SCEC Award Number 23178 View PDF
Proposal Category Individual Proposal (Data Gathering and Products)
Proposal Title Active-Source Adjoint Tomographic Update of CVM-H 15.1 in Salton Trough
Investigator(s)
Name Organization
Folarin Kolawole Columbia University Rasheed Ajala Columbia University
Other Participants One (1) graduate student (summer research internship)
SCEC Priorities 4b, 4c, 3b SCEC Groups GM, SAFS, CXM
Report Due Date 03/15/2024 Date Report Submitted 03/15/2024
Project Abstract
 Several decades have elapsed since the inception of full waveform inversion. Yet, the earthquake hazard community has not been able to benefit from this leap in seismic optimization. The main culprit behind the stagnation of progress is the computational cost required to generate Earth models with frequency content relevant to deterministic hazard analysis by structural engineers. Fortunately, recent developments in multiscale model synthesis have introduced a paradigm shift. We can now focus on the most risk-prone regions, i.e., high-population density cities, areas with expensive infrastructure, and sedimentary basins, and develop local high-frequency models in a more computationally tractable manner. Thus far in California, model synthesis studies, including ones that have created new merging approaches, have focused on relatively low-frequency Earth models. However, to realize the full utility of model synthesis, we need to develop local high-frequency models exceeding 1 Hz in frequency content. To achieve this goal, we are inverting active-source seismic data in Salton Trough to produce the first 3D 5 Hz Earth model in Imperial Valley. We use results from our previous research in model merging and space exploration to build a good starting velocity model. Due to challenges with land seismic data, such as noise and high-frequency content, existing models in the region will produce large kinematic errors in a 5 Hz wavefield inversion. Therefore, we adopt the optimal transport misfit functional with better convexity and convergence properties in our optimization. Once developed, our new model will significantly benefit earthquake science research in California.
Intellectual Merit We aim to develop the first 3D high-frequency Earth model in Salton Trough that utilizes active-source waveform data with frequencies up to 5 Hz. The results will have details that will be useful for seismic hazard assessment by satisfying structural engineers' frequency requirement and will also be imperative in further understanding the nature of continental extension in the region. The project's research goals align with SCEC science milestones A1-1, A2-1, and A3-6.
Broader Impacts The SCEC project provided support for two black early-career geoscientists. The research activities in the work have also assisted the postdoctoral Co-PI in training and professional development toward a tenure-track position in geophysics. We are committed to open science. All our previously funded SCEC projects have been disseminated to the community via peer-reviewed publications and computer programs. Once our results are finalized, we will do the same for the current research.
Exemplary Figure Figure 3. Summary of the convexity properties of some misfit measures as a function of kinematic model errors. (a) Comparison between an observed SSIP trace and synthetic trace filtered in 0.1 – 5 Hz showing the huge mismatch in the P arrival. (b) Comparison of the SSIP trace with snapshots of the negative time-shifted synthetic trace at selected time intervals. (c) Graph of the normalized misfit relative to the time shift for different misfit functionals. L1 – L1 norm, L2 – L2 norm, CC – Cross-correlation coefficient rescaled to the interval [0 1], W1 – Wasserstein norm (Optimal transport metric) with L1 transport cost, W2 – Wasserstein norm with L2 transport cost. IW Linear – Integral Wavefield Linearization. (d) Graph showing the results of using the quadratic Wasserstein metric with different operators used to satisfy the data nonnegativity condition.

Credit: Rasheed Ajala