Shear-wave splitting is one of the fundamental tools in seismic imaging - but with OBS data, it can be hard to get good results. There have been some great successes, some mixed results, and some non-reports from large experiments. This work is from an active proposal and the code is available on my github page.

So what did I do differently?

- Realistic models of anisotropy in the splitting operator. Shear-wave splitting typically relies on a splitting operator that is exact for a constant anisotropy along the ray path. Variations can be inferred from variations in the “apparent” splitting results from different propagation paths. I’ve constructed an approach for SKS phases (the standard phase used in teleseismic studies) to do two things 1) multiple layers are in the operator, not inferred on the backend, 2) rotations in the fast direction within layers. See this paper for why even small rotations can matter.

- Bayesian statistics handle the waveform fitting and account for the covariance in the noise. See this paper for the idea I worked off.

- The splitting operator is inferred from many waveforms at once, not as independent station-event pairs. This matters a lot for low signal-noise ratios by constraining the source-time function for a common event.

So does it work better? Here’s a comparison with the minimum energy method’s recovery of splitting parameters at different signal-to-noise ratios. Panel A shows examples of data at different signal-to-noise ratios, and the other two panels are the averaged recovery over twenty trials with both methods. Verticle lines are the minimum signal-to-noise ratios for good splits proposed by this paper. Orange dots are above the blue dots - this new approach really helps!

This project is active and the first paper is currently in preparation for the NoMelt site. Two sets of results were also be shown at the 2023 Fall AGU meeting -

  • Lau Basin to test if there is aligned melt in the mantle

  • The Old Orca site, to test how plate motion relates to flow in the asthenosphere

Results from the NoMelt site will be in review soon.