1,485ms 1,530ms-1

Thus, ray-theoretical considerations indicate that noise from the surface cannot arrive within about 14° of the horizontal axis of the receiver, at least under the environmental and geometrical conditions assumed here. These conditions, however, are typical of many deep-ocean environments that support deep-sound channel propagation.

The apparent paradox created by the discrepancy between range-independent ray theory and observations can be explained in terms of range-dependent propagation effects. In particular, two effects can

Sound speed (ms 1)

Sound speed (ms 1)

contribute to filling the noise notch through conversion of near-surface shipping and weather noise into shallow-angle propagation paths near the sound channel axis. These mechanisms are slope conversion and horizontal sound-speed variations (e.g. Anderson, 1979). These mechanisms were addressed previously in Chapter 6. Slope conversion considers the reflection and scattering of sound into the deep-sound channel from ships transiting the continental shelves and slopes of the ocean basins. Horizontal soundspeed variations versus latitude cause the axis of the sound channel to shoal at high latitudes, thus allowing far northern (Kibblewhite et al., 1976) or far southern (Bannister, 1986) shipping and weather noise to enter the deep-sound channel and to propagate for long distances. Dashen and Munk (1984) examined this problem and suggested that diffusion may also play a role.

Carey and Wagstaff (1986) reviewed low-frequency (<500Hz) physical-noise models and measurements. They confirmed that coherent signals from surface ships were a dominant characteristic of the horizontal ambient-noise field. The vertical directionality of the noise showed a broad angular distribution centered about the horizontal axis at lower frequencies (<200 Hz), and a dual-peaked (±10° to ±15° off the horizontal) distribution at frequencies between 300 and 500 Hz. In the frequency range 20-200 Hz, the data exhibited a smooth variation with frequency. However, the tonal nature of surface ship radiated-noise spectra would suggest a spiky spectral variation. Smooth spectral variations in the vertical directionality of the noise field are also characteristic of low-shipping areas. Carey and Wagstaff (1986) thus suggested that environmental noise sources such as wind-driven noise, in addition to shipping, might be required to explain the broad angular and frequency characteristics observed in the data. Carey et al. (1990) conducted additional simulations using a parabolic equation (PE) model. These simulations confirmed the role of downslope conversion as a low-pass filter in determining the vertical noise field at mid-basin.

Hodgkiss and Fisher (1990) made a series of direct measurements that confirmed the contribution of the downslope conversion mechanism to the near-horizontal noise distribution. They noted that the effects of absorption appeared to diminish the near-horizontal energy with increasing distance from the coast, and that these effects were more pronounced at higher frequencies.

Models of beam noise statistics use a statistical approach to model the low-frequency ambient-noise field in the ocean. To be of practical use to large-aperture, narrow-beam passive sonar systems, these models must include sonar-specific beam pattern characteristics in addition to point-to-point representations of TL.

The probability measures of beam noise depend on array configuration, orientation, location and season. For detection predictions, the measure of interest is the total power in selected frequency bands. For prediction of false alarm rates, the desired measure is a characterization of the narrowband components of shipping noise (Moll etal., 1979). Only detection predictions will be addressed here.

Using the formalism developed by Moll et al. (1979), an expression for the total noise power in a specified band at the beamformer output can be presented. As a consequence of the principle of superposition of the instantaneous pressures of sound from multiple point sources, the averaged noise power at the beamformer output (Y) can be expressed as:

i=l j=i k=l where m is the number of routes in the basin, n the number of ship types, Aij the number of ships of type j on route i (a random variable), Sjjk the source intensity of the kth ship of type j on route i (a random variable that is statistically independent of the source intensity of any other ship), Zjk the intensity transmission ratio from ship ijk to the receiving point and Bjk the gain for a plane wave arriving at the array from ship ijk.

The probability density function for Y can then be obtained from its characteristic function.

