Electronic Theses and Dissertations


Jiayan Tan



Document Type


Degree Name

Doctor of Philosophy


Earth Sciences

Committee Chair

Charles Langston

Committee Member

Eunseo Choi

Committee Member

Christine A Powell

Committee Member

Mitchell M Withers

Committee Member

Randel T Cox


Crustal velocity structure is critical to understanding near-surface geological features and seismic wave propagation of shallow earthquakes. We implement a traditional method, the joint inversion of ambient noise dispersion and teleseismic P wave receiver functions, to explore Oklahoma's 3-D shear wave velocity structure. The use of short-period surface wave dispersion and high-frequency receiver functions constrains upper crust shear wave velocity structure. The inverted result correlates well with geological provinces, and model-generated synthetic waveforms have a good match with observed seismograms for a regional event in Oklahoma. We develop a shapeDTW (shape Dynamic Time Warping) waveform inversion method to investigate 1-D crustal velocity structure in Tanzania, East Africa, and the Cherokee Platform in Oklahoma using broadband waveforms from regional earthquakes. This method pairs corresponding seismic phases and estimates the optimal time shifts between compared waveforms as a function of time position in the waveforms. Cycle-skipping issues in waveform inversions can be effectively alleviated by reducing the estimated time shifts in warped waveforms. We prove that the shapeDTW method performs better in waveform alignments than classic DTW. Results show that this method can get reasonable solutions from inaccurate initial models, and model-generated synthetics agree well with observed waveforms. Use of these methods opens up the possibility of using the full displacement field of regional earthquakes to infer crustal structure to supplement active source data. We employ a differentiable DTW method, softDTW, to the misfit functions used in full-waveform inversion to explore 3-D velocity structure. The application of adjoint sources built using the gradients of the softDTW misfit function can help update numerous model parameters in 3-D models. Although cycle-skipping problems induced by inaccurate initial models are significant challenges for conventional waveform inversion methods, the large mismatch in time and amplitude between compared waveforms can still be effectively reduced by softDTW full-waveform inversions as determined by two tests. The use of shape descriptors also shows a better performance for waveform inversions. The test results suggest that this DTW method is a robust way to align seismic waveforms and has a good potential for imaging complex velocity structures.


Data is provided by the student

Library Comment

Dissertation or thesis originally submitted to ProQuest.


Embargoed until 2024-07-19

Available for download on Friday, July 19, 2024