Adjoint Waveform Tomography of the Crust and Uppermost Mantle of the Western United States for Improved Waveform Simulations and Source Characterization

Arthur J. Rodgers, Claire D. Doody, Andrea Chiang, Lion Krischer, Christian Boehm, Michael Afanasiev, & Nathan Simmons

Submitted August 12, 2021, SCEC Contribution #11288, 2021 SCEC Annual Meeting Poster #010

We are developing a model of three-dimensional radially anisotropic seismic structure of the crust and uppermost mantle of the Western United States (WUS) with adjoint waveform tomography. This region spans the diverse geologic/tectonic regions from the Pacific Ocean to the Great Plains and Mexico to Canada. We use the Salvus software from Mondaic, which includes a spectral-element method solver for forward and adjoint simulations and integrated python-based workflow to facilitate computationally intensive and repetitive tasks. We use the global SPiRaL model (Simmons et al., 2021) for the starting model, which fits long-period waveforms very well. We follow a multiscale approach by first inverting long-period waveforms in the period band 50-120 seconds, and then reducing the minimum period to 40, 30, 25, and 20 seconds. We perform many (20 or more) iterations within each period band, stopping when the waveform misfit is no longer reduced. Within each period band we start with strong smoothing of gradients (sensitivity kernels) then relax smoothing to allow more detailed structure to be inferred. The model (still in progress) infers large-scale 3D structure that is broadly similar to previous studies and correlates with major physiographic features (e.g. Sierra Nevada, Basin and Range, Colorado Plateau, Snake River Plain and Yellowstone Hotspot). Inferred 3D structure is broadly similar to the SPiRaL model but has much stronger heterogeneity in the crust and uppermost mantle and more spatially concentrated anomalies. The resulting model provides significant improvement to waveform fits in the period band 20-120 seconds over the starting model for independent validation data. We show how the model can be used for source moment tensor inversions with 3D Greens functions and result in better waveform fits and higher confidence solutions. Our WUS model is used as a starting model for a smaller-scale more detailed model California and Nevada (California and Nevada Adjoint Simulations, CANVAS, Doody et al., this meeting). Our WUS model could provide a background starting model for large-scale long-wavelength structure for other sub-domains.

Key Words
adjoint waveform tomography

Rodgers, A. J., Doody, C. D., Chiang, A., Krischer, L., Boehm, C., Afanasiev, M., & Simmons, N. (2021, 08). Adjoint Waveform Tomography of the Crust and Uppermost Mantle of the Western United States for Improved Waveform Simulations and Source Characterization. Poster Presentation at 2021 SCEC Annual Meeting.

Related Projects & Working Groups
SCEC Community Models (CXM)