Improved Characterization of Ultralow-Velocity Zones Through Advances in Bayesian Inversion of ScP Waveforms

Ultralow-velocity zones (ULVZs) have been studied using a variety of seismic phases; however, their physical origin is still poorly understood. Short period ScP waveforms are extensively used to infer ULVZ properties because they may be sensitive to all ULVZ elastic moduli and thickness. However, ScP waveforms are additionally complicated by the effects of path attenuation, coherent noise, and source complexity. To address these complications, we developed a hierarchical Bayesian inversion method that allows us to invert ScP waveforms from multiple events simultaneously and accounts for path attenuation and correlated noise. The inversion method is tested with synthetic predictions which show that the inclusion of attenuation is imperative to recover ULVZ parameters accurately and that the ULVZ thickness and S-wave velocity decrease are most reliably recovered. Utilizing multiple events simultaneously reduces the effects of coherent noise and source time function complexity, which in turn allows for the inclusion of more data to be used in the analyses. We next applied the method to ScP data recorded in Australia for 291 events that sample the core-mantle boundary beneath the Coral Sea. Our results indicate, on average, similar to 12-km thick ULVZ with similar to 14% reduction in S-wave velocity across the region, but there is a greater variability in ULVZ properties in the south than that in the north of the sampled region. P-wave velocity reductions and density perturbations are mostly below 10%. These ScP data show more than one ScP post-cursor in some areas which may indicate complex 3-D ULVZ structures. Plain Language Summary Studies of the Earth's deep interior using seismic energy generated by earthquakes show the presence of small-scale structures at the core-mantle boundary (CMB) known as ultralow-velocity zones (ULVZs). The seismic waves traveling through these features are inferred to be slowed down by as much as 50% with respect to normal mantle material. However, it is challenging to determine the physical composition of these ULVZs and how they are linked to other structures in the Earth's interior because we don't have precise knowledge on their physical properties. To understand their origins, we need to estimate their physical properties (i.e., velocity and density) as accurately as possible. This study focuses on the development and application of a probabilistic approach that can provide an unbiased estimate of seismic velocity and density. Our new approach can take data from multiple events simultaneously and incorporates data noise and possible absorption of seismic energy along the path. Application of this approach to real data that sample the CMB beneath the Coral Sea show a wide spread ULVZ with much less lateral variation in material parameters in comparison to solutions using single events. We also find observations that may be linked to 3-D ULVZ structure.