Sound Speed Profile Research Articles

Seabed type and source parameters predictions using ship spectrograms in convolutional neural networks.

Broadband spectrograms from surface ships are employed in convolutional neural networks (CNNs) to predict the seabed type, ship speed, and closest point of approach (CPA) range. Three CNN architectures of differing size and depth are trained on different representations of the spectrograms. Multitask learning is employed; the seabed type prediction comes from classification, and the ship speed and CPA range are estimated via regression. Due to the lack of labeled field data, the CNNs are trained on synthetic data generated using measured sound speed profiles, four seabed types, and a random distribution of source parameters. Additional synthetic datasets are used to evaluate the ability of the trained CNNs to interpolate and extrapolate source parameters. The trained models are then applied to a measured data sample from the 2017 Seabed Characterization Experiment (SBCEX 2017). While the largest network provides slightly more accurate predictions on tests with synthetic data, the smallest network generalized better to the measured data sample. With regard to the input data type, complex pressure spectral values gave the most accurate and consistent results for the ship speed and CPA predictions with the smallest network, whereas using absolute values of the pressure provided more accurate results compared to the expected seabed types.

An equivalent source CVIS method and its application in predicting structural vibration and acoustic radiation in ocean acoustic channe

The acoustic wave propagating in ocean is severely distorted due to the sound speed profile of sea water and the reflection and absorption of seabed. The FEM/BEM is a preferable method to solve the fluid-structure interactive vibration and sound radiation problem of underwater structure in ocean acoustic environment, since the Green function can be applied in the integral equation. The simple source integral method is more preferable compared with Helmholtz integral because the calculation of normal derivative of Green function is avoided. In order to remove the irregular frequency problem, Zhou (Zhou et al., 2010) proposed a CVIS theory, which however introduced the normal derivative of Green function on CVIS surface and limited its application into ocean acoustical environment. Therefore, an equivalent source formulation of CVIS is proposed based on the definition of normal derivative so that the normal derivative of Green function on the integration of CVIS is replaced. The radiated SPL of spherical shell is calculated in shallow and deep sea with various sound speed profiles. It is shown that the equivalent source CVIS can be used to eliminate the irregular frequency problem of the far field radiated sound pressure of underwater elastic structure in ocean acoustic environment.

Reconstruction Method of Ocean Front Model Based on Fuzzy Cluster Analysis of Sound Speed Profile

Ocean fronts are mostly determined according to the horizontal temperature gradient, and various selection criteria may lead to different frontal zones. This study proposes a frontal zone delineation method based on sound speed profile. In this paper, the sound speed profile of the Kuroshio intrusion front in the Luzon Strait is classified via fuzzy cluster analysis based on transitive closure and fuzzy C-means algorithm. At the same time, considering that ocean fronts affect sound propagation, we verify the effectiveness of the two algorithms in terms of sound transmission loss. The results show that in different depth ranges, the effectiveness of the two algorithms differs in identifying the frontal zone. At depth below 300 m, the sound speed profile changes drastically, and fuzzy cluster analysis based on the transitive closure is more effective. At 300–700 m depth where the sound speed profile tends to be stable, the fuzzy C-means algorithm is more effective. At 700–1000 m depth near the SOFAR axis (the depth of the minimum sound speed), fuzzy cluster analysis based on the transitive closure is more effective. Finally, based on the sound speed profile, we combine both algorithms for the first time to obtain a three-layer geometric optimized structure of the Kuroshio intrusion front in the Luzon Strait. The classification of sound speed profiles via the two fuzzy cluster algorithms provides a new way to judge the frontal zone in different depth ranges and reconstruct geometric models of ocean fronts.

Reconstructing Sound Speed Profile From Remote Sensing Data: Nonlinear Inversion Based on Self-Organizing Map

By establishing a linear regression relationship between the projection coefficient of the empirical orthogonal function (EOF) of the sound speed profile (SSP) and remote sensing parameters of the sea surface, the single empirical orthogonal function regression (sEOF-r) method was used to reconstruct the underwater SSP from satellite remote sensing data. However, because the ocean is a complex dynamical system, the parameters of the surface and the subsurface did not conform to the linear regression model in strict sense. This paper proposes a self-organizing map (SOM)-based nonlinear inversion method that used satellite observations to obtain anomalies in data on the sea surface temperature and height, and combined them with the EOF coefficient from an Argo buoy to train and generate a map. The SSP was then reconstructed by obtaining the best matching neuron. The results of SSP reconstruction in the northern part of the South China Sea showed that the relationship between the parameters of the sea surface and the subsurface could be adequately expressed by the nonlinear neuronal topology. The SOM algorithm generated a smaller inversion error than linear inversion and had better robustness. It improved the average accuracy of reconstruction by 0.88 m/s and reduced the mean-squared reconstruction error to less than 1.19 m/s. It thus offered significant promise for acoustic applications.

