Resolution Analysis and Jointly Inverted P- and S-Wave Models of Japan’s Crust and Upper Mantle.

Wednesday, 17 December 2014
Petros Bogiatzis and Miaki Ishii, Harvard University, Cambridge, MA, United States
Seismic tomography is the most powerful tool for imaging the internal structure of the Earth, and the deployment of dense arrays over the last decade provides opportunities for investigating the interior with exceptionally high resolution. We use travel-time data from regional earthquakes recorded by the Hi-net array in Japan to constrain the elastic properties of the crust and the upper mantle. The large number of high-quality recordings allows refined resolution through the increase of the number of model parameters. However, large amount of data also gives rise to significant challenges. It makes manual picking and reviewing impractical, and these picks are neither errorless nor objective. We address this problem by developing and applying a wavelet-based automatic algorithm to pick the arrival time of both compressional and shear waves and to obtain estimates of the picking uncertainty. Additionally, both forward and inverse calculations require significant computational resources. We use parallel computing implemented in a distributed memory cluster. Forward problem is solved under the high-frequency approximation of wave equation by calculating P and S wave ray paths using a combination of graph theory and pseudo-bending method. We incorporate finite frequency effects by calculating the first Fresnel volume of each travel-time measurement based upon its dominant frequency that is also provided by the automatic picking algorithm. The inverse problem similarly presents difficulties as the number of model parameters increases. Among the most prominent issues is the computational difficulty of explicitly calculating the model and data resolution matrices. These matrices are of great importance in quantifying the spatial resolution of the method and in designing the inverse problem, but they are computational expensive, if not almost impossible, to determine. We attempt to remedy these problems by taking advantage of recent computational developments and distributed resources. Based upon these data and approaches, we jointly invert for three-dimensional variations in both P- and S-wave speeds beneath Japan to understand detailed structure, dynamics, and chemistry of the subduction process.