Inversion of Time-Lapse Surface Gravity Data for Detection of 3D CO$_2$ Plumes via Deep Learning

TL;DR

This paper presents three innovative deep learning algorithms for inverting time-lapse surface gravity data to accurately and rapidly detect 3D CO₂ plumes, enhancing monitoring of CO₂ sequestration with high resolution and efficiency.

Contribution

It introduces three novel algorithms combining deep learning and physical modeling for 3D CO₂ plume detection from gravity data, outperforming traditional methods.

Findings

Achieved Dice scores up to 0.8 for plume geometry

Produced near-perfect data misfit in microgals

Enabled near real-time, high-resolution 3D reconstructions

Abstract

We introduce three algorithms that invert simulated gravity data to 3D subsurface rock/flow properties. The first algorithm is a data-driven, deep learning-based approach, the second mixes a deep learning approach with physical modeling into a single workflow, and the third considers the time dependence of surface gravity monitoring. The target application of these proposed algorithms is the prediction of subsurface CO$_2$ plumes as a complementary tool for monitoring CO$_2$ sequestration deployments. Each proposed algorithm outperforms traditional inversion methods and produces high-resolution, 3D subsurface reconstructions in near real-time. 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 4D surface gravity monitoring with deep learning techniques represents a low-cost, rapid, and non-intrusive method for monitoring CO$_2$ storage sites.