Bidimensional Ray Tracing Model for the Underwater Noise Propagation Prediction

An increasing attention has recently been paid to the effect of the underwater noise field generated by ship activities on the marine environment. Although this problem is widely discussed in international treaties and conventions, it has not yet found a consolidated technical-scientific treatment capable of quantifying the level of underwater noise emissions produced by naval systems. As part of a national research collaboration, a novel code has been developed to predict noise propagation according to the Ray Tracing approach. Such optical geometry-based technique allows for calculating the Transmission Loss (TL) trend in its respective contributions: geometrical loss (due to the distance between the source and receiver), dissipation loss (due to the characteristics of the propagation environment), and reflection loss (due to the surfaces that delimit the field). The simulation requires as input parameters the source info as spatial position, frequency, and sound pressure level (SPL) as well as the sea properties like seabed depth, the speed of sound profile, the layers thickness the water column is divided into, the sea salinity, temperature, and pH. The simulation code provides the SPL spatial distribution useful as a fast industrial tool in the future studies addressed to identify the emission limits for the protection of marine wildlife.

Modelling and Simulation on Acoustic Channel of Underwater Sensor Networks

The reliability of the modelling system mainly depends on the simulation of underwater acoustic channel characteristics. The reliability of simulator is improved because of the use of Bellhop in NS‐Miracle. World Ocean Simulation System (WOSS) can retrieve the data of marine environment by accessing the database of seabed depth, sound speed profile and seabed sediment, and transmit them to Bellhop simulator automatically, so that the model is closer to the actual underwater acoustic transmission channel than not using WOSS. In order to verify the reliability of NS2/NS‐Miracle simulation system with WOSS, a centralized underwater sensor network with five nodes is simulated on the integrated simulation system. The characteristic empirical model, Bellhop Ray‐Tracing model, and WOSS combined with Bellhop model are, respectively, adopted to simulate underwater acoustic channel. The results of three types of simulation, such as average throughput, average delay, and packet error rate, and simulation time are very close under the same condition. It proves that the accuracy of integrated simulation system is as excellent as that of NS‐Miracle. However, WOSS can automatically acquire the actual sea environment parameters and provide them to the simulator, which can improve the authenticity of the simulation system. Furthermore, three MAC protocols, Aloha‐CS, CSMA/CA, and DACAP, are simulated on the integrated simulation system under the same condition including ocean environment, network topology, and parameters. The results show that the performance of CSMA/CA is greater than the other protocols in such networks. It also proves that the integrated simulation system can accurately simulate the relevant characteristics of the MAC protocol.

Lower turning point convegence zone in deep water with an incomplete channel

In a deep sea sound channel, rays will bend due to the sound speed profile, and convergence zone will occur when the rays are intensive. Transmission loss in the convergence zone is smaller and it is conducive to acoustic detection and communication. Therefore the study of acoustic characteristics in convergence zone is always the focus of deep-sea acoustics. A long-range sound propagation experiment is conducted in the South China Sea. An equivalent broadband explosive sound source of 1 kg is placed at a depth of 200 m, and the hydrophone receives the data at 3146 m far. The processing and analysis of the experimental data indicate that there is a convergence zone below the sound channel axis in the incomplete deep channel. Compared with the upper turning point convergence zone near the surface, this convergence zone has a high convergence gain at a long distance. The caustic lines of refracted type and refracted surface-refleted type are determined by means of ray-normal mode theory. It is found that the location of the deep convergence zone observed in the experiment is consistent with the position of the refracted caustic line. It is proved that the convergence zone is a lower turning point convergence zone formed by the superposition of a large number of normal modes in the same phase, and it has a convergence effect at a certain depth below the sound channel axis in the deep sea. The formation conditions of the convergence zone and the influence of sound source depth on the caustic structure of the convergence zone are studied. The comparisons of the transmission loss and the width between the upper and lower turning point convergence zone at a long distance aremade. The analysis shows that the convergence gain in the seventh lower turning point convergence zone is still no less than 10 dB. The influence of the vertical structure of sound velocity on the lower turning point convergence zone is studied. The theoretical analysis results are in good agreement with the experimental data.

