The integrity of critical facilities and networks during an earthquake is a vital component for the sustainable development of resilient societies and environments. When quantifying the risk of complex extended structures, spatial interactions between the different components of an engineered structure or among different structural elements within a system can play a crucial role in the integrated probabilistic seismic hazard and loss assessment. In the recent years, a significant research effort has been devoted to seismic hazard and structural vulnerability assessment, but the effect of the spatial variation of ground motion is generally not accounted for. Though, it is now increasingly recognized that the spatial variability of seismic load has an impact on the dynamic response of extended and multi-supported structures such as bridges, embedded galleries, pipelines and more generally on infrastructure networks or industrial plants comprising several adjacent buildings. Many past earthquakes have outlined spatially heterogeneous damage distribution over short distances (tens to hundreds of meters; e.g. Loma Prieta earthquake in 1989, Northridge in 1994,) even for similar engineering structures (e.g. Boumerdes, 2003 and Christchurch, 2010). Also, the post-processing of seismic data measured at dense arrays has shown that seismic motion exhibits considerable spatial variability even at local scale (Argostoli in Greece, Pinyon Flat & Parkfield in the US, Chi-Chi & Hualien in Taiwan, St Guérin in France…). The origin of such spatial variation of ground motion is ranging from near-fault rupturing effects, regional variations and site effects due to topography, sedimentary basins and stratigraphy as well as local spatial variability of soil properties. In order to account for variability in at local scale, it is necessary to describe the correlation structure of soil properties which is still an unresolved issue. The local variability of ground motion is generally described by a coherency function. At larger scales, the correlation of seismic intensity measures is considered. The integration of spatial variability into a comprehensive model, accounting for variability at different scales, from local to regional, remains a challenge. It is the ambition of this project to develop a methodology based on simple physically constrained models and efficient numerical tools for the generation of ground motions and seismic load at regional scale that incorporates the different sources of spatial variability. Such models allow assessing the reliability of the infrastructures including the local and regional spatial ground-motion variability and will be implemented in the integrated risk analysis chain.

The site for experimental assessment of spatial variability at local scale is the Argostoli basin situated in the Cephalonia Island in Western Greece. This site has been extensively instrumented over the last years within the framework of EU NERA (2011-2014) and ANR SINAPS@ (2014-2018) projects. More specifically, a dense seismological experiment comprising 62 velocimeters with interstations varying from 5 to 50 m and distributed from one edge of the basin to the other has allowed to record around 3000 earthquakes (with a high signal-to-noise ratio) to be recorded from 2011-2012. This very unique dataset has outlined large spatial variability of ground motion throughout the basin (Imtiaz, 2015; Theodulidis et al., 2017, Cultrera et al 2014). Taking advantage of the available site characterization and large dataset, EXAMIN intends to explore new methodologies to estimate the variability of soil properties as well as the seismic ground motion at the Argostoli valley based on ambient vibration dense arrays.

The physical soundness and the feasibility of the approach will be demonstrated by the application to a case study. The site for the demonstrative risk assessment at regional scale is the Grenoble basin that has been extensively studied by researchers from ISTerre (e.g. Chaljub et al., 2010, a more complete list of references is given in the appendix). While the ground motion is correctly reproduced up to 1 Hz, the high frequency component of the ground motion (above 1-2 Hz) is dominated by wave motion propagation within the most superficial geological layers. Importance of geological surface layers on the spatial variability of seismic response was evidenced through ground motion modelling of a magnitude 5.5 earthquake (Causse et al., 2009): seismic response is spatially variable and predicted acceleration spectra may exceed EuroCode8 (EC8) response spectra between 1 and 5 Hz. Numerous geophysical experiments and geotechnical boreholes collection over the past years are now enabling to build a “high resolution” 3D velocity model including near-surface layers spatially variable properties.