SCEC Award Number 18121 View PDF
Proposal Category Individual Proposal (Integration and Theory)
Proposal Title Development and Implementation of Full 3-Dimensional Seismic Tomography
Investigator(s)
Name Organization
Thomas Jordan University of Southern California
Other Participants Alan Juarez (grad student)
SCEC Priorities 4a, 4b, 4c SCEC Groups CXM, Seismology, GM
Report Due Date 03/15/2019 Date Report Submitted 05/06/2019
Project Abstract
 We have assembled software elements for automating F3DT on high-performance computers using the Hercules Toolchain. The F3DT toolkit includes subroutines for automated seismic source inversion, data extraction and an iterative tomography algorithm. We report on progress in developing several toolkits, and we raise one issue regarding how our work can best be coordinated with SCEC CME objectives.
Intellectual Merit The waveform variety observed the 3D environment of Southern California is much richer than that described by the 1D taxonomy of body waves and surface waves. Our goal is create a “full-3D seismology” that uses prior knowledge of 3D structure to interrogate seismograms for coherent groups of seismic energy that can be accurately measured (“good waveforms”) and have coherent, path-localized Fréchet kernels (“good sensitivity”). The algorithm we are developing recombines wavelets of the decomposed seismogram using weights derived by optimizing an objective function that measures both forms of goodness. The example in Fig. 4 was obtained by summing the wavelets of Fig. 3 in a way that optimizes the sensitivity between 5-15 km depth.
Broader Impacts The success of this project will provide SCEC with F3DT capabilities critical to the CyberShake project. It is supporting Alan Juarez's thesis project.
Exemplary Figure Figure 2. GSDF data processing: (a) Synthetic and observed seismograms and isolation filter. The isolation filter was obtained from the seismogram decomposition in Fig. 1. (b) Synthetic and data cross-correlograms. The cross-correlation function has many pulses at different lag-times. (c) Windowed cross-correlograms. (d) Filtered windowed cross-correlograms. These signals can be approximated with a five-parameter Gaussian wavelet. (e) GSDF measure the difference between one phase on both seismograms.