Ground control point requirements for structure-from-motion derived topography in low-slope coastal environments

Vegetated coastal dunes grow as a result of feedbacks between vegetation and sand transport. Observing the coevolution of vegetation and the sand surface is therefore critical for unraveling the dynamics of coastal dune growth. Capturing synchronous topography and photography at high spatial resolution and high temporal frequency using traditional techniques (airplane-based aerial photography, LiDAR) is expensive and time-consuming. Structure-from-Motion combined with Multiview-Stereo, a photogrammetry workflow that uses low-cost, consumer-grade equipment, is an economical alternative to traditional collection methods. This workflow still requires the definition of ground control points (GCPs) — locations with known coordinates — to develop accurate digital surface models. In this contribution we address how the number of GCPs used impacts the accuracy of digital surface models. We flew a 9-foot single-line delta kite attached to a consumer-grade camera to photograph the beach and dune of Hog Island, VA, a site that contains 178 high precision GCPs over an area of ~0.025 km² (as part of an ongoing field experiment). We then processed the 318 photographs using Agisoft Photoscan and compared the elevation accuracy of digital surface models rendered using SfM, with varying GCPs, to points surveyed by a total station. Our results suggest that there is ‘diminishing returns’ when greater than 10 GCPs are used. Results from this study can be used to inform future Structure-from-Motion studies using UAVs or kites in flat, low-sloping coastal environments.