Broadband Geoacoustic Inversion using Single Hydrophone Data
R. Chapmana, A. Agarwala, J. Deserta, X. Demoulinb and Y. Stephanb
aSchool of Earth and Ocean Sciences, University of Victoria, PO Box 3055, Victoria, B.C. Canada V8W 3P6 bCMO-SHOM, B.P. 426, 29275 Brest, France
Inversion methods based on spatial coherence of multi-tone CW data have been applied extensively in underwater acoustics to estimate geoacoustic profiles. This paper presents results for a different approach based on broadband waveform inversion at a single hydrophone. The inversion is formulated as an optimization problem to determine the geoacoustic model that provides the best fit between calculated and measured waveforms. A hybrid simulated annealing algorithm is used to search the multi-dimensional model parameter space. The method is applied to data from the INTIMATE98 shallow water experiment to estimate the sound speed and density at the sea floor.
INTRODUCTION and uncertain knowledge of the experimental geometry precluded conventional matched field processing. The Matched field inversion methods that take advantage broadband chirp signals (bandwidth of 300 – 1000 Hz) of the spatial coherence of acoustic field information at were time-compressed using a template signal an array of hydrophones have been used successfully recorded near the source, and averaged over one to invert geoacoustic profiles in shallow water[1,2]. minute. The processed replica correlogram About 10 years ago, Frazer and Pecholcs introduced an approximated the impulse response of the waveguide, alternative approach to narrow band/wide aperture and was used as the data for the inversion. matched field methods and showed formally that similar inversion performance could be expected for broadband data and small aperture geometries[3]. More recently, this approach was applied in work INVERSION METHOD reported by Hermand[4] and Michalopolou[5]. This paper uses a time domain approach and presents The inversion method is based on waveform modeling results for geoacoustic inversion using broadband data in the time domain. The inverse problem is formulated measured at a single hydrophone. as an optimization problem to determine the geoacoustic model that provides the best match between the measured and modeled impulse responses of the shallow water waveguide. We assume that the INTIMATE’98 EXPERIMENT processed replica correlogram of the chirp signal is an adequate representation of the impulse response. In The method was applied to data from the this work, the modeled impulse response is calculated INTIMATE98 experiment that was carried out at three using the ORCA normal mode propagation model to sites on the continental shelf in the Bay of Biscay[6]. compute pressure fields at specific frequencies within The data considered in this paper were obtained on a the signal band[7]. The time series is generated by very short 8-element, 6-m vertical line array that was inverse Fourier transforming the spectral components. deployed at one of the sites at a depth of about 80 m. The cost function, E(m), for comparing the waveforms The array recorded 2-s and 4-s chirp signals that were is given as the cross-correlation function calculated transmitted every 12 s from a projector at a depth of over lags : about 90 m as the source ship closed range on the array to about 600 m. At this close range, the ocean N waveguide approximated a range-independent sand max p (t )r(t) i1 i sediment environment. The water sound speed was E(m) 1 (1) N N sampled by XBTs and CTD casts at regular intervals 2 2 pi (t) ri (t) during the experiment. The very small array aperture i1 i1 where P(t) is the measured impulse response, r(t) is the The estimated values for the experimental modeled impulse response and is the time delay that geometry agree closely with the experimental maximizes the cross-correlation between the two time measurements for range and source depth. Array series. For the perfect match, E = 0, otherwise the depth (for the topmost hydrophone) was known only value of E will lie between -1 and 1. A hybrid search within about 5 m during the experiment. Water depth algorithm that combines simulated annealing with a was known to be around 150 m from available local downhill simplex algorithm is used to search the bathymetric data, and the estimated values of sea floor multi-dimensional geoacoustic model parameter sound speed and density are consistent with the values space[8]. The simplex method has proved to be an expected for the medium-grain sandy sediment at the efficient search process for navigating parameter site. Although not shown in the table, the estimates for spaces in which parameters are correlated. water depth and source depth were strongly correlated. As might be expected, the estimates for source depth and receiver depth are negatively correlated. The correlations between the measured and the modeled INTIMATE98 DATA INVERSION impulse responses that were calculated using the values estimated from the inversions for each The inversion estimated the parameters of a simple individual hydrophone were around 90%. three-layer geoacoustic model consisting of a water layer with known sound speed profile over a thin The results presented in this paper demonstrate that sediment layer and a half space. The most sensitive broadband, single-hydrophone information is effective geoacoustic parameters were the water depth and the for matched field inversion, especially in this sediment sound speed at the sea floor. Since the experiment where the conventional approach was not experimental geometry was not well known, the source viable due to the very small array aperture and the depth and range, and the array depth were also uncertain experimental geometry. The inversion has included in the search process. A total of twelve achieved an overall low level of mismatch in the geoacoustic and geometric parameters were estimated. acoustic data, internal consistency among the estimates However, as might be expected for the high frequency from the array hydrophones, and close agreement with chirp signal, the sensitivity of all but the sea floor known a priori or ground truth information. geoacoustic parameters was very low.
The inversion was applied to data from the end of one of the approaching source tracks where the ship REFERENCES held position at a range of about 650 m. For this range, the impulse response consisted of three strong 1. N. R. Chapman and C. E. Lindsay, IEEE J. Oceanic multipath signals followed by much weaker secondary Eng., 21, 347-354, (1996). arrivals. The dominant multipath signal corresponded 2. J-P. Hermand and P. Gerstoft, IEEE J. Oceanic Eng., 21, to the direct path, single bottom reflection and single 324-346, (1996). surface reflection paths, and the bottom and surface reflected paths. Arrivals comprised of multiple 3. L. N. Frazer and P. I. Pecholcs, J. Acoust. Soc. Am., 88, bottom/surface reflections followed the dominant 995-1002, (1990). group and were about ten times weaker. 4. J-P. Hermand, IEEE J. Oceanic Eng., 24, 41-66, (1999). The inversion was applied separately to the same signal received on each of the array hydrophones. The 5. Z-H. Michalopulou, J. Acoust. Soc. Am., 108, 2082-2090, results are shown in Table 1 where the mean over the (2001). eight hydrophones and the ranges of the estimates are 6. Y. Stephan, T. Folegot, J. Leculier, X. Demoulin and J. shown. Small, Report 48/EPSHOM/CMO/OCA/NP (1999)
Table 1. Parameter estimates 7. E. K. Westwood, C. T. Tindle and N. R. Chapman, J. Parameter Mean Estimated value Acoust. Soc. Am., 100, 3631-3645, (1996). Water depth (m) 149 ± 3 Sediment sound speed (m/s) 1660 ± 5 8. M. R. Fallat and S. E. Dosso, J. Acoust. Soc. Am., 105, 3 Sediment density (g/cm ) 1.8 ± 0.2 3219-3230, (1999). Range (m) 651 ± 5 Source depth (m) 90.5 ± 2 Array depth (m) 79 ± 1 Coral reef and tropical shallow water soundscapes John R. Potter(a) and Eric Delory(b) (a) Acoustic Research Laboratory, Tropical Marine Science Institute, National University of Singapore, 10 Kent Ridge Crescent, Singapore 119260. [email protected] http://www.arl.nus.edu.sg (a) Unidad de Investigacion para la Conservacion de los Mamiferos Marinos, Departamento de Morfologia, Facultad de Veterinaria, Universidad de Las Palmas de Gran Canaria, Trasmontaña, 35416 Arucas, Gran Canaria, Islas Canarias, España. [email protected] Much is known about deep ocean ambient noise, rather less about the more highly structured and variable noise in temperate coastal waters, and rather little about the tropical shallow waters that include coral reefs and the biological diversity these ecosystems support. Interest in shallow water acoustics has been growing over the last decade, and more recently specific interest has bloomed in tropical shallow water and the features particular to these areas. As sonar systems and operations become more complex, moving towards the use of ambient noise and the need for communicating via networked acoustic modems in noisy environments, the soundscape of shallow waters has become more important. This paper presents some of the challenges and advantages of tropical shallow water acoustic environments, and makes some geographic comparisons. While some aspects of tropical ambient noise make life more difficult, such as the non-Gaussian energy distribution of snapping shrimp noise, other aspects are seen to be quite robust, such as the very similar pulse signatures of snapping shrimp over widely- separated regions. The development of signal processing tools based on wavelet and other compact transforms to deal with the detection, separation, and classification of overlapping tropical shallow water noise sources is discussed. commonly usable bandwidth (from a few kHz INTRODUCTION to over 350 kHz) We present a simplified characterisation of ambient noise conditions in shallow tropical SOME CHARACTERISTICS OF waters, excluding wind-driven noise from the SNAPPING SHRIMP SOUNDS sea surface. Wind and resulting wave noise has been more extensively studied and much The characteristics of snapping shrimp more is known about these sources than noise are remarkably consistent across the biological sources in warm shallow water. globe. This may perhaps be explained by the Surf noise can also be very important, but recent discovery that it is not the mechanical obviously only when near a surf line. This impact of the animals jaw parts that creates the paper concerns itself with ambient sources in sound, but the cavitation bubble