Calibration of Three-dimensional Upper Mantle Structure in Eurasia Using Regional and Teleseismic Full Waveform Seismic Data

We present progress in the development of a new approach to develop and evaluate earth models at the regional scale that utilizes full waveform seismograms. We have recently implemented a non-linear three-dimensional (3-D) Born approximation and constructed a 3-D shear velocity model of a subregion of Southeast Asia (longitude 75 to 150 degrees and latitude 0 to 45 degrees) based on this approach for the forward modeling part of the problem and linear Born kernels for the inverse part. Our initial model in the large region (longitude 30 to 150 and latitude -10 to 60 degrees) is derived from a large data set of teleseismic surface waveforms (fundamental mode and overtones in the period range 60-300 s) using the Nonlinear Asymptotic Coupling Theory (NACT), an approach used successfully for global and regional mantle tomography at Berkeley since 1995. In the subregion of study, our “N-Born” model is parameterized at relatively short wavelengths on the order of 200 km. The NACT approach assumes two-dimensional (2-D) sensitivity kernels, confined to the vertical plane containing the source and the receiver, and is adequate for the development of a smooth velocity model. We first developed a model for Southeast Asia using NACT. Using this model as a starting model, we performed one iteration using the N-Born approach. This approach computes synthetic seismograms including the effects of single scattering (linear) as well as a non-linear part, as in the standard “path average approximation,” that accounts for large accumulated phase delays on paths that sample large scale smooth anomalies. The inversion kernels are complete 3-D Born kernels. A regional version of the Spectral Element Method (SEM) code, RegSEM.1, has been completed. This version of the code accepts a non-conformal grid, uses PML (Perfectly Matched Layers) at the borders of the region, and includes general 3-D anisotropy, Moho and surface topography, ocean bathymetry, attenuation, and ellipticity. We are computing synthetics in our N-Born model and will perform one iteration of inversion using 3-D Born sensitivity kernels to produce the next generation 3-D model of upper mantle structure in the subregion defined above. In order to obtain finer velocity structures beneath the region of our study, we are performing forward modeling at shorter periods using the method of frequency-wave number integration (FKI). Using our N-Born 3-D model, we add sedimentary layers in the crust and change their thickness to improve the fit to the observed seismograms, and then, based on the refined velocity structures, we conduct our own moment tensor inversions, as intermediate steps. 29th Monitoring Research Review: Ground-Based Nuclear Explosion Monitoring Technologies