Sea Surface Sound '94

The following sections are included:

• Acknowledgements

• Contents

• Preface

In the last decade our understanding of ambient surface sound in the ocean has increased significantly. While important fundamental issues still remain to be resolved, our knowledge is such that we can now contemplate novel applications of ambient sound for the measurement and monitoring of the air-sea interface. Except in the most benign conditions, breaking waves are the dominant source of natural sea surface sound, so many of the applications of ambient sound are related to physical processes associated with breaking. These include dissipation of the surface waves, transfers of heat, mass (gas) and momentum. The sound generated by breaking waves may be used to ‘illuminate’ the water column, and its interaction with the sea bed may be used to infer seabed parameters. Another intermittent but significant source of sea surface sound is rain. In this paper recent progress in some of these areas will be reviewed and future applications will be discussed.

Recently we have derived a method which we call Matched-Phase Filtering for improving signal-to-noise ratios in noise dominated data, when the shape of the noise spectrum is approximately known. [J. Acoust. Soc. Am., 91 2444, 1992] [J. Acoust. Soc. Am., 92 2418,1992]. The recent work of Kennedy [J. Acoust. Soc. Am., 91 1974-82,1992] is suggestive of a universal form for the spectrum of wind wave generated noise. We extract from Kennedy's data a simple few-parameter model for the shape of the wind generated noise spectrum, and illustrate its utility with the matched-phase filter. The method assumes no a priori information about whatever signal is buried in the noise, except that its spectrum is distinct from that of the noise. We demonstrate algorithm performance in cases where one's knowledge of the noise spectrum is not precise. We also demonstrate that when the shape of the noise spectrum is known exactly, the matched-phase filter can reveal signals at a signal-to-noise ratio of less than -100 dB.

A remote sensing tool has been developed to determine the compressional and shear wave speeds of the seabed using the ambient noise field in shallow water regions. The technique is based on a simple two hydrophone measurement of the ambient noise field coherence in the vertical. A theoretical study has demonstrated that the structure of the vertical coherence is sensitive to the geoacoustic parameters of the seabed. This sensitivity allows the properties of the seabed to be determined by matching the two hydrophone measurements with a theoretical calculation of the coherence. Results of inversions for the sound speeds near the water/sediment interface are presented for several sites around the UK coast. The sound speeds determined from these inversions show good agreement with independent seismic studies of the same sites. The technique is being extended to include the effects of horizontal stratification of the seabed. Preliminary results of a multiple layer analysis are also presented.

The idea of using naturally occurring ambient noise in the ocean to detect a submerged object is investigated. Full-field scattering and noise theories are combined to formulate expressions for the total noise field covariance of a stratified waveguide with a submerged object present. Noise generated by atmospheric forcing at the ocean surface is modeled by a continuous and infinite sheet of stochastic sources below the free surface. In the absence of the object, this stochastic formulation leads to noise-field directional characteristics which agree well with experimental measurement. The noise field then is scattered from the object using a modification of a previously developed theory for scattering from an object in a waveguide. Expressions for the noise field covariance are then evaluated numerically by wavenumber integration for a spherical object in a shallow water waveguide. In order to determine the plausibility of detection, the Cramer-Rao upper bound on performance is computed for some illustrative examples using conventional sensing arrays. The results indicate that high frequency (~10 kHz) localization of small objects (1 m scale) is possible, but at very short ranges due to rapid fall-off in scattered to direct noise ratio. The idea may be of most practical use for remote detection of large objects (10 m scale) at low frequency (~100 Hz). Beamforming simulations indicate that detection is significantly enhanced in azimuthally inhomogeneous noise, for example near a coastline with surf. Imaging with ambient noise is most effective within the deep shadow range of the object. Here the total forward field will cause the least diffractive interference in imaging reflections from the object. For effective imaging outside the deep shadow range, directional variations in noise intensity incident upon the object, or variations in reflectivity of the object, must be increasingly large for increasing measurement range.

