A practical method for the inversion of magnetotelluric data for a layered earth

A numerical one‐dimensional magnetotelluric inversion technique based on the Nabetani‐Rankin algorithm for a layered earth is presented and illustrated with a noise‐free theoretical model and two examples with real field data. The inversion of noise‐free data is for practical purposes unique. The method is computationally efficient and removes most of the subjective biases which may be introduced into the interpretation of more or less real noisy data. A very satisfactory fit of the inverted result to field measurements is achieved using the least‐squares criterion applied to the real part of the Nabetani‐Rankin function V. The advantages of this function over others which have been suggested lie in its simplicity, its whiteness, and its relative freedom from noise. The Marquardt algorithm for the estimation of nonlinear parameters used for fitting the V curves overcomes the frequent failure of convergence in the Taylor series method and the slow convergence of the gradient method.