Improving external source models for mantle conducivity studies with observatory and satellite data (Invited)

Recent developments in modeling and inversion finally make 3D imaging of mantle electrical conductivity at the global scale practical, at least in principal. Conductivities of the principal mantle minerals have been measured in the lab, with general agreement among different groups for dry samples. Efforts to quantify the relatively substantial impact on conductivity of volatiles such as water and CO2, and to explore sensitivity to variations in mantle iron content and oxidation state are also making great progress. Prospects thus appear excellent for electromagnetic induction studies to provide an alternate, physically interpretable view of lateral heterogeneity in the mantle that can significantly complement what we have learned so far from seismic imaging. Initial results from 3D inversion of global observatory data for periods T > 5 day (under the assumption of a ring current source) suggest that there are significant lateral variations in conductivity at depths of 500-900 km, which have been interpreted as evidence for the transport of significant amounts of water into the deep mantle by subducting oceanic lithosphere. These results still need to be confirmed, and then refined and extended, e.g., filling in gaps in the southern hemisphere and beneath the oceans, improving resolution, and extending images to shallower and deeper depths. However, signals associated with 3D conductivity variations are fairly subtle in the magnetic variation data, and can only be interpreted accurately if external source spatial structure is sufficiently well constrained. This represents the major obstacle to further progress in global studies of mantle conductivity. External variations at daily variation periods, which will be essential for complete imaging of the critical mantle transition zone, are a particular challenge since the source fields are in the ionosphere, and thus have substantial power at short spatial wavelengths. I will discuss present efforts in our research group to improve empirical models of daily variation fields for induction studies. In the first stage of our analysis we apply robust frequency domain principal components analysis to extract the dominant spatial modes of variability from the observatory array. Our approach allows for significant amounts of missing data, allowing us to incorporate a large fraction of the available (recent and historical) observatory data in our analysis. In the second stage these modes must be interpreted in terms of internal conductivity variations (the same for all modes/periods) and external spatial structure. Our ultimate plan is to treat this as a large coupled inverse problem, with a priori constraints on external source structure derived from physics based models of ionospheric (and magnetospheric) current systems. Progress in this direction, including development of source inversions with a 1-D conductivity model (with a 3D thin sheet for the oceans), and development of a prior covariances for daily variation external sources (based on the TIE-GCM ionospheric model) will be discussed. Initial efforts to incorporate satellite data will also be briefly considered.