DI002-0013
Global observations of mantle discontinuities from ScS reverberations
Abstract:
In this study, we introduce a new large global dataset of ScS reverberations, compiled with an automatic waveform identification code which uses a Convolutional Neural Network (CNN). The CNN model was trained on a handpicked global dataset of SS phases, and here we show its application to the identification for other shear wave phases with similar frequencies. We build new topography maps for the MTZ and mid-mantle discontinuities using an adaptive stacking technique based on Voronoi tessellation, which automatically adjusts its parameterisation to account for topography of the discontinuities, noise, and data coverage. We consider our observations in the context of mineral physics predictions for various thermochemical models. The new geometries provided by this dataset supplement our existing maps of the MTZ and mid-mantle, built with SS and PP precursors, providing improved resolution in regions with previously poor data coverage. Our results offer new insight into the relationship between the MTZ discontinuities and mid-mantle reflectors, informing regarding the regionally diverse styles of mantle mixing.