Info

40 60

Range (nm)

40 60

Range (nm)

Detection coverage (nm)

Detection coverage (nm)

Figure 5.16 Effect of a cold-core eddy on acoustic propagation: (a) TL versus range; and (b) corresponding detection zones assuming an FOM of 90 dB. (Transmission loss curves are from Vastano and Owens, 1973; J. Phys. Oceanogr., 3,470-8; copyright by the American Meteorological Society.)

Sargasso Sea

Figure 5.16 Effect of a cold-core eddy on acoustic propagation: (a) TL versus range; and (b) corresponding detection zones assuming an FOM of 90 dB. (Transmission loss curves are from Vastano and Owens, 1973; J. Phys. Oceanogr., 3,470-8; copyright by the American Meteorological Society.)

submarine located in the colder interior of the eddy since the sound waves are refracted downward (away from the higher temperatures and higher sound speeds) and away from the ship's passive listening device. The submarine thus can successfully avoid detection by positioning itself inside known eddy locations.

5.6.7.2 Internal waves

Zhou et al. (1991) investigated the interaction of underwater sound with internal gravity waves (specifically solitons) in an attempt to explain the

Distance from eddy center (km)

Figure 5.17 Sound-speed contours through a cold-core eddy (Gemmill and Khedouri, 1974).

Distance from eddy center (km)

Figure 5.17 Sound-speed contours through a cold-core eddy (Gemmill and Khedouri, 1974).

anomalous behavior of low-frequency (~300-1,100 Hz) acoustic propagation conditions observed in some shallow-water areas. As a result of this investigation, it was noted that acoustic measurements could be employed in an inverse fashion to remotely sense internal wave activity in the coastal zone. Rodriguez et al. (2000) used a range-dependent normal-mode model (C-SNAP) to simulate the propagation of acoustic signals through soliton-like fields of temperature and sound speed. The simulation reproduced experimental observations of signal enhancement attributed to focusing effects correlated with peaks in current, temperature and surface tides. Chin-Bing et al. (1993a) numerically simulated the effects of a solitary internal wave on the low-frequency acoustic field in a shallow-water waveguide using two-way, range-dependent, finite-element models (FOAM, FFRAME and SAFE). These simulations focused on the refractive and backscatter effects

Figure 5.18 Schematic illustration of a submarine in a cold-core eddy avoiding detection by a surface ship equipped with passive sonar. Sound is refracted downward away from the ship by the effects of the colder water in the interior of the eddy.

of a single soliton wave packet in an effort to understand the cumulative effects of multiple resonances on acoustic propagation.

Oba and Finette (2002) simulated acoustic propagation in a shallow-water, anisotropic ocean environment at frequencies of 200 and 400 Hz for horizontal ranges extending to 10 km. Their simulation employed a 3D PE code based on differential operators representing wide-angle coverage in elevation and narrow-angle coverage in azimuth. The water column was characterized by random volumetric fluctuations in the sound-speed field induced by internal gravity waves superimposed on a thermocline. There was a localized contribution from a solitary wave packet (undular bore or soli-bore). The resulting simulation showed azimuthal filtering of the propagated field, with the strongest variations appearing when the propagation was parallel to the solitary wave depressions of the thermocline. The solitary wave packet was interpreted as a nonstationary oceanographic waveguide within the water column that preferentially funneled acoustic energy between the thermocline depressions.

Tang and Tappert (1997) used a broadband model (UMPE) to explain the lack of multipath replicas of a transmitted pulse in broadband acoustic experiments in the Straits of Florida. The observed single broad cluster was attributed to the effects of internal waves, which produced moving acoustic "footprints" on a rough sea floor. Tielburger et al. (1997) investigated the acoustic field properties in a shallow-water waveguide where the sound speed had a deterministic, time-independent component and two stochastic components induced by internal-wave activity. Simons et al. (2001) used a broadband normal mode model (PROSIM) to simulate variability in a shallow-water channel resulting from fluctuations in oceanographic parameters. The acoustic band of interest was 1-8 kHz.

5.6.8 Coupled ocean-acoustic modeling

Oceanographers have recently developed synoptic forecasting techniques appropriate for predicting the locations of frontal features such as currents and eddies (Mooers et al., 1982; Robinson et al., 1984; Robinson, 1987; Peloquin, 1992). These forecasts can be used by forces afloat to facilitate the efficient allocation of naval resources during ASW operations.

These same dynamical ocean models can be used in conjunction with underwater acoustic propagation models to generate timely forecasts of sonar performance in the vicinity of highly variable frontal features. When used in this fashion, the ocean models generate input variables necessary for initialization of the acoustic models. This synergistic arrangement is referred to as coupled ocean-acoustic modeling. Such coupling has been successfully demonstrated, for example, by Botseas et al. (1989), who interfaced an implicit-finite-difference PE model with ocean forecasts generated by the HOOM.

