East China Sea Experiment Research Articles

Source depth estimation with feature matching using convolutional neural networks in shallow water

A feature matching method based on the convolutional neural network (named FM-CNN), inspired from matched-field processing (MFP), is proposed to estimate source depth in shallow water. The FM-CNN, trained on the acoustic field replicas of a single source generated by an acoustic propagation model in a range-independent environment, is used to estimate single and multiple source depths in range-independent and mildly range-dependent environments. The performance of the FM-CNN is compared to the conventional MFP method. Sensitivity analysis for the two methods is performed to study the impact of different environmental mismatches (i.e., bottom parameters, water column sound speed profile, and topography) on depth estimation performance in the East China Sea environment. Simulation results demonstrate that the FM-CNN is more robust to the environmental mismatch in both single and multiple source depth estimation than the conventional MFP. The proposed FM-CNN is validated by real data collected from four tracks in the East China Sea experiment. Experimental results demonstrate that the FM-CNN is capable of reliably estimating single and multiple source depths in complex environments, while MFP has a large failure probability due to the presence of strong sidelobes and wide mainlobes.

Deep-learning geoacoustic inversion using multi-range vertical array data in shallow water.

A multi-range vertical array data processing (MRP) method based on a convolutional neural network (CNN) is proposed to estimate geoacoustic parameters in shallow water. The network input is the normalized sample covariance matrices of the broadband multi-range data received by a vertical line array. Since the geoacoustic parameters (e.g., bottom sound speed, density, and attenuation) have different scales, the multi-task learning is used to estimate these parameters simultaneously. To reduce the influence of the uncertainty of the source position, the training and validation data are composed of the simulation data of different source depths. Simulation results demonstrate that compared with the conventional matched-field inversion (MFI), the CNN with MRP alleviates the coupling between the geoacoustic parameters and is more robust to different source depths in the shallow water environment. Based on the inversion results, better localization performance is achieved when the range-dependent environment is assumed to be a range-independent model. Real data from the East China Sea experiment are used to validate the MRP method. The results show that, compared with the MFI and the CNN with single-range vertical array data processing, the use of geoacoustic parameters from MRP achieves better localization performance.

A noval method of constructing shallow water sound speed profile based on dynamic characteristic of internal tides

In order to provide constraint to the number of inversion parameters, sound speed profile is often modeled by empirical orthogonal functions (EOFs). However, the EOF method, which is dependent on the sample data, is often difficult to apply due to insufficient real-time <i>in-situ</i> measurements. In this paper, we present a novel basis for reconstructing the sound speed profile, which can be obtained by using historical data without real-time sample. By deducing the dynamic equations and the state function of water particle, the hydrodynamic mode bases (HMBs) can be calculated from historical data without real-time in-situ measurement, and a method of constructing the sound speed profile is established by using the dynamic characteristics of seawater. The use of the World Ocean Atlas 2013 (WOA13) can obtain the seasonal profiles of temperature and salinity, and then the HMB which represents the dynamic characteristic of internal tides is obtained and analyzed. Unlike EOF, the HMB and its projection coefficients are directly related to the sea dynamic features and have a more explicit physical meaning. According to the orthogonality analysis of hydrodynamic mode, the first-order coefficient can be used to describe the depth change of sound speed iso-lines and the second-order coefficient can be used to describe the change of sound speed gradient. Based on the conductance-temperature-depth profiles and broadband data from underwater explosion collected in the East China Sea experiment of the Asian Seas International Acoustic Experiment, the HMB is tested and compared with the EOF in the sound speed profile reconstruction and matched field tomography. The results show that the sound speed profile in shallow water area can be expressed by the HMB with proper precision. By means of the conventional matched field tomography, the valid sound speed profile can also be obtained in the form of HMB coefficients. The result of transmission loss prediction and tomography from HMB are as good as those from EOF, while the HMB has less dependent on real-time <i>in-situ</i> measurement. The HMB is easy to obtain and closely related to the physical characteristics of seawater, it can be used as an efficient alternative to EOF for monitoring the marine dynamic phenomena in sea areas with insufficient real-time in-situ measurement.

Statistical estimation of source location in presence of geoacoustic inversion uncertainty

A statistical estimation of source location incorporating uncertainty in ocean environmental model parameters is derived using a Bayesian approach. From a previous geoacoustic inversion, a posterior probability distribution of the environmental parameters that reflects uncertainty in the ocean environment is obtained. This geoacoustic uncertainty then is mapped into uncertainty in the acoustic pressure field and is propagated through the Bartlett matched-field processor for source localization. Using data from the ASIAEX 2001 East China Sea experiment, the estimated source location and variability over time are compared with the known source positions.

