基于交叉梯度约束的重力、磁法和大地电磁三维联合反演

为了降低单一地球物理方法反演的多解性及受噪声的影响程度,本文围绕重力、磁法和大地电磁法开展了三维联合反演的研究.重、磁采用基于对数障碍法的正则化反演算法,大地电磁使用limited-memory BroydenFletcher-Goldfarb-Shanno(L-BFGS)反演算法,引入交叉梯度函数实现了三种物性结构的相互耦合,最终开发出一套重磁电三维联合反演算法,并实现MPI并行加速计算.通过理论模型算例验证了算法的准确性,结果表明:不论是单棱柱体模型还是组合棱柱体模型,联合反演结果相较单独反演对于异常体的空间形态刻画以及物性数值恢复具有较好的提升;单棱柱体模型算例使得异常体的物性参数(密度、磁化率和电阻率)更加接近于真实的物性参数;组合棱柱体模型的联合反演结果不仅仅消除了围岩物性参数的假异常,而且还增强了异常体边界结构的恢复程度.

Three-dimensional joint inversion of gravity,magnetic and magnetotelluric data based on cross-gradient theory

In order to reduce the uncertainty of inversion from a single geophysical method and the influence of noise, this paper conducts three-dimensional joint inversion of the gravity, magnetic and magnetotelluric data. We use regularization algorithm for Gravity and magnetic inversions based on logarithm barrier method. We use limited-memory Broyden-Fletcher-Goldfarb-Shanno (L-BFGS) algorithm for Magnetotellurics inversion. In the joint inversion, we add cross-gradient function to achieve the structural coupling of these three kinds of physical models (density, magnetic susceptibility and resistivity). Finally, we developed three-dimensional joint inversion of gravity, magnetic and magnetoelectric data. Furthermore, we used the MPI parallel algorithm to speed up the calculation. The accuracy of the algorithm is verified by the inversions of several theoretical models. By inverting synthetic data from a single-prism model and a multiple-prism model, we conclude that joint inversion has better performance in recovering anomalous prism body's spatial morphology. The resistivity value of the single-prism is much closer to the true model in joint inversion. Besides, joint inversion not only removes electrical false anomaly of backgrounds to a certain degree, but also largely improves the resolution of density and magnetic susceptibility at deep depth in the multiple prism models. The results show that the joint inversion is better than the single inversion.