Measuring glacier velocities on the Kenai Peninsula, Alaska, using multispectral satellite imagery with feature tracking

Previous studies have shown that the glaciers of southern Alaska are receding and thinning, and contributing significantly to sea-level rise, and that temperate glaciers are sensitive indicators of climate change. Therefore, a method is developed to apply feature tracking software to moderate resolution multispectral satellite imagery to derive velocity fields for outlet glaciers of the Harding Icefield, Alaska, which will serve as a basis for future studies of mass balance, contribution to sea-level rise, and glacier response to climate change. The data produced by a feature tracking program are systematically sorted by location, direction, magnitude, correlation strength, and error estimates to eliminate errant vectors using customized IDL programs. A successful new method of systematically removing anomalous vectors within a velocity field via a feedback loop is introduced and tested. Point data are spatially interpolated and a vector field is plotted and color-coded to display ice surface velocities on a glacier, then velocity profiles are produced to gauge the influence of topography on ice velocity and to determine the primary mechanism of ice flow. Velocity fields for Tustumena and McCarty Glaciers are derived that show speeds ranging from < 50-300 m/yr along their lengths. The velocity fields are accurate to within 5.05 ± 3.66 m/yr (0.15 ± 0.11 pixel) or less in magnitude and within 4.5° or less in direction. In-depth analysis of results produced with the image correlation software Image Correlation (IMCORR) show that the magnitude of offset produced changes with varying reference chip size, kernel size used in filters, and direction of the applied filter. A comparison of results produced by IMCORR and a second image correlation program, Coregistration of Optically Sensed Images and Correlation (COSI-Corr), reveals COSI-Corr's frequential correlator is preferable to IMCORR in terms of consistency of results, ease of use, speed of computations, and accuracy. Velocity fields produced using 15 meter Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) imagery from July 2003 and August 2005, and 30 meter Landsat-5 TM imagery from September 2006, August 2007, and July 2008 reveal that the correlation techniques are sensitive to the amount of surface ablation that has occurred during the time interval between image pairs. The results show that feature tracking is capable of producing accurate velocity fields for Harding Icefield's outlet glaciers across a one year time span using moderate resolution multispectral satellite imagery.