Approaches to obtaining 1D Vs profiles at inversely dispersive and laterally heterogeneous sites using surface wave methods.

Abstract:

Metropolitan Vancouver in British Columbia, Canada, lies at the edge of the Late-Cretaceous Georgia sedimentary basin, 150 km East of the Cascadia subduction zone. As part of an ongoing Metro Vancouver seismic microzonation mapping project (metrovanmicromap.ca), earthquake site classification based on the average shear wave velocity (Vs) of the upper 30 meters (Vs30) is determined from invasive and non-invasive seismic methods. Surface wave array methods are used to generate dispersion estimates which can be jointly inverted with microtremor horizontal-to-vertical spectral ratios (MHVSR) to retrieve Vs depth profiles. The largest sources of Vs profile uncertainty arise at the data processing (dispersion curve) and inversion (model parameterization) stages. Velocity reversals can induce higher modes in dispersion estimates, and lateral heterogeneity such as dipping layers can generate variable dispersion estimates at 10-30 meter scales. These geologic complexities, coupled with the non-uniqueness and ill-posed nature of the inverse problem can affect the accuracy and reliability of Vs profiles. To mitigate impact to fundamental-mode dispersion estimates, we use surrounding Vs and stratigraphic data in conjunction with measured MHVSR amplification spectra to identify frequency bandwidths where the dispersion curve is susceptible to mode-mixing. Lateral heterogeneity is assessed by calculating gradients in MHVSR fundamental peak frequency across the site’s seismic arrays, allowing for multiple dispersion estimates in heterogeneous sites. Joint inversion of the verified fundamental-mode dispersion estimates and MHVSR amplification spectra are used to reduce Vs uncertainty with depth. We use a systematic inversion approach to retrieve optimal Vs profiles from multiple model parameterizations with their parameter bounds restricted using reference Vs profiles of three major stratigraphic units commonly found in Vancouver generated from a priori seismic data. Vs and Vs30 uncertainty is calculated from the suite of optimal Vs profiles. We find that the use of prior information at both processing and inversion stages reduces the overall complexity of surface-wave inversion and provides accurate (low variability) Vs30 estimates consistent with the expected geology.