formed by the somewhat deeper water, in the region of 5-50 rapidly ejected water jet [1]. Thus, while the m. exact physiological characteristics of the jaws may differ between slightly different species, WARM SHALLOW WATER cavitation bubble physics remains much the same. Differences in amplitude and some NOISE SOURCES changes in bandwidth may be expected from In the absence of wind and surface wave larger or smaller shrimp. A typical time series sources (or, if present, in addition to them), acoustic record (normalised to ±1) in warm this paper argues that the plethora of ambient shallow water is shown in the upper panel of noise sources in shallow water can be usefully Fig. 1. The lower panel shows the time- divided into four main classes, so separated frequency plane result for the same data. A because of their very different structures in the measure of the degree of non-Gaussian time-frequency plane. These classes may be behaviour can be made by calculating the described as: scintillation index, SI (normalised variance of intensity) given by · Near transients (such as snapping shrimp) 2 2 · Near tonals (such as shipping machinery SI = I 2 - I I Equation 1 noises) · Coloured background (smoothly varying where I is the acoustic intensity and <> in the frequency domain) indicates the ensemble average [2]. The value · Time-Frequency transients (biologic or of SI increases as extreme values of acoustic man-made). pressure become more common. For a zero- The Near-transients are of particular mean Normally distributed random variable, SI interest to us, since they often consist of = 2 irrespective of the mean power. We have snapping shrimp (at least in warm waters, analysed shallow-water acoustic records from above 11 degrees C) and dominate 99% of the several widely spread countries and calculated FIGURE 1. Time series and spectrogram of snapping shrimp from San Diego (courtesy of SIO, UCSD) the SI. In each case, the record was dominated must be relied upon for ANI. by snapping shrimp sounds, although with varying degrees of other noise sources also audible. The results are given in Table I. WAVELET SEPARATION OF SHALLOW WATER SIGNALS The results indicate that the records are all We have developed a shift-invariant highly non-Gaussian and also highly non- stationary (the standard deviation of the method for separating near-transients based on scintillation index is often comparable to the a first-level wavelet decomposition and index value itself). Nevertheless, the pseudo-entropy test [3]. If the information cost of representing the data at the first level consistency of bubble pulse transient across decomposition exceeds the zeroth level, then wide geographical areas is striking, and may significant near-transients are present and a be of fundamental importance to reef health and shallow water ecosystems; it has been cross-correlation between the two first-level suggested that fish fry navigate to and from decomposition coefficients identifies the common features. These can then be removed reefs to find food and avoid predators by and the signal reconstituted. A similar homing on the sound of snapping shrimp [Eric method, using Cosine Packet Transforms, can Wolanski, personal communication]. be used to remove near-tonal signals. Smooth- Fig. 1 also shows that there is generally a spectrum coloured noise is then removed by considerable interval between snaps compared soft thresholding a conventional biorthogonal to snap length (typically 30 ms). A Poisson wavelet packet decomposition, leaving only model applied to data from several locations the time-frequency transients. An acoustic indicates that there are 0.01-0.25 snaps per signal can thus be neatly split into these four square m of seabed per second (depending on classes, with each class separately terrain), and that typically some 30 ms lapses reconstituted for further analysis. For between snaps that are 20 dB or more above example, near-transients may be used for ANI, the mean power level. This temporal structure and time-frequency transients for the passive permits deterministic ambient noise imaging detection of whales or submarines. (ANI) to be performed for ranges up to 50 m or so. At larger ranges, statistical techniques ACKNOWLEDGEMENTS Table I. Scintillation Index values for several We thank the many helpful colleagues who shallow-water sites around the world. have provided acoustic time series records from around the globe in support of this work. Location Scintillation Sigma of SI Index, SI estimates REFERENCES 1. Versluis, M. Schmitz,B, Von der Heydt, A Domenican 427 132 Republic and Lohse, D. Science 289 2020-2021, (2001). Eilat, Israel 88 83 2. Uscinski, B.J. “The elements of wave Johore, Malaysia 48 15 propagation in random media”, McGraw Hill, London, p 62, (1977). San Diego, USA 223 114 3. Delory, E. and Potter, J.R. DSO/NUS joint Singapore 33 10 seminar series, 1999. Parabolic equation techniques for scattering in a waveguide J. F. Lingevitcha, M. D. Collinsa and M. J. Millsb aU. S. Naval Research Laboratory, Washington, DC 20375 bMitre Corporation, McLean, VA 22102-3481
Parabolic equations are an efficient technique for solving wave propagation problems in range-dependent media. The method is based upon a factorization of the wave equation into incoming and outgoing waves. Wide propagations angles are handled by implementing higher order rational approximations of the factored operator. In gradually varying range-dependent media, the outgoing energy dominates the backscattered component and the outgoing wave satisfies a parabolic equation which can be solved accurately and efficiently with a range marching algorithm. Accuracy of the outgoing wave is maintained by conserving energy flux across vertical interfaces. In cases where backscattered energy is significant, the single scattering approximation is used to derive the two-way parabolic equation. This method has been applied to diffraction gratings and the Sommerfeld diffraction problem. Multiple scattering can also be incorporated into the parabolic equation method by implementing an iterative sweeping method whereby incoming and outgoing energy is coupled. We demonstrate these methods and compare with analytic results and benchmark solutions generated by a coupled normal mode model.
PARABOLIC EQUATION METHOD In a slowly varying range-dependent environment where the backscattered energy is negligible, an accurate The linearized wave equation for the pressure p in a solution is obtained with the parabolic equation by con- fluid can be expressed as an elliptic equation of the form: serving energy flux across vertical interfaces. A linear
condition for conserving energy flux across vertical inter- ∂2
2 faces is given by
¦ §
k 1 X £¥¤ p 0 (1)
¢ ¡
∂x2 ¡ 0
1 2 1 4
¨
¦ §
ρ 1 X £ p 0 (6) ¢ where k0 is a reference horizontal wavenumber. The ¡
depth operator X is given by, where denotes the jump across a vertical interface [7]. ∂ 1 ∂ The narrow angle approximation of Eq. (6),
2 ρ 2 © 2 ¨ ¤ §
X ¦ k k k (2) 0 ∂ ρ ∂ ¡ 0
z z p ρ ¦ 0 (7)
1 c
ρ ¦ ω where is the ambient density, k c ¨ is the medium ω wavenumber, c is sound speed and is circular fre- is sufficient for many applications in ocean acoustics. quency. In a range-independent environment the operator
in Eq. (1) factors,
∂ ∂ TWO-WAY PARABOLIC EQUATION
©
1 2 1 2
¤ £ ¤ ¦ §
ik 1 X £ ik 1 X p 0 ¢
0 ¢ 0
¡ ¡ ∂x ¡ ∂x (3) For problems involving strong backscatter, the two- into incoming and outgoing waves. In this case the out- way parabolic equation has been developed to enforce going wave satisfies a first order equation in range continuity of pressure and displacement at a vertical in- terface in the single scattering approximation [8], ∂p
1 2
£ § ¦ ik 1 X p (4)
0 ¢ ¡ ∂ pi pr ¦ pt (8)
x ¡
∂ ∂
pi pr pt § with a solution [1] that can be marched outward in range ∂x ¦ ∂x (9)
according to, § where the subscripts i § t r refer to the incident, transmit-
1 2 ted and reflected fields respectively. The interface condi-
∆ £ £ £
p x ∆x £ ¦ exp ik x 1 X p x (5)
¢ ¢ ¢
¢ 0 ¡ ¡ tions are implemented by replacing the range derivatives The operators in Eqs. (4,5) are approximated by ratio- with depth operators derived from the outgoing and in- nal functions [2, 3, 4]. Operator approximations of the coming factors of Eq. (3), square root operator based on rotations of the branch cut
pi pr ¦ pt (10)
are useful in stabilizing the iteration formulas to be dis- ¡
© § cussed below [5, 6]. L p p £¦ L p (11) A ¢ i r B t −4 MULTIPLE SCATTERING −3
−2 In problems where multiple scattering is important, the
−1 two-way parabolic equation can be generalized to an iter- ative sweeping algorithm over the computational domain. 0 y (m) Consider a problem involving N range-independent re- 1 gions, where the vertical interface between regions j and 2
j 1 is located at x ¦ x j and the source is located in the ¡
3 first region at x ¦ 0. To avoid stability problems associ-
¤ © 4 ated with the evanescent modes, we define p j to be the −2 −1 0 1 2 3 4 5 6 © x (m) outgoing field at the beginning of region j and ¥ p j to be the incoming field at the end of region j. The vertical FIGURE 1. Parabolic equation solution for the Sommerfeld
knife edge problem with Dirichlet conditions on the knife edge. interface conditions become,
¤ ¤
© © ©
The incident field is a plane wave. ©
¥ ¥ E p p ¦ p E p (13)
j j 1
j j j 1 j 1 ¡ ¡
¤ ¢ ¢ ¤
© © © ©
© ©
1.2
¦ ¥ ¥ £
D E p p £ D p E p (14)
j j j 1 j 1 j j j 1 j 1
1 1 2
ρ ¨ £ D ¦ ik 1 X (15) j 0 ¢ j
j ¡
0.9
¢ ©
1 2
¦ £ £
E exp ik x x 1 X £ (16) ¢
j 0 ¢ j j 1 j
¡ ¨
0.6
¤
© ©
Amplitude
¦ ¥ The unknowns include p for 2 ¦ j N and p for
¤ j j
© ©
©
¥
¦ ¦ 0.3 1 j N 1; p 1 and p N are specified by source and ra- diation conditions respectively. Iteration formulas for the
0.0 outgoing and incoming fields can be solved efficiently by -4 -3 -2 -1 0 1 2 3 4 y (m) repeatedly iterating out and back in range.
FIGURE 2. Quantitative comparison of amplitudes of the parabolic equation solution (solid) and exact solution (dashed) at x 5 m for the Sommerfeld diffraction problem shown in ACKNOWLEDGMENTS Figure 1.
Work supported by ONR. ¡
1 2 §
where L 1 X £ and the subscripts A B refer to the ¢
¡ REFERENCES incident and transmitted medium respectively. An itera- tion formula for the transmitted field follows from elimi- 1. M. D. Collins, J. Acoust. Soc. Am. 93, 1736-1742 (1993). nating the reflected field in Eqs. (10,11), 2. J. F. Claerbout, Fundamentals of Geophysical Data Pro- cessing McGraw-Hill, New York, 1976 pp. 206-207.