The utilization of noise models requires a specialized database containing information on shipping routes, shipping density by merchant vessel type and radiated noise levels by vessel type. Vessel types are usually differentiated according to freighters, tankers and fishing craft. Further distinctions can be made according to gross tonnage. A ship-count database that can be automatically accessed by ambient-noise models is the historical temporal shipping (HITS) database, which is incorporated in AUTOSHIPS (see Chapter 10). The resolution of HITS is 1° latitude-longitude squares by ship type. Spatial coverage is essentially worldwide (Estalote et al., 1986).

The complexity of the databasing task can be better appreciated by examining Figure 7.8, which shows generalized shipping routes from Solomon et al. (1977). The routes are generalized in the sense that the lines connecting the various ports are not necessarily indicative of the exact paths followed by the merchant traffic. (The numbering of the various routes in Figure 7.8 served as a bookkeeping mechanism for tracking port destinations.) Factors such as seasonal climates and severe weather patterns can alter the exact routes followed by the ships at any particular time.

Mapping transformations between rectangular (for the ship positions and routes) and polar (for noise model sector geometry applications) further complicate model implementation. Figure 7.9 illustrates one particular implementation that relates polar sector geometries to Cartesian latitude and i=l j=i k=l

Figure 7.8 Generalized shipping routes (Solomon et al1977).

Figure 7.10 Ambient noise level (spectrum level, dB re 1 |xPa) variations in the Mediterranean Sea as simulated by a noise model. Simulation is valid for low frequency (<100 Hz) noise for a shallow receiver depth during the winter season. The contour interval is 5 dB and H signifies areas of high noise levels.

Figure 7.10 Ambient noise level (spectrum level, dB re 1 |xPa) variations in the Mediterranean Sea as simulated by a noise model. Simulation is valid for low frequency (<100 Hz) noise for a shallow receiver depth during the winter season. The contour interval is 5 dB and H signifies areas of high noise levels.

longitude divisions. Here, a sector is defined in terms of sector angle (SECANG), sector range (SECRNG), angle increment (ANGINC) and range increment (RNGINC).

An example of the spatial variability of ambient-noise levels generated by a noise model is presented in Figure 7.10 for the Mediterranean Sea. At low frequencies (<100 Hz), most noise is due to shipping traffic. The regions of high noise levels (denoted by H) coincide with established shipping routes (compare with Figure 7.8). Noise variations of as much as 20 dB can occur over relatively short distances.

A summary of available noise models is presented in Table 7.1. This summary segregates the models according to the categories of ambient-noise and beam-noise statistics. Models in the category of beam-noise statistics are further segregated according to analytic and simulation approaches. Numbers within brackets following each model refer to a brief summary and appropriate documentation. Model documentation can range from informal programming commentaries to journal articles to detailed technical reports containing a listing of the actual computer code. Abbreviations and acronyms are defined in Appendix A. This summary does not claim to be exhaustive.

Table 7.1 Summary of underwater acoustic-noise models

Ambient noise

Beam-noise statistics

Analytic

Simulation

AMBENT [1] ANDES [2] CANARY [3] CNOISE [4] DANES [5] DINAMO [6] DUNES [7] FANM [8]

Normal mode ambient noise [9] RANDI -1 / II / III [10]

BBN shipping noise [11] BTL [12]

USI array noise [13] Sonobuoy noise [14]

Notes

1 AMBENT calculates the ambient-noise level for a cylindrically symmetric beam generated by a uniform distribution of surface radiators. Propagation effects are computed by NISSM (McConnell, 1983; Robinson and McConnell, 1983). AMBENT is intended to determine the noise level due to wind, or rain or both.

2 ANDES (Version 4.2) addresses issues related to shallow-water ambient-noise modeling including upgrades to the shipping-density and sound-speed databases, in addition to a new capability to model fluctuations in noise directionality due to changes in wind speed and the movement of discrete sources through the TL field (Renner, 1986a,b, 1988, 1995a,b).

