Past Event: CSEM Student Forum

Optimal Transport for Seismic Source Inversion

Tyler Masthay, CSEM PhD Student, Oden Institute, UT Austin

Abstract

Least-squares full waveform inversion has a highly nonconvex optimization landscape and thus suffers from sensitivity to the initial model. This has spurred literature on alternate misfits that offer a more convex optimization landscape. The Wasserstein-2 metric, a specific case of an optimal transport metric, is particularly attractive due to its convexity with respect to shift and partial amplitude change. We present results for a model earthquake location problem in the western United States using the USArray data set. We compare the least-squares optimization landscapes directly and their respective accuracies for this model problem.

Biography

Tyler Masthay is a CSEM PhD student working with Dr. Bjorn Engquist on the application of optimal transport to seismic source inversion. He graduated from the University of Dayton in mathematics and computer science in 2017 before joining the Oden Institute.