Interpreting the GyPSuM Tomography Model in Terms of Thermal Heterogeneity and Major Oxide Composition

Global seismic mantle tomography has advanced greatly over the past two decades with improved methods and increasing data coverage. However, interpreting these velocity distributions as thermochemical distributions remains challenging, and is the focus of our work here. Several groups of researchers [e.g., Forte & Mitrovica 2001, Cobden et al. 2009] have calculated the sensitivity of density and seismic velocities to temperature and compositional variations to estimate thermochemical heterogeneity in the lower mantle. Building on this work, we have assembled new expressions to calculate these derivatives throughout Earth's mantle, and included five major oxides: SiO2, MgO, FeO, Al2O3, and CaO. To do this, we constructed a layered 1-D compositional Earth model, and calculated its seismic and density profile using the mineral physics software package BurnMan [Cottaar et al. 2014]. Likewise, we calculated a series of profiles by systematically perturbing the compositional structure within individual layers and the temperature. From these profiles, we calculated the temperature and compositional derivatives, and determined the relative perturbations to shear velocity, bulk sound velocity, and density in terms of temperature and mineral concentration. We applied these new expressions to the 3-D GyPSuM tomography model [Simmons et al. 2010] in order to reinterpret the model's heterogeneity in terms of large-scale temperature and major element perturbations. The GyPSuM model is ideal because it incorporates P and S seismic travel times, geodynamic observations related to mantle density (and also includes initial, simplified mineral physical constraints) to simultaneously invert for global distributions of density and P and S velocities. The importance of this analysis is underscored by the ongoing debate concerning the interpretation of the Large Low Shear Velocity Provinces and their dynamic relationship to the convecting mantle. In addition to gaining insight into major-element chemical heterogeneity, this work will provide the basis for a reinterpretation of the magnitude and pattern of lateral viscosity variations in the mantle [e.g. Glisovic et al. 2015]. We will thereby obtain new insights on the importance of lateral heterogeneity in composition and rheology for mantle convection dynamics.