Acoustic Daylight is a new technique for creating pictorial images of undersea objects from the acoustic illumination provided by the ambient noise field. As in conventional photography, the directionality of the illumination affects the contrast in the image through the shadows that are cast. Three aspects of the directionality, or anisotropy, of ambient noise in the ocean are discussed in this paper, in the context of Acoustic Daylight imaging: firstly, some observations of the horizontal anisotropy of the ambient noise around Scripps Pier are reported, which indicate that the pier itself is a significant source of noise; secondly, a theoretical model of the acoustic contrast under differing degrees of noise anisotropy is described; and finally, a numerical simulation algorithm that generates Acoustic Daylight images is used to illustrate pictorially the effects of shadowing when the illumination is a representative shallow water noise field.

For a half century underwater acousticians, supported principally by Defense Department funds, and driven by military needs, have made expedient measurements of noise at sea over a range of locations and ocean conditions. Unfortunately, until recently those measurements were guided by only a vague understanding of the natural sources of sea surface sounds. In that era, rain noise was incorrectly ascribed to simply the raindrop impact, and Knudsen sea state noise was attributed to the wind. To the contrary, recent simple laboratory experiments have conclusively shown that the raindrop impact sound is generally insignificant, and what has been called “wind noise” can be generated by swell, with no wind at all! Transiently oscillating microbubbles are the dominant source of both of these sounds for frequencies over 500 Hz. Most recently, again through laboratory experiments, it has become clear that a significant part of the underwater sound during heavy rainfall is caused by scattered aerosols that are ejected by the splash of terminal speed large raindrops. In addition, the sound of these complex sea-surface interactions is now known to include a component due to the rising bubbles which become transient glide frequency, micro Helmholtz resonators as they burst at the surface.

The predominance of wind dependent ambient noise over a broad frequency band and a wide range of conditions allows the noise in the ocean to be predicted from wind speed measurements (and vice versa), and possibly to be predicted from weather forecasts. The accuracy of such predictions depends on how well the noise is correlated with available wind speed estimates. This paper presents some measurements of noise and wind speed in Australian waters where traffic noise is low, and discusses the wind-wind and wind-noise correlation over various distance scales up to 55 km. Measurements were made using systems comprising a hydrophone on the sea floor and an anemometer mounted on a buoy moored nearby. Comparisons are also made with wind speeds measured at coastal weather stations.

Results of underwater sounds analysis related to the presence of a powerful thundercloud during its approaching to and moving away from the receivers are presented. It is shown that the thundercloud presence manifests itself at large ranges in the appearance of infrasound bursts that are ~ 3 sec in duration and with two peaks in power spectrum within 10–30 Hz. At closer ranges the impulsive sounds in the band 100–300 Hz begin to appear. At close vicinity (5–10 km) to the research vessel, a great variety of different impulsive sounds are observed, which can be associated with the presence of lightning strikes. Different hypothesis on origin of impulsive signals appearing in presence of a thundercloud are discussed.

Sound propagation in shallow water is mainly influenced by the small-scale roughness of the sea surface especially in cases of high wind speed. Breaking waves, rain impact, and spray which are also closely related to ambient noise influence the surface roughness. Thus, in-situ measurements of rain and spray impact are necessary to study the effects of precipitation on the sea surface.

Accurate measurements under open sea conditions of rain momentum fluxes and drop size distributions are a complex problem, especially on buoys. A novel technique is introduced which is based on the measurement of the momentum transfer from impacting drops on a hydrophone surface.

Laboratory studies using defined drops as well as natural rain measurements are described. Based on simultaneous rain measurements using a Joss - Waldvogel Disdrometer and a hydrophone, an analytical function has been derived which relates drop size and hydrophone voltage output. Hydrophone measurements of different rainfall intensities are presented. Drop size distributions are shown for drops with diameters from about 1 mm to 3.3 mm. Additionally a comparison between ambient noise spectra and rain condition is given.

