smith Glacier

Tuesday, 16 December 2014
David Lilien, Ian R Joughin and Benjamin Eaton Smith, Applied Physics Lab, University of Washington, Seattle, WA, United States
Glaciers along the Amundsen Coast are changing rapidly, which has drawn substantial scientific and public attention. Modeling and observation suggest warm-water intrusion and consequent melting as the cause of observed changes, and that unstoppable retreat may have already been triggered in some drainages. While Pine Island and Thwaites Glaciers are losing the most mass and have been the predominant objects study, other systems, particularly Smith, Pope and Kohler Glaciers and the corresponding Dotson and Crosson Shelves, are changing more rapidly relative to their size. Though smaller, these glaciers still have potentially large implications for overall regional dynamics as their beds connect below sea level to surrounding basins. In particular, the long, deep trough of Smith Glacier nearly links to the large eastern tributary of Thwaites, potentially causing rapid changes of Smith to have significant impact on the continuing retreat of Thwaites.
We implemented a numerical model in Elmer/Ice, an open-source, full-Stokes, finite-element software package, to investigate the response of the Smith/Pope/Kohler system to different initial conditions. We use various parameterizations of sub-shelf melting with constant magnitude to examine the sensitivity of overall dynamics to melt distribution. Because melt distribution affects lateral buttressing and upstream grounded areas, it is potentially an important control on ice shelf and outlet glacier dynamics. Through comparison to the most recent velocity data, we evaluate the ability of differing melt parameterizations to reproduce the behavior currently seen in Smith/Pope/Kohler glaciers. In addition, we investigate the effect of using different years of velocity data with constant elevation input when initiating model runs. By comparing results over the satellite record to initiation with synchronous observations, we assess the accuracy of the often necessary practice of using differently timestamped datasets.