Deep learning-based oceanic internal wave detection from SWOT satellite imaginary observations.

Area of Science:

• Oceanography
• Remote Sensing
• Artificial Intelligence

Background:

• Oceanic Internal Waves (OIWs) are crucial for ocean dynamics but challenging to detect in satellite imagery.
• Subtle signatures and interference from noise complicate accurate OIW identification.

Purpose of the Study:

• To develop a novel hybrid deep learning framework for accurate and efficient detection of Oceanic Internal Waves (OIWs).
• To improve OIW detection from SWOT satellite data, addressing challenges like noise and subtle wave patterns.

Main Methods:

• A hybrid deep learning framework combining Convolutional Neural Networks (CNNs) and Transformer encoders with Spatial Reduction Attention (SRA).
• Utilizing a multi-scale fusion strategy to integrate hierarchical features for robust OIW classification.
• Validation on a large SWOT-based dataset of 31,495 annotated patches.

Main Results:

• Achieved state-of-the-art performance with 96.2% accuracy in OIW detection.
• Demonstrated superior performance compared to existing baseline methods.
• Enabled real-time inference at 35 ms per patch and full-swath processing within seconds due to SRA efficiency.

Conclusions:

• The proposed CNN-Transformer framework offers a scalable and accurate solution for operational ocean monitoring.
• Enhances maritime safety and contributes to long-term internal wave climatology studies.
• Highlights the effectiveness of integrating CNNs and Transformers with SRA for complex pattern recognition in remote sensing data.