Work

Knowledge-informed 3D Stochastic Geometry Gravity Inversion with FFT Acceleration

Year: 2026
Type: article
Abstract: Reconstructing three-dimensional (3D) geometric features is important for mineral exploration and development. However, most geophysical inversion studies focus on recovering physical properties, with less attention paid to geometric characteristics and associated uncertainty of the resulting models. Although the stochastic inversion framework has been widely used for uncertainty quantification, it still faces significant computational challenges. To address these issues, a 3D stochastic geometry gravity inversion framework accelerated by an FFT-based strategy is developed. Geological interfaces are represented by an implicit level-set function, and spatially correlated perturbations are generated in the frequency domain to improve sampling efficiency. Gravity observations, drillhole lithological constraints, and geological diagrams are integrated into a unified probabilistic objective function. Synthetic experiments show that the method reconstructs complex intrusion geometries with clearer boundaries than property-based inversion and provides spatial uncertainty estimates. Frequency-domain sampling achieves reconstruction accuracy comparable to spatial-domain sampling while reducing computation time to approximately one third. A robustness test further indicates that the framework can recover the main target geometry even when the geological diagram is biased. Application to the Olympic Dam deposit demonstrates that the method can construct geologically plausible 3D models and identify regions of higher structural uncertainty. This framework provides an efficient way to combine geophysical data and geological knowledge for uncertainty-aware geometry inversion.
Language: English
Subfield: Geophysics
Open Access status: gold
Estimated APC paid: $3,416
OpenAlex ID: W7236373433
FWCI:
Cites:
Cited by:
Related to: 10