The Effect of S-velocity Heterogeneity in the North American Mantle on Waveforms of Regional Surface Waves from the February 2008 Nevada Earthquake

We compare observed waveforms recorded at the USArray Transportable Array stations from the 02/21/08 Nevada earthquake (Mw 6.0) with synthetics through three-dimensional (3D) velocity models such as S20RTS and NA04. We calculate synthetics down to 17 s through 1) mode summation for a 1D model obtained by averaging velocity variations in the 3D model between the epicenter and the station (path-average method), and 2) the Spectral Element Method. We include the effects of lateral variations in crustal structure, mantle structure, Moho depth, and topography, but keep Q model fixed. Because the Nevada earthquake occurred almost at the center of the Transportable Array stations at that time, distances between the epicenter and stations are typically less than 1000 km. As expected, we find that body waves are more affected by mantle structure and surface waves more by crustal structure. In general, the 3D models predict the observed waveforms better than a reference 1D model, but some waveform fits need to be significantly improved by better 3D models of velocity heterogeneity and discontinuity depth, particularly of the Moho. We find that for global 3D models the spectral element method and path-averaging method produce similar waveforms but for higher-resolution continental-scale models the spectral element waveforms are markedly different from the path-averaging ones. We utilize the Nevada earthquake to explore these differences in an attempt to quantify the modeling error made when the path-averaging method is used to invert observed waveforms for 3D velocity heterogeneity.