Approaches to model non-uniqueness and site complexity for non-invasive shear-wave depth profiling
Abstract
Shear-wave velocity (VS) depth profiling and associated seismic site classifications were performed at 15 sites across Metro Vancouver, British Columbia using passive seismic and surface wave methods. Inversion model parameters are constrained at each site using nearby geodata in combination with developing regression models of shear-wave velocity with depth for three primary stratigraphic units. Statistical methods such as a Bayesian Information Criterion are applied post-inversion to evaluate models between and within varying parameterizations. Data evaluation metrics, including the use of microtremor horizontal-to-vertical spectral ratios (MHVSRs), are applied to identify two common deviations from the simple case of normally dispersive laterally homogeneous soils typically associated with surface-wave methods: lateral variations and velocity inversions. Lateral site variability is overcome by using the spatial variability in MHVSR peak frequency to sub-divide the site into quadrants for which quadrant-specific dispersion curve inversion provides reliable site classification for each quadrant. Velocity inversions are captured by performing inversions using partial fundamental mode dispersion curves. Partial inclusion of apparent-mode dispersion estimates is a reasonable compromise to modelling velocity inversions, providing a site classification between that of removing the apparent mode estimates (minimum velocity inversion modelling) and wrongfully treating apparent mode estimates as the fundamental mode (maximum velocity inversion modelling). These accessible approaches overcoming lateral site variability and apparent-mode dispersion estimates related to velocity inversions are proposed to obtain reliable seismic site classifications.