Analysing feature-tracking methods to derive flow rates at John Evans Glacier, Canadian High Arctic
Item statusRestricted Access
Harcourt, William David
Deriving estimates of glacier velocity over interannual timescales is a major constraint on understanding recent changes to glaciers cross the globe. In the Arctic, there is an even greater need for such analyses due to the effect of Arctic Amplification, and the accelerated reduction in glacial mass. Glacier velocity is estimated by measuring the relative displacement of image patches across an image scene, and this may be computed using a number of image matching techniques. Current algorithms and software compute this step using a range of different methods, the most common of which are Normalised Cross Correlation (NCC), Orientation Correlation (OC), and Phase Correlation (PC). This study aims to compare these image matching techniques, and their implementation within three existing software; ImGRAFT; COSI-corr, and CIAS, each chosen because it has been recently created (ImGRAFT), or it has been widely used in the literature (COSI-corr and CIAS). It was found that COSI-corr performs the best when measuring deviation over assumed stable ground, whereas PC compared most accurately to field data. NCC performed the worst out of the algorithms, and ImGRAFT performed the worst out of the software, although ImGRAFT achieved greater coherence on the ice surface. The results presented here will be useful in understanding biases in feature-tracking algorithms, and to approach a feature-tracking analysis with caution due to the potential difference with another algorithm. Future research should investigate the new pre-processing methods, and ways to combat the reduced number of Landsat images at the start of the 21st Century.