GNGTS 2019 - Atti del 38° Convegno Nazionale

90 GNGTS 2019 S essione 1.1 Bruno et al. , 2010 b, 2013; Villani et al. , 2015a, b, 2017, 2018b). Models are parameterized through velocity grid nodes and traveltimes are computed by a finite-difference technique, which accounts for transmitted, diffracted and head waves. The best-fit P-wave velocity model is searched by a Monte Carlo inversion method. We performed a second refraction tomography based on a linear inversion approach and on the SIRT image reconstruction technique (Gilbert, 1972). The theoretical travel times are computed via ray-tracing method, using a starting velocity model that is iteratively upgraded to minimize the misfit between observed and compute travel times. The data for MASW analysis were extrapolated from acquired dataset using common receiver configurations with 24 geophones and a minimum offset of 1 m. Each common shot data was analyzed generating a frequency-phase velocity spectrum and picking the fundamental mode. The dispersion curve was successively inverted to generate a S-wave 1-D vertical profile. To constrain the space of the model-search, a range of parameters are defined a-priori . At last, a pseudo-2D shear-velocity section was made by the interpolation on aligning 1D models at the midpoint of each CSG (Yinhe et al. , 2009). By using the Vs pseudo-section and the Vp tomographic model, we computed the Vp/Vs ratios and the Poisson coefficient to infer the degree of saturation of the soils and in the fault zone and to facilitate comparison of velocity and stack sections with the electrical resistivity data. Seismic and resistivity models were also processed through the k-means algorithm, that performs a cluster analysis for the bivariate data set to individuate relationships between the two sets of variables (Bernardinetti et al. 2017). The result is an integrated model with a finite number of homogeneous clusters, which therefore helps to distinguish and interpret different geophysical facies. Results. Both reflection image and velocity models clearly show a WSW-dipping normal fault zone, located at 40-80 m that matches the VF fault surface rupture. Analyzing the depth converted reflection section, the VF fault is clearly detected by an abrupt reflection truncation in the 5-80 m depth range, while synsedimentary activity is inferred by shallow reflectors that show back-tilt increasing with depth in the fault hangingwall. Reflection truncations indicate a cumulative fault throw, within the fan deposits, of about 30 m, which agrees with the cumulative throw since Middle-Late Pleistocene calculated by Villani et al. (2017). Furthermore, an unknown NW-dipping normal fault is detected in the eastern part of the profile, located at 120- 150 m. The reflection image clearly points out the occurrence of alternating high-amplitude reflections that we interpret as sandy-gravels layers related to different phases of alluvial fan accretion. Those layers show variable dip and incremental steepening at depth due to recent fault activity. The two Vp tomographic sections (i.e. linear and non-linear) have a resolution depth of about 50 m, show comparable features, but the non-linear velocity model better describes Fig. 2 - acquisition site. The red arrows indicate the 30 October 2016 coseismic rupture.