Transdimensional Bayesian Joint Inversion of Surface Wave Dispersion and Horizontal-to-Vertical Spectral Ratio Curves for the Southern Korean Peninsula

The seismic noise mainly consists of surface waves, including Rayleigh and Love waves, which carry information on velocity structure beneath the earth. Dense seismic arrays provide opportunity to obtain the data and constrain entire structures of not only the crust and upper mantle but also the shallow local subsurface. Data retrieved from the surface waves, namely the surface wave dispersion (SWD) curve and Horizontal-to-Vertical Spectral Ratio (HVSR) curve, are utilized to build the velocity models in the southern Korean Peninsula. However, the inherent non-uniqueness, meaning that different models may produce similar misfit, exists in standard inversion algorithms where only one typical model is obtained. Based on Bayesian inference, transdimensional inversion methodology provides an alternative way to jointly invert the SWD curve and HVSR curve, where the number of layers is treated as unknown to allow for transition during the inversion. Therefore, the data automatically determine the complexity of the model displaying a parsimonious feature that the simpler model is more preferable. Then, more rigorous uncertainty can be estimated together with the hierarchical framework to account for data uncertainties. With this parameterization and the hierarchical scheme, models are sampled by the reversible jump Markov chain Monte Carlo algorithm and an ensemble of these samples represents the posterior probability distribution. We apply the inversion algorithm to synthetic SWD and HVSR curves to validate the transdimensional algorithm. In the future, these two kinds of field data will be jointly inverted to obtain Vs and Vp profiles simultaneously beneath broadband stations in the southern Korean Peninsula. Results will then be compared with previously published models of crust and upper mantle.