Methodology for producing hybrid Earth models with improved waveform predictions near the southern San Andreas fault and updates to SCEC UCVM API
Rasheed Ajala, Patricia Persaud, & Alan JuarezPublished August 15, 2020, SCEC Contribution #10660, 2020 SCEC Annual Meeting Poster #189 (PDF)
We evaluated the latest SCEC community models and hybrid versions of these embedded with hi-res basin-scale models at low frequencies utilizing medium-magnitude events not used in developing any of the models in the assessment. We benchmark the community models by performing a manual neighborhood search. The resulting model subspace includes topography, Ely geotechnical layer, and attenuation, or the lack thereof. For CVM-S 4.26 developed without topography, we experiment with different topographic representations – a default model that pulls the model up to the free surface, a 1D model extension, and a linearly-increasing extension. For the viscoelastic model, we test several Olsen attenuation ratios. Afterward, we smoothly embed our basin models developed using active-source seismic data into the best versions of the derived community models. We then similarly benchmark each hybrid model. The hybrid model subspace comprises a suite of models with varying embedding volume and smoothing boundary ratios. Using the relative waveform misfit functional to evaluate all model subspace, we find that our hybrid models outperform CVM-S 4.26 and CVM-H 15.1. Our hybrid model currently represents the best model in the Salton Trough available for higher frequency updates and hazard assessment. These research findings also support the development of local Earth models that are becoming increasingly popular with dense seismic arrays. Local models can be produced with relatively lower computational resources and embedded into the community models to create hybrid models with better waveform prediction ability.
To facilitate our research, we developed a C model blending library that enables the definition of standard window functions like the Cosine taper or Ormsby filter on arbitrary support polygons defined on regular grids. With this new tool, we can smoothly embed local models into regional models in non-cuboid volumes using regular gridded data formats. The volume of the embedded model can be, in principle, arbitrarily shaped to a certain degree. We implemented our library to UCVM and updated the API, thereby fostering the reproducibility and rapid construction of complex hybrid Earth models to improve earthquake hazard assessment.
Key Words
hybrid Earth models, SCEC UCVM development, ground motion prediction
Citation
Ajala, R., Persaud, P., & Juarez, A. (2020, 08). Methodology for producing hybrid Earth models with improved waveform predictions near the southern San Andreas fault and updates to SCEC UCVM API. Poster Presentation at 2020 SCEC Annual Meeting.
Related Projects & Working Groups
SCEC Community Models (CXM)