Based on the scattering field theory

the forward algorithm of the horizontal-to-vertical spectral ratio (HVSR) establishes a connection between the surface HVSR and the properties of the site’s soil layers. By combining this with ground motion observations at the surface

the inversion of the velocity structure of the site’s soil layers can be achieved. However

most current HVSR inversions employ traditional deterministic inversion methods

which lead to significant non-uniqueness in the inversion results and make it difficult to assess their uncertainties. This study proposes a Bayesian inversion method for soil layer velocity structures to enable the assessment of uncertainties in the inversion parameters. This method combines Bayesian principles with the forward algorithm of the earthquake horizontal-to-vertical spectral ratio (EHV)

using the S-wave components of strong ground motion observations as the data source to achieve the inversion of the site’s soil layer structure. The effectiveness and applicability of the proposed method are verified through numerical examples. The results indicate that the proposed Bayesian inversion method can effectively identify the velocity structure of the site’s soil layers and comprehensively assess the uncertainties of the parameters in the inversion model.