Effect of earth curvature on long range sound propagation

Since the earth is approximately a sphere, the sound speed equivalent surface on the range-depth plane is not a parallel plane, but a concentric sphere at a long distance. Therefore, for a long-range sound propagation, the effect of the curvature of the earth cannot be ignored. In this paper, conformal mapping is used to propose a method of earth curvature correction without changing the existing sound field calculation model, and has the characteristics of good portability and simple calculation. The simulation results in a typical environment show that because of the earth curvature, the location of the convergence zone moves toward the sound source, and its movement can reach 10 km at 1000 km in distance. Before and after the earth curvature correction, the transmission loss difference can reach up to 15 dB at a particular location. Through the analysis of the simulation results in four typical sound speed profiles in Northwest Pacific Ocean, it is found that the movement of the convergence zone and the distance change are approximately linear, and the movement of the convergence zone increases by about 1km for every increase of the propagation distance by 100 km. For deep-sea channel propagation, the earth curvature will cause the arrival structure to move forward as a whole on the time axis, and the degree of the forward motion will gradually increase with the distance increasing. At 1000 km, the amplitude of the forward motion can reach 136 ms. In addition, the earth curvature will also cause the depth and time extension of the arrival structure. For the received time-domain waveform at a certain depth, there is a big difference between before and after the earth curvature correction. The modified results from different earth approximation models show that the accuracy of the calculation can be satisfied by approximating the earth as a standard sphere.

Learning location and seabed type from a moving mid-frequency source.

While source localization and seabed classification are often approached separately, the convolutional neural networks (CNNs) in this paper simultaneously predict seabed type, source depth and speed, and the closest point of approach. Different CNN architectures are applied to mid-frequency tonal levels from a moving source recorded on a 16-channel vertical line array (VLA). After training each CNN on synthetic data, a statistical representation of predictions on test cases is presented. The performance of a single regression-based CNN is compared to a multitask CNN in which regression is used for the source parameters and classification for the seabed type. The impact of water sound speed profile and seabed variations on the predictions is evaluated using simulated test cases. Environmental mismatch between the training and testing data has a negative impact on source depth estimates, while the remaining labels are estimated tolerably well but with a bias towards shorter ranges. Similar results are found for data measured on two VLAs during Seabed Characterization Experiment 2017. This work shows the superiority of multitask learning and the potential for using a CNN to localize an acoustic source and detect the surficial seabed properties from mid-frequency sounds.

Study of high-frequency acoustic current tomography

<p indent="0mm">The sea-land interface is the most concentrated region of human marine development activities, and online monitoring of river current is very important for material flux and ecosystem evolution. In this study, the high-frequency acoustic current tomography of sea-land interface was studied. Based on the principle of acoustic reciprocity, the acoustic inversion algorithm of current was investigated. The experiment was carried out on the Jiulongjiang sea-land interface in Zhangzhou. The high-frequency transducer, with a center frequency of <sc>60 kHz</sc> and bandwidth of <sc>40–80 kHz,</sc> was used. According to the sound speed profile and underwater topographic survey, ray acoustic simulation was performed to reveal high-frequency sound propagation. The experimental results showed that the acoustic signal has high-signal amplitude and peak value of cross-correlation, which can be used to reconstruct the current of the river. Real-time monitoring of the river current during the typhoon “haibeisi” was achieved. The results showed that high-frequency acoustic tomography can be applied to the study of hydrodynamics of sea-land interface, hydrological measurement, and monitoring of extreme events, such as flood peak and heavy rainfall.

An Integrated Positioning and Attitude Determination System for Immersed Tunnel Elements: A Simulation Study.

Immersed tunnel elements need to be exactly controlled during their immersion process. Position and attitude of the element should be determined quickly and accurately to navigate the element from the holding area to the final location in the tunnel trench. In this paper, a newly-developed positioning and attitude determination system, integrating a 3-antenna Global Navigation Satellite System (GNSS) system, an inclinometer and a range-measurement system, is presented. The system is designed to provide the absolute position of both ends of the element with sufficient accuracy in real time. Special attention in the accuracy analysis is paid to the influence of GNSS multipath error and sound speed profile. Simulations are conducted to illustrate the performance of the system in different scenarios. If both elements are very close, the accuracies of the system are higher than 0.02 m in the directions perpendicular to and along the tunnel axis.

