Transmission Loss Field Research Articles

Relative entropy of machine-learning for long-range sound propagation

Machine-learning algorithms can provide fast surrogate models when trained on sampled predictions from physics-based numerical models. However, it is unknown whether surrogate data generated from propagation through single realizations of atmospheric turbulence, or from an ensemble of multiple realizations of turbulence, is most suitable for machine-learning models. This study examines the use of a Crank-Nicholson parabolic equation (CNPE) for generating surrogate data, and Latin hypercube sampling for the CNPE input. Two separate datasets are generated for 5000 samples of governing parameters. In the first dataset, each transmission loss field is computed for a single realization of turbulence, whereas the second uses transmission loss from an ensemble of 64 realizations. Errors for various machine learning approaches are evaluated against experimental observations of long-range (out to 8 km) sound propagation. Between experimental and surrogate differences in transmission loss at four ranges, the average deviation from zero relative entropy was quantified. As a measure of dissimilarity between two distributions it is 0.17 for single realizations of turbulence, and 0.29 for 64 realizations of turbulence. Surrogate data generated from single realizations of turbulence agree better with experimental data.

Long-range reverberation in an Arctic environment: Effects of ice thickness and elasticity.

A full-field perturbation approach is modified for an ice-covered ocean and applied to estimating narrowband long-range reverberation caused by roughness of the ice-water interface. First-order approximation of the approach is used which requires the roughness amplitudes be small compared to the acoustic wavelength. To obtain the zeroth-order Green's function and transmission loss field used in the reverberation model, elastic parabolic equation solutions are generated in range-independent environments. Ice is represented by an isospeed layer on top of a linear transition layer. Effects of ice properties are discussed and demonstrated by comparing reverberation calculated for different ice layer thicknesses and wave speeds for typical ice values.

Estimating the influence of ice thickness and elasticity on long-range narrow-band reverberation in a representative Arctic environment

A full-field perturbation approach [A. Ivakin (2016), J. Acoust. Soc. Am., 140(1), 657-665] is modified for an ice-covered ocean and applied to estimating narrow-band long-range reverberation caused by roughness of the ice-water interface. First-order approximation of the approach is used which requires the roughness amplitudes be small compared to the acoustic wavelength. An upward refracting sound speed water column profile typical for Arctic conditions is assumed. To obtain the zeroth-order Green’s function and transmission loss field used in the reverberation model, elastic parabolic equation solutions [J. Collis, S. Frank, et al. (2016), J. Acoust. Soc. Am., 139(5), 2672-2682] are generated in range-independent environments. Ice is represented by two layers. The first approximates acoustic properties of a relatively thin, water-saturated transitional ice layer and is described as a stratified fluid with sound speed increasing with the distance from the ice-water interface. Second layer is introduced as a homogeneous isotropic elastic medium with fixed complex shear and longitudinal wave speeds. Effects of ice properties are discussed and demonstrated by comparing reverberation calculated for different ice layer thicknesses and shear speeds varying from zero (for a fluid model) to typical ice values. [Work supported by ONR.]

Exploiting correlation in passive sonar Bayesian localization techniques.

Low frequency acoustic signals in shallow water are strongly affected by interference between multiple paths resulting from boundary interactions. As the acoustic source moves through this interference pattern, the spatial variation in transmission loss can result in strong temporal modulation of the received signal, which can be used to localize the source. Acoustic propagation models can produce accurate transmission loss field predictions, but are sensitive to ocean environmental parameters such as bottom composition, bathymetry, and sound speed profile. If the uncertainty in the undersea environment can be described by probability density functions of these parameters, Monte Carlo forward models can be used to produce an ensemble of possible transmission loss realizations. A probabilistic model representing this ensemble must include a density function of transmission loss at each location, as well as correlation of transmission loss between locations. In addition, the choice of probabilistic model has a large impact on the form of the resulting Bayesian localization algorithm. Previous results have shown that including spatial correlation of transmission loss can result in improved localization. This talk will introduce a non‐Gaussian probabilistic model for representing uncertainty in transmission loss predictions that includes correlation, and discuss the resulting recursive Bayesian localization algorithm.

Optimal temperature sampling with SPOTS to improve acoustic predictions

The Modular Ocean Data Assimilation System (MODAS) uses optimal interpolation to assimilate data (e.g., XBTs), and to create temperature nowcasts and associated uncertainties. When XBTs are dropped in a uniform grid (during surveys) or in random patterns and spaced according to resources available their assimilation can lead to nowcast errors in complex, littoral regions, especially when only a few measurements are available. To mitigate, Sensor Placement for Optimal Temperature Sampling (SPOTS) [Rike and DelBalzo, Proc. IEEE Oceans (2003)] was developed to rapidly optimize placement of a few XBTs and to maximize MODAS accuracy. This work involves high-density, in situ data assimilation into MODAS to create a ground-truth temperature field from which a ground-truth transmission loss field was computed. Optimal XBT location sets were chosen by SPOTS, based on original MODAS uncertainties, and additional sets were chosen, based on subjective choices by an oceanographer. For each XBT set, a MODAS temperature nowcast and associated transmission losses were computed. This work discusses the relationship between temperature uncertainty, temperature error, and acoustic error for the objective SPOTS approach and the subjective oceanographer approach. The SPOTS approach allowed significantly more accurate acoustic calculations, especially when few XBTS were used. [Work sponsored by NAVAIR.]

The importance of including the direct field contribution in transmission loss field testing

The contribution of the direct field has been traditionally ignored in transmission loss testing. For field testing where there is a moderate amount of absorption in the receiving room, this practice results in an underestimate of the actual transmission loss of the partition. Part of the ancient 5 dB difference between lab tested and field tested results may be attributable to this phenomena.