Source identification in the self-potential method and its connection to Stokes type systems

TL;DR

This paper introduces a new method for identifying source currents in the self-potential method, splitting the problem into scalar source reconstruction and divergence equation solution, with applications demonstrated through numerical experiments.

Contribution

It presents a projection-based algorithm for scalar source identification that avoids Fredholm equations and reduces inverse problem size, along with a PDE approach for divergence equation solving.

Findings

Effective source reconstruction with exact and noisy data

Reduced computational complexity in inverse problem

Successful implementation in 2D physical space

Abstract

This paper develops a novel approach to the problem of source current identification for the diffusion equation in connection with geophysical self-potential measurements. The problem is split into two subproblems: (a) the scalar source identification, and (b) solution of the divergence equation. For subproblem (a), we design an algorithm for reconstructing the scalar source function, which does not require solving the Fredholm integral equation of the first kind. Instead, the problem is reformulated as a linear operator equation, which is solved by a projection method. The dimension of the subspace, in which the source function is sought, is independent of the dimension of the forward problem, leading to reduction of the size of the inverse problem. Numerical experiments with exact and noisy data are presented. For subproblem (b), the divergence equation was posed as a minimization problem, which, by means of Lagrangian formalism, was reduced to a system of partial differential equation of Stokes type with a unique solution. To demonstrate how this framework can be used in a practical application, we implemented the algorithm in the two-dimensional physical space, using a finite-different discretization on staggered grids.

Figures