Assessment of CO2 storage potential in global marine sediments based on machine learning methods

Abstract

The ocean, as a vast potential carbon sink, provides a promising avenue for long-term CO₂ storage through the formation of stable hydrate formation zones (HFZ) or negative buoyancy zones (NBZ) in subseabed sediments under high-pressure and low-temperature conditions. Despite its potential, CO₂ sequestration in marine sediments remains at an early stage due to technical challenges, insufficient understanding of subseabed conditions, and the absence of a comprehensive global storage potential assessment. A key factor in evaluating regional HFZ and NBZ thicknesses is the distribution of marine geothermal gradients. Unlike conventional methods that assume constant geothermal gradients or use simplistic interpolations, this study integrates diverse global geological and geophysical datasets with advanced machine learning regression techniques, including Random Forest Regression, Support Vector Regression, and Extreme Gradient Boosting Regression, to predict the high-resolution (0.1° × 0.1°) global distribution of marine geothermal gradients. Utilizing these predicted gradients, along with CO₂ hydrate phase equilibrium and CO₂ density curves, we present the first global maps of HFZ and NBZ thicknesses in marine sediments. The results highlight significant spatial variability in HFZ and NBZ thicknesses, offering valuable insights into ocean thermal structures and establishing a robust scientific basis for advancing large-scale CO₂ sequestration in marine sediments.