1 ¢
© 1
¦ ¨ pt pi 1 L LB £ pt (12) 3. A. Bamberger, B. Engquist, L. Halpern and P. Joly, SIAM
¡ A 2 J. Appl. Math. 48, 129-154 (1988). In the limit of gradual range dependence, the two-way 4. M. D. Collins, J. Acoust. Soc. Am. 86, 1097-1102 (1989). parabolic equation yields solutions consistent with the en- 5. F. A. Millinazzo, C. A. Zala and G. H. Brooke, J. Acoust. ergy conserving solution [6]. Soc. Am. 101, 760-766 (1997). A similar method has been developed for solving scat- 6. J. F. Lingevitch and M. D. Collins, J. Acoust. Soc. Am. 104, tering problems with Neumann or Dirichlet conditions 783-790 (1998). on the scatterer [9]. In this case the incident and scat- 7. M. D. Collins and E. K. Westwood, J. Acoust. Soc. Am. 89, tered fields may be of comparable amplitude. For a scat- 1068-1075 (1991). terer that is a delta function in range, the single scattering method should yield the exact solution. The parabolic 8. M. D. Collins and R. B. Evans, J. Acoust. Soc. Am. 91, 1357-1368 (1992). equation solution for scattering of an incident plane wave from a knife edge (the Sommerfeld diffraction problem) 9. M. J. Mills, M. D. Collins and J. F. Lingevitch, Wave Mo- is shown in Figure 1. Figure 2 shows that the amplitude of tion 31, 173-180 (2000). the parabolic equation solution is in excellent agreement
with the exact solution at x ¦ 5 m [9]. Acoustical signal fluctuations in time–varying shallow–water environments P. L. Nielsen and M. Siderius SACLANT Undersea Research Centre, Viale S. Bartolomeo 400, 19138 La Spezia, Italy
Oceanographic and acoustic data collected in two benign shallow-water areas in the Mediterranean Sea are presented. These data are processed to assess the extent of time variability in received acoustic signals transmitted over fixed propagation path over a period of several hours. The variability in received energy levels and de-correlation time of the acoustic signals is clearly affected by this weakly time-varying ocean environment.
SOUND-SPEED STRUCTURE INTRODUCTION The sound-speed structure along the acoustic prop- agation track was obtained by a towed Conductivity- Sound propagation in shallow-water is strongly Temperature-Depth (CTD) chain. The left panel in Fig. dependent on the actual underwater environment, e.g. 1 shows the CTD-chain data for the PROSIM’97 exper- seabed properties, water-column sound speed, sea iment (total of 4 tows over a 36 h period) and the right surface state etc. Recently the effect of a time-varying panel for the ADVENT’99 experiment (total of 18 tows ocean on sound propagation in shallow-water has been over a 18 h period). assessed experimentally [1, 2]. A strong filtering effect ( 20 dB) in received energy levels at around 600 Hz was PROSIM’97 (m/s) ADVENT’99 (m/s) 1514 seen in the Yellow Sea (China) [1], and a de-correlation 0 1522 time of received acoustic signals down to a few minutes 20 1519 1513 at a frequency of 224 Hz over a 40-km propagation path 40 1516 1512 was observed in the SWARM experiment off the coast of 1511 60 1513 New Jersey (USA) [2]. These fluctuations in the acoustic Depth (m) 1510 1510 80 signals are most likely due to strong time-dependent 1507 1509 internal waves altering the sound-speed structure along 0 5 10 0 5 10 Range (km) Range (km) the acoustic propagation path. This strong internal wave activity in the two experimental areas was known a priori FIGURE 1. Sound-speed structures measured by a towed CTD- to the acoustic measurements. The time variability of the chain along the acoustic propagation track for the PROSIM’97 water-column sound speed is often neglected in sound (left panel) and ADVENT’99 (right panel) experiments. propagation prediction although these oceanographic The PROSIM’97 CTD-data clearly shows a warm fluctuations can have significant impact on stability of (high sound speed) surface layer with a thermocline ex- arrival time and received energy levels. Also, the uncer- tending down to a depth of 30 m. Hereafter the sound tainty in inferred bottom properties from Matched-Field speed is almost constant with depth. The depth of the geo-acoustic inversion increases with the presence of thermocline varies along the track and with time, and a water-column sound-speed fluctuations [3]. front-like feature is seen at a range between 5 and 10 km. The ADVENT’99 CTD-chain data shows an almost iso- velocity water column with only 5 m/s variation in sound SACLANTCEN conducted the PROSIM’97 (May speed from sea surface to the bottom. However, the 1997) and ADVENT’99 (April/May 1999) experiments acoustic track is divided into a low and high sound-speed in the Mediterranean Sea in order to assess acoustic signal region. The position of the boundary between these 2 re- fluctuations in benign shallow-water areas. Broadband gions and the shape of the high sound-speed part along acoustic signals (200-850 Hz) were transmitted over fixed the track (> 5 km) change with time. propagation paths: PROSIM’97 for 36 h over a 15-km path, ADVENT’99 for 18 h along a propagation path of ACOUSTIC DATA 10 km. The acoustic signals were received on a bottom- moored vertical array, and extensive oceanographic mea- The acoustic data were processed to obtain time- surements were performed during the acoustic experi- averaged transmission loss (TL) and standard deviation as ments in order to correlate the changes in the environment a function of depth at 15 and 10 km for the PROSIM’97 with changes in the received acoustic signals. and ADVENT’99 experiments, respectively. In addi- R=15km, F=350Hz R=15km, F=700Hz PROSIM’97 ADVENT’99 0 0 1 30 50 50 20 0.5 100 100 Depth (m) 10 Geo−time (h) 0 0 R=10km, F=350Hz R=10km, F=700Hz 400 600 800 400 600 800 0 Frequency (Hz) Frequency (Hz)
20 FIGURE 3. Normalized Bartlett power of received acoustic 40 signals as a function of frequency and transmission time for PROSIM’97 (left panel) and ADVENT’99 (right panel).
Depth (m) 60 cantly to within less than 1 h of transmission for both 80 80 70 60 80 70 60 experiments. Notice the significant decrease in correla- Loss (dB) Loss (dB) tion around a particular frequency over the entire trans- FIGURE 2. Time-averaged TL (black signature) and standard mission period. This decrease in correlation appears at deviation (gray shading) over depth at ranges of 15 km for a frequency around 500 Hz for PROSIM’97 and around PROSIM’97 (upper panels) and 10 km for ADVENT’99 (lower 650 Hz for ADVENT’99. The behavior of the Bartlett panels). The TL is shown for centre frequencies of 350 and correlator can be explained by the time-varying sound- 700 Hz averaged over a 10 Hz band. speed structure. Numerical modelling of sound propaga- tion for ADVENT’99 including all 18 measured sound- speed structures shows similar behavior as observed in tion, the normalized Bartlett correlator was applied co- the data. herently over hydrophone depth for individual frequen- cies representing the coherence of the acoustic signals SUMMARY over time. The TL (low level) and Bartlett correlator The two fixed acoustic propagation path experiments (high level) represent two levels of processing to assess PROSIM’97 (15-km track) and ADVENT’99 (10-km the fluctuations of the received acoustic signals over time. track) show weakly range- and time-varying sound-speed The TL over depth at 350 and 700 Hz (10 Hz band aver- structures along the propagation path. This fluctuating aged) from PROSIM’97 (upper panels) and ADVENT’99 water column causes a maximum variation in the re- (lower panels) is shown in Fig. 2. In general the standard
ceived energy levels of ¢ 5 dB at higher frequencies. The deviation of the TL increases with frequency for both ex- correlation time between received signals varies from periments as higher acoustic frequency is more sensitive several hours at low frequency but decreases to < 1 h to changes in the environment. The deviation is high-
¡ at higher frequencies. The data show a clear transi- est for the PROSIM’97 experiment ( 5 dB) which may tion at a particular frequency where the correlation time be caused by: (1) longer propagation range, (2) trans- changes significantly. This behavior is attributed to the mission over 36 h compared to 18 h or (3) that the envi- changing environment. The results from the experiments ronmental fluctuations have more impact on the acoustic clearly demonstrate the complexity of sound propagation signals in the PROSIM’97 experiment. The correlation in shallow-water regions. (normalized Bartlett power) of the received acoustic sig- nals across the array is shown in Fig. 3 for PROSIM’97 REFERENCES (left panel) and ADVENT’99 (right panel), respectively. 1. J. Zhou, X. Zhang, and P.H. Rogers, “Resonant interaction One signal received early in the transmissions is denoted of sound wave with internal solitons in the coastal zone”, as the reference signal, and all subsequently received sig- J. Acoust. Soc. Am., 90, 2042-2054 (1991). nals over time are correlated with this reference signal. 2. J.R. Apel et al, “An overview of the 1995 SWARM The correlation has a value of 0 for totally uncorrelated shallow-water internal wave acoustic scattering experi- signals and 1 for completely correlated signals. In gen- ment”, IEEE Journal of Oceanic Engineering 22, pp. 465- eral a high correlation is obtained for several hours at 500 (1997). frequencies below 500 Hz. However, after around 10 h 3. M. Siderius, P.L. Nielsen, J. Sellschopp, M. Snellen and D. the low frequency signals from the PROSIM’97 experi- Simons, “Experimental study of geo-acoustic inversion un- ment start to de-correlate while the correlation from the certainty due to ocean sound speed fluctuations”, Accepted ADVENT’99 seems stable over the 18 h of transmission. for publication in the Journal of the Acoustical Society of At higher frequencies the correlation decreases signifi- America (2001). A Method for Rapid Bathymetric Assessment Using Reverberation from a Time Reversed Mirror B. Edward McDonalda, Charles Hollandb,c aNaval Research Laboratory, Washington DC 20375, USA bSaclant Undersea Research Centre, La Spezia, Italy cPenn State University, State College PA 16804, USA
Analysis of a recent experiment (FAF99 near Elba) suggests a method for wide area rapid assessment of gently varying bathymetry. Transmissions from a vertical source/ receive array (SRA) were brought to focus at a range of several km using a time reversed signal from a 3.5kHz probe source. Adiabatic mode theory and propagation models indicate that if the focal range were set along a given gently varying shallow water path, then transmission into a slightly more upslope (downslope) environment would tend to reduce (increase) the focal range. In FAF99 we initiated probe source signals 1m from the ocean bottom, so that the time reversed SRA signal resulted in a bottom- interacting focal annulus near the probe source range. Enhanced reverberation from the focal annulus was measured at the SRA location. As illustrated in simulation studies, the gentle bathymetic variation appropriate to FAF99 implies azimuth- dependent focal shifts of order 200m for a 3km focus. A properly placed azimuth- resolving array should then be able to reveal azimuthal variation in reverberation from the focal annulus. Variations in reverberation travel time then yield bathymetric trends as a function of azimuth.
