Geologic Publications for Mount Rainier

A semi-automated approach to derive elevation time-series and calculate glacier mass balance from historical aerial imagery

[ back to previous page ]

Author(s): Erin N. Whorton, Alexander O. Headman, David E. Shean, Evan McCann

Category: PRESENTATION
Document Type: Abstract #C53D-07
Publisher: American Geophysical Union, Fall Meeting 2017
Published Year: 2017
Volume:
Number:
Pages:
DOI Identifier:
ISBN Identifier:
Keywords:

Abstract:
Understanding the implications of glacier recession on water resources in the western U.S. requires quantifying glacier mass change across large regions over several decades. Very few glaciers in North America have long-term continuous field measurements of glacier mass balance. However, systematic aerial photography campaigns began in 1957 on many glaciers in the western U.S. and Alaska. These historical, vertical aerial stereo-photographs documenting glacier evolution have recently become publically available. Digital elevation models (DEM) of the transient glacier surface preserved in each imagery timestamp can be derived, then differenced to calculate glacier volume and mass change to improve regional geodetic solutions of glacier mass balance. In order to batch process these data, we use Python-based algorithms and Agisoft Photoscan structure from motion (SfM) photogrammetry software to semi-automate DEM creation, and orthorectify and co-register historical aerial imagery in a high-performance computing environment. Scanned photographs are rotated to reduce scaling issues, cropped to the same size to remove fiducials, and batch histogram equalization is applied to improve image quality and aid pixel-matching algorithms using the Python library OpenCV. Processed photographs are then passed to Photoscan through the Photoscan Python library to create DEMs and orthoimagery. To extend the period of record, the elevation products are co-registered to each other, airborne LiDAR data, and DEMs derived from sub-meter commercial satellite imagery. With the exception of the placement of ground control points, the process is entirely automated with Python. Current research is focused on: one, applying these algorithms to create geodetic mass balance time series for the 90 photographed glaciers in Washington State and two, evaluating the minimal amount of positional information required in Photoscan to prevent distortion effects that cannot be addressed during co-registration. Feature tracking and identification utilities in OpenCV have the potential to automate the georeferencing process. We aim to develop an algorithm suite that is flexible enough to enable its use for many landscape change detection and analysis problems.

View Report:
Sorry, no PDF is available - Contact Scott to get a copy

Suggested Citations:
In Text Citation:
Whorton and others (2017) or (Whorton et al., 2017)

References Citation:
Whorton, E.N., A.O. Headman, D.E. Shean, and E. McCann, 2017, A semi-automated approach to derive elevation time-series and calculate glacier mass balance from historical aerial imagery: Abstract #C53D-07, American Geophysical Union, Fall Meeting 2017,