Ground Control Point Survey Strategy for Snow Depth Retrieval Using Drone-based Structure-from-Motion Photogrammetry

The drone-based structure-from-motion photogrammetry is recently gaining its popularity in the mapping of snow depth, owing to the efficiency and simplicity of this technique in the process of data collection. To retrieve snow depth accurately, a survey of Ground Control Points (GCP) for the snow accumulation field is often necessary. However, to the best of the authors knowledge, the influences of the GCP elevation accuracy and GCP survey strategies on retrieval of snow depth have not been investigated in depth. In this study, two sets of GCPs were surveyed using two different instruments (i.e., low-accuracy handheld GPS and high-accuracy total station). These two sets of GCP measurements were then used individually to generate Digital Surface Models (DSMs) before and after snowfall. We also divided each set of GCPs into two groups. For each set, one group of GCPs was used to generate the before-snowfall DSM and the other group was used for the after-snowfall DSM. The drone-based snow depth was then retrieved by subtracting the two DSMs. By comparing drone-based snow depth measurements with the concurrent in-situ measurements, we found that when using the same GCPs to generate the before and after snowfall DSMs, the accuracy of the snow depth retrieved with the total station was slightly higher than that of the snow depth retrieved with the handheld GPS. However, the contrast was marked when the GCPs for generating the DSMs before and after snowfall were different. The results suggest that when GCPs with high elevation accuracy are not available (e.g., when in a field with an ordinary handheld GPS unit), a better strategy for deriving accurate drone-based snow depth is to use the same GCPs for both before and after snowfall DSMs.