Inversion of Time-Lapse Surface Gravity Data for Detection of 3-D CO2 Plumes via Deep Learning

We introduce two algorithms that invert simulated gravity data to 3-D subsurface rock/flow properties. The first algorithm is data-driven, deep learning (DL)-based approach, and the second is also data-driven but considers the temporal evolution of surface gravity events. The target application of these proposed algorithms is the prediction of subsurface CO2 plumes as a complementary tool for monitoring CO2 sequestration deployments. Each proposed algorithm outperforms traditional inversion methods and produces high-resolution, 3-D subsurface reconstructions in near real-time. In addition, our proposed methods achieve Dice scores of up to 0.8 for predicted plume geometry and near-perfect data misfit in terms of mu Gals. These results indicate that combining 4-D surface gravity monitoring (low-cost acquisition) with DL techniques represents an effective and nonintrusive method for monitoring CO2 storage sites.

Automated summary

This paper introduces two algorithms that can invert simulated gravity data to 3D subsurface rock/flow properties. The first algorithm is a data-driven, deep learning (DL)-based approach, while the second algorithm is also data-driven but considers the temporal evolution of surface gravity events. The target application of these algorithms is the prediction of subsurface CO2 plumes for monitoring CO2 sequestration deployments. Both algorithms outperform traditional inversion methods, producing high-resolution 3D subsurface reconstructions in near real-time. Additionally, the proposed methods achieve high scores for predicted plume geometry and data misfit, indicating the effectiveness of combining 4D surface gravity monitoring with DL techniques for monitoring CO2 storage sites.