Three-dimensional magnetotelluric inversion based on multigrid algorithm

In order to promote the practical applications of three-dimensional (3D) Magnetotelluric (MT) inversion, we propose in this paper an efficient multi-scale 3D inversion algorithm by integrating the Multigrid (MG) with the Limited memory Broyden-Fletcher-Goldfarb-Shanno (LBEGS) method. The method starts with a fine mesh and gradually coarsens to generate the coarsest mesh. Then the inversion is started from the coarsest mesh and the coarse inversion results are used step by step to provides a relatively reasonable initial model for the upper-level fine grids. This can accelerate the iterative process of inversion for fine meshes and obtain multiscale inversion results from coarse to fine at the same time. In this paper, the chessboard model, and the undulating interface and intrusive zone models are first tested to theoretically verify the efficiency of the algorithm. The results show that the computational efficiency of this algorithm is 2-3 times that of traditional inversion. Then, the algorithm is used to invert the measured data in Cascadia area of the United State, to further verify the practicality of the algorithm. The results showed that the computational efficiency of the algorithm was nearly twice that of traditional inversion. Hope that this study will provide an efficient tool to 3D MT inversions.