Geometric and Level Set Tomography for Interface Detection in the Near Surface

Jack B. Muir, & Victor C. Tsai

Submitted August 12, 2018, SCEC Contribution #8399, 2018 SCEC Annual Meeting Poster #105

Seismic travel time tomography is a key component of geophysicists' basic toolbox for the characterization of the near surface. Traditional seismic travel time tomography seeks to optimize a field of velocity parameters to fit observed data; this is fundamentally an ill-posed infinite dimensional inverse problem, necessitating smoothing regularization to ensure stability of the inversion. In many instances however, a priori geological knowledge suggests that the subsurface would be better parametrized by subhorizontal (e.g. layered media) or subvertical (e.g. buried fault scarps) interfaces delineating discrete geologic units. Introducing these features into a formal seismic inverse problem in 2 or 3 dimensions requires a flexible and expressive parametrization of the Earth that can incorporate our geologic knowledge, as well as inversion algorithms that can handle this parametrization. We present results using an ensemble Kalman inversion scheme to invert for sharp interfaces in a typical near-surface seismic refraction tomography geometry. We use a mixture of geometric primitives (fault offsets) and the level sets of anisotropic Gaussian fields (undulating layer interfaces) to parametrize the material properties of the subsurface. Synthetic tests on geometries similar to cross-sections perpendicular to the Landers fault show that recovery of sharp interfaces is improved compared to traditional tomographic modeling, without recourse to explicit forward refraction modeling.

Citation
Muir, J. B., & Tsai, V. C. (2018, 08). Geometric and Level Set Tomography for Interface Detection in the Near Surface. Poster Presentation at 2018 SCEC Annual Meeting.


Related Projects & Working Groups
Seismology