Validating 3D Seismic Velocity Models Using the Spectral Element Method

As seismic instrumentation, data storage and dissemination and computational power improve, seismic velocity models attempt to resolve smaller structures and cover larger areas. However, it is unclear how accurate these velocity models are and, while the best models available are used for event determination, it is difficult to put uncertainties on seismic event parameters. Model validation is typically done using resolution tests that assume the imaging theory used is accurate and thus only considers the impact of the data coverage on resolution. We present the results of a more rigorous approach to model validation via full three-dimensional waveform propagation using Spectral Element Methods (SEM). This approach makes no assumptions about the theory used to generate the models but require substantial computational resources. We first validate 3D tomographic models for the Western USA generated using both ray-theoretical and finite-frequency methods. The Dynamic North America (DNA) Models of P- and S- velocity structure (DNA09-P and DNA09-S) use teleseismic body-wave traveltime residuals recorded at over 800 seismic stations provided by the Earthscope USArray and regional seismic networks. We performed systematic computations of synthetics for the dataset used to generate the DNA models. Direct comparison of these synthetic seismograms to the actual observations allows us to accurately assess and validate the models. Implementation of the method for a densely instrumented region such as that covered by the DNA model provides a useful testbed for the validation methods that we will subsequently apply to other, more challenging study areas.