A method for sub-pixel scale feature-tracking using Radarsat images applied to the Mertz Glacier Tongue, East Antarctica

Abstract Tracking features between pairs of satellite synthetic aperture radar (SAR) images using image cross-correlation methods is a powerful tool for defining regions of image movement such as ice flow of Antarctic glaciers. Unlike optical images, processed SAR images have various instrumental effects such as view angle distortions and lateral displacement errors caused by slant range mapping. Co-registration issues are problematic and need to be addressed if there are few/no suitable stationary tie points at or near sea level within the images. Methods are presented that use Radarsat ScanSAR images to reliably measure sub-pixel displacements with errors of ≲ ± 0.25 pixels but this estimate is target dependent. The techniques and issues are illustrated using examples from a long-term study of the Mertz Glacier Tongue in East Antarctica. For image pairs only a few days apart, this extensive structure forms an excellent grid-like test pattern which provides up to 5000 valid correlation matches. Images separated by 100's of days can then be processed in the same way to reveal many subtle and some large changes over time. The technique described is robust and may be widely applicable.