Resolution and Error Analysis in Non-Linear Inversion of Potential Field Data

In geophysical inverse problems with increasing number of data and model parameters or increasing level of data errors (noise)، we have many problems about model instability and inversion of bad condition matrixes. Progressive methods can be use for achieving desire convergence and precision in mentioned systems base on real geological reality. Most geophysical inverse problems can be solved by minimizing a misfit function between observed data and theoretical estimations. In inverse problems studying model stability and independence of model parameters can be done by Error analysis and resolution matrix. Other important factors for evaluating inverse methods are rate of convergence، running time of computer codes and number of iterations. This paper had focused on Marquardt-Levenberg (ridge regression) as one of conventional and efficient geophysical inverse method against subspace inversion to invert Magnetic data. In subspace inversion، Spanning original model parameters space to some subspaces can be done in two essential manner respect to physical dimension of model parameters. The efficiency of subspace method is demonstrated by comparing quality of resolution matrixes and error analysis in Marquardt-Levenberg with subspace inversion method. Comparison results show that the subspace method has fast and stable convergence compare to conventional method.