The characteristics of microseisms measured by seismometers near the shore of Lake Ontario and Great Slave Lake are analyzed. For Lake Ontario the rms levels in the 1 to 3 Hz band are coherent between stations widely separated around its western basin indicating a common generative mechanism. A distinct onshore intermittent flux of Rayleigh-like wave energy was detected at the onshore sites for both lakes. Microseismic energy in this band is correlated with the wind speed. The correlation improves as the winds are averaged into the past until an optimum is reached corresponding to the time constant of water wave generation by changing wind speed. For a given fixed wind speed, the microseismic energy correlates with the average fetch of the wind over the lake. The sensitivity to fetch effects is similar for both onshore and offshore stations indicating that shoaling is probably not a source. Niagara Falls which also can have a wind-dependent flow from Lake Erie causes measurable effect to at least 25 km but does not noticeably affect stations at a distance of 150 km.

It is suggested that the microseismic flux provides a natural, relatively inexpensive way to monitor the water wave field on such large lakes. Further, such seismic observations may provide useful insights into wave generation mechanisms, in particular a lake's response to variable wind speed, the onset of rough flow and the spatial variability of the wave field. Additionally a large lake may well prove to have a stronger source strength of microseisms than an ocean.

Spilling breakers are the most common type of breaking wave and hence are a dominant source of underwater sound in the ocean. The dynamics of short, spilling breakers is discussed, and it is shown that the process begins with a basic instability of the wave crest, which has has only recently been recognised. This leads to the generation of parasitic capillary waves, which in turn generate significant amounts of vorticity. The theoretical predictions are closely borne out by recent observations of gently spilling breakers in a laboratory wave tank. At a later stage of development, a mixing-layer is formed beneath the water, with or without entrainment of air, according to the local Weber number.

A simple model is described for the initial distribution of bubble sizes due to air entrainment. The model is based on the idea of a random splitting of a given volume of air. This leads to a unimodal but asymmetric form for dP/d(lna), where P(a) is the size distribution. Both the width and the skewness of the theoretical curve are in rough agreement with observations. However, there is evidence in the data of some bimodality, which is so far unexplained.

Most laboratory and field work on the sound generated by breaking waves has been done at frequencies less than 20 kHz. Two examples of transient noise emissions from ocean breaking waves, taken from recent measurements at 20–50 kHz and 240 kHz, are presented. In magnitude (generally 6–15 dB above ambient levels) and duration (1–2 s), the emissions resembled those produced by spilling containers of seawater during calm periods in the course of the same experiment.

Breaking wave statistics were observed during the Surface Wave Processes Program, using a broadband hydrophone array. It is found that the source level of breaking waves is well correlated with their travel speed, implying that dynamical intensity of breaking is related to breaking wave speed. Therefore we investigate the possibility of calculating breaking wave speed based on directional wave spectra through a Monte-Carlo simulation. The simulation involves synthesizing the surface wave field and tracking breaking regions in the field selected with a breaking criterion. The linear wave field was synthesized by Fourier transforming time-dependent directional wave spectra. Weak nonlinearity of surface waves was also incorporated based on an approach developed by Creamer et al. (J. Fluid Mech., 1989). Directional wave spectra were simultaneously measured using Doppler sonar, and used as input for the simulation. In this paper, we present a comparison of the simulation results with the acoustic observations.

Measurements of the ambient noise spectrum level N with simultaneous, coincident wind and wave measurements were made from RP Flip in Fall 1991. The measurements were designed to investigate the correlation between the ambient noise and surface wave parameters. The results suggest that wave parameters related to the incidence of wave breaking correlate well with the ambient noise level. The correlation between N and the RMS wave amplitude a was found to be poor but that between N and the RMS amplitude of the local wind sea a_w was comparable to that between wind speed U and N. Similar good correlations were found between the RMS wave slope s and N, and the higher frequency surface wave spectral levels and N.

Correlations between the surface wave dissipation estimates D based on the Hasselmann (1974) and Phillips (1985) models and the ambient noise were comparable to those between the wind speed and N. The mean square acoustic pressure was found to be proportional to Dⁿ with n in the range 0.6–0.8. The implications of these results for monitoring surface waves and air-sea fluxes are discussed.

