Overview of today’s joint inversion methodology and its future into the machine-learning era for geophysical imaging

Monday, October 22, 2018 - 9:00am - 9:50am
Keller 3-180
Monica Maceira (Oak Ridge National Laboratory)
When an earthquake or underground explosion occurs, the seismic waves that are generated propagate through the earth, reflecting and refracting at different interfaces and illuminating its 3D structure, while also carrying the signature of the source. The wavefield recorded for many events at many stations around the world can be used to image the structure of the Earth using tomographic approaches. Seismic tomography, first introduced 50 years ago, is still a rapidly developing field and provides the most important constraints to unravel our planet's present and past dynamics and the driving forces for plate tectonics. In the last few years, with the advent of new numerical methods, unprecedented data sets and more commonly available high performance computing resources, we have seen an outburst of 3D geophysical models and techniques for seismic imaging.

This talk will first present recent developments and the application of advanced multivariate inversion techniques to generate a realistic and comprehensive 3D model of the seismic structure of the crust and upper mantle. This model satisfies several independent geophysical datasets- surface wave dispersion measurements, gravity data, teleseismic receiver functions, and seismic body wave travel times. We will conclude with a work-in-progress section dedicated to application of unsupervised and supervised machine learning methods to 1D and 2D imaging.