Ship source level estimation and uncertainty quantification in shallow water via Bayesian marginalization.

Area of Science:

• Ocean acoustics
• Marine bioacoustics
• Statistical signal processing

Background:

• Accurate estimation of underwater ship noise is crucial for marine ecosystem monitoring and naval applications.
• Shallow water environments present challenges due to complex sound propagation and unknown environmental parameters.
• Traditional methods struggle with uncertainties in seabed characteristics and source localization.

Purpose of the Study:

• To develop and apply a robust non-linear Bayesian marginalization approach for ship spectral source level estimation.
• To address challenges posed by unknown seabed properties and uncertain source depth in shallow water acoustics.
• To quantify the uncertainty in source level estimations for marine vessels.

Main Methods:

• Utilized a trans-dimensional Bayesian matched-field inversion to sample over unknown seabed models.
• Employed Metropolis-Hastings sampling to integrate over multiple source depths and ranges.
• Derived source levels and their uncertainties from marginal distributions of source strength.
• Applied the method to container ship radiated noise data from a horizontal array.

Main Results:

• Successfully estimated ship spectral source levels in a shallow water environment.
• Quantified the uncertainty in source level estimations, with an average of 3.8 dB/Hz for tonal frequencies.
• Demonstrated the effectiveness of the Bayesian marginalization approach in handling environmental and source uncertainties.

Conclusions:

• The developed non-linear Bayesian approach provides reliable ship source level estimation in complex shallow water conditions.
• This method offers improved accuracy and uncertainty quantification compared to conventional techniques.
• The findings contribute to better understanding and monitoring of underwater radiated noise from shipping activities.