Oral Presentation

The SCEC Community Velocity Model (CVM) offers 3D descriptions of seismic wave speeds in Southern California's crust and upper mantle. One is based on a sedimentary basin description from stratigraphic layering rules (CVM-S) which is heavily iterated by full waveform tomography incorporating ambient noise (Magistrale et al., 2000; Lee et al, 2014). Another is a geostatistical description of basin sediments combined with waveform and teleseismic tomography of the crust and upper mantle (CVM-H, Tape et al., 2010; Shaw et al., 2015). Here, we focus on enhancing the CVM-H, on current developments of tomographic models, and updates to the software (UCVM, Maechling et al., 2021) used to organize access to the CVM and related models. UCVM has the capability to combine different models - a critical function to support development of next generation tomographic models and CVMs.

The CVM-H includes a GTL, basement and Moho surfaces, with velocities described by parameterized 1D functions, kriging of dense well and stacking velocity data, and 3D full waveform, adjoint tomography. The result is a largely self-consistent, data rich model. However, seismic validation exercises show that such models must be iterated in order to achieve best waveform fits. This has been performed for CVM-S but not CVM-H. Thus, we seek to develop a new CVM that incorporates elements from both. For CVM-H. we extracted nine high-quality, detailed basin models with the goal of embedding them into the latest iteration of the CVM-S and other new starting models (including statewide CVM’s). These models would then be subsequently iterated to improve their fits with observed waveforms. New basin models are available via UCVM, and are prepared for hosting by IRIS. UCVM advances include four additional regional models (Albacore Offshore, Wasatch Front, SSIP Imperial and Coachella), a web-based model viewer, and steps to accommodate other community models (e.g., GFM).

Using joint body wave-surface wave tomography (Fang et al, 2016) we made advances in embedding models and in iterating composite starting models. The joint inversion code is being made suitable for the strong gradients present in CVM models. We gained an ability to embed smaller, well resolved tomographic models into the larger CVM-H and could successfully generate synthetic waveforms using such embeddings. Adjoint tomography infrastructure was greatly enhanced by open-source code (pyatoa, specfem3d) and data format updates (netcdf).