3 CANARY is a ray-based model of ambient noise and noise coherence that is used to estimate the performance of sonars in range-dependent and azimuth-dependent environments. CANARY treats noise sources as surface distributions rather than as points (Harrison, 1997a,b; Harrison et al., 1999, 2001). (Also see the DINAMO model.)

4 CNOISE predicts ambient noise due to shipping. TL versus range files must be generated externally (Estalote, 1984).

5 DANES generates noise levels and horizontal directionality estimates for shipping traffic and wind noise (Osborne, 1979; Lukas et al., 1980b).

6 DINAMO models three-dimensional noise directionality and array performance for operational applications (Harrison, 1998). DINAMO is closely related to CANARY, which was designed for research use. While CANARY first calculates the correlation matrix for the array and then sums these terms to form the array response, DINAMO performs a straightforward integral over all solid angles of the calculated noise directionality multiplied by the array's beampattern.

7 DUNES provides estimates of omnidirectional, vertical, horizontal and three-dimensional directional noise versus frequency. The model includes high-latitude and slope-enhanced wind noise effects. The model emphasizes calculation of noise due to the natural environment. Therefore, shipping contributions are entered explicitly and not via extensive shipping databases (Bannister et al., 1989).

8 FANM uses a simplified (range-independent) ocean environment together with shipping and wind speed databases to predict ambient noise at a fixed receiver location (Lasky and Colilla, 1974; Cavanagh, 1974a,b; Baker, 1976; Long, 1979).

9 Normal mode ambient noise calculates the relative noise level versus depth in addition to the coherence of the noise field at any two points. This calculation is performed using a normal-mode representation of the acoustic field (Kuperman and Ingenito, 1980).

10 RANDI-I/II/III - The original version, RANDI-I (Wagstaff, 1973) calculates and displays the vertical and horizontal directionalities of ambient noise in the frequency range 10 Hz-10kHz. RANDI-II (Hamson and Wagstaff, 1983) was constructed at the SACLANTCEN to account for the special nature of ambient noise in shallow-water environments. RANDI-III (Version 3.1) predicts ambient-noise levels and directionalities at low-to-mid frequencies in both shallow and deep waters. Shipping noise can be calculated for highly variable environments using either the finite-element or the split-step PE method. Local wind noise is computed using the range-independent theory of Kuperman-Ingenito, including both discrete normal modes and continuous spectra. US Navy standard and historical databases are used to describe the environment (Schreiner, 1990; Breeding, 1993; Breeding et al., 1994, 1996). Version 3.3 is a modified version of RANDI 3.1 for use in shallow water. This version provides the user with the option to supply the model with measured or estimated environmental information in areas where the US Navy standard databases may not provide coverage (Pflug, 1996). (The RANDI model is discussed in detail in Section 7.4.)

Beam-noise statistics

11 BBN shipping noise calculates probability density functions of the beam noise power envelope using acoustic source-level data for classes of surface ships, shipping routes and traffic density along those routes (Mahler et al., 1975; Moll et al., 1977, 1979).

12 BTL provides statistical descriptions of shipping noise for low-frequency, horizontally beamed systems (Goldman, 1974).

13 USI array noise numerically estimates the ensemble and time-averaged, one-dimensional statistical probability density function of beam noise (Jennette et al., 1978).

14 Sonobuoy noise was developed for sonobuoy applications (McCabe, 1976). It considers both the temporal correlation of ship-generated noise and the spatial correlation of average intensities for distributed sensors.

15 BEAMPL computes random ambient noise time series within a user-specified beam by statistically taking into account the motion of ships along user-specified routes (Estalote, 1984).

16 DSBN generates beam-noise time series from component submodels for surface ships, TL and receiver (Cavanagh, 1978a,b).

17 NABTAM computes the response of a linear array of hydrophones to wind-sea interactions, surface ships and designated target vessels (W. Galati, E. Moses and R. Jennette, unpublished manuscript).

Was this article helpful?

## Post a comment