High-accuracy UAV photogrammetry of ice sheet dynamics with no ground control

Unmanned aerial vehicles (UAVs) and structure from motion with multi-view stereo (SfM–MVS) photogrammetry are increasingly common tools for geoscience applications, but final product accuracy can be significantly diminished in the absence of a dense and well-distributed network of ground control points (GCPs). This is problematic in inaccessible or hazardous field environments, including highly crevassed glaciers, where implementing suitable GCP networks would be logistically difficult if not impossible. To overcome this challenge, we present an alternative geolocation approach known as GNSS-supported aerial triangulation (GNSS-AT). Here, an on-board carrier-phase GNSS receiver is used to determine the location of photo acquisitions using kinematic differential carrier-phase positioning. The camera positions can be used as the geospatial input to the photogrammetry process. We describe the implementation of this method in a low-cost, custom-built UAV and apply the method in a glaciological setting at Store Glacier in western Greenland. We validate the technique at the calving front, achieving topographic uncertainties of ±0.12 m horizontally ($\sim \mathrm{1.1}\times$ the ground sampling distance) and ±0.14 m vertically ($\sim \mathrm{1.3}\times$ the ground sampling distance), when flying at an altitude of ∼ 450 m above ground level. This compares favourably with previous GCP-derived uncertainties in glacial environments and allows us to apply the SfM–MVS photogrammetry at an inland study site where ice flows at 2 m day⁻¹ and stable ground control is not available. Here, we were able to produce, without the use of GCPs, the first UAV-derived velocity fields of an ice sheet interior. Given the growing use of UAVs and SfM–MVS in glaciology and the geosciences, GNSS-AT will be of interest to those wishing to use UAV photogrammetry to obtain high-precision measurements of topographic change in contexts where GCP collection is logistically constrained.