Probabilistic estimation of shear-wave velocity profiles and site classes from surface wave surveys for accelerometer stations in the southern Korean Peninsula

Shear-wave velocity (VS) profiles are presented using a probabilistic inversion framework for 21 permanent accelerometer stations in the southeastern Korean Peninsula. This region experiences relatively high seismicity. Accurate site-correction of recordings at these stations is essential for reliable seismic hazard analysis for nearby critical infrastructures including nuclear power plants. Surface wave dispersion data from multichannel analysis of surface waves (MASW) and microtremor array measurement (MAM) were jointly inverted using an ensemble-based stochastic approach that accounts for measurement errors, frequency gaps, and inversion parameterization. The resulting VS profiles characterize subsurface structures, extending from the surface to depths of up to 1000 m. Validation of the VS profiles was performed using direct borehole measurements and fundamental resonance frequencies derived from one-dimensional transfer functions (1DTF) and horizontal-to-vertical spectral ratios (HVSR). In most cases, the derived profiles closely match borehole data, and the resonance frequencies derived from 1DTF and HVSR also show strong agreement. From these profiles, we estimate site parameters including average VS of the top 30 m and the soil layer, and bedrock depth with their uncertainties. Using these parameters, probabilistic site classes were determined, ranging from soft soil (Class D) to rock (Class A), along with the interrelationships among the parameters themselves. By quantifying uncertainty through ensemble modeling, this result provides advantages over conventional deterministic methods. It effectively addresses the non-uniqueness in VS profile and site parameter estimations. Furthermore, realistic uncertainties can be incorporated into probabilistic seismic hazard analyses to improve reliability in ground motion predictions.