OCCAM-nonlinear Conjugate Gradients Algorithm For 2-D Magnetotelluric Inversion

We have developed a two-dimensional occam-nonlinear conjugate gradients inversion algorithm of magnetotelluric(MT).This inversion algorithm combines the storage and computational efficiency of nonlinear conjugate gradient inversion algorithm with the OCCAM scheme for regularization.During the inversion process,this inversion algorithm does not need to calculate and store the Jacobian matrix.It only requires two pesudo-forward to updates the model,the first pesudo-forward is to compute the Jacobian matrix multiplying an vector and the second one is to compute the transpose of the Jacobian matrix multiplying another vector.The choice of Lagerange multiplier of occam scheme can then be used to this inversion algorithm,which can accelerate inversion and get the smoothest model in the meantime.We modified the selected method of Lagerange multiplier in Occam's inversion.In the search progress,the data misfit between the calculated data are added and compare with the target data level.This can decrease the number of forward and then accelerate the efficiency of inversion.We illustrate the occam-nonlinear conjugate gradients inversion algorithm on simple problems,2D geoelectric model,using synthetic data.The target structure can be well resolved and the efficiency of the occam-nonlinear conjugate gradients inversion algorithm is better than the nonlinear conjugate gradients inversion algorithm.We also use this inversion algorithm to invert some real MT data.From the trial inversion with the synthetic and real MT data,the practicability,the reliability and the efficiency of occam-nonlinear conjugate gradients inversion algorithm is verified.