Source depth estimation using spectral transformations and convolutional neural network in a deep-sea environment.

Multiple approaches for depth estimation in deep-ocean environments are discussed. First, a multispectral transformation for depth estimation (MSTDE) method based on the low-spatial-frequency interference in a constant sound speed is derived to estimate the source depth directly. To overcome the limitation of real sound-speed profiles and source bandwidths on the accuracy of MSTDE, a method based on a convolution neural network (CNN) and conventional beamforming (CBF) preprocessing is proposed. Further, transfer learning is adapted to tackle the effect of noise on the estimation result. At-sea data are used to test the performance of these methods, and results suggest that (1) the MSTDE can estimate the depth; however, the error increases with distance; (2) MSTDE error can be moderately compensated through a calculated factor; (3) the performance of deep-learning approach using CBF preprocessing is much better than those of MSTDE and traditional CNN.

Decadal Spatiotemporal Halocline Analysis by ISAS15 Due to Influx of Major Rivers in Oceans and Discrepancies Illustrated Near the Bay of Bengal

The discharge from rivers is one of the major factors of regional salinity perturbations in addition to precipitation, evaporation, and circulation of the ocean, whereas simulations regarding the marine environment are dominantly affected by ocean salinity. Moreover, perturbations in the timing and quantity of freshwater cause salinity fluctuations, which in turn, affect the communities of both plant and fauna. In this regard, the study ingeniously employs In Situ Analysis System-15 (ISAS15) data, which is freely available online, to ascertain the salinities in proximity of the major rivers around the globe. Such computations are multilayered, i.e., for 1, 3, 5, and 10 m, and conducted along major freshwater influxes, i.e., the Amazon River, Bay of Bengal (BoB), and Yangtze River, on decadal scales, i.e., in 2004 and in 2014. Depending upon the location and availability of ISAS-15 data, the area in proximity of the Amazon is analyzed horizontally, vertically, and obliquely, whereas the areas in proximity of the BoB and Yangtze estuary are analyzed vertically and obliquely. Similarly, the study analyzed the freshwater influx at the aforementioned locations both for the maxima and minima, i.e., during the particular months that observed the maximum and minimum influx into the ocean from the above-mentioned freshwater sources in 2004, as well as in 2014. The detailed analysis proved the outcomes to be conforming with the documented literary data along the Amazon and Yangtze estuaries. However, the computed analysis illustrated the anomalous values in proximity of the BoB. The study proceeds to discuss an ingenious approach of computing, as well as extrapolating, the salinities, temperatures, and sound speed profiles (SSPs) by employing in situ deep Argo data in order to counter such anomalies, as well as conjoin it with ISAS data, to investigate such regions with broader spatiotemporal capabilities for the future course of action. For this particular study, this method is employed on certain Argo buoys in order to prove the efficacy of the aforementioned novel approach.

Effect of environmental uncertainty on source localization from mid-frequency tonals using convolutional neural networks

In ocean acoustics, the lack of labeled training data creates the need to train on simulated data. The simulated training data need to represent the environmental variability in the anticipated area where acoustic data are collected. Further, the manner in which environmental uncertainty should be incorporated is a significant question facing all ocean acoustics applications of machine and deep learning. To begin addressing this question, a case study is presented using mid-frequency tonal data recorded on a VLA. The time-varying tonal levels are input to a convolutional neural network (CNN) which finds the source range, depth, and speed and seabed type. The CNN is trained on simulated data and then tested on simulated data with different sound speed profiles and seabed types than used in training. The impact of this mismatch on the CNN predictions highlights the need to carefully account for environmental variability during training in order to provide robust machine and deep learning applications in ocean acoustics. [Work supported by the Office of Naval Research.]

Modeling and measurement of relative intensity of acoustic arrivals within the beaufort duct

The Beaufort Duct, located at approximately 175 m depth in the Canada Basin, allows for long-range acoustic propagation from sources moored within the duct. As part of the Canada Basin Acoustic Propagation/Glider Experiment (CANAPE/CABAGE), LFM sweeps with frequencies around 250 Hz were transmitted by broadband moored sources and received by two Seagliders traversing between them. In addition to receiving acoustic data, these Seagliders measured temperature, pressure, and salinity, providing measurements of the sound speed profile throughout the CANAPE study region. These profiles were used as inputs for broadband acoustic parabolic equation predictions. Measured data receptions within the duct show intense peak arrivals prior to the final cutoff that are higher in relative intensity than predicted. The duct properties of the measured sound speed profiles are parameterized and acoustic predictions using actual profiles as well as idealized profiles are used to study the predictability of these intense peaks within the duct.

