Deriving glacier surface velocities from repeat optical images

Abstract

The velocity of glaciers is important for many aspects in glaciology. Mass accumulated in the accumulation area is transported down to the ablation area by deformation and sliding due to the gravitational force, and hence gla­cier velocity is connected to the mass balance of glaciers. It also contributes directly to the mass balance of calving glaciers because it is an important control of the ice discharge rate for such glaciers. Changing glacier velocities is an indicator of instable glaciers, and monitoring velocity over time can make people aware of possible hazards that may arise from instable glaciers. The movement of glaciers is also important for transporting material and for eroding the landscape. The focus of this thesis is to further develop image matching within glaciology. In image matching, images from two di.erent times are compared us­ing correlation techniques to derive glacier displacement over the time period. Most studies have concentrated on using image matching to derive glacier velocities instead of developing this method further. To be able to derive the densest possible velocity grids for all glaciers in the world, image matching methods over glacier surfaces have to be explored further. So far all images that have been used to derive velocity in glaciology have been high or medium spatial resolution images. Low resolution images cover large sections in one image, and this makes them suited for investigating the velocity of large areas such as Antarctic ice shelves. We derive velocities for Antarctic ice shelves using MODIS images with a spatial resolution of 250 m to test whether these images are suited for deriving ice shelf velocity. Because the accuracy is about one fourth of a pixel, and it is possible to use images acquired several years apart due to the low surface transformation, MODIS images are well suited for deriving velocity of Antarctic ice shelves and also to monitor their changes over time. We found when comparing di.erent image matching methods over dif­ferent glacier surfaces that the most commonly used method, normalized cross-correlation, generally performs worse compared to orientation correla­tion and the matching part of the program COSI-Corr. The only situation where normalized cross-correlation outperforms the two other methods are on narrow glaciers where small window sizes are needed. COSI-Corr per­forms best overall, but orientation correlation performs almost as well. In addition orientation correlation is the only method that manages to match striped Landsat images after the failure of the Scan Line Corrector. Both orientation correlation and COSI-Corr are considered to be methods well suited for global glacier velocity mapping. Normalized cross-correlation can supplement these two methods on narrow glaciers. The effort that has been put into developing image matching in glaciology since the start of this study, both in this study and in other studies, makes it possible to derive glacier velocities over large regions, and only computer processing time hinders automatic matching of glacier velocities worldwide. Global glacier velocities can give valuable insights. We show in this thesis that it can give information about how glaciers respond to climate change. Glacier velocity of .ve regions of the world with negative mass balance is derived, and in all regions the general glacier speed is decreasing over the last decades. In addition global glacier velocities can be used to understand glacier dynamics, and predict glacier hazards. It can be tested against gla­cier inventory parameters, and it can be used to estimate erosion rates and transport times.