Coupled ocean-acoustic forecast systems comprise three basic components: an oceanic forecast scheme, a coupling scheme and an ocean acoustic propagation scheme (Robinson etal., 1994). These systems can also be used to generate nowcasts and hindcasts. Nowcasts are estimates of the present state of a system. They are based on a combination of observations and dynamical modeling. Hindcasts are a posteriori forecasts. They are useful in evaluating modeling capabilities based on historical benchmark data (e.g. Martin, 1993).

Requirements for oceanographic data to support coupled ocean-acoustic forecast systems often exceed observational capabilities. Therefore, data assimilation, which introduces data generated by feature models, is used to achieve accurate synoptic realizations. Feature models are statistical representations of common synoptic structures in the ocean such as fronts and eddies (Robinson et al., 1994).

The optimum thermal interpolation system (OTIS) (Clancy et al., 1991a) forms the basis for an upgraded tactical ocean thermal structure (TOTS) system (Hawkins, 1992) designed for use aboard ships equipped with the tactical environmental support system (TESS). OTIS assimilates real-time observations from multiple sources into a complete 3D representation of the oceanic thermal field. A corresponding 3D representation of the salinity field is derived using empirical techniques. To compensate for sparse subsurface measurements, OTIS supplements actual observations with synthetic data derived from an empirical orthogonal function (EOF) representation of water masses, from ocean feature models, and from subjective interpretation of satellite imagery. The resulting product provides an accurate 3D representation of the temperature and salinity fields together with a realistic depiction of fronts and eddies. Such information is useful in organizing naval assets as well as in planning oceanographic field work. OTIS has also been used to generate high-quality data for the evaluation of oceanographic forecasting models (Lai et al., 1994).

The coupling of oceanographic mixed layer models with surface duct propagation models was discussed earlier in this chapter (Section 5.2.3). The subject of marine modeling, which properly embraces the numerical modeling of ocean physics, geology, chemistry and biology, has been summarized by Goldberg et al. (1977) and by Kraus (1977).

5.6.9 Acoustic tomography

Ocean-acoustic tomography (Munk and Wunsch, 1979) is an inverse technique that measures perturbations in travel times between fixed acoustic sources and receivers (see Figure 5.19). This technique is analogous to the medical procedure called tomography (from the Greek word for "slice" or "cut"), and also has elements in common with conventional seismology (Menke, 1989). A recent book by Munk et al. (1995) provided s s

Figure 5.19 Schematic illustration of an ocean tomographic geometry. Perturbations in acoustic travel time from any source s to any receiver r (dashed paths) are used to estimate sound speed perturbations in an arbitrary grid area formed by the solid lines. (Munk and Wunsch, 1979; Deep-Sea Res., 26,123-61; copyright by Pergamon Press Plc.)

Figure 5.19 Schematic illustration of an ocean tomographic geometry. Perturbations in acoustic travel time from any source s to any receiver r (dashed paths) are used to estimate sound speed perturbations in an arbitrary grid area formed by the solid lines. (Munk and Wunsch, 1979; Deep-Sea Res., 26,123-61; copyright by Pergamon Press Plc.)

a comprehensive review of the oceanography and mathematics necessary to understand and develop ocean-acoustic tomographic systems.

This technique has been proposed for large-scale monitoring of ocean basins. Specifically, the perturbations in travel times between sources and receivers can be used to deduce sound-speed (and thus water density) fluctuations in the interior of the oceans. Moreover, if reciprocal transmissions are used between sources and receivers, the differences in travel times can be used to compute the mean ocean currents along the acoustic path (Spiesberger, 1989). Sea surface spectra can also be estimated using acoustic tomography (Miller et al, 1989).

The temporal resolution required to distinguish multipath arrivals has been estimated at 50ms. Existing low-frequency (<200Hz) broadband sources appear to satisfy these requirements. The real limiting factor will probably be the lack of spatial resolution owing to practical limitations on the number of sources and receivers. Consequently, there will usually be insufficient information from which to determine a unique solution to the inverse problem.

A field test of ocean-acoustic tomography was conducted in the North Atlantic Ocean in 1981 (Cornuelle et al., 1985). The test was conducted over a 2-month period in a 300 km area centered at 26° N, 70° W using nine acoustic deep-sea moorings with sea floor transponders. The acoustic sources operated at a center frequency of 224 Hz, with a 5.4 Hz rms bandwidth. The 3D sound-speed field obtained by inversion techniques compared favorably with direct observations of sound-speed profiles. Tomographic maps constructed at 3-day intervals over the test period revealed a pattern of eddy structure in general agreement with that deduced from direct observations. Mapping errors were attributed to noise variance in travel-time measurements.

