Reconstructing a 3D model from geophysical data for local amplification modelling: The study case of the upper Rhone valley, Switzerland

The geometry of three-dimensional subsurface structures plays an important role in determining local seismic site effects as in the case of alpine valleys. Detailed knowledge of these structures is fundamental in seismic hazard and risk studies. In this study we investigate an area in the upper Rhone valley around Visp, in the southwestern part of Switzerland. A large dataset of geological and geophysical data, consisting of borehole logs, microtremor horizontal to vertical spectral ratios and shear-wave velocity measurements, was compiled to build a detailed 3D model of the subsurface. By combining fundamental frequency information from noise recordings and shear-wave velocity profiles, three main geophysical discontinuities were identified and their physical properties constrained through a stepwise process. First, the bedrock depth was estimated; in a second step a generic velocity model was defined and finally, combining all the available geological and geophysical information, we developed a 3D geophysical model. The model was compared with a local 3D geological model and a model derived from gravimetric data. The study area is a complex alpine valley where 2D/3D wave propagation phenomena occur. In such case a purely 1D response assumption is considered to be invalid. In order to test the 3D model, we modelled different ambient-vibration wave fields and compared observed and synthetic H/V spectral ratios. We slightly modified our 3D geophysical model in some areas based on this comparison. Finally, a good match between simulated and empirical spectral ratios corroborated the model. The results suggest that the use of ambient vibration techniques are a powerful and cost-effective tools to reconstruct three-dimensional models of the subsurface. Finally, we used the 3D model to predict amplification of earthquake ground motion in the basin. Again, the match between observed and modelled amplification at the locations of the seismic stations is good. This allows us to map amplification inside the study area.