Enhancing regional seismic velocity model with higher-resolution local results using sparse dictionary learning

Hao Zhang, & Yehuda Ben-Zion

Published September 10, 2023, SCEC Contribution #12831, 2023 SCEC Annual Meeting Poster #005

We use sparse dictionary learning to develop transformations between seismic velocity models of different resolution and spatial extent. Starting with data in the common region of both models, the method can enhance a regional lower-resolution model to match the style and resolution of the local higher-resolution results while preserving its regional coverage. The method is demonstrated by applying it to two-dimensional and three-dimensional and local and regional velocity models in southern California. The enhanced reconstructed regional models exhibit clear visual improvements, especially in the reconstructed ratios, and better correlations with geological features. Moreover, the reconstructed regional models outperform the original ones in comparison of simulated earthquake waveforms to observations. The improved fitting to the observed waveforms extends beyond the domain of the overlapping region. The developed dictionary learning approach provides physically interpretable results and offers a powerful tool for additional applications for data enhancement in earth sciences.

Key Words
Seismic Velocity Models, Multi-scale, Machine Learning, Dictionary Learning, Tomography

Citation
Zhang, H., & Ben-Zion, Y. (2023, 09). Enhancing regional seismic velocity model with higher-resolution local results using sparse dictionary learning. Poster Presentation at 2023 SCEC Annual Meeting.

Related Projects & Working Groups
Seismology