Application of Image Cross-Correlation to the Measurement of Glacier Velocity Using Satellite Image Data
Image-to-image cross-correlation software is applied to pairs of digital satellite images to map the velocity field of moving ice. This technique uses small-scale glacial surface features, such as crevasse scars and snow dunes, as markers on the surface of the moving ice. Displacements of the surface features are mapped by selecting small image areas centered on distinct features, or by dividing a large area of densely featured glacial surface into a grid of areas, and searching a subsequent image for matching areas using a cross-correlation algorithm. Interpolation of the peak correlation values allows the displacements to be measured to subpixel accuracy, resulting in very precise velocity measurements. Cross-correlation is also applied to provide image coregistration in areas devoid of bedrock exposures. In such areas, subtle large-scale topographic undulations in the ice surface, related to underlying bedrock structure, may be correlated by using large image areas and low-pass filtered images. Both types of applications are demonstrated, using Ice Stream D and Ice Stream E in West Antarctica as test areas. A high-resolution map of the velocity field of the central portion of Ice Stream E, generated by the displacement-measuring technique, is presented. The use of cross-correlation software is a significant improvement over previous manually-based photogrammetric methods for velocity measurement, and is far more cost-effective than in situ methods in remote polar areas.