Calibration of correction factors for source, attenuation and site effects to predict ground shaking across Italy

This doctoral research focuses on improving the spatial prediction of ground shaking across Italy by developing empirical correction factors for source and attenuation effects and creating a new national-scale Vs30 map to account for site conditions. These tools are designed to support and enhance the generation of empirical ground-shaking scenarios at both regional and local scales. The work is framed within the fields of applied seismology and seismic hazard assessment. Its goal is to improve the accuracy of ground motion predictions and reduce the uncertainty associated with the classical ergodic assumption still widely used in empirical Ground Motion Models (GMMs). Ground motion is the result of the combination of three fundamental components: the earthquake source, the wave propagation path, and local geological conditions (site effects). While traditional empirical GMMs allow for fast and reliable ground motion estimates, they rely on the ergodic assumption, which implies that variability observed across multiple sites is representative of the variability expected at a single site over time. However, practically, this assumption tends to overestimate uncertainty and fails to capture regional geological and tectonic peculiarities. In this context, the regionalization of source and attenuation effects is based on the analysis of residuals, defined as the logarithmic differences between ground motion observations and model predictions, and their decomposition into repeatable components following the methodology of Al Atik et al. (2010): between-event residuals (δBe), representing source effects not captured by the GMMs; site-to-site residuals (δS2S), associated with site effects not explained by the reference model; and within-event residuals (δWes) corrected for source and site, which may include additional effects not accounted for by the model, such as regional propagation phenomena or source directivity. In this thesis, the δBe and δWes components are spatially modeled to identify homogeneous zones and derive regional correction factors for source and attenuation. The entire process is supported by a robust dataset, in particular the ITACAext database (Brunelli et al., 2022), which includes more than 37,000 accelerometric and velocimetric recordings from over 1,800 earthquakes across Italy. As a predictive model, the ITA18 GMM (Lanzano et al., 2019), developed for shallow crustal earthquakes in Italy, is selected as the national reference. A large part of the work of this thesis has been spent on the evaluation of site effects, through a new nationwide map of the shear wave velocity in the uppermost 30 m (Vs30), constructed through parametric linear regressions and geostatistical analysis. To this end, approximately 15,000 shear wave velocity (Vs) profiles were collected and organized into a georeferenced archive, derived from regional microzonation studies and surveys conducted by freelancers. Furthermore, several site proxies, both raster and vector-based, were compiled and analyzed for their correlation with Vs30 at multiple spatial resolutions. A new lithological classification was developed and used, together with slope, as predictors in the final Vs30 model. The result is a comprehensive framework for generating regionalized empirical ground motion scenarios applicable to various purposes such as the generation of ground shaking maps, the reconstruction of historical events without recordings, and the production of scenarios for possible future events. The methodology represents a robust and efficient alternative to physics-based simulations, providing faster computation and improved consistency with observations. In conclusion, this thesis contributes significantly to the evolution of ground motion prediction models, promoting a data-driven, non-ergodic, and scientifically sound approach capable of leveraging Italy’s rich geological and seismological data to meet the challenges of more accurate and realistic seismic modeling.