REVERBERATION AS A PROBE studies conducted for FAF99 [3, 4] indicate that the near- bottom focus is shifted away from the SRA on azimuths Boundary reverberation in the ocean is considered a around the focal annulus where the ocean is deeper than problem to sonar performance. When the properties of at the probe source, and shifted toward the SRA where the the boundary ensonification can be controlled, however, ocean is shallower. A theoretical explanation of this effect reverberation may help to probe boundary properties. is easily derivable from adiabatic mode theory and the One method by which boundary interaction may be partly WKB approximation applied to waveguide invariants [5]. limited to a desired region is the use of a time reversed mirror [1, 2]. A probe source operating around 3500Hz was placed near the ocean bottom in some of the work Theory conducted by Saclantcen and the Scripps Oceanographic Institution in the Focused Acoustic Fields (FAF) experi- We express the acoustic field far from a source in shal- ments north of Elba in July, 1999. The probe signal was low water as a sum of modal responses:
recorded on a vertical send - receive array (SRA) at var-
r
i ¥ 0 kndx
¢¤£ ¢ ¡ ious ranges between 3 and 11km from the probe source. p r¡ z ∑Anψn z e (1) The retransmitted time reversed signal resulted in a nearly n circular focal annulus centered on the SRA and passing where A and ψ are only weakly dependant on r. The through the probe source (FVLA in Figure 1). Modeling n n
acoustic intensity is then expressed as
¦ ¦ ¦
¦ r
§ ¨ ¥
2 2 i k k dx
0 n m
ψ ψ ψ p ∑ An n 2Re ∑ AnAm n me (2)
n n © m For nearly horizontally propagating modes in shallow wa- ter of constant depth D, we can derive simple parame- ter estimates using pressure release surface and bottom boundary conditions. The modal wavenumbers are then
ω2 nπ 2 k2 § (3)
n 2
c D
¨ ¢
If we perturb the depth D £ D δD r we find
FIGURE 1. Bathymetry for FAF99. For the SRA at F4, the
2
δD r ¢ nπ § focal shift from the dashed circle is schematically illustrated. δkn r ¢ (4)
knD D Table 1. Simulation vs. theory for bathymetric focal shift. the RAM code were rounded off in bins of length 8m, and the theory leading to equation (5) has introduced several Propagation Depth (m) Shift (m) Shift (m) approximations. The higher difference for site F4 is con- Path at Focus RAM Theory sistent with the sampling of more variable bathymetry. F2 - FVLA (3.0km) 112.0 0 0 F2 - SW 111.0 -24 -24 F2 - W 117.0 144 136 SUMMARY F2 - NW 116.7 144 129 F2 - N 114.1 64 56 We find from the data in Table 1 that modeled bathy- F2 - NE 112.2 16 6 metric changes in round trip propagation paths to the fo- F2 - E 111.2 -24 -22 cus have an rms value of 158m for site F2, and 1709m F2 - SE 109.8 -56 -56 for site F4. The corresponding rms travel time changes are 0.10 sec and 1.14 sec. respectively. If the reverber- F4 - FVLA (7.4km) 112.0 0 0 ant return from the focal annulus could be resolved in F4 - SW 111.0 792 538 azimuth, it is quite reasonable that the F4 results could F4 - W 117.0 1552 1662 be related to the bathymetry by equation (5) in a statis- F4 - NW 116.7 1376 978 tically meaningful way. One possible payoff the method F4 - N 114.1 864 636 suggested here is rapid bathymetric assessment of many F4 - NE 112.2 288 293 square km of an unknown shallow water body. In fact a F4 - E 111.2 -200 -196 probe source moving away from a fixed SRA could be F4 - SE 109.8 -200 -196 used to map out concentric circles from which the two dimensional bathymetry could be obtained by differenti- ation in radius. Future refinements of the method could assist in locating shoals endangering near shore craft. The shift in focal range is derived from requiring that
the phase integral in equation (2) be stationary for each
¢
modal pair n ¡ m . The result is ACKNOWLEDGMENT
r
δk δk ¢ dx
§ n m δr 0 Work supported by Saclantcen and the US Office of kn km Naval Research.
r
¢¡
∂δkn ∂kn
¢¤£ ¢ £ dx (5) 0 ∂n ∂n
r ¡
¥ 2 δDdx REFERENCES D 0 1. H. C. Song, W. A. Kuperman, and W. S. Hodgkiss, A time- The last expression in equation (5) is independant of reversal mirror with variable range focusing, J. Acoust. mode number, so it expresses the shift of the total acous- Soc. Am. 103, 3234-3240 (1998) tic field intensity due to small bathymetric variation. 2. W. S. Hodgkiss, H. C. Song, W. A. Kuperman, T. Akal, A long-range and variable focus phase-conjugation exper- iment in shallow water, J. Acoust. Soc. Am. 105, 1597 Comparison with Numerical Results (1999) 3. Charles Holland and B. E. McDonald, Shallow Water Re- Two way propagation along the eight principal com- verberation from a Time Reversed Mirror, Saclantcen Re- pass directions from SRA sites F2 and F4 of Figure 1 was port SR-326 (December, 2000) calculated with the RAM code [6]. The complex acoustic field near the ocean bottom was decomposed into upgoing 4. B. E. McDonald and Charles Holland, Shallow Water Re- verberation from a Time Reversed Mirror: Data- Model and downgoing waves. The intensity of the downgoing Comparison, J. Acoust. Soc. Am. 109, 2495 (2001) wave from RAM was examined for maxima near the focal range. The range shift of the focal maximum relative to 5. G. A. Grachev, Theory of acoustic field invariants in lay- the probe source range is tabulated along with the predic- ered waveguides, Acoust. Phys. 39, 33-35 (Akust. Zh. 39, 67-71), (1993) tion of equation (5) in Table 1. One finds that focal shifts for the SRA at site F2 have a root mean square difference 6. M. D. Collins, Generalization of the Split-Step Pade Solu- of 9.5%, and for F4 an rms difference of 22%. This is ad- tion, J. Acoust. Soc. Am. 96, 382-385 (1994) equate agreement considering that the focal ranges from Matched field detection and localization by a ‘’ shape array GONG XianYi, GE Huiliang, LI Ranwei
State Key Lab. Of Oceanic Acoustics, Hangzhou Applied Acoustics Research Institute, P.O.Box 1249,Hangzhou, China
A ‘’ shape array, which is consisted of respectively 60-element vertical and horizontal arrays, is used for matched field detection and localization in a shallow-water experiment carried out in East China Sea in June early, 2001. This paper presents a preliminary matched-field analysis and puts emphasis on the weak-signal detection and its performance compared with that by use of just the vertical array or the horizontal array in terms of threshold SNR normalized on number of elements.
angle estimations, and signals received by HLA/VLA and INTRODUCTION monitored/measured are simultaneously acquired into the multichannel recording system. Matched Field Processing has been extensively studied The projecting array with the different depths deployed since last two decades and mostly concentrated on vertical from transmitting vessel, which moves to the positions line arrays(VLA) or horizontal line arrays(HLA)[1]. The being 5, 10, 20 and 40km away from the receiving vessel, range and depth of the source can be estimated by use of radiates signals with different forms, for example, CW, VLA, but the its bearing can not be obtained except the PCW, LFM, and PRN. In the paper only signals of 630Hz environment of acoustic propagation is of significant CW and at range 40km are matched-field-analyzed for azimuthal variation and modeled. On the other hand, the source detection and localization. source’s bearing can be given using HLA, but its range and depth are often difficult to be estimated accurately except MATCHED FIELD ANALYSIS for endfire direction. Besides, VLA or HLA themselves have no or poor horizontal or vertical resolutions, The matched field analysis used in this paper is linear respectively, resulting in detection performance Bartlett estimator: [2] degradation. Booth et.al. describe the detectability of low B(r, z, ) p* (r, z, )Rp(r, z, ) (1) o level signals using a tilted line array (45 ) and demonstrate where R is cross-spectral matrix of the sound data sampled the performance improvement compared with that of the by the ‘’ shape array, p(r, z, ) is the modeled pressure VLA with same length and number of elements. vector which is calculated with Porter’s model KRAKEN. This paper presents the preliminary detection and For comparison, the matched field result of VLA is localization (bearing, range and depth) results using ‘’ also calculated by: array which is consisted of a VLA and HLA in a shallow- * water experiment carried out in ECS in June early, 2001. B(r, z) p (r, z)Rp(r, z) (2) The following Sec. II describes the ECS acoustical The range/depth, range/bearing, depth/bearing experiment data used for matched field analysis of ‘’ ambiguity surfaces for the matched field results of the ‘’ array, Sec. III contributes to the data analysis and the final shape array is given in Fig. 1, 2 and 3, respectively. The section (Sec. IV) concludes the paper. maximum of the ambiguity surface is located at r=41.6km , z=39m and 28 . The result is very near to the true ECS ACOUSTICAL EXPERIMENT DATA source location which is at range 40km and depth 40m. The ambiguity surface for the matched field results of Matched field analysis using ‘’ shape array is performed the VLA is given in Fig. 4. One of the maximum of the with data from ECS acoustical experiment which was ambiguity surface is located at r=37km , z=37m. It is conducted from June 2 to 7, 2001 in the site centered at E shown that the accuracy of localization (r,z) using VLA is degraded from that using ‘’ array and there is no 126 49.02> and N 29 40.52> . The ‘’ array deployed from azimuthal resolution in VLA. Moreover, due to many the receiving vessel is consisted of a drift 60-element, sidelobes in the ambiguity surface, detectability of VLA is 147.5m long HLA and a suspended 60-element, 56m long also poor than that of ‘’ array, and its is expected that the VLA. The deployed depth of HLA is 10m and the array threshold SNR normalized on number of elements of just shape is monitored by base-line method. The depth of the first element of VLA is about 3.4m and the array’s tilt angle VLA or HLA is higher than that of ‘ ’ array. is measured by the depth/tilt sensors for array shape and tilt (b) FIGURE 3. (a) Depth/bearing ambiguity surface for ‘’ shape array (b) A slice of the ambiguity surface at z=39m
FIGURE 1. Range/depth ambiguity surface for ‘’ shape array.
FIGURE 4. The ambiguity surface for VLA
IV SUMMARY
The matched-field analysis of ‘’ array for source detection and localization presented in the paper is preliminary. The future works are directed to effects of geoacoustic features and range-dependent environment and (a) coherent broadband processing.
