C24B-02:
The Next Step in Ice Flow Measurement from Optical Imagery: Comprehensive Mapping Of Ice Sheet Flow in Landsat 8 Imagery Using Spatial Frequency Filtering, Enabled by High Radiometric Sensitivity

Tuesday, 16 December 2014: 4:15 PM
Mark A Fahnestock1, Theodore A Scambos2 and Marin J Klinger2, (1)Univ. of Alaska/GI, Fairbanks, AK, United States, (2)National Snow and Ice Data Center, CIRES University of Colorado, Boulder, CO, United States
Abstract:
The advent of large area satellite coverage in the visible spectrum enabled satellite-based tracking of ice sheet flow just over twenty years ago. Following this, rapid development of techniques for imaging radar data enabled the wide-area mapping and time series coverage that SAR has brought to the documentation of changing ice discharge. We report on the maturation of feature tracking in visible-band satellite imagery of the ice sheets enabled by the high radiometric resolution and accurate geolocation delivered by Landsat 8, and apply this to mapping ice flow in the interiors of Antarctica and Greenland. The high radiometric resolution of Landsat 8 enables one to track subtle patterns on the surface of the ice sheet, unique at spatial scales of a few hundred meters, between images separated by multiple orbit cycles. In areas with significant dynamic topography generated by ice flow, this requires use of simple spatial filtering techniques first applied by Scambos et al. 1992. The result is densely sampled maps of surface motion that begin to rival the coverage available from SAR speckle tracking and interferometry. Displacement accuracy can approach one tenth of a pixel for reasonable chip sizes using conventional normalized cross-correlation; this can exceed the geolocation accuracy of the scenes involved, but coverage is sufficient to allow correction strategies based on very slow moving ice. The advance in radiometry, geo-location, and tracking tools is augmented by an increased rate of acquisition by Landsat 8. This helps mitigate the issue of cloud cover, as much of every 16-day orbit cycle over ice is acquired, maximizing the acquisition of clear-sky scenes. Using the correlation techniques common to IMCORR and later software, modern libraries, and single-cpu hardware, we are able to process full Landsat 8 scene pairs in a few minutes, allowing comprehensive analysis of ~1K available ice sheet image pairs in a few days.