Evaluation of SfM for surface characterization of a snow-covered glacier through comparison with aerial lidar

Abstract

The combined use of unmanned aerial vehicles (UAVs) and structure-from-motion (SfM) is rapidly growing as a cost-effective alternative to airborne laser scanning (lidar) for reconstructing glacier surfaces. Here we present a thorough analysis of the precision and accuracy of a photogrammetric point cloud (PPC) constructed through SfM from UAV-acquired imagery over the spring snow surface at Haig Glacier, Alberta, Canada, the first of its kind in a glaciological setting. An aerial lidar survey conducted concurrently with UAV surveys was used to examine spatial patterns in the PPC accuracy. We found a median error in the PPC of −0.046 ± 0.067 m, with a 95% quantile of 0.218 m. Mean precision of the PPC was 0.199 m, with large spatially clustered outliers. We found an association between high-error, low-precision, and high-surface roughness in the PPC, likely due to illumination characteristics of the snow surface. Glacier surface reconstructions are important for geodetic mass balance measurements, giving key insights into changing climate where in situ measurements are difficult to obtain. The PPC errors are small enough that they would have minimal effects on total mass balance, should the technique be applied across the glacier.

Assessing UAV-based laser scanning for monitoring glacial processes and interactions at high spatial and temporal resolutions

“…In order to assess the accuracy of the setup in the present study, seven ground control points (deployed for the UAV-SfM surveys described in 3.5) were used to compare to the processed UAV-LS data. Due to the limited sample size of GCPs a comparison of the median and Normalised Median Absolute Deviation (NMAD) errors are reported for each day, as well as the median, NMAD, 68.3%, and 95% confidence intervals for all the data, as outlined by Höhle and Höhle (2009) and Bash et al (2020), to better analyse the non-normal distribution of error.…”

“…Each image set was then imported into Agisoft Metashape for image alignment, with a sparse point cloud first generated. Dense point clouds were then constructed using the "aggressive" depth filtering setting, as is common in glacial research (e.g., Tonkin et al, 2014;Jouvet et al, 2019;Bash et al, 2020), before producing DEMs for each survey day. No subsequent mesh or dense cloud smoothing was performed.…”

Comparison of UAS-Based Structure-from-Motion and LiDAR for Structural Characterization of Short Broadacre Crops

“…While the dominant GSD was ~1.8-2.6 cm, the author reported an RMSE of 0.3-0.4 m; however, that study also found that that the SfM accuracy varied by different surface features, with vegetation, water, and other textures causing more errors. Bash et al [73] analyzed the precision and accuracy of a UAS SfM point cloud versus airborne LiDAR data over a spring snow surface at Haig Glacier; with a GSD of 2.4 cm, most of the errors were within a range of −0.049 ± 0.111 m . Our M3C2 distance between UAS-based LiDAR and SfM point clouds resulted in distance RMSEs of 0.05-0.18 m, which seemed to be slightly better than those from previous studies.…”

Terrestrial and Airborne Structure from Motion Photogrammetry Applied for Change Detection within a Sinkhole in Thuringia, Germany

“…However, in densely vegetated areas (dense forest, bushes, or high grass), SfM-MVS can only provide points of the uppermost surface from the sensor's viewpoint, resulting in a surface model rather than a terrain model that is more easily derived from laser scanning data [4,12]. Finally, SfM has been applied at different scales, landscapes, and landforms such as volcanoes and lava movements, large planar regions, glaciers, rock glaciers, badlands, sinkholes, landslides, rivers, burned and aeolian landscapes, as well as at laboratory experiments to detect and analyze surface changes (e.g., [2,[17][18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33]).…”