Bayesian source tracking via focalization and marginalization in an uncertain Mediterranean Sea environment

Area of Science:

• Ocean acoustics
• Geophysics
• Signal processing

Background:

• Accurate underwater source localization is critical for various applications.
• Environmental uncertainties and data limitations pose significant challenges to acoustic tracking.
• Previous studies have focused on specific aspects of source tracking, but a comprehensive approach considering multiple factors is needed.

Purpose of the Study:

• To apply and evaluate Bayesian source tracking methods in an uncertain marine environment.
• To investigate the impact of data information content and prior information on track accuracy.
• To compare two Bayesian tracking approaches: Focalization-tracking and Marginalization-tracking.

Main Methods:

• Simultaneous inversion of acoustic data from multiple time segments for towed sources.
• Incorporation of environmental uncertainties (water-column, seabed properties) as nuisance parameters.
• Application of Focalization-tracking (maximizing posterior probability density) and Marginalization-tracking (integrating over environmental parameters).
• Inclusion of source-velocity constraints in both tracking approaches.

Main Results:

• Bayesian source tracking effectively handles environmental uncertainties in shallow-water acoustics.
• Data information content (number of segments, frequencies, SNR) and prior information significantly influence track uncertainty.
• Both Focalization-tracking and Marginalization-tracking provide valuable insights into source trajectories and uncertainties.
• The study demonstrates the robustness of the methods with towed-source acoustic data from the Mediterranean Sea.

Conclusions:

• Bayesian source tracking offers a robust framework for localizing underwater acoustic sources in complex environments.
• Understanding the interplay between data quality, environmental priors, and tracking performance is crucial for reliable source localization.
• The developed methods are applicable to real-world scenarios, such as tracking marine seismic sources or submerged vehicles.