REFERENCES
1. Gong, X.Y., Matched-field processing for source detection and localization, Lecture(in Chinese), 1997 2. N. O. Booth, A.T. Abawi, et.al. IEEE J. Oceanic Eng. 25,296-313 (2000) (b) FIGURE 2. (a) Range/bearing ambiguity surface for ‘’ shape array (b) A slice of the ambiguity surface at r=41.6km
(a) Modal Travel Time Methods for Shallow Water Geoacoustic Inversions
M. I. Taroudakis
Department of Mathematics, University of Crete, 71409 Heraklion, Crete, Greece and Institute of Applied and Computational Mathematics, FORTH, P.O.Box 1527, 711 10 Heraklion, Crete, Greece
The paper presents briefly the procedures developed at FORTH for treating the problem of geoacoustic inversions using the information on the modal structure of the acoustic field due to a known source. The configuration considered is that of a single source-receiver pair and the emitted signal is modelled as a gaussian pulse. Its arrival pattern is used for the inversions. A two- stage algorithm is proposed. First stage involves the characterization of the peaks of the signal as modal arrivals based on the information at the central frequency of some "candidate" environment defined through a systematic search over the search space. Second stage involves an analysis of the identified modal peaks based on the individual candidate environments, which under an optimization procedure leads to an environment closer to the actual one. An iterative linear inversion scheme based on a perturbation analysis over the observed modal peaks could eventually fine tune the recovered environment towards the actual one.
INTRODUCTION For the definition of the inverse problem, we assume that a single source receiver pair is available Modal travel time inversions have been applied by and that the sound source is broad-band. several researchers over the last ten years for Measurements of the acoustic field in the time domain underwater acoustic identification problems with good will be the input data. The characteristic feature of the results, under the assumption that there is good a-priori signal is that its magnitude contains peaks, some of knowledge of the parameters to be recovered which are characterized as "modal peaks" or "modal (reference environment) which leads to the application arrivals" corresponding to energy propagating at the of linear inversion techniques. individual modes
In our approach a relatively wide search space is The core of the inversion procedure has two stages: considered. The inversion procedure adopted is hybrid Modal peak identification and inversion. The modal in the sense that an optimization procedure for the identification is performed on the basis of associating estimation of the most probable reference environment predicted modal travel times with arrival times of the is combined with a linear inversion method to estimate signal arrival pattern peaks. The prediction of the the unknown parameters. Both stages are supported by modal arrivals is done using standard group velocity an identification process to assign modal arrivals at the theory and a reference environment chosen peaks of the measured signal while each one can stand systematically among the environments defined within alone as an inverse scheme for the problem in hand. the search space. If a multidimensional search space is to be examined, an appropriate search algorithm should be used to accelerate the procedure which THE INVERSION PROCEDURES otherwise is time demanding. The predicted modal arrivals are compared with the peaks of the signal and Our study is referred to the problem of a peak of the signal is considered “identified” if the simultaneous inversion for the sound velocity in the difference between its arrival time and the predicted water column and the compressional velocity in the modal travel time is minimum. The whole procedure is sediment layer. This is a special case of tomographic described in [1]. Among the various candidate problems, which in underwater acoustics are normally environments used for modal identification, we keep related to the recovery of the sound speed in the water the ones leading to the maximum number of identified column only. In shallow water, these problems modal peaks. normally include the simultaneous recovery of the A statistical analysis performed over the set of bottom parameters, as the seabed influence on the environments thus defined leads to the “most acoustic field is of profound importance. probable” environment corresponding to the given set of measurements. This environment could be Table 1. The acoustic parameters of the test waveguide. alternatively defined as the one giving minimum value Actual values and search bounds values. With bold italics we to a norm defined on the basis of the travel time indicate the recoverable parameters differences between predicted modal travel times and actual travel times of the identified peaks. Sometimes Actual Search it is adequate to stop the inversion procedure at this bounds point. In some other cases it is desirable to fine-tune Sound speed at the sea 1490 1460-1500 the results using an appropriate local search. This is surface (m/sec) C1(0) accomplished using an already developed method for Sound speed at the 1500 1490-1520 linear modal travel time inversions, using the most water/bottom interface probable environment defined above as the reference C1(h) environment for performing the inversions [2]. Sound speed at the upper 1680 1600-1750 part of the sediment layer C2(h) AN EXAMPLE OF APPLICATION Sound speed at the lower 1700 1680-1800 part of the sediment An inversion toolbox consisting of the layer C2(h+d) identification process and the two-stage inversion Water depth (m) 100 scheme has been developed at FORTH. The input of Sediment thickness 40 the toolbox is the tomographic signal measured in the Sediment density (kg/m3) 1500 time domain. Normally it is the average of several Substrate sound speed 1800 receptions recorded at the receiver. The procedure has Substrate density 1700 been validated so far using synthetic data. For the case presented here, linear variation of the sound speed in Table 2. Inversion results the water and the sediment layers has been considered Actual Most Estimated and thus the unknowns at these layers are the sound probable after local speed values at the top and the bottom of the search corresponding layers. Further we assume for simplicity C1(0) 1490 1480.52 1489.94 that the substrate parameters are known. The density in C1(h) 1500 1497.15 1500.14 the sediment and its thickness is also considered C2(h) 1680 1690.42 1682 known. This is a realistic assumption taking into C2(h+d) 1700 1700 1700 account that the thickness is normally derived using alternative inversion methods.
The acoustic source is considered gaussian with central frequency 150 Hz and bandwidth 50 Hz. Source and receiver depths are the same 50 m. The receiver has been placed at the distance of 10 Km. Table 1 presents the environmental parameters for the actual environment along with the search bounds for the unknown parameters. The inversion results, using the linear inversion approach based on the reference environment defined as the most probable one defined after a suitable analysis of the candidate environments as the one ensuring the fastest convergence of the linear scheme is presented in Table II. The performance of the method is remarkably good. REFERENCES
1. M.I.Taroudakis J. Comput.Acoustics, 8, 307-324 (2000). FIGURE 1. Arrival pattern of the actual and the “most probable”environment. 2. E.C. Shang, J.Acoust.Soc.Am., 85, 1531-1537 (1989). Observed Internal Solitons on a Shelf and their Effects on Sound Propagation
A. N. Serebryanya and A. I. Belovb
aN. N. Andreyev Acoustics Institute, Shvernik Str. 4, Moscow 117036, Russia bWave Research Center, General Physics Institute, RAS, Vavilov Str. 38, Moscow 117942, Russia
Up-to-date review on intense internal waves in sallow water based on long-term field observations on shelves of the former Soviet Union are presented. These waves frequently referred to as “solitons” are often not stationary waves with equilibrium between dissipation and nonlinearity, but are strongly nonlinear in nature. A variety of important nonlinear effects in observed internal waves, including vertical and horizontal asymmetry of wave profile, change of internal wave polarities, and others, are considered. Some observed properties of waves identified with soliton features are demonstrated. The impact of the internal soliton packets on acoustic propagation on a shelf for different hydrological conditions (strong and weak thermocline) is analysed on the basis of field observations and numeric modelling.
SPECTRUM OF INTERNAL WAVES our own results which were obtained on the basis of ON A SHELF almost 20-year investigation in the field. During this time we carried out 15 experiments on internal wave study on shelves of the former Soviet Union Study of shallow water internal waves and acoustic (observations were made in the Caspian, Black and effects produced by them is a popular topic during last Barents Sea, the Sea of Japan, and on Pacific coast of decade (see, for example [1,2]). In the frame of this Kamchatka) [3,4]. Fig.1 shows typical frequency paper we have no possibility to make overview of all spectrum of internal waves on a shelf. There are clear last achievements in this topic. Our task is to present seen two main peaks (semidiurnal tidal and short- period waves, shown by arrows) on averaged spectrum of vertical displacement of thermocline. During an approach of intense short-period internal waves the spectral levels in high frequency band became significantly higher while peaks became more prominent. Our observation showed that the peak of short -period waves is formed by non-linear internal waves. NONLINEAR FEATURES OF OBSERVED INTERNAL WAVES
We will summarise main features of intense internal waves, which were revealed in our long-term observations on a shelf as follows: 1. An arrangement of experimental points (parameters of observed waves) above the first mode dispersion curve of the linear internal waves, a typical feature for waves consisted in trains, is connected with non-linearity of the waves; 2. Next non-linear effects are widespread phenomena on a shelf: - the effect of vertical asymmetry of internal . waves (wave depressions take place in the case of FIGURE 1. Frequency spectrum of internal waves thermocline close to the sea surface, while wave obtained from measurements on shelf of the Sea of elevations – in the case of thermocline close to the Japan. Averaged spectrum (solid line); spectra bottom, see Fig.2); - the effect of change of internal calculated for trains of intense waves (1-3) and G-M wave polarity, which takes place for internal waves spectrum (broken line). propagating in the shoreward direction from deep to FIGURE 2. Vertical asymmetry of internal waves: trains with opposite polarities (consist of wave elevations and depressions) propagated to the shore. Observation was made on shelf of the Sea of Japan on August 1982 [5,6]. shallow water and passing through “turning point”; - the effect of horizontal asymmetry (asymmetry owing to difference in the slope of the leading and following ACKNOWLEDGMENTS edges) is widespread both for depression and elevation waves; 3- it was observed in some cases a good The research work described in this publication was agreement with KdV soliton theory (manifestation of made possible in part by Award N RP2-2255 of the soliton properties were revealed, in particularly, the U.S. Civilian Research & Development Foundation for evidences of the dependency of internal wave speed on the Independent States of the Former Soviet Union amplitude were obtained); 4 - internal wave– (CRDF). predecessor, small-amplitude wave ahead of a train of intense waves (behaviour of wave which is in a disagreement with theory of KdV equation) can be explain by including into consideration an effect of REFERENCES horizontal turbulent viscosity and a shoaling effect. 1. J. Zhou, X. Zhang and P.H.Rogers,J. Acoust.Soc.Amer., SOME EFFECTS OF INTERNAL 90, 2042-2054 (1991). WAVES ON SOUND PROPAGATION 2. J.R. Apel et al, IEEE J.Oceanic Eng., 22, 465-500 (1997). Effect of internal waves on sound propagation on shelf we studied both by carrying out field experiments and 3. A.N. Serebryany, Short-period internal waves on a shelf.(In Russian). Candidate Sc. Dissertation (PhD), by numeric modelling, utilising data of our observation Moscow: N. N. Andreyev Acoustics Institute, 1987, on internal waves [7]. The important result we 220 p. obtained is that in the case of weak thermocline and normal crossing of the acoustic track by a short 4. A.N. Serebryany, Nonlinear internal waves on a shelf internal wave train, the amplitude and phase variations and near bottom rises in the ocean.(In Russian). Dr. Sc. of sound pressure (low-frequency CW signal) are Dissertation, Moscow: N.N. Andreyev Acoustics similar to the form of internal wave train. In the case Institute, 2000, 263 p. of sharp thermocline, the amplitude and phase variations of sound pressure exhibit arising of higher 5. A.N. Serebryany, Oceanology., 25, 744 -751 (1985). frequency oscillations, which can be explained by 6. A.N. Serebryany, Izv. AN SSSR. Atmos. Ocean. Phys., mode coupling. In the paper some new results of 26, 285-293 (1990). numeric simulations of acoustics effects connected with crossing of a track by internal wave depressions 7. A.I. Belov, A.N. Serebryany and V.A. Zhuravlev, and internal wave elevations will be present. “Observations of internal wave effects on acoustic signals in a shallow sea with a weak thermocline”, in Shallow water acoustics, edited by R. Zhang and J.-X. Zhou, Beijing: China Ocean Press, 1997, pp. 283-288. A Unified Model for Reverberation and Scattering from both Stationary and Moving Objects in a Stratified Ocean Waveguide Nicholas C. Makris, Purnima Ratilal and Yi-San Lai Massachusetts Institute of Technology. 77 Massachusetts Avenue Room 5-222 Cambridge, MA 02139 Phone: 617-258-6104 Fax: 617-253-2350 email: [email protected] A unified model for reverberation and submerged object scattering in a stratified medium is developed from full-field wave theory. The unified approach enables the first consistent predictions to be made of target-echo-to-reverberation ratio. The model is applied to determine whether submerged targets can be detected above seafloor reverberation for a variety of watercolumn and sediment types, measurement geometries and target shapes and sizes. The model is also applied to deterimine conditions necessary for discrete geomorphological features of the seafloor and sub-seafloor to be detected above diffuse seafloor reverberation. A spectral formulation for the 3-D field scattered by an object moving in a stratified medium is also derived using full-field wave theory. The derivation stems directly from Green’s theorem and accounts for Doppler effects induced by target motion as well as source and receiver motion. A normal mode formulation that is more computationally efficient but less general is also derived from first principles. The Doppler effects are illustrated through a number of examples in a Pekeris waveguide.
