Broad-band ambient noise surface wave tomography: Technique development and application across the United States

In recent years, it has been shown that surface wave signals can be extracted from high-quality empirical Green functions (EGF) obtained through cross-correlation of long ambient noise timeseries. Early work showed that Rayleigh wave components of EGFs could be created in a narrow period band under certain background noise characteristics. Such Rayleigh wave signals were used to develop shear wave tomography models of several geographic regions with unprecedented high resolution. However, questions remained regarding the robustness of these signals and their range of applicability. My work focuses on two problems. The first is researching the best method for computing, measuring and selecting high-quality EGFs. The second is to use this new technique to create a three-dimensional (3D) velocity model of the continental United States. Testing a variety of temporal and spectral normalization techniques yields an optimal method of creating EGFs. These signals are evaluated for robustness in a variety of noise environments effectively broadening the bandwidth from 7.5-20 s period to 6-100 s period. An automated dispersion measurement technique is presented as well as a preferred method of measurement selection and certain "best practices" are proposed for future study. Applying this method across the continental United States I develop Rayleigh and Love wave group and phase speed dispersion maps from 8-70 s period. The resulting set of dispersion maps possesses unprecedented high resolution and bandwidth for continental scale surface wave investigations and unites diverse tectonic regions into a coherent model. I invert the dispersion maps for a 3D shear velocity model with resolution from the surface to 150 km depth using a two-step procedure. First is a linearized inversion for the best fitting velocity model. Second is a Monte-Carlo re-sampling to develop an ensemble of models of sufficient quality and to generate uncertainty estimates at all points. The resulting velocity model allows identification of prominent features in the crust and mantle and sheds light on topics such as topographic compensation, crustal heterogeneity and radial anisotropy in the crust.