Estimation of shallow S-wave structures from the quasi-transfer spectrum and Bayesian inversion using microarray data at a deep drilling site at Chungnam National University, Korea

Estimating shallow S-wave structures is crucial for assessing the level of strong ground motion amplification. In this study, we applied a transdimensional and hierarchical Bayesian inversion to surface wave dispersions obtained from a passive f-k analysis of geophone microarray data. The dispersion data obtained from the microarray f-k analysis have uncertainties resulting from array conditions, possibly including subsurface lateral heterogeneities, array sizes, source distributions, and measurement lengths. The Bayesian approach provides complete posterior probability distributions of model parameters, enabling a robust investigation that explains the uncertainties. Moreover, the results from a quasi-transfer spectrum (QTS) analysis and geological core information were used during the interpretation of our inversion results. Ambient noise data were collected at a deep drilling site at Chungnam National University, South Korea. The estimated S-wave structure indicated a rapid transition in velocity increase at depths of 4–30 m from ∼ 200 m/s to ∼ 3,600 m/s, corresponding to a rapid variation from sandy soil to hard rock in the core samples. Despite relatively less constrained features, the QTS peak frequencies provided a reliable range of bedrock-depth estimations. We also tested and compared dispersion data and noise characteristics with different array configurations in time and space to better understand possible causes of data variations and present a practical guideline for future experiments in sites with similar conditions. Together with possible effects on dispersion data due to the variability, this study shows a comprehensive procedure to interpret shallow S-wave structures and their uncertainties accounting for the data uncertainties throughout the Bayesian inversions.