2022 年 87 巻 795 号 p. 487-497
It is important in the spectral inversion analysis to take into account the local inhomogeneity of the attenuation structure. In this paper, we present a two-step spectral inversion method based on a Bayesian model to estimate source parameters, local attenuation structure, and site effects from observation records. It is expected that the introduction of a Bayesian model allows us to estimate appropriate solutions even for a small number of observation records and to evaluate local attenuation structure. The applicability of the proposed method is discussed through an experiment using actual observation records and numerical estimates.