S023-08
Inversion for Shallow Elastic Structure using Wind, Pressure and Seismic Data

Wednesday, 9 December 2020: 16:30
Virtual
Toshiro Tanimoto, University of California Santa Barbara, Santa Barbara, CA, United States and Jiong Wang, University of California, Santa Barbara, Department of Earth Science, Santa Barbara, CA, United States
Abstract:
When surface pressure is large, seismic noise between about 0.01 Hz and 0.05 Hz is mostly excited by wind-related surface pressure changes. Pressure and seismic data in this frequency band are useful for deriving shallow elastic structure in the upper 50-100 m of the Earth. For example, we demonstrated that we can estimate VS30 by this approach (Wang and Tanimoto, 2020; Tanimoto and Wang, 2020).

However, there exist inherent conceptual difficulties in modeling this seismic-wave excitation process because winds consist of a mean shear flow and turbulence which result in two different parts for the excitation source. The mean flow should act like a moving pressure source (e.g., Sorrells, 1971) while turbulent parts should act like a stochastic source (e.g., Sorrells and Goforth, 1973) as the pressure field becomes heterogeneous.

We show that wind data provide key information that can clarify this situation because we can identify time intervals when winds are nearly uni-directional. For such time intervals, we are justified to use the moving pressure source model.

We show that our previous assumption that high-pressure time intervals are coincident with strong, uni-directional wind time intervals is not necessarily supported by wind data. Therefore, we modify our approach by first identifying time intervals of uniform wind directions from wind data and then apply our inversion algorithm (Tanimoto and Wang, 2019, 2020).

Comparison of our inversion approach with and without wind data shows, however, that the key observables for inversion do not differ very much. It appears that, with a level of uncertainties in current pressure and seismic data, the benefit of using wind data does not show up explicitly in practice. It still provides confidence as we can justify the use of a moving pressure source model and minimize the complexity associated with turbulence.