Three-dimensional conjugate gradient inversion of magnetotelluric impedance tensor data

Abstract

We developed a three-dimensional (3D) conjugate gradient inversion algorithm for inverting magnetotelluric impedance tensor measurements. In order to show the importance of including diagonal components of magnetotelluric impedance tensor in 3D inversion, synthetic data were inverted using the 3D conjugate gradient inversion, and the inversion results were compared and analyzed. The results from the 3D inversion of synthetic data indicate that both the off-diagonal and the diagonal components are required in inversions to obtain better inversion results when there are no enough data sites to recover the target resistivity structure. These examples show that lots of information about 3D structure is also contained in the diagonal components; as a result, diagonal components should be included in 3D inversions. The inversion algorithm was also used to invert the impedance tensor data acquired in the Kayabe area in Japan. Inversions with the synthetic and real data demonstrated the validity and practicability of the inversion algorithm.

A NLCG 3‐D Inversion Method of Magnetotellurics with Parallel Structure

“…In Fig. 8, NLCG is the inversion results of the algorithm in this paper; REBOCC is the results of Numerical domain inversion method proposed by Weerachai (2006) [12] ; CG is the conjugate gradient inversion results proposed by Lin C H (2010) [13] ; RRI is the rapid relaxation inversion results proposed by Tan H D et al (2003) [14] . Through the comparison of 4 inversion methods, we can summarize that the results of the 4 inversion methods are similar.…”