The dominant large scale structure in Earth’s lower mantle

Over the past two decades a variety of published tomography models derived from global seismic data have consistently imaged large-scale structures in the lower mantle under the Pacific and African plates that are characterised by strongly reduced seismic shear velocities: the so-called "low shear-velocity provinces" (LSVP). Efforts to directly constrain the thermo-chemical and dynamical properties of these LSVP have relied on tomography-based mantle flow models that require additional assumptions on the mineral physical relationship between seismic anomalies and density and on the rheological structure of the mantle. We developed a series of tests that probe the possible existence of compositionally distinct material in the central core of the LSVP under the African and Pacific plates using tomography-based mantle flow models that employ several independently-derived viscosity profiles (Mitrovica & Forte 2004, Behn et al. 2004, Steinberger & Calderwood 2006, Forte et al. 2010). In applying these tests we furthermore consider four global tomography models independently derived from seismic shear velocity data alone (Grand 2002, Panning & Romanowicz 2006, Kustowski et al. 2008, Ritsema et al. 2011). The possible combinations of mantle viscosity and tomography models yield 16 different tests for the presence of compositional heterogeneity inside the LSVP. In all tests we begin with a mineral physical scaling between lower-mantle shear velocity and density anomalies that is based on the assumption that thermal effects are dominant (e.g. Karato & Karki 2001) everywhere, including within the LSVP. We find that it is not possible, in any of the tests, to obtain a satisfactory fit to surface geodynamic data (especially the global, long-wavelength gravity anomalies) and space-geodetic inferences of excess CMB flattening with a purely thermal interpretation of lower-mantle heterogeneity. If, however, we allow for the presence of compositionally-distinct material in the central portions of the LSVP, all tests show a notable improvement in the fit to the gravity anomaly and CMB ellipticity data. These tests also show that an optimal reconciliation of the gravity and CMB data is obtained by allowing compositional heterogeneity to extend upwards to mid-mantle depths (between 1500 and 2000 km depth), with maximum-amplitude in the seismic D"-layer and exponentially tapering off to negligible values in the mid-mantle. We explore the dynamical implications of this deeply-rooted compositional heterogeneity by calculating the 3-D convective flow in the lower mantle, for each of the 16 test cases. All test cases reveal two dominant upwellings in the Pacific lower-mantle under the South-Eastern Pacific, below the axis of the East Pacific Rise (EPR), and under the Caroline Islands in the Western Pacific. Under the African plate all tests show two dominant lower-mantle upwellings: one under the southern half of the East-African Ridge and another under the Cape Verde Islands. The EPR "plume" has a special significance because of its association with Earth's dominant spreading ridge and plate-kinematic inferences of strong lateral fixity of this ridge over the past 83 Ma (Rowley et al. 2016).