Datensatz

Zusammenfassung

Earthquake simulations, geodetic and seismological data analyses, and field surveys of fault outcrops indicate that the slip distribution on the coseismic fault is self-similarity. This characteristic can be used for fault slip estimation as prior information for the regularization of an ill-posed problem. In this study, we introduced the characteristic using the von Karman autocorrelation function and developed the method to simultaneously estimate the fault slip and its correlation length, which has been treated as a hyperparameter from the ground displacement. The method was constructed based on Bayesian inversion using Hamiltonian Monte Carlo to evaluate parameters' uncertainty quantitively. We conducted numerical experiments using a random slip distribution whose correlation length is assigned to validate our method. Under the assumption of a dense observation network, the input slip and the correlation length were retrieved correctly. Furthermore, the experiment assumed that the observations are distributed at only the fault's down-dip side. As a result, although, the longer correlation length has been explored than the result under the dense observations, its posterior distribution peaked near the assumed value. This presentation will also show the results of an application to the 2016 Kumamoto earthquake (Mw 7.0). In addition, we will discuss an application for a numerical test of a megathrust in the Nankai trough subduction zone, which has a more realistic, 3D-distributed fault system.