Title: Seismic site-response variability : from site-classification to soil non-linear behaviour
Abstract: Local geology can strongly affect seismic ground motion at the surface. These so-called site-effects can be evaluated either numerically by simulating the seismic wave propagation or empirically using earthquake recordings analyses or statistical correlations between site parameters and site effects. This thesis concerns the improvement of site effect evaluation and the consideration of the seismic site response variability. This work is mainly based on the analysis of earthquake recordings. First, we analyzed the site response variability between sites for similar ground motion incident intensity (weak motion) and then, we analysed the site response variability between events in one site caused by non-linear soil behaviour. Finally, we used inversion method to find the depths where soil non-linear behaviour mostly occurs.We analyzed various earthquake recordings from the KiK-net database in Japan (more than 46 000), which is composed of more than 688 surface/borehole instruments. The vertical array configuration allows the computation of the empirical site response in borehole condition. This database was chosen because of its large amount of instrumented sites located at sediments, its large amount of accelerometric data and the existence of characterisation of the shear and compressive wave velocity profiles down to the borehole depth. The analysis of the seismic response variability per site classes indicated which parameters must be measured to improve the site-effect assessment. In the linear range, the parameters are the Vs profile gradient calculated until 30 m (B30) or 100 m depth (B100) and the fundamental resonance frequency of the site (f0). These additional parameters to Vs30 can be used to reduce the between-sites surface motion variability (such as in GMPEs, Ground Motion Prediction Equations).Besides, the analysis of inter-event site response variability caused by non-linear soil behaviour showed the relevant parameters for analysis of site effects in non-linear range: B30, The maximal amplification (Apred) and the associated frequency (fpred). This analysis showed as well: - that the PGA (Peak Ground Acceleration) is a relevant parameter for non-linear site effect assessment. - that, whatever the site, non-linear soil behaviour affects the site response curves from moderate acceleration (75 cm/s2 at the down-hole station). - that only information on the non-linear soil behaviour of the superficial layers is enough to fully assess the non-linear site responses. This last conclusion may have a large impact for non-linear soil characterisation. It has been confirmed by inversion of linear and non-linear borehole site responses and comparison of the obtained Vs profiles. At the same time, the present work showed : - that the depth from which non-linear soil behaviour has no influence on site response depends on the site and on the intensity of the seismic input motion. - that careful attention needs to be taken when inverting data from vertical arrays. Sensitivity analyses are a powerful tool to evaluate the resolution of the inversion considering the available information and the soil parameters that are well solved during the inversion. - That the combined used of sensitivity analysis with comparison of transfer function, without inversion, could be enough to assess the depth where the non-linear soil behaviour mostly take place