3-D inversion of potential field anomaly data

Abstract

We used a compact inversion algorithm to model a subsurface gravity/magnetic volumetric body in 3D space from the observed data. Because 3D modeling basically follows the underdetermined, ill-posed least squares approaches with more unknowns than observed data, we solve the damped least squares problems with a priori (or favorable) information for the use of the weighting functions for the depth change, data misfit and the model sensitivity. Depending on the choice of these regularization inputs to the minimum functionals, we obtain knowledge on how these choices affect the modeling images in 3D space as well as on the data misfits. In particular, the equivalent point source algorithm and Euler deconvolution method are implemented to better constrain model geometry and depth weighting in this iterative inversion. Different types of the sources such as dykes, sheets, and cavities are used for this study to help discern the proper choice of the regularization information.nowns than observed data, we solve the damped least squares problems with a priori (or favorable) information for the use of the weighting functions for the depth change, data misfit and the model sensitivity. Depending on the choice of these regularization inputs to the minimum functionals, we obtain knowledge on how these choices affect the modeling images in 3D space as well as on the data misfits. In particular, the equivalent point source algorithm and Euler deconvolution method are implemented to better constrain model geometry and depth weighting in this iterative inversion. Different types of the sources such as dykes, sheets, and cavities are used for this study to help discern the proper choice of the regularization information.