Speaker: Susan Minkoff, Brookhaven National Laboratory
Title: "Extension Methods to Mitigate Cycle Skipping in Seismic Full Waveform Inversion"
Abstract:
Parameter estimation studies such as seismic inversion which seek models of subsurface mechanical parameters suffer from a phenomenon known as "cycle skipping". Full waveform inversion minimizes a data fidelity term that measures the misfit between observed or recorded data and data predicted by a mathematical model of wave propagation in the Earth. Repeatedly solving the wave equation over realistic-sized domains is expensive, and hence one generally relies on local gradient-based optimization rather than global optimization techniques. However, to prevent these techniques from stalling in non-geologic local minimizers, one must use a very good initial guess for the model parameters. In fact, the ill-posed nature of these problems means that several different Earth models can fit the measured data equally well. Cycle skipping occurs when an initial model is used that predicts wave arrival times in error by more than half a wavelength. I will discuss a few ways to "extend" the problem by including additional degrees of freedom to the model in an attempt to convexify the objective function and mitigate cycle skipping, namely extended source inversion and matched source waveform inversion. These extension methods allow the inversion to make large model adjustments while maintaining data fit and so they reduce the chance of local optimization iterates stagnating at non-informative model estimates. This work is joint with Bill Symes, Rice University, and Huiyi Chen, University of Texas at Dallas.
Bio:
Sue Minkoff is the Chair of Applied Mathematics at Brookhaven National Laboratory. From 2012-2024, she was a professor of Mathematical Sciences and an affiliated professor in the departments of Sustainable Earth Systems Sciences and Science and Mathematics Education at the University of Texas at Dallas. From 2000-2012, she served on the faculty in the Department of Mathematics and Statistics at the University of Maryland, Baltimore County. Minkoff’s research interests include inverse problems, uncertainty quantification, AI/ML, digital twins modeling, Earth science, and photonics. She received her doctorate in Computational and Applied Mathematics from Rice University in 1995. From 1995-1997, she was a National Science Foundation-Industrial postdoc joint between the University of Texas at Austin and British Petroleum, and, from 1997-2000, she held the von Neumann Fellowship in the Mathematics Department at Sandia National Laboratories (New Mexico). In 2000, Minkoff was promoted to senior member of the technical staff in Sandia’s Geophysics Dept.
In person attendance at this seminar is only open to Columbia University affiliates. External guests are welcome to attend remotely. Please contact [email protected] if you need the Zoom link for this seminar.