The possibility is examined that a breaking wave field is self-organized and critically maintained. It is well-known that instabilities associated with a critical vertical acceleration, which themselves are related to the exceedance of a critical local wave slope, for given assumptions of the probability distributions governing the occurrence statistics of these instabilities, control the extent of wave breaking. The concept that there exists synergism by which the generation of instabilities is controlled in some sense by the wave breaking is examined here. It is shown that the necessary ingredients for self-organized criticality are present - an instability mechanism, spatial and energetic scaling, and control directly linked to a phase transition. The existence of a percolation threshold, which essentially indicates the presence of an extended connectivity for a critical local bonding threshold, provides additional evidence of self-organization. A constitutive relationship for a breaking field is developed. It is argued that these developments can be directly merged with recent results on the creation, flux and loss of information.

Measurements of the ambient sound generated by breaking waves over the range 100–20000 Hz are described. Waves were generated by a computer-controlled plunging type wavemaker and propagated along a 12.7m long channel where they were made to break at the mid-surface of a 3m × 3m × 2.8m anechoic water tank. The wavemaker parameters were controlled to produce spilling- and plunging-type breakers with various breaker intensities. The individual bubbles and bubble clouds entrained by the breaking wave provide a mechanism for sound production. The bubble size distributions inside the cloud were measured enabling us to obtain measurements of the average void fraction in the cloud. These observations reveal that the sources of sound in laboratory spilling breakers are due primarily to single bubble oscillations that can have frequencies as low as 400 Hz. In the case of plunging breakers, it appears that both individual bubbles and bubble clouds can contribute to the acoustic emissions. The average sound spectra reveal that the peak frequencies of the spectra shift from a few kHz (weak, spilling breaker) to few hundred Hz (plunging breaker), and the high frequency portions have slopes approximately 5–6 dB/octave which are the slopes observed from the noise spectra of the ocean. The active noise measurements involve the scattering of sound from bubble clouds as determined by their shapes, bubble concentration, and size distributions. This procedure was conducted in the presence of waves with and without breakers to isolate the acoustic scattering strength of bubble clouds from the effects of the surface roughness. The results of the scattering strength of a typical bubble cloud with average void fraction of 0.10% as a function of frequency are discussed.

The following sections are included:

• Introduction

• Precipitation Noise
  • Unresolved Issue: What is the role of bubbles in the noise produced by precipitation

• Capillary waves
  • Unresolved issue: What is the role of capillary-wave encapsulated gas bubbles in ocean ambient noise

• Breaking Waves
  • Unresolved Issue: Do bubbles produced within breaking waves account for the majority of the ocean ambient noise

• Turbulence Amplification of Low Frequency Noise
  • Unresolved Issue: What is the origin of the low frequency (< 100 Hz) ambient noise in the ocean?

• Surface and Volume Mode Coupling
  • Unresolved Issue: What role does mode coupling between surface and volume oscillations play in ocean ambient noise?

• “Birthing Wails” versus “Adult Screams”
  • Unresolved Issue: Does adult bubble excitation play a role in ocean ambient noise

• The effect of surface-active contaminants on bubble stabilization and dynamics
  • Unresolved Issue: What role do surface contaminants play on bubble stability and dynamics in the ocean

• Summary and Conclusions

• Acknowledgment

• List of References

Stable bubble clouds can be produced in polymer solutions. Two primary acoustic properties of bubble clouds, the acoustic velocity and attenuation, were found to be unaffected by the viscosity buildup by the polymer addition. It is expected that the scattering property of the artificial bubble clouds to be similar to the natural ones. The acoustic properties of the artificial cloud can be easily tuned by varying the void fraction, bubble size distribution, and cloud size and geometry. Applications for acoustic research and remote sensing of oceanographic parameters are suggested.

A novel optical technique is introduced to measure the size distribution of air bubbles entrained into water by breaking waves. Bubbles are visualized by a light blocking technique, and images are taken with a video camera. Using a small depth of field and a depth from focus technique, a virtual measuring volume that is roughly proportional to the size of the bubble, can be defined. With an adequate low-level image processing technique, the correct size of a bubble and its distance from the focal plane can be obtained from parameters such as the apparent (blurred) area of the object and the mean brightness in this area. The technique is based on a precise knowledge of the 3-D point spread function and requires objects of uniform brightness and simple shape.

