Refining the cratonic upper mantle: modeling North American upper mantle and crustal structure using the Spectral Element method

A recent 3D tomographic model of azimuthal anisotropy (SVEMum_NA2), developed by joint inversion of long period teleseismic waveforms and SKS splitting measurements, reveals the presence of three anisotropic layers throughout the stable part of the North American cratonic upper mantle (Yuan and Romanowicz, Nature, 2010). While in the bottom asthenospheric layer the fast axis direction is parallel to the current plate motion direction, the top two lithospheric layers have distinct fast axis directions, with ancient suture zone trending directions in the top layer and a general north-south direction in the bottom layer, respectively. The boundary between the two lithospheric layers, as defined by systematic changes in the direction of azimuthal anisotropy, correlates well with the sharp mid-lithospheric negative velocity boundary found by available S-wave receiver functions measurements. This spatial correlation suggests that the two boundaries, found by the surface wave inversion and receiver functions, may share a common origin. To further explore the nature of this mid-lithospheric boundary, and to better constrain the absolute values of shear velocities of the lithospheric layers, we propose a refined 3D tomographic inversion of the cratonic North America. The new inversion utilizes shorter period waveforms (down to 40 s) from recent local and regional events to improve both horizontal and vertical resolution, and a regional Spectral Element code, RegSEM (Cupillard, 2008) to compute the forward synthetics. RegSEM includes ellipticity, attenuation, arbitrary anisotropy, easy mapping of discontinuities, and can accurately represent scattering and focusing/defocusing effects caused by the 3D Earth structure. We first apply an iterative inversion approach that uses approximate but computationally efficient 2D finite frequency kernels based on the path-average approximation of normal mode perturbation theory. We also test a much more time-consuming but accurate inversion scheme in which we consider Frechet kernels computed using RegSEM and the adjoint formalism. In constructing the starting 3D model for forward RegSEM computations, we test the combination of our previous 3D shear velocity upper mantle model, SVEMum_NA2, with different 1D and 3D crustal models. Preliminary comparison of the RegSEM synthetics and corresponding data shows the importance of 3D Moho topography/crustal velocity, and the need for crustal anisotropy. We verify the appropriateness of a synthetic crustal model with radial anisotropy, which is obtained by jointly inverting short period group velocity dispersion measurements and long period surface waveforms. This synthetic crustal model has a uniform 40-km crustal thickness in North America, which not only greatly improves the computation efficiency of RegSEM simulations, but also allows for further linear perturbations of the Moho topography in our inversion. Our new results show consistent upper mantle structure with SVEMum_NA2, with, however, more details in the uppermost cratonic upper mantle.