Leveraging satellite observations and machine learning for underwater sound speed estimation

Area of Science:

• Oceanography
• Marine Acoustics
• Climate Science

Background:

• Accurate underwater sound speed is critical for oceanography, climate science, and sonar systems.
• Current methods for mapping sound speed are limited by data availability and computational power.
• Real-time, high-resolution 3D sound speed mapping remains a significant challenge.

Purpose of the Study:

• To develop an integrated approach for rapid and accurate estimation of 3D underwater sound speed fields.
• To overcome limitations of traditional methods in mapping vast ocean regions.
• To enable real-time, high-resolution sound speed prediction for diverse oceanographic applications.

Main Methods:

• Combined remote sensing (sea surface temperature, salinity) with machine learning algorithms.
• Integrated spatial and temporal variables for enhanced prediction accuracy.
• Validated the model using in-situ profiles and Argo float data.

Main Results:

• Successfully predicted 3D underwater sound speed with high accuracy across different seasons and regions.
• Demonstrated the model's capability for detailed, real-time ocean mapping.
• Validated accuracy against independent datasets, confirming reliability.

Conclusions:

• The integrated approach significantly advances underwater sound speed prediction capabilities.
• The model offers a powerful tool for acoustic propagation modeling, underwater detection, and communication.
• This method provides a foundation for improved ocean monitoring and resource exploration.