Towards constraining mantle flow through imaging of radial anisotropy, with full uncertainty quantification

The seismic imaging of radial anisotropy can be used as a proxy for the direction of mantle flow. Previous studies have imaged radial anisotropy throughout the mantle on a global scale and are starting to show some consistent features. However, the interpretation of existing models is hindered by the lack of uncertainties provided from the employed inversion method. To address this, we build a new global radially anisotropic model of the Earth’s upper mantle which consists of two main stages. Firstly, we build global Rayleigh and Love wave phase and group velocity maps using ~47 million measurements, including fundamental mode and up to 5^th overtone measurements, and compute their associated uncertainties. Weights according to similar paths and data uncertainties are employed in the inversions. We construct a total of 310 2D maps, at periods between T16-375 s, expanded in spherical harmonics up to degree lmax=60 and observe many relevant small-scale structures, such as e.g. the curvature of the Tibetan plateau at T~40s (fundamental mode). As expected, uncertainties are higher in regions of poor data coverage (e.g., southern hemisphere and oceans). Then, we invert for 1D profiles of radial anisotropy using two Monte Carlo based inversion methods: the Neighbourhood Algorithm (NA) and the Reversible-Jump Markov Chain Monte Carlo algorithm (RJMCMC). The NA has been widely used for seismic inversion, as it efficiently explores the parameter space. However, the advantage of the RJMCMC is that in addition to constraining e.g. radial anisotropy, it can also constrain e.g. layer thickness. We compare the 1D profiles from both methods, and their associated uncertainties, which will lead to a new global 3D model of radial anisotropy.