Towards Higher Resolution Imaging of Earth's Lower Mantle

Earth's deep mantle has been shown to be heterogeneous at a wide spectrum of scales. At larger scales, recent tomographic models have resolved large-scale features in the lower mantle, such as the large low shear velocity provinces (LLSVPs). However, differences are present between models, especially at shorter scales, possibly due to differences in data (type and coverage) and regularization. This work aims to provide a dense sampling of the mantle by building a global dataset of S waves (SH), with a motivation that better constraints on deep mantle heterogeneity hold promise to improve our understanding of mantle flow, composition and possibly smaller-scale structures like plumes and ULVZs.

We have developed an Empirical Wavelet (EW) based method to document travel time and waveform information of every visible SH wave in a global data set of earthquakes and stations. For each event, an EW is iteratively built from S-waves, which are then adaptively fit to each of six phases (S, SS, SSS, ScS, ScSScS, Sdiff) in all records in a fully automatic approach, which are subsequently human-reviewed. A dataset of 250K records with comprehensive quality measurements was built and made publicly available. For less common (lower amplitude) phases, such as higher multiples of S and ScS, we developed a geographic bin stacking of neighboring stations, which we call virtual stations. This was applied to SSSS (S4), S5, S6, ScS3, ScS4, and ScS5, for both minor and major arc data (as well as major arc SSS). The virtual data provide unique path sampling and coverage to the dataset, vastly improving sampling in the southern hemisphere.

Our current work involves a ray-based iterative layer-stripping forward tomography approach, which iteratively updates a starting tomographic model by mapping the travel time residuals from the surface down to the lowermost mantle. This forward approach allows us to investigate the differences in lower mantle heterogeneity amplitude and scale in starting tomography models. We will present updated models and discuss the degree to which different models converge after iterative updating. We will discuss final models in terms of deep mantle phenomena (e.g., LLSVPs) and relationship to surface features (e.g., subduction, hotspots).