Variations of seabed parameters, particularly the acoustic compressional velocity relative to that of the water column, have a very strong effect on the predicted magnitude of the propagated acoustic field. In many cases, the sensitivity of the predicted losses to the assumed seabed parameters is sufficiently severe to render the predictions meaningless. In this article, coupled-mode theory is employed to study how the presence of lateral seabed inhomogeneities affects this sensitivity. The dependence of predicted propagation losses on sediment sound speed is first examined for horizontally stratified sediments; then a rough layered structure of clay-silt interbedded with sand is assumed. With the introduction of rough sub-bottom layering, coupling occurs between modes, so that the conversion of energy into progressively higher-order modes, which attenuate rapidly, becomes an important loss mechanism. The extent to which acoustic energy penetrates the seabed and interacts with the sub-bottom inhomogeneities is governed by the sediment sound speed. Hence, the mode coupling that is induced is simply another loss mechanism dependent on sediment sound speed, leading to an increase in sensitivity to this parameter. Studies to determine the effect of inhomogeneous sediment layering on the transverse horizontal spatial coherence of the propagated field reveal that, even in the presence of mode coupling, the coherence remains high over many acoustic wavelengths, in agreement with measured data trends.

A model for the plane-wave reflection coefficient from a layered elastic solid bounded on either side by a fluid half-space is developed and applied to study environmental factors affecting the reflectivity of smooth arctic ice. Experimental measurements of the internal friction in ice and snow are reviewed and applied to compute realistic attenuation profiles. An examination of the effect of ice layers conforming with measured temperature profiles in floe ice shows that the use of average values for sound speed and attenuation is an acceptable approximation for modeling purposes. A study of the reflectivity due to the absorption of shear and compressional waves demonstrates that shear wave attenuation is the most important loss mechanism from 20 ø to 60 ø incidence. The effect of an additional snow layer is to produce more attenuation without shifting the pattern of reflection nulls. Major results are presented for a frequency of 2 kHz. Data comparisons are performed from 0.5-3 kHz that show a limited qualitative agreement with the model.

This letter defines linear, linearized, and nonlinear measures of environmental sensitivity for ocean acoustic propagation that account for realistic uncertainties in various environmental parameters ͑water-column sound-speed profile and seabed geoacoustic properties͒. Simple interpretations of sensitivity are based on the implicit assumption of a linear relationship between parameter sensitivity and parameter uncertainty. This assumption is examined by comparing the three sensitivity measures over a range of parameter uncertainties about the actual assumed environmental uncertainty. Sensitivity range and depth dependencies are illustrated for realistic geoacoustic uncertainties and oceanographic variability of the sound-speed profile.

The scattering of sound from a sinusoidal pressure release boundary has long been the subject of study among researchers. Recently, R. L. Holford [J. Acoust. Soc. Am. 70, 1116-1128 (1981)] outlined an exact solution to a variant of the Helmholtz integral equation that permits calculation of the unknown velocity potential on the surface by the solution to a matrix equation. Application of this solution is made to the sinusoid to generate plane wave reflection coefficients. The reflection coefficients are then compared with two sets of model tank data as well as with the approximate solutions of Kirchhoff, Eckart, Rayleigh, and Brekhovskikh. The exact solution compares most favorably with the data, particularly on the rougher surfaces. The principal errors produced by the approximate solutions are a shift in the location of the reflection nulls, an inability to correctly predict unity when all side orders are evanescent, and the absence of Rayleigh reflection anomalies. These errors appear regardless of the roughness parameter range or radius of curvature, even when the solution is theoretically within its region of validity. The reflection coefficients are also compared with a previous solution that utilized the same basic method of solution for a different formulation of the scattering problem.

This article examines the effects of spatial field shifts in ocean acoustic environmental sensitivity analysis. Acoustic sensitivity studies are typically based on comparing acoustic fields computed for a reference environmental model and for a perturbed model in which one or more parameters have been changed. The perturbation to the acoustic field due to the perturbed environment generally includes a component representing a spatial shift of the field (i.e., local field structure remains coherent, but shifts in range and/or depth) and a component representing a change to the shifted field. Standard sensitivity measures based on acoustic perturbations at a fixed point can indicate high sensitivity in cases where the field structure changes very little, but is simply shifted by a small spatial offset; this can conflict with an intuitive understanding of sensitivity. This article defines and compares fixed-point and field-shift corrected sensitivity measures. The approaches are illustrated with examples of deterministic sensitivity (i.e., sensitivity to a specific environmental change) and stochastic sensitivity (sensitivity to environmental uncertainty) in range-independent and range-dependent environments.