Robust matched-field processor in the presence of geoacoustic inversion uncertainty

This presentation examines the performance of a matched-field processor incorporating geoacoustic inversion uncertainty. Uncertainty of geoacoustic parameters is described via a joint posterior probability distribution (PPD) of the estimated environmental parameters, which is found by formulating and solving the geoacoustic inversion problem in a Bayesian framework. The geoacoustic inversion uncertainty is mapped into uncertainty in the acoustic pressure field. The resulting acoustic field uncertainty is incorporated in the matched-field processor using the minimum variance beamformer with environmental perturbation constraints (MV-EPC). The constraints are estimated using the ensemble of acoustic pressure fields derived from the PPD of the estimated environmental parameters. Using a data set from the ASIAEX 2001 East China Sea experiment, tracking performance of the MV-EPC beamformer is compared with the Bartlett beamformer using the best-fit model.

Validation of statistical estimation of transmission loss in the presence of geoacoustic inversion uncertainty

Often the ocean acoustic environment is not well known and sonar performance prediction will be affected by this uncertainty. Here, a method for estimating transmission loss (TL) is proposed which incorporates these environmental uncertainties. Specifically, we derive an approach for the statistical estimation of TL based on the posterior probability density of environmental parameters obtained from the geoacoustic inversion process. First, a Markov chain Monte Carlo procedure is employed in the inversion process to sample the posterior probability density of the geoacoustic parameters. Then, these sampled parameters are mapped to the transmission loss domain where a full multidimensional probability distribution of TL as a function of range and depth is obtained. In addition, TL is also characterized by its summary statistics including the median, percentiles, and correlation coefficients. The approach is illustrated using a data set obtained from the ASIAEX 2001 East China Sea experiment. Based on the geoacoustic inversion results, the predicted TL and its variability are estimated and then compared with the measured TL. In general, there is a good agreement with the percentage of observed number of data points inside the credibility interval.

Ambient noise directivity in the East China Sea: Model/data comparison

In general, the vertical directivity pattern of high-frequency ambient noise is anisotropic. For measurements made beneath a shallow-water mixed layer, the observed noise directivity often displays a trough in the horizontal called the ambient noise notch. This notch develops because noise sources near the sea surface generate sound that propagates at relatively steep grazing angles. The directivity pattern is also generally asymmetric about the horizontal notch. This asymmetry develops because the seabed is an imperfect reflector of sound relative to the sea surface. In the present work, ambient noise data taken in the 1- to 5-kHz band during the 2001 East China Sea Experiment are analyzed and compared to models. The noise notch model integrates three components: an internal wave model, a noise model, and an acoustic propagation model that uses transport theory. It is shown how the depth and width of the observed noise notch are consistent with model predictions when internal waves of moderate strength are included in the calculation. The observed asymmetry in the directivity pattern is consistent with predictions when the correct bottom reflectivity is used. Future ambient noise measurements and tests for the models are described. [Work supported by ONR.]

Data error covariance matrix for vertical array data in an ocean waveguide

Information about the data errors is essential for solving any inverse problem. The likelihood function plays a critical role in describing the data uncertainties in geoacoustic inversion. The choice of likelihood function depends on the statistics of the errors (the difference between observed and estimated fields). In all work to date, the likelihood function has been derived based on an assumption of Gaussian errors. Typically, the errors are assumed to be independent, identically distributed with equal variance (referred to as the error variance), and the error variance is estimated as part of the optimization. Recently, there has been interest in estimating a more full data error covariance matrix. To estimate a truly full data error covariance matrix, we adopt a maximum-likelihood approach based on ensemble averages using the observed errors over many inversions. The approach is illustrated using data obtained during the ASIAEX 2001 East China Sea experiment. The parameter uncertainties resulting from incorporating the full data error covariance matrix are compared with those obtained from the simplified data error covariance matrix characterized by the error variance alone.

Validation of statistical estimation of transmission loss in the presence of geoacoustic inversion uncertainty

Often the ocean acoustic environment is not well known and sonar performance prediction will be affected by this uncertainty. Here, a method for characterizing transmission loss (TL) is proposed which incorporates these environmental uncertainties. The statistical estimation of TL based on the posterior probability density of environmental parameters obtained from the geoacoustic inversion process is the main theme of this study. First, a Markov chain Monte Carlo procedure is employed to sample the posterior probability density of the geoacoustic parameters. Then, these parameter uncertainties are mapped to the transmission loss domain, where a full multidimensional probability distribution of TL as a function of range and depth is obtained. In addition, TL is also characterized by its summary statistics including the median, percentiles, and correlation coefficients. The approach is illustrated using a data set obtained from the ASIAEX 2001 East China Sea experiment. Based on the geoacoustic inversion results, the predicted TL and its variability are estimated and then compared with the measured TL. In general, the results show that approximately 80% of the observed TL data fall within 90% of the TL probability distribution.