A classification approach to the characterization of seabed geoacoustic profiles via deep learning

Acoustic source localization algorithms in the shallow ocean require information about the ocean seabed. While past approaches have characterized seabed geoacoustic profiles via continuous physical parameters, a classification approach approximates the distribution of seabed types as a discrete set of seabed classes. To exclude acoustically indistinguishable seabeds from the discrete classes, the acoustic separation between two seabeds may be evaluated by the Pearson correlation between their characteristic transmission loss curves. This process yields a subset of discrete seabed classes. To distinguish these classes, we train a convolutional neural network classifier on the pressure time-series of SUS charges, which are simulated using the geoacoustic profiles of the seabed classes. This network successfully predicts seabed class, when tested on a holdout set where the sound speed profiles of the ocean are adequately similar to those in the training set. An acoustic separation measure between sound speed profiles, analogous to the one for seabeds, directly relates to the generalization error of the network. [Work supported by the Office of Naval Research Contract No. N00014-16-C-3065 and by the NSF REU program, Grant No. 1757998.]

Perturbing sound speed field estimates to study quality of acoustic ranging

Acoustic ranging is clearly dependent on the sound speed field c(x,y,z), where the points (x,y,z) are constrained to the two-dimensional region defined by a submerged source at a certain point in the ocean and a submerged receiver at a different but unknown point in the ocean. An in-situ estimate of c(x,y,z) based solely on a sound speed profile taken at the source of known location and one taken at the receiver at a unknown location has been used to obtain reasonable ranging estimates over large source-receiver separation [Sarma etal., IEEE UASP (2019)]. We discuss an approach to perturb this estimate of c(x,y,z) in a physically sound manner consistent with internal wave and other ocean processes in order to study the effect on the accuracy of ranging methods. The approach can indeed be applied to any estimate of c(x,y,z). The ultimate motivation for this work is in-situ determination of the inherent variability expected for a specific range estimate from any range estimation technique.

Representing sonar target spectral features using a two-dimensional Gabor wavelet

We present an alternate representation of sonar spectral features with popular machine learning techniques for automated target classification. Sonar target recognition suffers from feature uncertainties due to unpredictable environmental parameters: changing sound speed profiles, inhomogeneous propagation of acoustic waves, interference and delayed echoes due to multipath effects, ambient ocean noise and reverberation, and signal deformation due to reflections from the seafloor, sea surface, and the Doppler effect. These parameters are further dependent on target orientation and may combine in a nonlinear fashion, making target classification near impossible. We propose a representation of sonar spectral features using the two-dimensional Gabor wavelet as a kernel filter. This extracts highly informative features specific to a target, regardless of orientation. We validate the robustness of our representation against feature uncertainties by comparing the classification performance of three machine learning techniques—a support vector machine, random forest tree, and a neural network—trained on the raw spectrogram (acoustic color) versus the Gabor-filtered feature space through resulting confusion matrices. We will provide classification results of public domain experimental field data. [This research is funded by the Office of Naval Research under Grant No. N00014-19-1-2436.]

Using constructed impulse responses to study the influence of sound speed profiles on outdoor acoustic propagation

Time-domain predictions can be accomplished by either explicitly calculating within the time domain, such as with the Finite-Difference, Time-Domain method, or by processing frequency-domain calculations into impulse responses. While the FDTD method has been used extensively in outdoor sound propagation studies, the requirement for extensive computational resources greatly limits the computational domain. The impulse response method, while well-utilized in underwater sound propagation has seen limited use in outdoor sound studies. For a given environmental scenario, the impulse response may be obtained from the inverse Fourier transform of transmission loss simulations in the frequency-domain. When it is convolved with the time variation of the source, the receiver signal may be readily predicted. One may re-use the impulse response for any new source, so long as the sampling criteria are met, and the source and receiver positions and environment match those of the modeled situation. A discussion of the implications of frequency resolution and example comparisons to measurements are presented. [This work was funded by the Assistant Secretary of the Army (Acquisition, Logistics, and Technology) [ASA(ALT)] with portions funded under 0602784/T40/24 and 0602146/AR9/01. Distribution Statement A: Approved for public release; Distribution is unlimited.]

Study on the influence of sound speed profiles on the optimum depth of shallow water acoustic propagation

Study on the influence of sound speed profiles on the optimum depth of shallow water acoustic propagation