GNGTS 2019 - Atti del 38° Convegno Nazionale

52 GNGTS 2019 S essione 1.1 with a triangular slip-rate function whose activation time from the origin time will depend on the source position with respect to the specific receiver according to the directivity function Cd (Ben-Menahem, 1961). As for the inverse problem we implemented the Metropolis-Hastings sampler approach to investigate the model space parameter. For a given model m , the next candidate point is generated as m t = m t-1 + z where z is an increment random variable from a proposal distribution. The components of m are the rupture velocity v r , the rise-time t and slip distribution at the given control points. The best model parameter corresponds to the model that maximize the posterior distribution of the model space parameters, which is given by (1) Where d is data vector and m is the model vector selected in the model space Ω, r(m) is the priori distribution and f(d | m) is the likelihood function given by (2) and, (3) Where Nstaz is the number of available stations and Nt is number of points of the source time functions S i . Results and conclusions. The preferred EGF is the foreshock occurred on May 19, 2012, at 23:13:27 GMT with a Mw 3.8 earthquake. The STFs display the largest amplitudes and frequencies, between N200° and N230°, at stations located toward SW of the epicentre. Smoother functions are observed to the N-NE (Fig. 2). However, the complexity of the STFs as observed at various azimuths suggests a complex style of rupture propagation. We performed a direct modelling of the retrieved moment rates by trial-and-error approach. The best results are characterized by two main rupture propagation directions, along N225° and N103°, respectively. Fig. 2 - Observed STFs (black lines) and synthetics (red line) corresponding to the best solution obtained from the Bayesian approach. In each frame the station code, the source-to-station azimuth, and the correlation coefficient (bold) are reported.