Gravity Gradient Inversion Based on Multiscale Recursive Adaptive TV-Misfit Optimization

This is the authors' abstract. We don't add key points for this paper.

Gravity gradient exploration has become an important method for mineral resource exploration and fine crustal structure research. It responds with high resolution to tiny subsurface density changes and effectively suppresses regional field interference. This study builds a multiscale gravity gradient joint inversion method based on adaptive total variation (TV)-Misfit, using an improved alternating direction method of multipliers (ADMM) algorithm to solve the objective function. The method calculates the optimal focusing factor μ by introducing a TV-Misfit hyper- L curve. It also uses a multiscale recursive strategy to search at different grid scales, improving computational efficiency and algorithm stability. We use the ADMM to solve the inversion objective function, strictly constraining physical bounds and spatial consistency. We introduce a multidimensional quantitative evaluation system, including the root-mean-square error, structural similarity index measure, Dice coefficient of the target area, and centroid depth offset, to evaluate the experimental results of complex models. We compare our results with typical parameter μ =0.5 of the hybrid regularization method and L₁ focused inversion. The comparison shows that our method has certain advantages in boundary characterization and deep vertical resolution. Finally, we apply this method to the measured gravity gradient data of the Vinton dome. It successfully recovers the spatial morphology of the high-density caprock and low-density salt body, proving the feasibility of this method in real data inversion.