Tele-effect of geomorphological change on the spatial variability of the precision of SfM-MVS 3D point-cloud models

Earth-surface processes research is increasingly using the SfM-MVS (Structure from Motion and Multiple-View Stereophotogrammetry) method to model land surface change over time at a very fine-scale. However, the role of topographic change on the error calculated from “stable and fixed” Ground Control Points is under-documented and as far as the authors are aware, it has not been evaluated as yet. Therefore, the present study is an analysis of the variability inherent to the SfM-MVS method used for 3D terrain modeling, in a semi-controlled environment, comparing repeats of measurements, and repeats including topographic change in the laboratory scene, in order to assess the role of elevation change in the scene on the space that remains unchanged. The methodological framework involves varying the terrain morphology by adding 50 and 100 ml of sand to an originally horizontal sandbox, creating a mount in the centre. Then, the authors compared the different experimental surfaces and their repeats acquired by SfM-MVS, and using Gaussian Kernel Density Estimation (KDE). Results demonstrated that under stable and uniform flat surface conditions, the SfM method yields relatively consistent results (standard deviation variety less than 0.027 mm). However, when the experiments included the 50 ml and 100 ml mount of sand, the variability between repeats increased, even for location where no topographic change had occurred. The authors argue that the topographic variability is spreading the error, increasing it compared to the flat experiment. By extension, this consideration is essential, especially for research investigating topographic change such as landslide and other erosion and deposition processes, because the error propagation varies with the surface change, and relating erosion/deposition to topographic change needs to be done carefully.