In ocean acoustics, many types of optimizations have been employed to locate acoustic sources and estimate the properties of the seabed. How these tasks can take advantage of recent advances in deep learning remains as open questions, especially due to the lack of labeled field data. In this work, a Convolutional Neural Network (CNN) is used to find seabed type and source range simultaneously from 1 s pressure time series from impulsive sounds. Simulated data are used to train the CNN before application to signals from a single hydrophone signal during the 2017 Seabed Characterization Experiment. The training data includes four seabeds representing deep mud, mud over sand, sandy silt, and sand, and a wide range of source parameters. When applied to measured data, the trained CNN predicts expected seabed types and obtains ranges within 0.5 km when the source-receiver range is greater than 5 km, showing the potential for such algorithms to address these problems.

Marginal probability distributions for parameters representing an effective sound-speed structure of a fine-grained sediment are inferred from a data ensemble maximum entropy method that utilizes a sparse spatially distributed set of received pressure time series resulting from multiple explosive sources in a shallow-water ocean environment possessing significant spatial variability of the seabed. A remote sensing seabed acoustics experiment undertaken in March 2017 off the New England Shelf was designed so that multiple independent analyses could infer the statistical properties of the seabed. The current analysis incorporates the measured horizontal variability from interpretations of a subbottom profiling survey of the experimental area. An idealized range- and azimuth-dependent parameterization of the seabed is derived from identification of horizons within the seabed that define multiple sediment layers. A sparse set of explosive charges were deployed on circular tracks with radii of about 2, 4, and 6.5 km with an acoustic array at the center to correlate a set of random measurements to physical acoustic processes that characterize the seabed. The mean values of a surface sound speed ratio and a linear sound speed gradient for the fine-grained sediment layer derived from 12 data samples processed in the 25–275-Hz band provide an estimate of the effective sound-speed structure in a 130-km $^2$ area. The inferred sediment sound speed values are evaluated by predicting measured time series data not used in the statistical inference, and are also compared to historical measurements. Finally, the low-frequency maximum entropy estimate of the sediment sound speed along with physical measurements derived from piston core measurements are utilized to estimate the sediment grain bulk modulus.

This paper considers the use of broadband noise from a ship-of-opportunity in statistical inference for estimating geoacoustic parameters of a layered mud-sand seabed model via trans-dimensional (trans-D) Bayesian matched-field inversion, with applications to data collected with a bottom-moored horizontal array in the 2017 Seabed Characterization Experiment conducted on the New England Shelf. The trans-D approach applied here samples probabilistically over possible model parameterizations (different numbers of seabed layer interfaces), and provides quantitative uncertainty estimates of seabed geoacoustic profiles. Inversions are carried out for acoustic data sets collected both when the ship-of-opportunity (a container ship) was oriented with its bow and with its stern towards the array. A third inversion involved combining data from a series of segments along the ship track. Inversion results image an upper sediment layer 3–7 m thick with low-sound speed (close to the water sound speed) over higher speed sediment, with indication of a transition layer above the interface. Sediment parameter estimates from the inversions are in good agreement with direct measurements from sediment cores and other geophysical data collected in the experiment area.

Merchant ship-radiated noise, recorded on a single receiver in the 360–1100 Hz frequency band over 20 min, is employed for seabed classification using an ensemble of deep learning (DL) algorithms. Five different convolutional neural network architectures and one residual neural network are trained on synthetic data generated using 34 seabed types, which span from soft-muddy to hard-sandy environments. The accuracy of all of the networks using fivefold cross-validation was above 97%. Furthermore, the impact of the sound speed and water depth mismatch on the predictions is evaluated using five simulated test cases, where the deeper and more complex architectures proved to be more robust against this variability. In addition, to assess the generalizability performance of the ensemble DL, the networks were tested on data measured on three vertical line arrays in the Seabed Characterization Experiment in 2017, where 94% of the predictions indicated that mud over sand environments inferred in previous geoacoustic inversions for the same area were the most likely sediments. This work presents evidence that the ensemble of DL algorithms has learned how the signature of the sediments is encoded in the ship-radiated noise, providing a unified classification result when tested on data collected at-sea.

Numerical and experimental studies were conducted to investigate the advantage of utilizing absolute travel time information in Bayesian geoacoustic inversions of broadband acoustic data. It is shown that inversions using absolute travel time can yield smaller uncertainties compared to inversions using relative arrival time and maximum-likelihood estimates for clock time synchronization. Experimental data collected in the Seabed Characterization Experiment on the New England Mud Patch in the Middle Atlantic Bight were used for real data demonstration, and it is shown that inversions using relative arrival times have greater uncertainty in estimating source distance, which consequently affects the overall posterior probability distribution of inverted parameters. Numerical study enables investigation of performance dependence on the signal-to-noise ratio, and it is found that absolute travel time information may have more profound advantages when the signal-to-noise ratio is low.

The papers in this special section focus on the exploration of fine-grained sediments that were collected from the Seabed Characterization Experiment in 2017 (SBCEX2017), an international, multidisciplinary, multi-institutional research project.