We report on the results from two field experiments conducted on the measurement of the sound speed and volume fraction of air (hereafter void-fraction) near the ocean surface. Simultaneaous time-series of the sound speed and void-fraction at several depths show the presence of large fluctuations over time periods on the order of minutes or less which are caused by the formation of bubble plumes or the passage of bubble clouds. Simultaneous measurements with an upward-looking sonar show bubble clouds extending down to a depth of 3m and greater while the void-fraction measurements above the instrument noise of 2×10⁻⁷ are confined to the first meter below the surface for wind-speeds up to 8m/s. The (20min) time-averaged void-fraction and sound -speed anomaly (sound speed reduction) profiles are found to be shallow and the average void-fraction is found to increase with wind speed. Both variables obeyed exponential probability distributions.

A “tipping bucket” experiment was conducted to investigate the influence of salinity on the characteristics of a bubble plume in the UCONN Whitecap Simulation Tank IV. A submersible video microscope system was deployed 100mm beneath the water surface and at the center of the bubble plume, and the size distribution of bubbles with radii from 180μm to 5mm was measured. The mechanical action of the tipping bucket remained constant throughout the experiment while the water in the tank was diluted several times to salinities of 20‰ to 0‰ (fresh water). The experimental results indicate that the bubble concentration is a few orders of magnitude higher in brackish water at a salinity of 20‰ than in fresh water, and that the bubble plume in brackish water has a longer lifetime than the freshwater bubble plume. Void fraction at the top of the bubble plume was estimated by integrating over the bubble spectrum. The maximum void fractions are about 20% for both the fresh water and brackish water cases, but the void fraction in brackish water has a much slower decay rate than that in fresh water, due to the presence of many tiny bubbles. From this void fraction estimation, the calculated temporal variations of sound velocity show significant differences between fresh water and brackish water.

Wave breaking and subsequent formation of whitecaps are known to be the major contributor to the wind dependent ambient noise levels in the ocean in the low frequency range (100Hz-1kHz). A theoretical model that can account for the noise emissions from randomly distributed bubble clouds is developed in this study. The model assumes that individual whitecaps produce bubble plumes that grow as a result of air entrainment at the ocean surface. The injection of bubbles at the base of this plume excites the bubble cloud. The underwater ambient noise level is calculated by integrating contributions from bubble clouds of all sizes with minimal experimental input. The results are in good agreement with the field measurements.

The presence of bubbles in the ocean is an important phenomenon, and studies into a range of effects (atmosphere/ocean gas flux, near surface acoustic propagation, etc.) often require knowledge of their size, number and distribution. Such information is also important for studying bubbles in industrial or clinical systems. Because bubbles are excellent scatterers of sound, with well-defined acoustic resonances which (to a first approximation) are inversely proportional to their size, these measurements lend themselves towards the use of acoustics. At large amplitudes an asymmetry is introduced into the pulsation of the bubble wall, and this nonlinearity is used to detect resonant bubbles. The results presented are of a technique which uses two sound fields incident on the bubble - one high fixed imaging frequency and another lower frequency that is adjusted to match the resonant frequency of the bubble. The nonlinearity gives rise to sum-and-difference coupling of the imaging frequency with the bubble resonance, and with harmonics, subharmonics and ultraharmonics of this resonance. From these the bubble radius can be determined. This paper gives details of investigations into the suitability of this method to actively size bubbles of unknown radius and distribution, and discusses its accuracy and limitations. In addition, the feasibility of automated high-resolution bubble sizing is examined using specialised signal processing and heterodyning techniques.

Ambient noise production by bubbles presupposes air entrainment below the ocean surface. The purpose of this paper is to illustrate in a few idealized cases the basic mechanics of this process.

The Naval Research Laboratory (NRL) has developed a measurement system called the Acoustical Resonator, which can determine in-situ size spectra of bubbles with radii ranging from about 30 to 1200 microns. This acoustical bubble system had been calibrated against two other optical bubble sensors in laboratory-controlled situations and subsequently deployed in blue-water field experiments in the past three years.

