Insights into the conditions of application of noise-based spectral ratios in a highly industrialized area: a case study in the French Rhone Valley

Abstract

SUMMARY The local ground motion amplification related to the geology at a specific site (i.e. the so-called site effects) may be classically quantified through the SSR (standard spectral ratio) technique applied on earthquake recordings. However, such a quantification might be challenging to carry out in low-to-moderate seismicity regions. Methods based on background ambient noise, such as noise-based standard spectral ratio (SSRn), might be of great interest in these areas. But noise-derived amplification is particularly sensitive to local anthropogenic sources, which may introduce biases in the evaluation of site effects by dominating the geological effects, especially for frequencies higher than 1 Hz. A hybrid approach (SSRh), developed to reduce biases in noise-based spectral ratios by combining classical earthquake-based spectral ratio (SSR) and SSRn, was recently introduced and relies on a site reference. We here investigate the applicability of SSRn and SSRh in a heavily industrialized environment in the Tricastin Valley (south-east France), where critical facilities are located. We continuously recorded ambient noise from 2020 February to March on a 400-sensor seismic array covering an area of about 10 km by 10 km. We demonstrate that SSRn and SSRh computed below 1 Hz are able to reproduce amplification factors provided by SSR. By contrast, at frequencies higher than 1 Hz, SSRn strongly deviates from SSR. SSRh shows closer results to SSR but presents a dependence to the choice of the local site reference, thereby questioning the possibility to use SSRh blindly to estimate the local amplification in our context. These discrepancies reflect the impact of local anthropogenic sources. We therefore introduced a two-step workflow to mitigate the influence of local sources. The first step is to define a characteristic time window to optimally isolate significant transient signals. The second step consists in selecting the time segments that do not contain these transients with a clustering-based approach. By applying this workflow, we were able to remove some strong anthropogenic transient signals likely to be generated by local sources at some sites and therefore to locally improve the amplification assessment through noise-based spectral ratios. However, stationary sources, whose impact cannot be removed through our procedure, remain a major issue. This study provides some insights into the application of SSRn and SSRh in noisy industrialized areas, especially regarding the impact of local noise sources. It illustrates the difficulty of having a procedure for mitigating the impact of these sources that is efficient everywhere inside such a complex anthropized environment, where different types of sources (including stationary sources) cohabit.