Simultaneous Bayesian estimation of multisegment fault geometry and complex slip distribution: application to the 2024 Noto Peninsula earthquake

SUMMARY For the inversion of crustal deformation data, where both the fault geometry and slip distribution need to be estimated, attempts have been made to estimate them simultaneously within the framework of Bayesian inference. In these methods, an analytic expression for the posterior distribution of the parameters cannot be obtained depending on the settings of the probability distribution. The Monte Carlo method is often used to approximate the probability distribution by several samples. However, Monte Carlo sampling becomes computationally expensive as the dimensionality of the parameters increases, making it challenging to apply these methods to inverse analyses targeting large-scale earthquakes with complex rupture patterns, where the parameters to be estimated are high-dimensional. In this study, we developed an efficient algorithm for the simultaneous Bayesian estimation of fault geometry and slip distribution that is robust against high-dimensional parameters, facilitating the estimation of multisegment fault geometry and complex slip distribution. When comparing the developed method with a conventional method in a numerical experiment where the parameters were high dimensional, it was confirmed that the accuracy and convergence of the estimation results were improved while reducing computational costs. As an application example of the developed method, we estimated fault geometry and slip distribution from the crustal deformation observed during the 2024 Noto Peninsula earthquake, considering models with one-, two- and three-fault segments. By comparing the results, we demonstrated that the three-segment model is the most plausible, emphasizing the importance of improving the estimation methods to accurately estimate multisegment fault geometry with a complex slip distribution.