Over the past ten years, there has been an explosion in both the quantity and quality of satellite and airborne imagery available in glaciology. Traditional methods of extracting information from such images are often labour-intensive and error-prone. Fortunately, there has also been a growth in automated methods for processing imagery, using tools borrowed from computer vision.
Ice front tracking
While at the Scott Polar Research Institute, I was involved in a project which used edge detection algorithms to track the front position of a large number of tidewater glaciers in eastern Greenland. By using daily images from the MODIS instruments, we were able to achieve near-daily time series of front positions over a ten year period, for over thirty glaciers. The results, as published in a paper in the Journal of Geophysical Research, showed that there is an apparent boundary around 69°N, with glaciers north and south of this exhibiting distinct patterns of behaviour.
Recently, I have been collaborating with Chris Williams on methods for mapping crevasse fields near glacier fronts. By studying the evolution of crevasses with time, we hope to learn more about the processes leading up to the calving of icebergs. Using a Fourier-based method, we have obtained promising results from a wide variety of data sources, including satellite radar, airborne lidar, and more conventional visual images. We have even applied the technique to the south polar region of Enceladus, one of the icy moons of Saturn.