Acoustic tomography experiments have been conducted to detect climatic trends of temperature in the oceans with basin-scale resolution (Munk and Forbes, 1989; Spindel and Worcester, 1990a,b, 1991; Spiesberger and Metzger, 1991a; Spiesberger et al., 1992; Forbes, 1994). Since acoustic travel times are inversely related to water temperature, tomography provides a method for ascertaining the bulk temperature of the oceans. This bulk temperature is potentially a more reliable metric for global warming than are conventional atmospheric measurements which, for example, can be influenced by the heat island effect that is often associated with urban temperature records. Moreover, basin-scale averages of water temperature suppress the intense mesoscale eddies that would otherwise dominate the climatic variations.

The Heard Island feasibility test (HIFT) established the limits of phase-coded, electrically driven sound sources (instead of explosive sources) in tomographic experiments (Munk, 1994). Figure 5.20 illustrates the ability of an acoustic source situated at Heard Island to sample vast volumes of the oceans over ray paths unimpeded by bathymetric obstructions. These ray

80° 60° 40° 20° 0° 20° 40° 60° 80° 100° 120° 140° 160° 180° 160° 140° 120° 100° W E W

Figure 5.20 Illustration of bathymetrically unimpeded ray paths emanating from Heard Island with the indicated launch angles. All rays are refracted geodesies along the surface containing the sound channel axis (see Figure 2.8). White boxes indicate locations of oceanographic research stations capable of receiving acoustic signals. (Munk and Forbes, 1989; J. Pbys. Oceanogr, 19, 1765-78; copyright by the American Meteorological Society.)

80° 60° 40° 20° 0° 20° 40° 60° 80° 100° 120° 140° 160° 180° 160° 140° 120° 100° W E W

Figure 5.20 Illustration of bathymetrically unimpeded ray paths emanating from Heard Island with the indicated launch angles. All rays are refracted geodesies along the surface containing the sound channel axis (see Figure 2.8). White boxes indicate locations of oceanographic research stations capable of receiving acoustic signals. (Munk and Forbes, 1989; J. Pbys. Oceanogr, 19, 1765-78; copyright by the American Meteorological Society.)

paths represent refracted geodesics along the surface containing the sound channel axis (see Figure 2.8). A geodesic path is the shortest distance between two points on the surface of an elliptical (versus spherical) Earth. Geodesics would be the proper ray paths in a uniform sound-speed field; however, the refractive effects of lateral sound-speed gradients alter these paths.

HIFT is a prerequisite to a program for ATOC. The goals of ATOC are: (1) to verify the acoustically measured changes in travel time by comparison with those inferred from traditional oceanographic measurements along the transmission path in the North Pacific Ocean; (2) to study the coherence of climatic scale variability in the North Pacific gyre; (3) to test long-range acoustic transmission characteristics using 40-element vertical receiver arrays; and (4) to test specially designed ATOC sound sources. The potential effects of acoustic transmissions on marine mammals have also been investigated in related efforts (Green, 1994; National Research Council, 1994). A series of 17 technical papers addressing various aspects of HIFT was recently published. These papers were introduced by Munk and Baggeroer (1994).

Long-range acoustic propagation issues relevant to tomographic investigations include bathymetric interactions. These aspects have been investigated by Munk (1991), Munk and Zachariasen (1991) and Heaney et al. (1991). Forward scatter by islands and seamounts was found to produce significant scattered arrivals (Munk and Zachariasen, 1991). Such arrivals can complicate interpretation of the recorded acoustic travel times if they are not properly differentiated from the direct arrivals.

Gaillard (1985) demonstrated that mobile sources or receivers can also be used in tomographic experiments, but that this mobility increases the sensitivity to noise. More efficient use of a priori knowledge of the ocean reduces this sensitivity by decreasing the number of degrees of freedom. Moving ship tomography (MST) is a method for obtaining high-resolution, quasi-synoptic, 3-D maps of the oceanic temperature field over large areas (The Acoustic Mid-Ocean Dynamics Experiment Group, 1994). All previous work in ocean acoustic tomography has been performed with fixed or moored instruments. MST is part of the larger acoustic mid-ocean dynamics experiment (AMODE), which has the goal of determining the measurement limits of moving ship tomography. The larger number of acoustic ray paths generated by a moving receiver will improve measurements of advect-ing fronts and interacting eddies. This increased volumetric resolution will enable more rigorous testing of the predictive capabilities of dynamical ocean circulation models.

