UAV imagery and in-situ measurements for structure-from-motion snow depth mapping over the Laurichard rock glacier, France - surveyed in 2017
Unmanned aerial veichle (UAV) imagery and in-situ field measurements at the Combe de Laurichard, France (45.01ºN, 6.37ºE, 2500 m a.s.l.) were collected to explore uncertainties in mapping snow depth with structure-from-motion and multi-view stereo 3D reconstruction in an alpine area. Repeat UAV surveys were flown on each survey date to create multiple elevation models to determine the precision (i.e., repeatability) of SFM-MVS elevation models. This measure of uncertainty can be used to determine the precision of SFM-MVS snow depths using a model of error propagation. It also can illustrate how uncertainty in the SFM-MVS snow depths and elevation models vary spatially. The repeated snow-cover (snow-on) elevation models (6 in total) were acquired on June 1, 2017, and the snow-free (snow-off) elevation models (7 in total) on October 5, 2017. These elevation models were derived by performing SFM-MVS reconstruction using Agisoft PhotoScan. The UAV imagery was surveyed using a DJI Phantom 4, which flew in pre-programmed parallel flight paths with 75% side and top image overlap. The flying height of the UAV was approximately 60 m above ground level. Artificial targets were used for SFM camera calibration and georeferencing using the RGF93 / Lambert-93 projection and the NGF-IGN69 vertical datum (EPSG::5698). Validation data was collected by measuring topographic heights (i.e., check points) using a real-time-kinematic (RTK) global navigation satellite system (GNSS) survey with an accuracy < 2 cm (at 1 σ). At each check point location, snow depths were measured using an avalanche probe to a maximum 3 m depth. This RTK-GNSS survey was also used for collecting ground control points (GCPs). The position of the base-station was corrected using the PUYA reference station, which is located approximately 19 km from the study area. Additionally, to compare the uncertainty of SFM-MVS snow depths in stable and active deforming terrain (i.e., rock glacier creep), a mask of the rock glacier area was mapped using the UAV derived imagery and elevation models.