Inversion of 2D Magnetotelluric Data Using The Complex Singular Value Decomposition Method

The numerical solution to the two dimensional (2D) magnetotelluric (MT) forward problem can be obtained in many ways involving the solution of integral or differential equations and the inverse problem is generally solved using regularised iterative inversion techniques. However, the existing inverse schemes effect matrix computations in real domain due to operational simplicity. A common problem encountered when dealing with real matrices is their huge size in 2D regularised inversion. To partly overcome this problem, a new inversion strategy using complex singular value decomposition (CSVD) techniques have been developed. The use of analytical partial derivatives and a variety of problem regularization measures ensure that the scheme is stable and rapidly convergent. In this method, instead of using the Cagniard apparent resistivity and phase, the frequency normalised impedance (FNI) is adopted as the interpretative data functions to improve model resolution. Applications to synthetic and field data from Parnaiba Basin in Brazil are presented and it is shown that the complex form of the data-space and parameter-space eigenvectors contain information on parameter resolution.