Waveform inversion of large data sets for radially anisotropic Earth structure

Abstract

SUMMARY The amount of high-quality seismic data is expanding rapidly, and there is a need for algorithms that take advantage of classical methods to achieve high efficiency using widely available computing power. In this study, we develop a novel waveform inversion method to retrieve radially anisotropic Earth models that can be used to investigate deformation and flow in the mantle. Our method is comprised of two parts: (1) extraction and fitting of the fundamental mode and (2) fitting of the full synthetic waveform. The waveform inversion method results in path average model constraints with uniquely determined independent uncertainties. We demonstrate through synthetic testing that the method is able to retrieve radially anisotropic perturbations down to the mantle transition zone, and leakage effects due to ignoring P-wave anisotropy are minimal. We apply the method to ∼16 000 waveforms generated by earthquakes occurring in the East Sea (Sea of Japan) region, and we demonstrate that the subsequent linear inversion of radially anisotropic path constraints produces models that are similar to those resulting from full waveform adjoint tomography methods. We validate our model by predicting waveforms for earthquakes not included in our inversion, and we show that our method is able to extract structural information. Our results indicate low-velocity anomalies and weak radial anisotropy in NE Japan, which may be due to competing influences from ascending fluids and/or melts and horizontal flow in the lower crust and upper mantle. In the southern East Sea, we image low velocities and relatively high radial anisotropy, which may reflect high temperatures, shallow dehydration and olivine LPO in the upper mantle.