brennanbrunsvik/THBI_MCMC

Joint Inversion of Receiver Functions and Surface Waves Using Markov Chain Monte Carlo

This may not be the most up-to-date version of the code. Please contact us at brennanbrunsvik@ucsb.edu for more information.

This code performs a joint inversion of receiver functions and surface wave data using a Markov Chain Monte Carlo (MCMC) inversion. This is a Trans-dimensional, Hierarchical, Bayesian Inversion (THBI).

If you use this code, please cite Brunsvik et al. (2024) and Eilon et al. (2018). See also Golos et al. (2024) and Petruska and Eilon (2022).

Running the examples

1. Run the example synthetic joint inversion by executing example_github/RUN_example.m.
2. Once the synthetic inversion is running, you can modify the eastern US example, applied to real data, found in the ENAM folder.

Dependencies

Required depending on your data types

• Surface wave phase velocity: Mineos (contact author for access). Working on integration with Josh Russell's public version, but it's not ready yet: Mineos_synthetics.
• Surface wave ellipticity: HV code from Tanimoto and Tsuboi (2009, GRL). Public release pending.
• Receiver functions: PropMat. Alternatively, Telewavesim is partially supported.
• $H-\kappa$ stacks, including radial anisotropy: hk_anis.

Other dependencies

These were placed in the ~/MATLAB/ folder. Refer to a0_STARTUP_BAYES.m for setup details.

• borders
• brewermap
• fastBSpline
• Seizmo
• seis_tools
• myFUNCTIONS
• timeout: install on Mac with "brew install coreutils". Already on Linux.

Setup steps

1. Update paths in a0_STARTUP_BAYES.m and obtain necessary dependencies.
2. Compile codes.
   1. All Fortran forward modelling codes.
   2. mex code in the fastBSpline folder to speed up spline calculations (optional).
3. Run the synthetic test in the example_github folder.
4. For real data applications, use the ENAM folder as a starting point. Modify "matguts" and evdata1_database.m to load your data from your computer.
5. HPC
   1. Use the "transfer" folder to migrate code from your work machine to your HPC.
   2. Modify ENAM/batch code (Slurm scripts for UCSB clusters) for your HPC setup.
   3. Use retrieve_results_example to get results from the HPC and format them (expect ~100 Mb/station).
6. The 3d_models folder creates a 3D model from the many 1D station PDFs.

Expected behaviors

• IO optimization: For systems with many cores, Fortran IO operations need a folder in RAM ("ramDriveTHBI") for efficiency. This should mount automatically.
• Parameter updates: update_bayes_inv_parms.m will modify your inversion parameters depending on your STAMP.
• run_one_chain.m is complex and hard to change because it handles many cases where Mineos fails. Test thoroughly if modified.

For any further assistance or to obtain dependencies, feel free to contact the author.

References

1. Brunsvik, B., Eilon, Z., & Lynner, C. (2024). Plate‐scale imaging of eastern US reveals ancient and ongoing continental deformation. Geophysical Research Letters, 51(12), e2024GL109041. https://doi.org/10.1029/2024GL109041
2. Eilon, Z., Fischer, K. M., & Dalton, C. A. (2018). An adaptive Bayesian inversion for upper-mantle structure using surface waves and scattered body waves. Geophysical Journal International, 214(1), 232–253. https://doi.org/10.1093/gji/ggy137
3. Golos, E. M., Brunsvik, B., Eilon, Z., Fischer, K. M. (in revision). A new view of the lithosphere-asthenosphere system in the Southwestern United States. JGR: Solid Earth.
4. Petruska, J., & Eilon, Z. (2021). Distributed Extension across the Ethiopian Rift and Plateau Illuminated by Joint Inversion of Surface Waves and Scattered Body Waves. Geochemistry, Geophysics, Geosystems, e2021GC010179.