Artificial intelligence for carbon sequestration: A multi-scale review of nature-based and engineered pathways

Area of Science:

• Environmental Science
• Artificial Intelligence
• Climate Change Mitigation

Background:

• Global carbon neutrality goals drive carbon sequestration technology development.
• Natural and engineered carbon sequestration methods face challenges in modeling, optimization, and mechanistic understanding.
• Complex carbon systems exhibit nonlinearity and multi-source heterogeneity.

Purpose of the Study:

• To systematically classify key carbon sequestration pathways (nature-based and engineered).
• To propose a structured "task-data-algorithm" mapping framework for AI model selection and benchmarking.
• To synthesize AI's role in modeling and optimizing carbon sequestration systems.

Main Methods:

• Systematic review and classification of carbon sequestration pathways.
• Comparative analysis of AI models (RF, SVM, CNN, RNN) for carbon system applications.
• Development of a "task-data-algorithm" mapping framework.

Main Results:

• AI models demonstrate specific strengths: CNNs for remote-sensing carbon mapping, RFs for soil organic carbon (SOC) prediction.
• The framework facilitates AI model selection and benchmarking for diverse carbon sequestration use cases.
• AI enables high-resolution prediction, feedback integration, and system optimization in both nature-based and engineered systems.

Conclusions:

• The proposed framework provides a strategic reference for improving AI model generalization and interpretability in carbon sequestration.
• This study lays a systematic foundation for intelligent, AI-enabled carbon sequestration systems.
• The findings support future policy and deployment of effective low-carbon strategies.