HR: 16:30h
AN: GP14A-02    [Abstracts]
TI: A Non-Linear Inversion for the Global 3-D Electrical Conductivity Distribution in the Upper to Mid-Mantle
AU: * Kelbert, A
EM: anya@ocean.cf.ac.uk
AF: Cardiff University, School of Earth, Ocean and Planetary Sciences Cardiff University Main Building Park Place, Cardiff, CF10 3YE United Kingdom
AU: Schultz, A
EM: adam@coas.oregonstate.edu
AF: Oregon State University, College of Oceanic and Atmospheric Sciences Oregon State University 104 COAS Admin Bldg, Corvallis, OR 97331-5503 United States
AB: The case for substantial heterogeneity in mantle conductivity has stimulated the development of methods for solving Maxwell's equations in a heterogeneous conducting sphere. A global 3-D frequency domain forward solver has been devised (Uyeshima & Schultz, 2000), accurate and efficient enough to be an attractive kernel of a practical inverse method. The solver employs a staggered-grid finite difference formulation in spherical coordinates. The induced fields are found as a solution to the integral form of Maxwell's equations, while the system is solved using stabilised biconjugate gradient methods. A single, accurate forward solution takes approx. 4 minutes on 5 GFLOP (peak) processor. The aim of our present research is to produce an inverse solver, to be applied to the Fujii & Schultz (2002) data set of globally-distributed EM response functions, which would reconstruct the 3-D electrical conductivity distribution in the upper to mid-mantle. Geophysical inversion is an ill-posed problem, therefore the aim is to apply suitable parameter constraints and a nonlinear search algorithm to identify candidate minima, then to apply local gradient methods around those minima. Our specific target involves designing a fast enough global optimisation routine that would allow us to produce at least one fully 3-D starting model, optimal with respect to the RMS misfit between the data and the forward solutions. A new and very flexible inverse solver has been developed utilizing parallel optimisation routines to obtain a starting model that satisfies the data. 3-D simulations have been run, the parametrization based on a spherical harmonic representation of a chess board model of varying degree and order. The inversion has demonstrated accurate fidelity in reproducing resolvable features of the test model. A study has been made of the reduction in fidelity as the number and distribution of observatory sites on the Earth's surface is degraded. An inversion of the Fujii & Schultz (2002) geomagnetic data set is underway. We also discuss implementing a linearised sensitivity analysis as part of the inversion.
DE: 3914 Electrical properties
DE: 3260 Inverse theory
DE: 1500 GEOMAGNETISM AND PALEOMAGNETISM
DE: 1515 Geomagnetic induction
DE: 0639 Nonlinear electromagnetics
SC: Geomagnetism and Paleomagnetism [GP]
MN: 2004 AGU Fall Meeting