Modeling 3-D radial anisotropy in the Indian Ocean upper mantle from fundamental and higher mode surface waves and a non-linear hierarchical transdimensional Bayesian approach

We aim at constraining lateral variations in upper mantle and mantle transition zone (MTZ) shear-wave speed and radial anisotropy beneath the Indian Ocean. We developed a fully non-linear transdimensional hierarchical Markov Chain Monte Carlo approach to obtain three-dimensional tomographic models and quantitative uncertainties, and applied it to published fundamental and higher mode phase and group velocities. We tested the effect of depth parametrization as well as that of including prior constraints on the 410- and 660-discontinuities topography on the models.

We found that the topography at the MTZ discontinuities does not affect the results significantly. We also found that if V_SV and V_SH have the same parameterization in joint inversions of Rayleigh and Love waves, the uncertainty of the less resolved parameter becomes significantly larger than if the parameters are allowed to have different parametrization. This is due to the forcing of matching structure between the them as pointed by Gao and Lekic (2018). In addition, joint inversions of Rayleigh and Love wave data with different depth parameterizations gave similar results as running separate inversions of Rayleigh and Love waves. We thus eventually inverted the two types of data separately, which was also computationally advantageous.

The non-linear inversion results show velocity anomalies that are generally well resolved and in agreement with previous work. The Central, Southwestern and Southeastern Indian Ridges are marked with shallow low velocity regions. The Rodrigues channel, a proposed region of flow from a mantle plume under La Réunion to the Central Indian Ridge, is visible as a low velocity area. Madagascar is associated with a shallow high velocity anomaly. In agreement with Visser et. al. (2008), there is a strong V_SV < V_SH signal near the Central Indian Ridge at 50km, and the data generally favor models with anisotropy in the transition zone. However, at most locations and depths, radial anisotropy uncertainties derived from the posterior distribution are significantly large, and we cannot confidently rule out the presence of vertical trade-offs in the anisotropy signal despite the inclusion of higher mode phase velocities and group velocity data.