Project title: Illuminating the Earth's inner core with Bayesian machine learning
PI: Paula Koelemeijer
The inner core (IC) is the final frontier of deep Earth seismology. It provides energy for the generation of the planetary magnetic field which protects life on the surface from space radiation, and a major factor for satellite technology driving modern society. To create images of the IC using seismic waves (like a CT scan), assumptions are made about the manner in which the waves travel. In recent decades, research groups have put forward various hypotheses about the structures that affect seismic wave propagation in the IC with no clear consensus. This project will exploit advances in AI/ML to quantitively assess the support the data have for each hypothesis. The Bayesian Evidence provides this assessment but has historically been difficult to calculate and thus largely ignored in seismic imaging. AI/ML techniques can now be used to estimate the Evidence with greater stability and less uncertainty. I will investigate further developments for AI-based Evidence estimators and use them for the first time in a seismic imaging context. I will perform a comprehensive comparison of proposed forms of seismic structure in the IC, thereby illuminating the deepest parts of the Earth and furthering our understanding of planetary evolution.
