A Novel Mesoscale Eddy Identification Method Using Enhanced Interpolation and A Posteriori Guidance

Area of Science:

• Oceanography
• Marine Science
• Deep Learning Applications

Background:

• Mesoscale eddies significantly impact marine environments.
• Accurate identification of eddies is crucial for oceanographic research.
• Current methods for identifying anomalous eddies are often segmented and require secondary analyses.

Purpose of the Study:

• To develop an advanced deep learning model for enhanced mesoscale eddy identification.
• To improve the accuracy and stability of identifying both normal and anomalous eddies.
• To integrate multi-source data and attention mechanisms for more comprehensive eddy detection.

Main Methods:

• A deep learning model was developed integrating multi-source fusion data.
• A Squeeze-and-Excitation (SE) attention mechanism was incorporated to enhance feature learning.
• Comparative ablation experiments were conducted to validate the model's performance.

Main Results:

• The proposed deep learning model demonstrated enhanced accuracy in identifying both normal and anomalous mesoscale eddies.
• The integration of multi-source data and the SE attention mechanism proved effective.
• Ablation studies confirmed the model's superior performance compared to existing approaches.

Conclusions:

• The developed deep learning framework offers a promising approach for nuanced, multi-source, and multi-class mesoscale eddy identification.
• This method advances the field by providing a more integrated and accurate way to study eddy dynamics and effects.
• The study highlights the potential of deep learning in complex oceanographic data analysis.