Image stereoscopic models for the advanced analysis of landslide deformation and thickness

The analysis of surface displacement and velocity fields from time series of terrestrial optical images is a useful tool for monitoring gravitational instabilities. It allows to define the state of instability of the slope, its evolution in time, its spatial coverage and to identify if the movement is progressing. Other types of information can also be extracted from landslide surface velocity fields such as the tangential and normal deformation or the strain fields that highlight areas of compression/extension (Travelletti et al., 2014) that can even allow to assess the mechanical properties of the moving mass (Baum et al., 1998). However, applying such advanced approaches necessitates to be able to compute the 3D displacements and deconvolute the normal and tangential displacements.

Landslide ground motion can be measured by various geodetic techniques either in-situ and point-based, or remote and giving access to spatially distributed information. In this study, we privileged a low-cost remote sensing method based on the use of a Single Lens Reflex (SLR) cameras. We acquired data at high frequency (i.e., time-lapse photography) from two fixed cameras at the Montgombert landslide.

The velocity fields were extracted from a time series of 13 images by applying the TSM (Tracing Surface Motion; Desrues et al. 2019) code. To detect tangential and normal displacements, we developed a methodology to construct the 3D displacements directly from the correlation results from the pairwise combination of the two monoscopic velocity fields, and further conducted a deformation analysis.

To estimate the thickness of the moving mass from the 3D displacements derived from the stereoscopic optical images, we propose a methodology based on the law of mass conservation (i.e., displacement incompressible) by invoking the rheology of the material involved (Booth et al., 2013). In order to take into account, in this model, a more complex slip geometry, we introduced a disbonding parameter that marks the presence of a dislocation area at the top limit of the moving mass which traduces a non-zero velocity at the sliding surface.

We present the methodology of reconstruction of the 3D displacements with a stereoscopic approach and of estimation of the landslide thickness by applying them to the Montgombert use case (Savoie, French Alps). The calculated displacement fields are consistent with in-situ data and the estimated depths, suggesting a shallow sliding, are consistent with geotechnical information.