Assessing the Accuracy of Glacier DSMs from UAV imagery
Overview
Digital Surface Models (DSMs) created from UAV-based photogrammetry are being increasingly used to quantify ice volume change and glacier motion. This paper by Gindraux et al. (2017) assesses the accuracy of the created models with specific reference to glacial surfaces.
Study Area & Acquisition
Three glaciers in the Swiss Alps were studied, across Summer, Winter and Autumn. Ground Control Points were measured with a dGPS, the mean accuracy being ~2cm, and the UAV was then flown over the glaciers at an approximate height of 115m, carrying a customised Canon camera, GPS and IMU.
Methodology & Workflow
Results
A vertical accuracy of 0.1-0.25m and horizontal accuracy of 0.03-0.09m can be achieved for DSMs when using an optimal number of Ground Control Points (GCP). Above this density, additional GCPs do not improve the accuracy of the DSM.
Greater distances to a GCP were found to have a detrimental impact on the accuracy at a rate of 0.09m per 100m, though there is large scatter and other factors are therefore important.
The lack of structure in fresh snow made it impossible to orient the images and therefore the creation of DSMs was not possible. However, snow that was one day old could be orientated in 19 of 20 images. Excluding fresh snow, the surface texture did not have any significant influence on accuracy.
Similar DSM accuracies can be achieved over all seasons.
References and Links
Gindraux, S., Boesch, R. & Farinotti, D. (2017) Accuracy Assessment of Digital Surface Models from Unmanned Aerial Vehicles’ Imagery on Glaciers. Remote Sensing, 9, (2), 186. Open Access Article here.
