A One-Degree Seismic Tomographic Model Based on a Sensitivity Kernel Database

Seismic tomography is instrumental in mapping 3D velocity structures of the Earth's interior based on travel-time measurements and waveform differences. Although both ray theory and other asymptotic methods have been successfully employed in global tomography, they are less accurate for long-period waves or steep velocity gradients. They also lack the ability to predict 'non-geometrical' effects such as those for the core diffracted phases (P_diff, S_diff) which are crucial for mapping heterogeneities in the lowermost mantle (D'' layer). On the other hand, sensitivity kernels can be accurately calculated with no approximations by the interaction of forward and adjoint wavefields, both numerically simulated by spectral element methods. We have previously shown that by taking advantage of the symmetry of 1D reference models, we can efficiently and speedily construct sensitivity kernels of both P and S wavespeeds based on the simulation and storage of forward and adjoint strain fields for select source and receiver geometries. This technique has been used to create a database of strain fields as well as sensitivity kernels for phases typically used in global inversions. We also performed picks for 27,000 S_diff, 35,000 P_diff, 400,000 S, and 600,000 P phases and 33,000 SS-S, 33,000 PP-P, and 41,000 ScS-S differential phases, which provide much improved coverage of the globe. Using these travel-times and our sensitivity kernel database in a parallel LSQR inversion, we generate an updated tomographic model with 1° resolution. Using this improved coverage, we investigate differences between global models inverted based on ray theory and finite-frequency kernels.