Regional seismic velocity models for Southern California based on travel time tomography with Poisson Voronoi cells parameterization

Abstract:

Improved models of seismic wavespeeds benefit many studies including deriving earthquake locations and focal mechanisms, simulating strong ground motion and various geomechanical applications. In this study we apply a recently developed seismic tomography method (Fang et al., 2020) with flexible parameterization based on Poisson Voronoi cells to derive seismic compressional (Vp) and shear (Vs) velocity models for Southern California. The analysis uses about 10 million first arrival picks of P- and S-waves from the SCEC data center, which we augment with more than 4 million S-wave picks between 2000 and 2019 using an automated procedure. The new tomography method avoids the need to explicitly regularize the often ill-conditioned system of equations that constitute the inverse problem by recasting it into a number of stochastic model subspaces. The procedure draws random samples from the dataset for each subspace to homogenize the data coverage and seamlessly implement a boot-strapping technique to estimate model uncertainties. Because the new method omits the damping and smoothing constraints required in conventional tomography, we also obtain a Vp/Vs ratio model by simply dividing the Vp and Vs results. The derived velocities correlate clearly with various geological features such as material contrasts across faults, sedimentary basins and batholiths. We validate the derived models by comparing simulations of wave propagation in the model, generated by tens of earthquakes located in various places, with observed waveforms. Detailed results will be presented in the meeting.

Fang, H.-J, Van der Hilst, R.D., De Hoop, M.V., Kotharis, K., Gupta, S., Dokmanic, I., Parsimonious seismic tomography with Poisson Voronoi cells: methodology and validation, Seismological Research Letters, vol. 91, p. 343-355, https://doi.org/10.1785/0220190141, 2020