Item

Abstract

Conventionally, $SKS$ splitting observations are analyzed individually for each seismic station. However, waveforms from high-density two- and three-dimensional seismic arrays facilitate the application of new data processing and inversion techniques, which exploit the coherent nature of the recordings. In this approach, measured splitting parameters, which are obtained under the assumption of a single homogeneous anisotropic layer beneath a station, are not directly interpreted. Instead, the splitting parameters are used as a parameterization of the observed waveforms which allows for an efficient (automated) comparison with results from forward modeling. The formal inversion of the measured splitting parameters employs an iterative approach using ({it i}) a local optimization technique i.e. the downhill Simplex method and ({it ii}) a (global) genetic-algorithm search. The forward-modeling part involves the calculation of $SKS$ waveforms by a complete finite-difference method to solve the anisotropic wave equation. The misfit function is calculated from a comparison of frequency-dependent splitting parameters obtained from the observed and synthetic waveforms. We apply these techniques to the analysis of $SKS$ phases recorded along a 100-km profile with an average station spacing of 1.2 km. The measured splitting parameters show gradual short-scale variations along the profile. The model used in the inversion consists of an anisotropic crust and a 100-km thick anisotropic mantle layer. Variations in anisotropy are accounted for by 2-dimensional block structures. We show the convergence of the two inversion methods by gradually increasing the complexity of the model (as defined by the number of anisotropic blocks). Additional tests based on synthetic models demonstrate the potential and resolution of the inversion approaches.