Scattering statistics of rock outcrops: Model-data comparisons and Bayesian inference using mixture distributions

Area of Science:

• Geophysics
• Acoustics
• Statistical modeling

Background:

• Seafloor acoustic scattering is crucial for sonar performance.
• Understanding the probability density function (PDF) of acoustic field amplitude is key.
• Previous models may not fully capture complex scattering environments.

Purpose of the Study:

• To determine the most appropriate statistical model for seafloor acoustic scattering.
• To compare single-component and two-component mixture models.
• To identify the best-performing model for a rocky Norwegian seafloor environment.

Main Methods:

• Utilized synthetic aperture sonar (SAS) to measure acoustic field amplitude.
• Interpreted measurements in terms of probability of false alarm.
• Applied Bayes' theorem to estimate PDF of mixture model parameters.
• Evaluated single-component, K-distribution mixture, and Rayleigh-generalized Pareto mixture models.

Main Results:

• Two-component mixture models outperformed single-component models.
• A mixture of two K distributions and a Rayleigh-generalized Pareto mixture showed the best performance.
• The K-K mixture exhibited significant parameter correlation.
• The Rayleigh-generalized Pareto mixture showed parameter correlation and multiple modes.

Conclusions:

• The mixture between two K distributions is the most applicable statistical model for this dataset.
• This model provides a robust representation of acoustic scattering in rocky seafloor environments.
• Findings contribute to improved sonar data interpretation and performance prediction.