Inverse modeling of Magnetic Data using Subspace Method

In this paper we used orthogonal basis functions and expansion coefficients for inverse modeling of magnetic data. The basis functions chosen are normalized eigenvectors of second derivation of the objective function (Hessian matrix) calculate for an initial model. Limited number of basis vectors obtained in this way defines a new subspace in model parameters space. A new objective function is defined in term of these new parameters and minimized in subspace of original space. As in geophysical inverse problems we need to inverse matrixes that are functions data and geometry of data and model parameters. The matrix inversion in new subspace of the original space will be better conditions due to less dimensionality in the inversion. Since the most significant eigenvectors corresponding the largest eigen values in Singular Value Decomposition (SVD) of matrixes. Others eigenvectors have less influence in fitting data or lead inversion procedures to local minima. With apply subspace method inversion will be fast and stable against the noise. The efficiency of the method is tested with synthetic and real magnetic data (acquired from Moghan area, north-west of Iran). The results proved fast convergence and stability of inversion against the noise.