3-D Magnetotelluric Inversion Using Non-linear Conjugate Gradients Method with Parallel Structure

Abstract

Summary In this study, a parallel inversion algorithm for three-dimensional(3-D) magnetotelluric sounding is presented. The algorithm is developed basing on the preconditioned nonlinear conjugate gradient method, with a improved preconditioning scheme. The efficiency and reliability of the code is verified with both synthetic data generated from theoretical models and field data from the KAYABE dataset in Japan. As the algorithm requires less in physical memory and CPU quantity than formal algorithms, the code can therefore easily run on ordinary personal computers, which makes the algorithm a very practical 3-D magnetotelluric inversion method.