Dynamics, flow and melt content of the Southern East Pacific Rise upper mantle from teleseismic tomography

The Mantle ELectromagnetic and Tomography (MELT) Experiment at the super-fast spreading Southern East Pacific Rise (SEPR) was designed to distinguish between passive upwelling in response to basal tractions from the lithosphere, and a more diapiric, focussed upwelling resulting from melt-enhanced buoyancy and viscosity reduction. To distinguish between these models I estimate the distribution of melt and temperature anomalies beneath the SEPR using teleseismic P and S phases recorded on the MELT seismic array. The non-linear tomographic inversion for VP, VS and anisotropy variations includes an a priori model for the mantle flow-induced seismic anisotropy, constraints on smoothness, minimum norm, coupling between VP and VS , and a grid search for the best anisotropy. Shear wave splitting measurements also constrain the anisotropy. The best fitting models have anisotropy with the symmetry axis horizontal or dipping shallowly to the west (<30°), and magnitudes of VP and VS anomalies approximately 3.7% and 4.7%, respectively. These translate to melt fractions no greater than .013 or temperature variations no greater than 150°C beneath the rise, using relations I estimated for this purpose. The broad distribution (>100 km) of low velocities favors models of passively driven mantle flow. The orientation of the anisotropy favors shallow return flow owing to a 32 mm/yr westward migration of the SEPR and proximity of the South Pacific Superswell.; To best infer melt fraction from seismic velocity, I developed quantitative models of seismic wave propagation through upper mantle partial melts. The elastic effects are estimated with finite element representations of grain-scale laboratory-derived melt geometries. Two- and three-dimensional deformation simulations show that the shear modulus is sensitive to melt inclusion geometry and organization. The anelastic effects are modeled analytically as ellipsoidal pores connected by tubes. Melt moves through the tubes in response to variable compression of the ellipsoids from seismic excitation. The results indicate that the relaxation occurs so rapidly that, in the seismic band, little or no attenuation results. Seismic velocities are reduced by at least 3.6% and 7.9% per percent melt for P and S waves, respectively.; This dissertation contains both my previously published and co-authored materials and unpublished co-authored materials.