Shear wave velocity imaging of the Avignonet landslide (France) using ambient noise cross correlation

The Avignonet landslide affects a 2 by 2 km area covered by clayey deposits. This paper presents the use of the seismic ambient noise cross-correlation technique to retrieve a 3-D model of the shear wave velocity of the area. Seismic ambient noise was recorded during 15 days at 13 stations located on the landslide. Cross correlations computed between the vertical components of all station pairs allow the retrieval of the Rayleigh wave Green's functions and the estimation of their group velocity dispersion curves in the 1.7-5 Hz frequency range. At frequencies lower than 1.5 Hz, the anisotropy of the wavefield strongly influences the apparent Rayleigh wave velocities. Moreover, the analysis of the convergence of the cross correlations shows that at frequencies higher than 5 Hz, the recording time length was not sufficient for the cross correlation to be stable. These 1.7-5 Hz passive group dispersion curves are complementary to the ones computed from shot signals in the 3-7 Hz frequency range. A tomographic inversion of the resulting 1.7-7 Hz Rayleigh wave group dispersion curves provides local group dispersion curves at each cell of the tomographic grid. These are inverted with a neighborhood algorithm to retrieve the 3-D model of the landslide. Despite the complex wave propagation in the eastern part of the landslide and the sparse ray coverage, estimated velocities and first-order features are in good agreement with previous investigations.