Estimation of the spatial variability of the New England Mud Patch geoacoustic properties using a distributed array of hydrophones and deep learninga)

Ariel Vardi ^1,2, Peter H Dahl ³, David Dall'Osto ³

⁰Department of Applied Ocean Physics and Engineering, Woods Hole Oceanographic Institution, Woods Hole, Massachusetts 02543, USA.

The Journal of the Acoustical Society of America +

|

December 24, 2024

View abstract on PubMed

Summary

Deep learning rapidly models seafloor sediment properties using impulse signals. This efficient method reveals spatial variations in sound speed, aiding underwater acoustic studies.

Area Of Science

Oceanography
Geophysics
Acoustics

Background

Geoacoustic inversion traditionally demands substantial computational power.
Understanding shallow water sediment properties is crucial for acoustic modeling.
Previous methods lacked efficiency for large-scale spatial analysis.

Purpose Of The Study

To develop and validate a deep learning (DL) based spatial environmental inversion scheme.
To model a single spatially-varying sediment layer using broadband impulse signals.
To analyze sediment sound speed and sound speed ratio in the New England Mud-Patch (NEMP).

Main Methods

Utilized deep learning (DL) with broadband impulse signals for inversion.
Applied the scheme to synthetic data simulating Seabed Characterization Experiment 2022 (SBCEX22) conditions.
Processed 1836 signals from bottom-moored hydrophones using a neural network for rapid inversion.

Main Results

Successfully predicted the lateral spatial structure of sediment sound speed and its ratio with water sound speed.
Validated the DL inversion scheme on a range-dependent synthetic test set.
Observed a southwest to northeast increase in compressional sound speed and sound speed ratio in the NEMP.

Conclusions

Deep learning offers an efficient alternative to traditional geoacoustic inversion methods.
The DL approach enables rapid processing of numerous signals for spatial analysis.
The findings are consistent with existing geological and acoustic data from the NEMP, highlighting DL's potential in shallow water geoacoustic inversion.