This one’s a new one for me.  Basically, the idea is to take the massive archive of oblique time-lapse imagery that the USGS has collected at Columbia Glacier, apply image processing techniques to derive velocity fields, correct these velocity fields using satellite-derived digital elevation models, and then feed the result into a model using inverse methods to look at the very short-term evolution of stresses near a tidewater terminus.  The banner image is my first crack at using the open source computer vision software OpenCV to tract features between image pairs.