Methods for computing and plotting tomographic inversions in ocean environments have been described by Nesbitt and Jones (1994a-c). In related work, Weickmann and Jones (1994) described computer programs used to perform ocean-acoustic tomography inversions based on a nonperturbative-inversion method. Harrison et al. (1998) described a localization technique that was an efficient approximation to the maximum a posteriori probability

(MAP) estimator intended for use in matched-field source-localization methods. Yaremchuk and Yaremchuk (2001) developed a nonlinear method for inverting ocean-acoustic tomography data. This method accounts for the quadratic term of the travel-time expansion in the powers of the reciprocal sound-speed perturbations. The higher-order correction to the travel-time model can be essential in regions of large spatial variability. This method was tested against data acquired in the Kuroshio extension region (near 32° N, 148° E). While this method does improve the quality of data processing, it increases the associated computational requirements. Dzieciuch et al. (2001) introduced a technique referred to as a "turning point filter" that permits a uniform treatment of long-range acoustic transmissions from the early ray-like arrivals to the late mode-like arrivals. This method may also be applicable to source localization problems.

As summarized by Spindel (1985), acoustic tomography is particularly attractive for application to the ocean for a number of reasons: (1) images of the ocean interior can be obtained from periphery measurements; (2) large ocean areas can be measured with relatively few instruments; (3) a single source-receiver pair samples the ocean vertically due to multipath propagation effects; and (4) the integrating properties of acoustic propagation produce average quantities not obtainable with point measurements.

5.6.10 Phase conjugation and time-reversal mirrors

Kuperman et al. (1998) experimentally demonstrated that a time-reversal mirror (or phase-conjugate array) could spatially and temporally refocus an incident acoustic field back to its origin. This work was extended by Song et al. (1998) to refocus an incident acoustic field at ranges other than that of the probe source. The basic idea of the approach was that the sound field maxima could be shifted to different ranges by appropriately increasing or decreasing the source frequency for a specific propagation environment.

Time-reversal acoustics can be applied in shallow water to focus energy back to a source location. This refocusing produces spatial intensification of the field (through removal of multipath spreading) as well as temporal convergence of the signal. These properties suggested potential applications in underwater acoustic communication systems (Abrantes et al., 1999).

5.6.11 Deductive geoacoustic inversion

Since direct measurements of seabed parameters are practical only over short distances, considerable attention has been devoted to the inversion of acoustic propagation measurements as a cost-effective alternative to measurements over large ocean areas (e.g. Diachok et al., 1995; Siderius et al., 2000). Using inverse techniques, the desired geoacoustic seabed parameters are extracted from the forward measurements of acoustic propagation in the oceanic waveguide.

Deductive geoacoustic inversion derives the desired sediment and substrate parameters from direct measurements of acoustic propagation. The desired parameters are deduced from the propagation measurements using, for example, simulated annealing or genetic algorithms (Gerstoft, 1994; Snellen et al., 2001). The resulting geoacoustic parameters represent averages over long ranges. Direct measurements over comparably long ranges are frustrated by sparse sampling (Hamson and Ainslie, 1998; Ainslie et al., 2001). Deductive geoacoustic inversion assumes that acoustic data are sensitive to different geoacoustic parameters at different frequencies. Sediment parameters of interest include sediment thickness, sound speed (and gradient), density, attenuation and shear speed. Substrate parameters of interest include density, attenuation, compressional sound speed and shear speed (Ainslie et al., 2000).

Siderius et al. (2001) used experimental data collected at a shallow-water site to demonstrate the impact of ocean sound-speed fluctuations on the quality of seabed properties deduced from geoacoustic inversion. The experimental TL, when averaged over frequency (0.2-3.8 kHz) and time (3 days), was modeled with sufficient accuracy. In fact, the mean modeled TL was within one standard deviation of the data for both short- (2 km) and longrange (10 km) propagation. However, the modeled standard deviation of the TL was in poor agreement with that of the data due to insufficient knowledge of the changing ocean environment. Fluctuations in the sound speed destroyed coherent processing beyond a few kilometers. Without reliable predictions of acoustic propagation, the inverted geoacoustic parameters showed erroneous time variability (erroneous since the seabed properties being deduced are expected to remain fixed over time). At ranges of 2 km, the temporal stability of the geoacoustic inversion indicated that the method was indeed sound. This demonstrated that, without sufficient temporal and spatial knowledge of the volumetric properties of the ocean, geoacoustic inversion at long ranges is subject to degradation.

Ainslie et al. (2001) compared results deduced from three different methods, all of which used a Bartlett processor: (1) the evolutionary search algorithm (ESA) used a PE forward model; (2) the genetic algorithm (GA) used a normal mode forward model; and (3) the deductive rapid environmental assessment model (DREAM) used a normal mode forward model. The results showed good agreement among the three methods, although differences in the computational procedures complicated definitive numerical assessments.

Recently, geoacoustic inversion of ambient noise has also been explored (see Chapter 6, Section 6.7). In particular, Harrison and Simons (2001) noted that inversion of the noise field could complement active-acoustic propagation inversion techniques in which sound-speed fluctuations might impede model matching.

0 0

Post a comment