AI for atmosphere-ocean sciences: advancements, challenges and ways forward.

Area of Science:

• Earth Science
• Atmosphere-Ocean Sciences
• Artificial Intelligence Applications

Background:

• Artificial intelligence (AI) presents transformative potential for addressing complex Earth science challenges.
• The integration of AI in atmosphere-ocean sciences is rapidly advancing.
• Current research focuses on leveraging AI for improved weather and climate insights.

Purpose of the Study:

• To review advances and challenges of AI in atmosphere-ocean sciences.
• To explore the application of AI in weather and climate forecasting.
• To outline future directions for AI in Earth science research.

Main Methods:

• Review of deep-learning methods and their application in forecasting.
• Examination of AI for phenomenon detection, data assimilation, and bias correction.
• Discussion on explainable AI and hybrid physics-AI modeling.

Main Results:

• Deep learning outperforms dynamical models in weather and climate forecasting accuracy and efficiency.
• AI enhances data assimilation, reconstruction, bias correction, and downscaling.
• Hybrid physics-AI models are crucial for generalizability and causal consistency.

Conclusions:

• Explainable AI is essential for building trust and extracting mechanistic insights.
• A framework for AI-based model intercomparison is needed for rigorous benchmarking.
• AI represents a paradigm shift for understanding and adapting to climate change, with applications in early-warning systems and green energy.