THE UNIFIED MODEL analytic expressions for the three-dimensional 3-D field scattered bistatically by both stochastic and deterministic A common problem in the active detection and local- objects from a source with arbitrary time function, as well ization of a radar or sonar target arises when scattered as the associated spatial and temporal covariances. This returns from the target become indistinguishable from re- enables realistic modeling of the moments of the raw re- turns from randomly rough boundaries, volume inhomo- verberant field received over extended spatial and tempo- geneities, or deterministic features of the environment. ral apertures as well as the output after subsequent pro- The goal of the present article is to investigate the extent cessing with standard beamforming and broadband sig- to which environmental reverberation limits the ability to nal processing techniques. In the present article, applica- detect and localize a target submerged in an ocean waveg- tions of the theory are restricted to systems which employ uide, where methods developed for the radar half-space the beamforming and temporally incoherent processing problem are inapplicable due to the added complications widely used in narrow-band signal reception. Analytic of multi-modal propagation and dispersion. expressions for the statistical moments of the scattered To this end, a unified model for 3-D reverberation field are obtained directly, but can also be obtained by and submerged target scattering in a stratified medium is sample averaging over realizations by Monte Carlo sim- devel- oped from wave theory. The model is fully bistatic ulations, as for example is done for rough surface scat- and stems directly from Green’s theorem, since it gener- tering in Ref. 7. The relative merit of either approach alizes Ingenito’s approach [1, 2] for harmonic scattering depends on the relative difficulty in evaluating the analyt- in a stratified medium by incorporating stochastic scat- ically obtained moments or performing the Monte Carlo terers and time-dependent sources. While it is consis- simulations for the given problem. The analytic approach tent with certain narrow-band results of previous “heuris- has proven to be more advantageous and insightful for the tic” [3] derivations [3-6] for shallow water reverberation illustrative examples of the present article. measured with an omni-directional receiver that are based In this paper, the single scatter theory is generalized on the work of Bucker and Morris [4], it offers more in- to include the effects of source, receiver and target mo- sight and generality since it is developed from first prin- tion. Analytical expressions are obtained for the field ciples with explicitly stated assumptions. For example, scattered to a moving receiver from a moving target in it clearly obeys reciprocity for source-receiver locations a stratified ocean waveguide by a moving source. The within a layered media, which is important in properly formulations are fully bistatic, and all the motions are as- modeling the absolute level of returns from targets or sur- sumed to be horizontal with constant velocities. Both the faces within the seafloor, and it allows absolute compar- expressions for a simple harmonic source and a source ison between reverberation and deterministic target re- with arbitrary time dependence are derived in this paper. turns. Such comparison led to inconsistencies in previ- Spectral and modal representations of the scattered field ous formulations as noted in Ref. 3. It also provides are derived from first principles using the time-domain formulation of Green’s theorem. The spectral representa- ergy returned from the target or scattering patch. This tion makes fewer assumptions and is more accurate than approach is used in the illustrative examples. A viewer- the normal mode representation at closer ranges, but the oriented reference frame is then adopted, translating from normal mode formulation provides a compelling physical the traditional target-oriented frame of waveguide scat- interpretation and can be used at longer ranges without ter theory, to incorporate the continuous distribution of significant loss of accuracy. The single scatter theory of scatterers encountered in waveguide boundary and vol- Ref. 1 and 2 then becomes a special case of the present ume reverberation. This enables analytic expressions to more general theory when the source, receiver and target be developed for the reverberant field returned bistatically are at rest. from seafloor within the resolution footprint of a typical active sonar system after narrow-band beamforming with a horizontal array. Analytical expressions for the 3-D field scattered by CONCLUSION a moving target from a moving source at a moving re- ceiver in a general stratified ocean waveguide are derived One of the greatest challenges to active sonar opera- from first principles using the time-domain formulation tions in shallow water arises when echo returns from the of Green’s theorem. Spectral and modal representations intended target become indistinguishable from reverber- of the Doppler-shifted scattered field for a simple har- ation returned by the waveguide boundaries and volume. monic source and a source with arbitrary time depen- To determine conditions in which a typical low-frequency dence are also obtained. The modal representation has active sonar system may operate effectively in a shal- a compelling physical interpretation exhibited by the fact low water waveguide, a unified model for submerged ob- that a simple harmonic source that excites N modes in ject scattering and reverberation is developed. The ap- the waveguide, for example, will excite roughly N2 dis- proach is to use a waveguide scattering model that follows tinct harmonic components in the scattered field due to directly from Green’s theorem but that takes advantage coupling between the incident modes and the scattered of simplifying single-scatter and far-field approximations modes. The spectral representation, however, is more that apply to a wide variety of problems where the source general and can be used at closer ranges to the target. and receiver are distant from the target. To treat rever- Simulations show that Doppler shifts induced in the beration from randomly rough boundaries and stochastic scattered field by target motion are highly dependent on volume inheterogeneties, the waveguide scattering model the waveguide environment, target shape and measure- is generalized to include stochastic targets. Analytic ex- ment geometry. For a highly dispersive waveguide that pressions for the spatial covariance of the field scattered supports many trapped modes, the frequency spectrum of from a stochastic target are then obtained in terms of the the field scattered by a moving target typically exhibits waveguide Green’s function and the covariance of the tar- significant distortion compared to that of a stationary tar- get’s plane wave scatter function. This makes the for- get or the same target moving in free space. mulation amenable to a wide variety of approaches for computing a stochastic target’s scatter function. For dif- fuse seafloor reverberation, two approaches are adopted, REFERENCES an empirical one of Lambert and Mackenzie and a funda- mental one based on first-order perturbation theory. It is 1. Ingenito, F., J. Acoust. Soc. Am. 82, 2051-2059 (1987). most convenient to describe the diffuse component of dis- 2. Makris, N. C., Ingenito, F., and Kuperman, W. A., J. tant seafloor reverberation with a modal formulation since Acoust. Soc. Am. 96, 1703-1724 (1994). the modes comprise the statistical entities of the field that the scattering surface may decorrelate. 3. Ellis, D. D., J. Acoust. Soc. Am. 97, 2804-2814 (1995). Since reverberation is measured in time but the waveg- 4. Bucker, H. P. and Morris, H. E., J. Acoust. Soc. Am. 44, uide scattering formulation is for harmonic field compo- 827-828 (1968). nents, the time dependence of the field scattered by a dis- 5. Zhang, R. H., and Jin, G. L., J. Sound. Vib. 119, 215-223 tant object from a source of arbitrary time dependence is (1987). derived analytically using the saddle point method. The 6. Lepage, K., J. Acoust. Soc. Am. 106, 3240-3254 (1999). resulting expression is given in terms of modal group ve- 7. Schmidt, H. and Lee, J., J. Acoust. Soc. Am. 105, 1605- locities, the frequencies of which vary as a function of 1617 (1999). time and source, receiver, and target position. A sim- 8. Mackenzie, K. V., J. Acoust. Soc. Am. 33, 1498-1504 pler analytic approach involving Parseval’s theorem can (1961). be applied when the integration time of the measurement system is sufficiently long to include the dominant en-
The Relationship Between Low-Frequency Phase Rate and Source-Receiver Motion in Shallow Water: Theory and Experiment
George V. Frisk
Department of Applied Ocean Physics & Engineering, Woods Hole Oceanographic Institution, Woods Hole, Massachusetts 02543 U.S.A. [email protected]
A theory is presented that relates the phase rate of low-frequency cw signals to the range rate between source and receiver in shallow water. The radiation condition and the paraxial approximation are incorporated into the method of normal modes to obtain a general result applicable to propagation in both range-independent and range-dependent waveguides. Specifically, it is shown that the leading-order behavior of the time rate-of-change of the phase is equal to the product of a typical wavenumber in the water column and the source-receiver speed. This result remains robust even for situations where the acoustic field magnitude displays a complex multimodal interference pattern. The theory accurately predicts the relative source-receiver speed from phase measurements obtained in several Modal Mapping Experiments. The experimental configurations consisted of a source transmitting several pure tones in the band 20-475 Hz to a field of freely drifting buoys, each equipped with a hydrophone, GPS navigation, and radio telemetry. Data were obtained in 50-150 m water depths for source-receiver speeds up to 2 m/s and ranges of 20 km in experiments off New Jersey and in the Gulf of Mexico. The applicability of the theory to long- range, deep-water scenarios is also discussed.
