Investigation of the deep subsurface of the Swiss Molasse basin using ambient vibrations and earthquake recordings

Abstract

Passive seismic methods represent a low-cost and non-invasive alternative to the use of active sources (e.g. explosive) and invasive drilling activities. Passive seismic methods aim at estimating the soil parameters as a function of depth, in particular the S-wave velocities. The data analysed by passive seismic methods include ambient seismic vibrations and earthquakes, recorded using single stations or arrays. Site-effect surveys and microzonation generally investigate the shallow subsurface using ambient seismic vibrations and earthquake recordings in the frequency range from 0.5 to 20 Hz. The depth of the estimated velocity profiles is between 200 and 300 meters depending on the site and the methods used. Tomography studies, using arrays deployed at regional and global scales, investigate the deep structure of the Earth down to 100 or more kilometres, but the velocity profiles have low or no resolution in the first kilometres. In this thesis, we aimed at filling the investigation gap between the velocity profiles for site-effect analysis and the ones for tomography by estimating the velocity profiles from the free surface down to several kilometres using ambient seismic vibrations and earthquake recordings collected at several sites in Switzerland. Ambient seismic vibrations were recorded at Herdern (Canton Thurgau) and Schafisheim (Canton Aargau), two sites within the Swiss Molasse basin, the Alpine foreland basin located north of the Alps. The Swiss Molasse basin extends from Lake Geneva (south-west) to Lake Constance (north-east). The collected seismic data was analysed using techniques developed for site investigation studies of shallow structures: single-station methods (horizontal-to-vertical spectral ratio and estimation of Rayleigh wave ellipticity) and array processing techniques (three-component high-resolution f-k, Modified SPatial AutoCorrelation and Wavefield Decomposition). The measured Rayleigh and Love wave dispersion curves were jointly inverted with the Rayleigh wave ellipticity angles to determine the S-wave velocity profiles down to large depths of several kilometers. The deployment in Herdern was performed using two arrays of increasing size with maximum interstation distances of 900 m and 29.6 km, respectively. At the second site in Schafisheim, three arrays were deployed with maximum interstation distances of 346 m, 2.4 km and 26.8 km, respectively. In Schafisheim, in addition to the array measurements, a dense single station survey was performed to map lateral variations of the structure in the investigated area. While site Herdern provided clear dispersion curves that could be used to invert the velocity structure, site Schafisheim showed different branches of the same mode due to the lateral variability of the structure. The survey of single-station measurements helped to separate the structure into two main parts, explaining the different branches of dispersion curves of the same mode. To verify the potential of earthquake recordings in investigating the subsurface, we analysed the seismic events recorded at 23 stations (short-period, broadband and strong motion sensors) of the Swiss Seismic Network. The data consist of local and regional earthquakes with magnitudes between 2.5 and 4.6. Using the signal after the shear-wave arrival, we computed the Horizontal-to-Vertical spectral ratio curves and used the Hybrid Heuristic Search algorithm (HHS - Nagashima et al., 2014) to invert the S-wave velocity profile. HHS is a combination of Simulated Annealing and Genetic algorithm. Constraints for the shallow layers are needed to obtain stable inversion results. The constraints are velocity profiles at the site of the seismic stations obtained from site-characterization measurements, performed by the Swiss Seismological Service (SED) in the past decade. We tested the inversion method by allowing more or less liberty to the initial near-surface velocity profile. Recurrent features of the performance of the inversion were found that we classified in three classes. The inverted velocity profiles from HHS, using different initial constraints and liberties, investigate the subsurface down to depths between 245 and 5930 m, depending on the site. For some sites, the velocity profiles of the shallow layers can be improved. For 16 out of 23 sites, we can conclude that the HHS inversion for H/V curve with a-priori constraints for shallow structure works well to determine S-wave velocity structure down to the seismological bedrock.