%0 Article
%J Bulletin of the Seismological Society of America
%D 2020
%T Effect of Random 3D Correlated Velocity Perturbations on Numerical Modeling of Ground Motion from the Source Physics Experiment
%A Scalise, M.
%A Pitarka, A.
%A Louie, J. N.
%A Smith, K. D.
%N 1
%P 139--156
%V 111
%1 10.1785/0120200160
%K SW4
%X Explosions are traditionally discriminated from earthquakes, using the relative amplitude of compressional and shear waves at regional and teleseismic distances known as the P/S discriminant. Pyle and Walter (2019) showed this technique to be less robust at shorter distances, in detecting small-magnitude earthquakes and low-yield explosions. The disparity is largely due to ground motion from small, shallow sources being significantly impacted by near-surface structural complexities. To understand the implications of wave propagation effects in generation of shear motion and P/S ratio during underground chemical explosions, we performed simulations of the Source Physics Experiment (SPE) chemical explosions using 1D and 3D velocity models of the Yucca Flat basin. All simulations used isotropic point sources in the frequency range 0-5Â~Hz. We isolate the effect of large-scale geological structure and small-scale variability at shallow depth (\<5 km), using a regional 3D geologic framework model (GFM) and the GFM-R model derived from the GFM, by adding correlated stochastic velocity perturbations. A parametric study of effects of small-scale velocity variations on wave propagation, computed using a reference 1D velocity model with stochastic perturbations, shows that the correlation length and depth of stochastic perturbations significantly impact wave scattering, near-surface wave conversions, and shear-wave generation. Comparisons of recorded and simulated waveforms for the SPE-5 explosion, using 3D velocity models, demonstrate that the shallow structure of the Yucca Flat basin contributes to generation of observed shear motion. The inclusion of 3D wave scattering, simulated by small-scale velocity perturbations in the 3D model, improves the fit between the simulated and recorded waveforms. In addition, a relatively low intrinsic attenuation, combined with small-scale velocity variations in our models, can confirm the observed wave trapping and its effect on duration of coda waves and the spatial variation of P/S ratio at basin sites.