3-D Joint Inversion of Gravity and Magnetic Data Based on a Level-Set Constraint

Abstract

Joint inversion of gravity and magnetic data is a crucial approach for mitigating the inherent nonuniqueness of geophysical inversion and enhancing the resolution of subsurface imaging. However, conventional joint inversion methods, when establishing relationships between density and magnetization, often rely either on empirical petrophysical formulas that lack universal applicability or on explicit parameterizations that suffer from insufficient flexibility in delineating complex geological boundaries. To address these limitations, we propose a 3-D joint gravity and magnetic inversion method based on a level-set constraint. This method introduces level-set functions to implicitly represent the shared geological boundaries of density and magnetization models simultaneously, thereby inherently achieving structural unity between the two physical properties. This approach not only provides a powerful capability for characterizing complex and irregular shapes but also effectively enhances the stability of the inversion and the accuracy of boundary identification by integrating the complementary information from gravity and magnetic data. Synthetic tests show that the proposed joint inversion reduces errors compared to separate level-set inversions. For Model 1, the Hamming distance (HD), (a measure of level-set inversion error) drops from 1.11% (gravity) and 1.24% (magnetic) to 0.90%, and the mean relative absolute error (MRAE), (a measure of physical property inversion error) from 2.44 and 2.69 to 2.10. For Model 2, HD decreases from 1.57% and 1.60% to 1.34%, and MRAE from 3.65 and 3.76 to 3.37. Furthermore, in the field data application from the McFaulds Lake area, the cross-plots of residual density and magnetization from the joint inversion exhibit better consistency than those from separate inversions, and the recovered boundaries are sharper compared with other joint inversion methods, demonstrating the practical value of the proposed method in real-world exploration scenarios.