Fast and accurate numerical modeling of gravity and magnetic anomalies is the basis of field-data inversion and quantitative interpretation. In gravity and magnetic prospecting, the computation and memory requirements of practical modeling is still a significant issue, which leads to the difficulty of using efficient and detailed inversions for large-scale complex models. A new 3D numerical modeling method for gravity and magnetic anomaly in a mixed space-wavenumber domain is proposed to mitigate the difficulties. By performing a 2D Fourier transform along two horizontal directions, 3D partial differential equations governing gravity and magnetic potentials in the spatial domain are transformed into a group of independent 1D differential equations wrapped with different wavenumbers. Importantly, the computation and memory requirements of modeling are greatly reduced by this method. A modeling example with 4,040,100 observations can be finished in approximately 28 s on a desktop using a single core, and the independent differential equations are highly parallel among different wavenumbers. The method preserves the vertical component in the space domain, and thus a mesh for modeling can be finer at a shallower depth and coarser at a deeper depth. In general, the new method takes into account the calculation accuracy and the efficiency. The finite-element algorithm combined with a chasing method is used to solve the transformed differential equations with different wavenumbers. In a synthetic test, a model with prism-shaped anomalies is used to verify the accuracy and efficiency of the proposed algorithm by comparing the analytical solution, our numerical solution, and a well-known numerical solution. Furthermore, we have studied the balance between computational accuracy and efficiency using a standard fast Fourier transform (FFT) method with grid expansion and the Gauss-FFT method. A model with topography is also used to explore the ability of modeling topography with our method. The results indicate that the proposed method using the Gauss-FFT method has characteristics of fast calculation speed and high accuracy.

We have developed a new 3D magnetotelluric modeling scheme in a mixed space-wavenumber domain. The modeling scheme is based on using a 2D Fourier transform along two horizontal directions to solve a vector-scalar potential formula derived from Maxwell’s equations based on the primary-secondary potential separation. The derived 1D governing equations in a mixed space-wavenumber domain are solved by using the finite-element method (FEM) together with a chasing method, and then the 2D inverse Fourier transform is used to recover the final solution of the electromagnetic (EM) fields in the 3D spatial domain. An iterative scheme is applied to approximate the true solution by repeating the previous steps because the governing equations cannot be solved directly due to an unusual primary-secondary potential field separation used. Nevertheless, the new method is capable of reducing the memory requirement and computational time in the mixed domain, and the 1D governing equations are highly parallel among different wavenumbers. For each of the 1D equations, the two- or four-node Gaussian quadrature rule can be used in both horizontal directions for Gauss fast Fourier transform. It is worth mentioning that the linear matrix equation to be solved is a fixed bandwidth system, and the chasing method is more efficient and convenient than solvers with preconditioners for the 1D matrix equations. The reliability and efficiency of the newly proposed method are verified with three synthetic 3D models by comparisons with a classical integral equation solution, an adaptive FEM solution, and a nonadaptive FEM solution. The proposed algorithm will be used in electrical resistivity tomography and controlled-source EM methods in future studies.

Gravity inversion plays a crucial role in estimating subsurface structural information. This process typically involves dividing the subsurface into numerous rectangular prisms and determining the parameters for each one. However, applying the method to large-area model parameters inversion is challenging due to high memory demands and computational costs associated with the extensive sensitivity matrix. To address this problem, a new gravity anomaly analytical expression has been developed to reduce the computational cost of the sensitivity matrix for gravity inversion. The compressed matrix technique, which eliminating the need to store and calculate redundant sensitivity matrix, further reduces storage and computation time. Additionally, depth and horizontal weighting functions have been introduced to alleviate inherent ambiguity and enhance the resolution of gravity inversion. The conjugate gradient approach is utilized as the fundamental solver leveraging the fast forward modeling that comes from the fast calculation of the sensitivity matrix and compressed matrix techniques. Model tests and real data applications demonstrate that the proposed approach effectively resolves large-scale precise gravity inversion challenges.