Refining Constraints on Seismic Discontinuities and Elastic Structure in the Earth's Upper Mantle

We propose a modification to the Non-linear Asymptotic Coupling Theory (NACT), a normal mode coupling method used for synthesizing seismograms and computing sensitivity kernels in 3D Earth. The modification is aimed to meet the computational challenges for NACT when approaching higher frequencies, which is required for obtaining finer scale images of the Earth's upper mantle. The new scheme is numerically validated. We present new constraints on the topography of the 410 and 660, two important seismic discontinuities in the Earth's upper mantle. The data used are the SS precursors recorded by the US Transportable Array. We first demonstrate a case of 3D mantle heterogeneity interplaying with discontinuity depth. We show observations from one event for which a large scale heterogeneity away from the SS precursor bounce point region produces an artificial "precursor". This new discovery raises a caution for identifying and interpreting SS (and perhaps PP) precursors. We then present high resolution maps of 410 and 660 discontinuity topography across a large area of the Pacific Ocean, derived from SS precursors. A new filtering tool called the Local Slant-Stack Filtered (LSSF) is employed to successfully clean up the SS precursor record sections, leading to robust precursor travel time measurements with increased spatial resolution. Good agreements are observed between our discontinuity images and a very recent upper mantle tomography model SEMum2. Besides the new constraints on the discontinuities, we also work on refining the elastic structure of the Earth's upper mantle in the context of regional/continental scale full-waveform tomography. We propose a framework for a hybrid adjoint tomography, in which the gradient of the misfit function is numerically computed using the adjoint method with the highly accurate Regional Spectral Element Method (RegSEM) code, while the Hessian is computed approximately using NACT. We present results from tests on a dataset for imaging the North American continent. Finally, we study the infra-gravity wave induced long-period noise on an ocean bottom broadband seismometer, MOBB, deployed offshore in the Monterey Bay, California, using ~10 years' continuous recording data. Strong correlation between the vertical component seismogram and the pressure record is observed. We define and calculate the transfer function between the two channels, and demonstrate that the transfer function is stable over time. We then utilize the average transfer function to remove pressure-correlated noise from the vertical component seismogram, and show that the cleaned MOBB waveforms help to better constrain the moment tensors for regional near-shore earthquakes on the San Andreas Fault system in Northern California