INTRODUCTION N iknr ikr N () = iφ = e ≈ e i kn −k r P Ae ∑ An ∑ An e . (1) Low-frequency sound propagation in shallow water n=1 k n r kr n=1 is characterized by multiple interactions with the surface and bottom in a waveguide whose properties Here A is the magnitude and φ is the phase of the field vary in depth, range, and azimuth. The complexities of which remain after shifting the measured signal to base the resulting propagation process are typically band. In Eq. (1), we have decomposed the field P into described, modeled, and measured in terms of their a sum of N radially propagating normal modes with effects on the acoustic field magnitude (or equivalently, eigenvalues kn and amplitudes An, where we have transmission loss). Far less attention is paid to the assumed that the source and receiver are separated by behavior of the phase, which we shall demonstrate has range r and remain at fixed depths during the course of simple and robust properties, even for situations where the experiment. In addition, we have explicitly the field magnitude displays a complex multimodal imposed the Sommerfeld radiation condition [1] interference pattern. Specifically, we will show both by factoring out the radiating cylindrical wave theoretically and experimentally, that for a source eikr kr associated with a typical wavenumber k=ω/c and/or receiver moving through a shallow-water and sound speed c in the water column (e.g., c=1490 waveguide, the leading-order behavior of the low- m/s). Finally, in Eq. (1), we apply the paraxial frequency phase rate is simply equal to the product of approximation [2], a typical wavenumber in the water column and the source-receiver speed. = − ε ≈ ()− ε ε << kn k 1 n k 1 n 2 for n 1 , (2)
THEORY which states that, although the source injects acoustic energy into the waveguide over a broad range of In the range-independent (horizontally stratified) angles, only a relatively narrow band of angles (less case, the measured pressure due to a time-harmonic than about 30° with respect to the horizontal) point source transmitting the frequency ω is given by -iω t dominates the radiated field, even in shallow water. Pe , where Equation (1) then becomes
ε eikr N −ik n r ≈ 2 P ∑ An e , (3) kr n=1 from which we can show that
dφ ≈ k + O()ε . (4) dr n
If the source and/or receiver are moving, so that r = r(t), we can use the chain rule and Eq. (4) to show that
dr 1 dφ ≈ , (5) dt k dt FIGURE 1. Comparison of GPS-derived range rate (solid line) with phase model results (dashed line) obtained using that is, we can determine the range rate dr dt between MOMAX I – 50 Hz data. the source and receiver from measurements of the phase rate dφ dt . Remarkably, the result in Eq. (5) CONCLUSIONS can be shown to hold true in long-range, range- We have presented a theory that relates the phase dependent, deep-water scenarios as well, simply by rate of low-frequency cw signals to the range rate using the radiation condition and the paraxial between source and receiver in shallow water. The approximation. radiation condition and the paraxial approximation were incorporated into the method of normal modes to EXPERIMENTAL RESULTS AND obtain a general result applicable to propagation in COMPARISONS WITH THEORY both range-independent and range-dependent waveguides. Specifically, it was shown that the During the past several years, three Modal leading-order behavior of the time rate-of-change of the Mapping Experiments (MOMAX) were conducted in phase is equal to the product of a typical wavenumber 50-150 m water depths off New Jersey and in the Gulf in the water column and the source-receiver speed. The theory accurately predicts the relative source-receiver of Mexico [3]. The experimental configurations speed from phase measurements obtained in several consisted of a source transmitting several pure tones in Modal Mapping Experiments. the band 20-475 Hz to a field of freely drifting buoys, each equipped with a hydrophone, GPS navigation, ACKNOWLEDGMENTS and radio telemetry. A key component of the method is the establishment of a local differential GPS system The support of the Office of Naval Research is between the source ship and each buoy, thereby gratefully acknowledged. The author greatly enabling the determination of the positions of the appreciates the contributions of his colleagues: Kyle buoys relative to the source with submeter accuracy. Becker, Laurence Connor, Barry Doust, James Doutt, During these experiments, data were obtained for Calvert Eck, Neil McPhee, Cynthia Sellers, Luiz source-receiver speeds up to 2 m/s and ranges up to 20 Souza, and Keith von der Heydt. km. An example of 50 Hz data obtained on one of the REFERENCES buoys (Shemp) during MOMAX I in 75 m of water off the New Jersey coast is shown in Fig. 1. Here the 1. G.V. Frisk, Ocean and Seabed Acoustics: A Theory of Wave Propagation, Prentice-Hall, Englewood Cliffs, NJ, range rate obtained directly from GPS measurements 1994. of the source and buoy positions is compared with the 2. F.D. Tappert, in Wave Propagation and Underwater range rate predicted from Eq. (5) using the 50 Hz Acoustics, J.B. Keller and J.S. Papadakis, eds., Springer- phase data. The agreement between theory and Verlag, Berlin, 1977, pp. 224-287. experiment is excellent. 3. G.V. Frisk, K.M. Becker, and J.A. Doutt, Proceedings of the Oceans 2000 MTS/IEEE Conference and Exhibition, Providence, RI, September 2000, Vol. I, pp. 185-188.
Internal Wave Fields in Shallow Water: Some Implications for Processing Acoustic Signals
S. Finettea, R. Obaa, M. Orra, B. Pasewarka, C. Stamoulisa, A. Turguta, S.Wolfa, J. Lynchb
aAcoustics Division, Naval Research Laboratory, Washington DC 20375 USA bWoods Hole Oceanographic Institution, Woods Hole, Massachusetts USA
An analysis of acoustic field data acquired in a shallow water region that exhibited significant internal wave activity illustrates inconsistencies between acoustic field properties in this environment and some basic assumptions concerning signal structure implied by phase sensitive signal processing schemes. Variability in the sound speed field induced by internal gravity waves is considered and some implications for acoustic signal processing are discussed in the context of an ocean medium that is both time-varying and horizontally anisotropic.
PROPAGATION IN A DYNAMIC crossed the propagation path at speeds of about .8 SHALLOW WATER ENVIRONMENT m/sec and often propagated through each other. Both the wave packet amplitudes and the number of packet oscillations changed as a function of range from their Sonar signal processing schemes are based on physical generation sites on the shelf break. While both assumptions concerning the spatial and temporal components contributed to the total sound speed structure of the acoustic field. Space-time perturbation as a function of space and time, data perturbations of the sound speed distribution within analysis indicated that the solitary wave contribution shallow ocean environments can alter the field played a more significant role in altering the structure structure, compromising the effectiveness of acoustic of the acoustic field. Instabilities and mixing signal processors by introducing inconsistencies associated with the internal wave packets randomized between these assumptions and the true signal the sound speed field at wave number scales larger structure of the experimental field data. than those associated with the internal wave spatial As an illustration, we refer to experimental and wave numbers. As a result, the sound speed field is numerical simulation results obtained from the neither stationary nor homogeneous in space or time, SWARM95 experiment, which was designed to address issues related to acoustic propagation through but intermittent and inhomogeneous. a random shallow water channel[1-7]. The experimental site was located on the continental shelf off the New Jersey coast in mildly sloping shallow SIGNAL PROCESSING ISSUES water. Oscillations of the strong density gradient resulted in significant depth-dependent sound speed Hydrophone arrays are used in conjunction with phase- variability, induced by internal gravity waves. Two coherent signal processing schemes to detect and fixed sources (center frequencies < 500 Hz) were localize acoustic sources in the presence of ambient deployed in 55 m of water. They were located 42 km noise. These techniques rely on a mathematical or shoreward from an NRL vertical line array spanning conceptual model of the signal and its differences from 21-85 m in water of depth 88 m. The internal wave interfering noise. Standard signal assumptions include field between the sources and receiving array consisted signal stationarity and signal coherence. Signal of two components. One component was a spatially stationarity was violated over integration times of diffuse random background field that was distributed minutes to several hours. Both short (5-30 min) and throughout the propagation region. The second long (semi-diurnal) time scales were significant in component consisted of high frequency solitary wave producing amplitude and phase variability in the signal packets that extended over some 10 km in range and structure. Amplitude variability resulted in signal-to- up to 30-35 km in cross-range. Several families of noise fluctuations of 20-30 dB, which can have these horizontally anisotropic, dispersive packets adverse effects for energy (i.e. signal) detection algorithms. Numerical simulations showed that mode components and their incoherence with the adiabatic coupling between the acoustic field and moving wave (uncoupled) term. Since the coupled-mode packets controlled the magnitudes and time scales of components have amplitudes which are of about the the fluctuations. In addition, adiabatic propagation same magnitude as the uncoupled terms, the success of along solitary wave crests is a significant cause of non- matched-field processing in such an environment will stationary changes in the intensity and phase of the require the prediction (or the suppression) of these wave front. This latter result is related to the horizontal stochastic components. Bartlett processing, minimum anisotropy of the internal wavefield. If the signal is constraint and minimum variance approaches were coherent (i.e. has a predictable spatial dependence of found to be ineffective, yielding errors in localization amplitude and phase) and that dependence is different of the order of the source range, even for short from the noise, then gain can be achieved through a ranges(< 2 km). Matched-mode processing yielded spatial matched filter such as a beamformer. For a improved localization. The horizontal wave fronts of normal mode representation of the field, the signal is coherently propagating acoustic signals may also be treated as a sum of perfectly coherent (in depth and perturbed by an internal wave field, affecting both cross range) modal components that propagate cross-range coherence and bearing estimation. The independently with individual phase and group speeds perturbation arises because local variations of the (which are functions of frequency, water depth, and sound speed profile create local perturbations of the other environmental parameters). In principle, the phase speed of an adiabatically propagating mode. coherence of such a signal can be exploited to achieve These (mode-order-dependent) phase speed variations gain against interference by processing on arbitrarily can, in turn, cause signal refraction--particularly if the large arrays. In addition, the signal can be resolved environmental perturbations are laterally anisotropic, into its modal components by vertical arrays and the as is the case with internal tides and solitary wave modal differences in phase speed (or group speed for packets. In some circumstances, particularly when the broadband signals) can be exploited to achieve range original signal is propagating in a direction nearly localization (e.g. by a matched field processor) as well parallel to wavefronts of these structures, model as further gains against noise that is spread in time. For calculations predict total internal reflection of the a given source and receiver position, the modal acoustic wave will be observed in the horizontal plane. structure in depth and delay (or phase) remains Phase speed changes accompanying mode coupling by constant in time, so processors can in principle correct extended internal wave fronts are also predicted to for this structure by using computed or data-adaptive induce lateral refraction and a loss of coherence. filters. Results of the SWARM95 data analyses demonstrate that the simple model described above is ACKNOWLEDGMENT strongly compromised if the acoustic signals propagate through an internal wave field. First, although the Work sponsored by the Office of Naval Research. depth dependence of the field's individual modal components is only slightly altered by the internal waves, the independence of propagation of these REFERENCES components is removed by mode coupling. The immediate consequence of the mode coupling is that 1. J. Apel et al., IEEE J. of Oceanic Engineering, 22, 465- the temporal structure of an individual mode of 500 (1997). propagation changes from a single pulse to a broadened pulse or a pulse train, which will overlap 2. J. Preisig and T. Duda, IEEE J. Ocean Engineering, 22, and coherently interfere with the pulses of other modal 256-269 (1997). components. This enriched time domain structure of 3. T. Duda and J. Preisig, IEEE J. Ocean Engineering, 24, the individual modal pulses is itself dynamic. The time 16-32 (1999). dependence has been shown by modeling the acoustic/internal wave interaction to be a consequence 4. R.H. Headrick.et al, J. Acoust. Soc. Am., 107, 201-236 of the motion of the internal waves that induce the (2000). modal coupling. Typical coherence times of the resulting complex structures in the complete signal 5 S. Finette et al., J. Acoust. Soc. Am., 108, 957-972 field are found to be of order a few minutes. In fact, (2000). the temporal coherence of pulse trains of single mode 6. A. Turgut, S. Wolf and D. Rouseff, in Oceans’99 fields was found to decay significantly over an interval MTS/IEEE, Vol. 2, 1052-1057 (1999). of a minute or so. The reduction of the single-mode pulse correlation has been shown numerically to be a 7. S. Finette and R. Oba, in Proc. of the Institute of consequence of the stochastic nature of coupled-mode Acoustics, 23, edited by T.G. Leighton, 321-326 (2001). Quantifying the Effects of Shallow Water Internal Waves on the Waveguide Invariant
D. Rouseff
Applied Physics Laboratory, University of Washington,1013 NE 40th Street, Seattle, WA 98105 U.S.A.
