Finite element based inversion of AEM data using stochastic optimization

Abstract

Modern data acquisition technology in airborne electromagnetic (AEM) produces huge data sets which cover areas of considerable extent. Both the large size of the domain of interest, in general a three-dimensional volume, and the large number of transmitters and receivers pose challenges to any type of modelling or inversion software. Solutions to the inverse problem requires repeated solves of the forward problem. The time to solve one forward problem in turn scales linearly with the number of transmitters. In this paper we examine scholastic optimization techniques for the solution of the inverse problem which essentially allows us to work with small subsets of transmitters/receivers and, thus, reduce the computational load significantly.