Using the well-known Wood's formula for bubbly flows, the bubble spectra measured by the above system can be used for deriving the corresponding low-frequency sound speed. It was found that the sound speed deficits from such combined measured-and-computed methods have reached more than 100 m/sec in the upper 1 to 2 m depth with the prevailing wind from 10 to 15 m/sec. In this paper, we shall present some of the results from a deep-water field experiment conducted in the Gulf of Alaska during Critical Sea Test (CST) 7.

In the ocean bubble layers play significant roles in sound propagation and generation. Most of early theoretical investigations dealt with acoustic properties of the bubble layers with sharp flat boundaries at the borders. However, the bubble layers with transition sublayers at the borders are observed to be more likely in the ocean. In this paper we investigated theoretically and experimentally sound speed variation due to the bubble size distributions and sound transmission through a plane bubble layer with transition sublayers. For normal incidence of sound waves the equal-size and Gaussian distributions of bubbles were considered for a bubble layer. The theoretical investigation shows that the transmission coefficients also essentially depend on the thickness of transition sublayers and the bubble layer with thicker transition sublayers has weaker resonance properties. The presented theory with transition sublayers gives better agreement for the experimental data than that with sharp flat boundaries.

A new method of nonlinear acoustic tomography for spatial bubble distribution in the ocean is presented. Quadratic and cubic nonlinear parameters of bubbly medium can be determined from sound speed variation at the junction that the tone burst probe and impulse pump waves intersect. Such sound speed variation of the probe wave at the junction induces the time shift that depends on bubble concentration and on impulse length and amplitude. From the time shift measurements the spatial distribution of bubble clouds in the ocean can be estimated. The presented estimation shows strong feasibility of the nonlinear acoustic tomography for spatial bubble distribution in the ocean.

Since the early contributions of Longuet-Higgins the theory governing the wave-wave interaction mechanism responsible for the main peak in the ULF noise spectrum has been refined progressively. In the development of the subject the source generation process and the resulting noise field have been treated separately. In traditional treatments the ocean is assumed to be infinitely deep, in which case the influence of the bottom can be ignored and the simple deep-water dispersion relation governing the ocean-wave field can be used. The validity of this approach clearly becomes questionable in shallow-water environments for several reasons. First it has been shown recently that for the traditional sum-frequency components of the nonlinear wave-wave interaction process the horizontal wave number has an intrinsic upper limit, which is related to the dispersion of the first-order wave field and thus to water depth. Secondly the primary pressure field (PF) and the inhomogeneous component of the double-frequency (DF) field both become more and more relevant as the water depth decreases. Thirdly by virtue of porosity the water-saturated sediments (sand) of the continental shelf can sustain shear and slow Biot waves of such low velocities that it is possible for an enhanced seismo-acoustic response to result from the inhomogeneous component of the wave-induced pressure field. When the wavelength of these source components is comparable with the thickness of the low rigidity layer, the DF component of the ULF field becomes a source of unique importance to inversion studies of sediment parameters. On general grounds therefore it is important to extend existing analyses to incorporate environments of any water depth. This paper provides such an analysis and presents ‘user friendly’ compilations of wind and depth-dependent ULF spectra for a variety of realistic geoacoustic environments.

The predictions of a theory of bubble fragmentation (Longuet Higgins 1992) are compared with experimental results. The theory, which models the fragmentation event as the simultaneous insertion of a number of randomly positioned dissecting planes through a cubical air body to predict the number and size of daughter bubbles produced can be formulated in one, two, or three dimensions, corresponding to the mutually perpendicular directions of the plane normals. The lower-dimensional events being more amenable to visual observation, the fragmentation of a nominally two-dimensional bubble (a 'disc' of air) was studied and found to be most accurately modelled by the one-dimensional theory. Subsequent high-speed video observation did indeed reveal one-dimensional characteristics in the collapse, though it found that it deviated from theory in that the dissecting planes are neither randomly positioned nor inserted simultaneously. The fragmentation of three-dimensional bubbles was studied by exploiting their acoustic emissions. Use of a Gabor transform in a time-frequency representation (TFR) of hydrophone data gave a time of formation and the natural frequency of each bubble entrained in a small freshwater waterfall, and also revealed the presence of a low-frequency cloud mode which had to be eliminated from subsequent comparison with theory. Once again the best fit was obtained for the model which utilised only one-dimensional fragmentation, a result in accord with the geometry of the waterfall. Data obtained through entrainment of bubbles by a liquid jet, impacting a liquid surface from above at various angles, could not be fit simply by the theory. The dimensionality of the best fit varied with the flow rate in, and angle of, the jet, suggesting more than one type of entrainment could operate. This was further indicated by multi-modal distribution in bubble size obtained at certain intermediate angles. Bi-modal distributions were also obtained using hydrophone data of light rainfall over the ocean on a calm day. Comparison of theory and experiment yields discourse on the differing roles of shape oscillations and surface waves on bubble fragmentation, and the issues involved with incorporating this distinction into the model.