In the second edition of Fundamentals of Ocean Acoustics, Brekhovskikh and Lysanov introduced the concept of a waveguide invariant to a larger audience. They showed how plots of acoustic intensity, mapped in range and frequency, often exhibit striations, contours of constant intensity. They defined a parameter ÒbetaÓ as a simple function of range, frequency and the slope of the striations. For certain special cases, beta is an invariant quantity. While beta is no longer literally an invariant for more general shallow water waveguides, the concept remains useful. In the present work, the waveguide invariant is formulated as a distribution. The effect of shallow water internal waves on this waveguide distribution is quantified.
INTRODUCTION In the present paper, the effects of realistic shallow water internal waves on the waveguide invariant are When plotted versus range and frequency, the examined by numerical simulation. First, a method for interference pattern that is observed far from an estimating the Òbeta contentÓ of measured intensity acoustic source often exhibits striations. Chuprov [1] data is developed. The waveguide invariant is related the slope of these striations, ddrω , to the modeled as a distribution rather than a scalar. An range r from the source and the frequency ω via the internal wave model is then outlined based on oceanographic and acoustic data taken in a recent scalar parameter β: shallow water experiment. Both background internal waves and ÒsoliboresÓ are considered. Finally, an rdω β ≡ . (1) overview of the numerical simulations is given. ω dr
Beta is often called the Òwaveguide invariant.Ó For an ideal shallow water waveguide with perfectly IMAGE PROCESSING ALGORITHM reflecting boundaries, one can show β = 1. For more realistic shallow water scenarios, the numerical value Assume the acoustic intensity I is measured in β of beta may change; Hodgkiss et al. [2] found = 1.4 range r over some finite aperture and in frequency ω for a case where both the acoustic source and receiver over some finite bandwith. The resulting windowed were located below the thermocline. Ir(,ω ) has the two-dimensional Fourier transform Brekhovskikh and Lysanov [3] caution that (1) Ä κτ applies only Òfor a group of modes.Ó As the range and I (,). By ParsevalÕs theorem, frequency change, Òthe sound field will be determined ∞ ∞ ∞ ∞ 2 by another group of modesÉresult[ing] in a change of ≡=ωω2 ()1 2 Ä κτκτ EIrdrdIdd∫∫∫ (, ) 2π ∫ (,) . β.Ó Similarly, range-dependence in the environment −∞ −∞ −∞ −∞ can cause the sound field to be determined by another ()2 group of modes even for an observer at a fixed range and frequency. Internal waves, for example, disturb In the polar coordinates κφ= K cos and τφ= K sin , the thermocline and introduce range-dependence in the environment. This range-dependence could be π 2 sufficiently strong as to cause the modes to couple and EEd= ∫ φ φ, (3) thereby change the value for β. As the internal wave −π 2 β. field changes in time, so could the value for where B 2 and a model for the strong, tidally-driven internal = πφφ−2 Ä EIKKKdKφ ()2 ∫ ( cos , sin ) . (4) wave packets (ÒsoliboresÓ) that were observed. −B Because the background thermocline had a The integral has been truncated at some maximum significant gradient below the mixed layer, the lower spatial frequency of interest B. The ramp filter K order acoustic modes had upper turning depths inside arises naturally from the change of variables. Note the water column. Typically, the value of β for Ä φφ that IK( cos , K sin ) is the transform evaluated interaction between these modes was greater than for along a line in Fourier space passing through the origin the higher order, sea-surface interacting modes. The at angle φ . With some manipulation, this angle can be mode coupling induced by the internal waves cause related to β. In this approach, beta is treated not as a energy to be interchanged between these different single number but rather as a distribution. The output types of modes. The background internal waves produce moderate mode coupling. The details of the of the processing is the waveguide distribution Eφ waveguide distribution vary in time, but the β value at plotted versus β. This distribution might be sharply which it peaks remains relatively unchanged. The β peaked around a single value of in which case the solibores produce strong mode coupling. The traditional notion of a scalar invariant would be observed Ir(,ω ), and consequently Eφ , changes reasonable. Other cases producing different sorts of rapidly in time rendering the concept of a waveguide distributions can also be studied. invariant problematic. Numerical calculations illustrating these results will be given in the accompanying oral presentation. EFFECT OF INTERNAL WAVES
In Rouseff and Spindel [4], the proposed image ACKNOWLEDGMENTS processing method was more fully developed. A formula was derived expressing the waveguide This work was supported by the United States Office distribution in terms of the acoustic modes for a range- of Naval Research. independent environment. For a range-dependent environment, the simple formula does not apply. In the present study, the effects of range dependence introduced by shallow water internal waves are studied. Realizations of a three-dimensional, time- REFERENCES evolving internal wave field are generated using an ocean simulation model. Acoustic propagation 1. S. D. Chuprov, ÒInterference structure of a sound field in a layered ocean,Ó in Ocean Acoustics. Current State, through vertical slices of the realization is simulated edited by L. M. Brekhovskikh and I. B. Andreevoi, using the parabolic equation method. The acoustic Moscow: Nauka, 1982, pp. 71-91. simulation is repeated at multiple frequencies to synthesize broadband output. The result is a map of 2. W. S. Hodgkiss, H. C. Song, W. A. Kuperman, T. Akal, acoustic intensity plotted versus range and frequency, C. Ferla, and D. R. Jackson, J. Acoust. Soc. Am. 105, Ir(,ω ). These images are processed using a discrete 1597-1604 (1999). version of the algorithm outline in the previous 3. L. M. Brekhovskikh, and Y. P. Lysanov, Fundamentals section. This gives a plot of the waveguide of Ocean Acoustics, 2nd ed. New York: Springer-Verlag, distribution Eφ plotted versus β at an instant in time. 1991, pp. 140-145. The internal wave field is allowed to evolve and the waveguide distribution tracked in time. 4. D. Rouseff and R. C. Spindel, ÒModeling the waveguide invariant as a distribution,Ó to appear in Ocean Acoustic The internal wave model used in this study was Interference Phenomena and Signal Processing, edited by W. A. Kuperman and G. L. D'Spain, New York: AIP based on measurements taken during the 1996 Coastal Press, 2002. Mixing and Optics Experiment [5]. The water was nominally 70 m deep and the averaged sound speed 5. T. D. Dickey and A. J. Williams, J. Geophys. Res. 106, profile showed the effect of the thermocline, typical of 9427-9434 (2001). conditions during the summer. The model has two primary components: a background internal wave representation valid for relatively quiescent periods, Interference Patterns in Shallow Water Acoustic Fields G. L. D’Spain, P. Lepper, J. A. Smith, E. Terrill, L. Berger, and W. A. Kuperman Marine Physical Laboratory, Scripps Institution of Oceanography, La Jolla, CA, 92093-0704 U.S.A.
When a source transmits a continuous, broadband signal in a waveguide, the received field displays patterns due to the construc- tive/destructive interference between multipath components. Many aspects of these patterns are predictable using simple techniques and provide useful information on the properties of the waveguide and the source. Results from several shallow water experiments off the west coast of the U.S. illustrate the impact on these patterns of range-dependent bathymetry and environmental fluctuations. The analysis also indicates the effects of multipath arrivals on the temporal and spatial coherence of the field. In addition, since ocean acoustic measurements in upcoming experiments will be made using an AUV, numerical modeling of the broadband, multiple scat- tering from the AUV body is performed to understand the effects on these interference patterns of the AUV itself. These additional effects also contain useful information, analogous to the use of interference patterns in human hearing.