Paper detail

Site effects in an alpine valley with strong velocity gradient: interest and limitations of the 'classical' BEM

Seismic waves may be strongly amplified in deep alluvial basins due to the velocity contrast (or velocity gradient) between the various layers as well as the basin edge effects. In this work, the seismic ground motion in a deep alpine valley (Grenoble basin, French Alps) is investigated through various 'classical' Boundary Element models. This deep valley has a peculiar geometry ("Y"-shaped) and involves a strong velocity gradient between surface geological structures. In the framework of a numerical benchmark [21-23], a representative cross section of the valley has been proposed to investigate 2D site effects through various numerical methods. The 'classical' Boundary Element Method is considered herein to model the strong velocity gradient with a 2D piecewise homogeneous medium. For a large incidence angle, the transfer functions estimated from plane SH waves are close to the one computed with shallow SH point sources. The fundamental frequency is estimated at 0.33 Hz (SH wave) and the agreement with previous experimental results (spectral ratios) is good. Comparisons between 1D and 2D amplification are then performed: the values of the fundamental frequency and the corresponding amplitude are larger in 2D. Converting frequency domain results into the time domain, we underline surface waves generated at the valley edges and the directivity effect for the amplified wave-field. In the time domain for plane SV-wave, we also computed the ground motion for a strong seismic event (M=6): time duration and peak ground velocity are found to be 3 times larger than for the input signal. Such 2D models involving basin effects are then capable to recover the high amplification level measured in the field. Nevertheless, to deal with complex 3D basins as also proposed in the "ESG" benchmark [21-23], the capabilities of the classical Boundary Element Method are limited. As shown recently [43,47,52], such improvements as the fast multipole formulation (FM-BEM) may be a promising alternative for future 3D simulations.