A SEMI-AUTOMATED METHODOLOGY FOR USING HISTORICAL AERIAL PHOTOGRAPHY TO CREATE HIGH-RESOLUTION ELEVATION AND ORTHOIMAGERY PRODUCTS OF SOUTH CASCADE GLACIER, WASHINGTON, USA

South Cascade Glacier has one of the longest vertical aerial photographic records of any glacier in the United States (1955-Present). Its 60 year annual and seasonal glacier mass balance record as part of the USGS Benchmark Glacier Program is theorized to be representative of alpine glacier change in mid-latitude maritime climates. Many historical aerial photographs of this glacier and others in Washington State have been scanned and made available as digital products. Recent efforts of the USGS Glaciers and Climate Program aim to create a semi-automated process to derive high resolution digital elevation models (DEM) and orthorectified imagery from these historical photographs.

Using the Python library OpenCV, a semi-automated method was developed to batch process historical aerial imagery. 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. Processed photographs are then passed to Agisoft Photoscan structure from motion (SfM) software through the Photoscan Python library to derive DEMs and orthoimagery. For ground control points (GCPs), photo-identifiable points are extracted from 2015 orthoimagery (0.22m vertical accuracy, 0.16m horizontal) to georeference the historical imagery. With the exception of the placement of GCPs, the process is entirely automated with Python.

We have created a time-series of surface elevation change from historical imagery spanning 1955 to 2001. Elevation products derived from satellite imagery (2008-2016) are also co-registered to the 2015 survey to extend the record period to the present. This time-series is used to calculate the geodetic mass balance for South Cascade glacier.

Preliminary results provide DEMs with 2 m spatial resolution and vertical accuracies of ~0.5m. We plan to collect better spatially-distributed GCPs in 2017 to decrease the products’ positional errors and uncertainties. Future endeavors and research include incorporating feature tracking and identification utilities from OpenCV to automate georeferencing. Using this method, we hope to expand this effort to other glaciers in the coterminous United States.