← Back to Sessions

Selection of a Starting Model for Adjoint Tomography of the Pacific Northwest

Description: 

High-resolution seismic tomography models are invaluable in characterizing subsurface geologic structures, quantifying seismic hazards, and understanding broader tectonic processes. Theoretical and computational advances over several decades have enabled full-waveform inversion techniques that extract more information from recorded seismograms than classical methods. In parallel, numerical methods have facilitated the rapid calculation of the forward and adjoint (time-reversed) wavefields in earthquake seismology, their interaction allowing the construction of the gradient of a chosen misfit function and the iterative improvement of the velocity model. It is well known that starting models play a crucial role in the success of full-waveform inversions. Here, the Stephenson et al. (2017) Cascadia seismic velocity model is systematically tested as a candidate starting model for future adjoint tomography of the Pacific Northwest. We use SPECFEM3D Cartesian (Komatitsch & Tromp, 1999) to calculate synthetic seismograms using the candidate starting model and a catalog of over 500 M4.0+ regional earthquakes recorded from 2012–2023. We assess starting model quality by comparing travel time anomalies between the synthetic and observed waveforms. We explore modifications to the starting model based on physical a priori constraints and make comparisons with a one-dimensional Earth model (ak135; Kennett et al., 1995). We focus on crustal and uppermost mantle structure which spans depth ranges (0–60 km) that are important to accurately estimate ground motions and seismic hazards, but may be poorly resolved in existing velocity models. While similar analyses with alternative candidate starting models remain necessary, this work will ultimately enable the development of future Pacific Northwest adjoint tomography models.

Session: 3D Wavefield Simulations: From Seismic Imaging to Ground Motion Modelling [Poster Session]

Type: Poster

Date: 5/2/2024

Presentation Time: 08:00 AM (local time)

Presenting Author: Haiyang

Student Presenter: No

Invited Presentation: 

Authors