Stochastic Inversion of Gravity Data Accounting for Structural Uncertainty

Conventional gravity inversion techniques have limited ability to quantify structural uncertainty in geologic models. In this paper, a stochastic framework is proposed that directly incorporates fault-related and density-related uncertainty into the inversion process. The approach uses Monte Carlo simulation to generate model realizations and the gradual deformation method to further refine models to match observed data. To guarantee that model realizations are structurally restorable, fault displacements are generated using a kinematic modeling approach in which fault model properties such as the number of faults, location, dip, slip, and orientation are considered uncertain. Using a synthetic case study problem, a reference gravity field was inverted to generate a suite of posterior model realizations. Analysis of the posterior models was used to create a fault probability map as well as quantify the distribution of slip and dip of faults in three zones of deformation. Uncertainty in density values was found to be greatly reduced in the top 250 m depth, suggesting limited sensitivity to deeper sources in this example. Following the synthetic case study problem, the inversion approach was applied to a field-observed gravity profile in Dixie Valley, Nevada, and the inversion results were compared to a previously published forward gravity model. By generating a suite of posterior models, structural uncertainty can be better assessed to make more informed decisions in a host of subsurface modeling problems.

Automated summary

A stochastic framework is proposed in this paper to incorporate fault-related and density-related uncertainty into the inversion process, using Monte Carlo simulation and gradual deformation method. The approach aims to assess structural uncertainty through generating model realizations and refining them to match observed data.