Magnetotelluric inversion of one- and two-dimensional synthetic data based on hybrid genetic algorithms

Abstract

We applied the technique of the genetic algorithms and a local methodology integrating the Gauss–Newton and Conjugate Gradient (GNCG) techniques to test one-dimensional inverse modeling of synthetic magnetotelluric data. The result of this modeling applied to a homogeneous and isotropic five-layer model led to the development a hybrid algorithm (GAGNCG), combining the aforementioned techniques, for inverse modeling of one-dimensional magnetotelluric data. The GAGNCG modeling of the synthetic data performs more efficiently than the local methodology in terms of both procedure and results. This showed that the hybridization procedure maximized the advantages of using the global search methodology and minimized the disadvantages of the local technique. Based on these results, we developed another hybrid methodology (GA2D), built from some characteristics of the genetic algorithm and the simulated annealing method, for the inverse modeling of two-dimensional magnetotelluric data. The results were satisfactory, and the GA2D algorithm was a good starting point for the inverse modeling of two-dimensional data.