Change-point detection for recursive Bayesian geoacoustic inversions.

Area of Science:

• Ocean acoustics
• Geophysical inversion
• Signal processing

Background:

• Geoacoustic inversion is challenging in low signal-to-noise ratio (SNR) conditions, often requiring extended observations and source/receiver motion.
• Environmental changes over time or space can alter underlying model parameters, complicating inversion efforts.

Purpose of the Study:

• To develop and demonstrate a robust inversion method capable of handling abrupt or gradual changes in geoacoustic model parameters.
• To introduce a change-point detection technique integrated with recursive Bayesian inversion for adaptive geoacoustic modeling.

Main Methods:

• A change-point detection algorithm is developed using importance samples and weights from recursive Bayesian inversion.
• The method adapts by restarting inversion after abrupt parameter changes or continuing until significant accumulated mismatch triggers change detection for gradual changes.
• Change-point detections inform heuristics for controlling coherent integration time in recursive Bayesian inversion.

Main Results:

• The proposed method successfully detects abrupt and gradual changes in geoacoustic model parameters.
• Simulations demonstrate the effectiveness of the change-point detection in managing parameter variations.
• The approach is validated using parameters from the Shallow Water 2006 experiment with low SNR linear frequency modulation pulses.

Conclusions:

• The developed change-point detection method enhances geoacoustic inversion in dynamic, low SNR environments.
• This adaptive strategy improves the reliability and accuracy of inversion by accounting for environmental variability.
• The findings have implications for underwater acoustic sensing and environmental monitoring.