Mapping site amplification with the dense recording of ambient vibration for the city of Lucerne (Switzerland) – comparison between two approaches

Abstract

Abstract Reliable site-specific amplification information can be retrieved using earthquake-based methods, involving the deployment of seismic networks, which, however, may need to operate for years. Hence, time- and cost-effective approaches using ambient vibrations are gaining popularity. Among such techniques, the hybrid standard spectral ratio method (SSRh) uses simultaneous recordings of ambient vibrations translated into a prediction of local seismic response with a limited number of instrumented sites with earthquake ground-motion data. The canonical correlation approach (CC) collates measured empirical amplification with its proxy computed from ambient vibrations (horizontal-to-vertical spectral ratios, HVSRn) at a training set of seismic stations, to predict site response at points without earthquake recordings. We apply both methods in the Lucerne area (Switzerland), located on a soft sedimentary basin, obtaining consistent results, comparable to amplification estimates from earthquake ground motions. The resulting linear amplification factors are significant (factor of 8–10 or more) at the fundamental frequency of resonance of the sediments (0.8-2 Hz). However, both techniques show systematic differences in the spatial and frequency domains. The CC method tends to underestimate the amplification at the fundamental frequency, while SSRh predicts higher amplification factors in the center and lower at the basin edges in comparison to the CC approach. The study discusses the impact of limited completeness of the calibration dataset, and variability introduced by factors like Vs30 model choice and inelastic behavior treatment for the CC method, as well as the influence of the measurement setup for the SSRh method.