TERRAIN CHANGE TIME MACHINE: STRUCTURE FROM MOTION OF ARCHIVAL AERIAL IMAGERY FOR FINE-RESOLUTION AND ACCURATE HISTORICAL DIGITAL ELEVATION MODEL GENERATION

Geomorphic, topographic, and landscape change analyses require detailed and accurate historic data: such analyses are crucial to characterizing anthropogenic impacts, to supporting conservation efforts, to hazard mitigation, and a variety of other applications. To derive information about the recent past (25-30-year timescale), geologists and geographers generally turn to Light Detection and Ranging (lidar) data, which provides 3D elevation information and whose archive generally reaches back into the 1990s. For long-term change detection (100-year timescale), another tool is needed. This tool is Structure from Motion (SfM) - an approachable and low-cost image analysis tool for constructing accurate 3D models from 2D imagery. When paired together with the U.S. government’s numerous aerial imagery holdings of various national, state, and local agencies, many of which can be found in the USGS Aerial Photo Single Frame (APSF) archive, SfM has the potential to unlock a treasure trove of high quality, fine-resolution 3D terrain data.

The perceived learning curve associated with aerial image analysis and SfM methods has deterred its wider utilization within the geosciences. While the science of SfM is rooted in traditional photogrammetric techniques, implementation of SfM methods requires only a rudimentary understanding of these concepts. Nevertheless, the quality of terrain data produced from historical aerial imagery depends largely on the application of specific photogrammetric techniques within the SfM methodological framework. This project synthesizes existing, disparate SfM workflows to standardize and document SfM best practices specifically for the generation of Digital Elevation Models from historical aerial imagery. The methodological findings of this study are highlighted, along with various lessons learned from several case studies.