A Study Of Two-dimensional Magnetotelluric Conjugate Gradient Inversion Based On Data Space

Magnetotelluric sounding method(MT), which use the natural electromagnetic field as the filed source, is a frequency domain electromagnetic method solve the geological problems. By observing the electric field component and magnetic field component, it calculates the cagniard resistivity and phase. Geophysicists from all over the world have conducted a variety of methods for the inversion problem and acquired good effects in practical use. However, they never stop to improving the methods and exploring the new ones.We accomplished the TE and TM mode of two-dimensional magnetotelluric forward by using the finite difference method and then Verified the accuracy and applicability of this algorithm using the typical geoelectric models. After that, The two-dimensional magnetotelluric inversion, which based on the data space conjugate gradient algorithm, has been accomplished. This algorithm updates the operator by using conjugate gradient algorithm model which based on the Occam algorithm. it can calculate the sensitivity from the data space instead of the model space, Greatly reduces the computational running memory, On the other hand, this algorithm calculates the JACOBI matrix or the matrix transpose and vector product, avoiding the storage and computing the JACOBI matrix directly.We inverse some typical geoelectrical model using the nonlinear conjugate gradient algorithm based on the data space and compared the results from the nonlinear conjugate algorithm. The results show that: DCG inversion algorithm is as accurate as the NLCG algorithm. The speed of convergence of DCG inversion algorithm is faster and has a higher vertical resolution than any other algorithms. The paper also discussed the inversion parameters such as the noise immunity, initial model, regularization factor. After that, we found that it is better to choose the “L” curve for regularization factor. This method has many good characteristics such as better anti-noise property, less dependence in the initial background resistivity and better adaptability.etc. It can provide a new way to interpretate the magnetotelluric data and study the three-dimensional inversion.