3D seismic tomography of the Harmaliére landslide (French Alps) by interferometry

Harmaliére, in southern France, is among the most active alpine landslides, posing a risk to the neighbouring settlements and infrastructures. A retrogressive slide in March 1981, that displaced a volume in the order of 100 m 3 , was followed in April 2016 by a landslide two order of magnitude larger, followed by minor reactivations in 2017 and 2018. Local bedrock paleo-topography and sedimentary structures within the glacio-lacustrine sediment layer are suspected to have a role in determining the dynamics of this slide, characterized by episodic large displacements as opposed to slow and continuous mass movements registered in neighbouring sites (e.g. Avignonet). However, current state of knowledge of the subsurface is limited to low-resolution volumes and local 1D layered S-wave profiles.

Within the RESOLVE project, in May-June 2021 a dense 3D array of 100 three-component geophones has been deployed to record continuously ambient seismic noise for a one-month period. This was complemented by the acquisition of an active dataset using 100 hammer-strike sources, with offsets ranging from 0 to 900 metres. The vertical component of the the active dataset has been used to obtain a 3D P-wave velocity model by first-arrival traveltime tomography. Particularly challenging field conditions, e.g. thick vegetation and surface water, along with the low-power of the hammer sources, required dedicated processing to enhance the signal-to-noise ratio and allow for confident first-arrival pickings.

Super-virtual interferometry (SVI) has been applied to improve the quality of offsets larger than 400 m, which contain head-waves key for the imaging of the sediment-bedrock interface. SVI enhances critically refracted arrivals by stacking the cross-correlations of traces pairs sharing a stationary-path in common-receiver gathers, and then convolving the resulting station-pair Green’s functions with the appropriate virtual sources in common-source gathers. An azimuth-varying approach has been developed to adapt SVI to the 3D problem, reducing the number of cross-correlations and mitigating artefacts resulting from non-stationary paths contributions. The dataset obtained by constrained automatic picking on the SVI dataset has been used for first-arrival traveltime tomography, yielding an improved-quality tomographic volume at depths larger than 50 m along with lower final data misfit, thanks to the greater number of reliable long-offsets picks compared to the pre-SVI dataset.

The P-wave velocities obtained within the sediment body, as well as the inferred bedrock topography, are sensible and appear to be consistent with independent geophysical data. In order to complete the elastic characterization of the site, a S-wave 3D model will be reconstructed from the empirical Green's functions obtained by interferometry on 1-month long noise recordings, opening the way towards a joint passive-active high-resolution 3D elastic remote characterization of the landslide volume, and thus an improved understanding of its controlling factors.