Two methods are described for determining the location of an acoustic source from data contaminated by noise generated at the sea surface. The multi-valued Bartlett processor is an eigen-processor that exploits the fact that energy due to discrete sources tends to partition into different eigenvectors of the covariance matrix and energy due to noise tends to be distributed among the eigenvectors. The noise-canceling processor uses information regarding the spatial coherence of the noise to localize sources buried in noise.

This paper describes ambient noise data recorded at several locations on and near the continental shelf off the west coast of Vancouver Island. The data were obtained in both shallow and deep-water with the Multi Element Vertical Array (MEVA) at various locations during a three year period. These data had not been examined to date; however, given the increased interest in shallow-water acoustic environments, it was considered timely to perform a detailed analysis of the available data. Four sites are examined, two located in deep water (2500-2600 m) just off the continental shelf, and two located in shallow water (400-500 m) on the continental shelf. One deep water site was monitored more or less continually over a four day period, and the data provide a good measure of the temporal variation present in the ambient noise. The analysis includes estimates of the omnidirectional noise level at the four sites as a function of both frequency and water depth (the depth dependency is obtained by examining a time series of the acoustic data recorded for selected individual hydrophones in the MEVA array). Directional estimates of the ambient noise field are also presented. The directional estimates are obtained by beamforming the recorded data over a selected time period and displaying the results in the form of coloured frequency-vs.-elevation angle surfaces.

Acoustic and seismic emissions from developing ice fractures are analyzed. It is shown that signals from these emissions allow the tracking and sensing of ice failure processes. The analysis reveals that the ice release stresses through intermittent crackings, and the final catastrophic failure is preceded by crack advancements of various scales ranging from meters to centimeters. Such scale variation causes frequency changes in radiated acoustic signals allowing inference of failure scales and, hence, the intrinsic rupture speeds from the corresponding acoustic spectra.

This paper presents the analysis and interpretation of measured and modelled Arctic ambient noise coherence in the frequency region of 10 to 30 Hz. The data were recorded on a vertical linear array of 22 hydrophones and a 7-hydrophone horizontal array. The arrays were suspended from the ice canopy in 420 m of water near the edge of the continental shelf in the Lincoln sea north of Ellesmere Island, Canada. The ambient noise coherence for distant ice ridging events is modelled using an adiabatic normal-mode propagation model. The number and direction of active ridges are estimated using the horizontal array. Broadband matched-field processing with uncertain bathymetry is then applied to the vertical array data to determine the source ranges. The model assumes three incoherent sources near the ocean surface. The search space includes as variables the source ranges as well as a model of the bathymetry in the direction of the sources. The bathymetry is modelled using multiple segments constrained within bounds based on published bathymetry charts. The best match between model and data is then sought using simulated annealing.

On one evening during the week of the workshop, a brain-storming session was held with a view to identifying important areas of research into sea surface sound that should be addressed in the future. Potential applications of sea surface sound were included in the discussion. Acting as chairman, Michael Buckingham (MB) introduced the session, which was attended by most of the participants at the workshop. The intention was to encourage the participants to explore, in an informal setting, the future of sea surface sound. A summary of comments and conclusions, compiled from MB's notes of the discussion, is presented below…

The following sections are included:

• Subject Index