Full-Waveform Validation of a 3D Seismic Model for Western US
Abstract:Since the initiation of tomographic studies in the 1970s, geoscientists have advanced the art of inferring 3D variations in the subsurface using collections of geophysical (primarily seismic) observables recorded at or near Earth’s surface. Advances have come from improvement and enhancement of the available data and from research on theoretical and computational improvements to tomographic and generalized inverse methods. In the last decade, utilizing dense array datasets, these efforts have led to unprecedented 3D images of the subsurface. Understandably, less effort has been expended on model validation to provide an absolute assessment of model uncertainty. Generally models constructed with different data sets and independent computational codes are assessed with geological reasonability and compared other models to gain confidence. The question of “How good is a particular 3D geophysical model at representing the Earth’s true nature?“ remains largely unaddressed at a time when 3D Earth models are used for both societal and energy security. In the last few years, opportunities have arisen in earth-structure imaging, including the advent of new methods in computational seismology and statistical sciences. We use the unique and extensive High Performance Computing resources available at Los Alamos National Laboratory to explore approaches to realistic model validation.
We present results from a study focused on validating a 3D model for the western United States generated using a joint inversion simultaneously fitting interpolated teleseismic P-wave receiver functions, Rayleigh-wave group-velocity estimates between 7 and 250 s period, and high-wavenumber filtered Bouguer gravity observations. Validation of the obtained model is performed through systematic comparison of observed and predicted seismograms generated using the Spectral Element Method, which is a direct numerical solution for full waveform modeling in 3D models, with accuracy of spectral methods.