Effect of shallow water internal waves on the ambient noise notch

Consider a vertical receiving array in shallow water that is recording ambient noise. If the recorded noise is beamformed, the resulting beam pattern often exhibits a notch at broadside to the array. Modeling of range-independent environments suggests that this ambient noise notch should become more pronounced as the sound speed gradient is increased. In the present work, the effect of range dependence on the ambient noise notch in the 1–5 kHz band is studied. Range dependence is introduced into the model in the form of random shallow water internal waves. Coupling between the propagating acoustic modes is calculated using the Dozier-Tappert formulation [J. Acoust. Soc. Am. 63, 353–365 (1978)] extended to consider bottom loss. The model is internally consistent as the buoyancy profile determining the internal waves is dictated by the sound speed gradient. Experimental results from the 2001 East China Sea Experiment are also reported. [Work supported by ONR.]

Matched-Field Geoacoustic Inversion of Low-Frequency Source Tow Data From the ASIAEX East China Sea Experiment

Geoacoustic inversion results based on data obtained during the Asian Seas International Acoustics Experiment (ASIAEX) 2001 East China Sea experiment are reported. The inversion process uses a genetic-algorithm-based matched-field-processing approach to optimize the search procedure for the unknown parameters. Inversion results include both geometric and geoacoustic variables. To gauge the quality of the inversion, two different analyses are employed. First, the inversion results based upon discrete source-receiver ranges are confirmed by continuous source localization over an interval of time. Second, separate inversions at many different ranges are carried out and the uncertainties of the parameter estimation are analyzed. The analysis shows that both methods yield consistent results, ensuring the reliability of inversion in this study.

Geoacoustic inversion results of low- to high-frequency source tow data from the ASIAEX East China experiment

During the 2001 ASIAEX East China Sea experiment, source tow data was collected by a 16-element, 75-m aperture, autonomously recording vertical line array in approximately 105-m-deep water. Transmissions from two similar 6-km-long tracks were transmitted. In the first, CW tonals at 95, 195, 295, 395, 805, 850, and 905 Hz were transmitted from a J-15 transducer at a nominal depth of 46 m. In the second, CW tonals at 1.6, 2.4, 3.5, and 4.4 kHz were transmitted from an ITC-2015 transducer at a nominal depth of 49 m. An environmental model in the study area initially was derived by matched-field geoacoustic inversion using the low-frequency track. In this study, the high-frequency data from the second track are included in the parameter assessment. A multiple-stage geoacoustic inversion scheme is applied. Inversion results for seafloor geoacoustic parameters from these transmissions will be presented. [Work supported by ONR.]

The effects of the ocean environmental variability on sound propagation and reverberation from ASIAEX (ECS)

During the duration of the Asian Seas International Acoustics East China Sea Experiment, the ocean environment (e.g., wind speed, sea state, internal wave) varied significantly. Propagation and reverberation data from explosive charges recorded by a vertical line array and horizontal line array at different times during the experiment show that the acoustic data have a strong fluctuation. In this paper, the recorded data are analyzed to estimate the relationship between the acoustic signal fluctuation and the environmental variability. A PE code is used to simulate the effect of the recorded ocean environmental variability on the sound signal fluctuation. The experimental data and numerical results indicate the strong dependence between the acoustic fluctuation and ocean environmental variability. A coupled normal-mode-based model is also used to explain some of the acoustic fluctuation. [Work supported by the National Natural Science Foundation of China and ONR.]

Source tow geoacoustic inversions from the 2001 ASIAEX East China Sea experiment

During the 2001 ASIAEX East China Sea experiment, source tow data were collected by a 16-element, 75-m aperture, autonomously recording vertical line array in 105-m-deep water. Transmissions from two similar 6-km-long tracks have been analyzed. In the first, CW tonals at 95, 195, 295, 395, 805, 850, and 905 Hz were transmitted from a J-15 transducer at a nominal depth of 46 m. In the second, CW tonals at 1.6, 2.4, 3.5, and 4.4 kHz were transmitted from an ITC-2015 transducer at a nominal depth of 49 m. Inversion results for seafloor geoacoustic parameters from these transmissions will be presented. [Work supported by ONR.]

Objective Functions for Ocean Acoustic Inversion Derived by Likelihood Methods

Selection of a suitable objective function is an integral part of the inverse problem, and poor selection can have a strong influence on the inverse result. Objective functions are here derived for many practical occasions such as for single frequency and broadband, with and without knowledge of source strength, and with and without the received signal phase. These objective functions are all derived from a unified approach based on maximum likelihood and additive Gaussian noise models. The assumptions for the objective function are thus clear and the resulting estimator has good properties. From a Bayesian point of view, the solution to the inverse problem is the a posteriori probability distribution of the unknown parameters, which can be found from the likelihood function. Thus using objective functions based on likelihood functions facilitates computing the a posteriori distributions.