In Simple Terms

Carbon dioxide (CO₂) is one of the main greenhouse gases causing climate change. To slow global warming, scientists are exploring ways to store CO₂ safely underground in deep saltwater reservoirs. But until now, predicting how CO₂ behaves under real geological conditions has been difficult and expensive. A new study shows that artificial intelligence (AI) can do this job faster, cheaper, and with remarkable accuracy opening the door to more reliable climate solutions.

How It Works

The research team led by Dr. Peyman Pourafshary (Nazarbayev University, Kazakhstan) applied advanced machine learning (ML) algorithms together with colleagues from Russia (Chemical Engineering Department, Ufa State Petroleum Technological University) and China (Institute of Unconventional Oil & Gas, Northeast Petroleum University, Daqing 163318), Random Forest, Gradient Boost Regressor, and XGBoost to predict how carbon dioxide dissolves and spreads in brine under reservoir conditions.

Using 176 high-quality experimental and simulation data points, the models analyzed the effects of pressure, temperature, and salinity on the diffusion coefficient (DC) of CO₂. The Random Forest model achieved the highest accuracy (R² = 0.95).

Why It Matters

• Climate impact: Secure CO₂ storage is essential for achieving global net-zero targets.

• Industry relevance: Reliable predictions improve the design of CCS projects and CO₂-Enhanced Oil Recovery (EOR).

• Public safety: Better models reduce the risk of CO₂ leakage, ensuring long-term environmental security.

Study Highlights

• Temperature is the strongest driver of CO₂ diffusion.

• High salinity reduces diffusion by up to 65%, slowing storage efficiency.

• The models cut the need for costly laboratory experiments and deliver faster, more accurate forecasts for CCS operations.

Why It Is Unique

Unlike earlier studies that relied on small datasets or narrow conditions, this research:

• Integrates a larger and more diverse dataset (176 points).

• Demonstrates superior accuracy with the Random Forest model.

• Provides interpretability through SHAP analysis, showing clearly which parameters matter most.

• Represents international collaboration across Kazakhstan, Russia, and China.