An Adaptive Conjugate Gradient Least-Squares Regularization (ACGLSR) Method for 3D Gravity Density Inversion

Gravimetric inverse modelling of subsurface mass density plays a key role in a gravimetric interpretation of the Earth’s inner structure. The conjugate gradient technique is an efficient numerical tool for solving large-scale matrix equation systems. However, numerical procedures of finding the optimal regularization parameter based on applying the L-curve or generalized cross-validation methods are very time-consuming. To improve the numerical efficiency, we develop the adaptive conjugate gradient least-squares regularization technique that combines the conjugate gradient iterative algorithm with the adaptive regularization parameter methods. The proposed algorithm is developed for the recovery of a three-dimensional (3D) density structure from gravity and gravity gradient data. The algorithm is verified by using synthetic and real data. Results indicate that the developed algorithm could realistically recover the density information using the depth of density interface and the dip angle of titled body.