Continuous forecasting of range-dependent ocean sound speed field: Diffusion model meets multi-output Gaussian...

Area of Science:

• Oceanography
• Acoustics
• Machine Learning

Background:

• Accurate forecasting of the range-dependent sound speed field (SSF) is crucial for environment-aware underwater acoustic detection and communication.
• Existing methods like Gaussian process regression (GPR) and conditional diffusion models have limitations in capturing spatial correlations and continuous forecasting.

Purpose of the Study:

• To develop an integrated model combining multi-output GPR and conditional diffusion models for enhanced SSF forecasting.
• To enable continuous, precise prediction of range-dependent SSFs for improved underwater acoustic applications.

Main Methods:

• Integration of multi-output Gaussian process regression (GPR) with conditional diffusion models.
• Careful design of diffusion noise, neural architecture, and training strategies for the integrated model.
• Validation using HYCOM hindcast datasets from the South China Sea.

Main Results:

• The proposed integrated model demonstrates superior performance compared to state-of-the-art baselines.
• Accurate forecasting of range-dependent SSFs at any given time.
• Improved prediction of associated underwater transmission losses.

Conclusions:

• The novel integrated model effectively addresses limitations of standard GPR and conditional diffusion models for SSF forecasting.
• This approach offers a significant advancement for real-time, environment-aware underwater acoustic systems.
• The model provides precise predictions essential for optimizing underwater acoustic detection and communication.