EFFECTS OF SOUND SPEED FLUCTUATIONS DUE TO INTERNAL WAVES IN SHALLOW WATER ON HORIZONTAL WAVENUMBER ESTIMATION KYLE M. BECKER Acoustics Group, The Pennsylvania State University, Applied Research Laboratory P.O. Box 30, State College PA 16803, USA E-mail: [email protected] GEORGE V. FRISK Dept. of Applied Ocean Physics and Engineering, Woods Hole Oceanographic Institution MS #11, Woods Hole MA 02543, USA E-mail: [email protected] There is considerable current interest in the influence of water-column variability on acoustic propagation and its effects on geoacoustic inversion. In general, the effect of sound speed fluctuations due to internal waves in the water column is to promote the coupling of energy between propagating acoustic modes. The effects of mode coupling include fluctuations in individual modal amplitudes and arrival times along with time spreading of the original pulses. In contrast to the broadband case, little research has been conducted on the effects of internal waves on cw modal-based inversion methods. These techniques require estimates of the propagating modal eigenvalues for a cw point source field as input data to the inversion algorithm. In much of the literature, wavenumber estimation is performed with the assumption that pressure is given by an adiabatic mode sum. Changes in modal content as a function of range are then attributed to local changes in the waveguide boundaries, specifically, the bottom. For a shallow-water waveguide including internal waves, the adiabatic assumption is violated and estimates of local wavenumber content is affected. This paper addresses the nature of the these affects on the wavenumber estimation problem. In particular, numerical studies of internal wave effects are conducted with respect to identification and bias of individual modes along with the ability to resolve closely spaced eigenvalues. Preliminary results for a weak internal wave field show that mode coupling leads to an enhancement of the wavenumber spectral estimates due to the energizing of weak modes that were previously not excited. 1 Introduction Sound propagation in range-independent shallow-water waveguides can be concisely represented by the exact Hankel transform relationship between complex pressure, measured as a function of range, and the depth-dependent Green’s function, expressed in the horizontal wavenumber domain >dH. Further, from the modal interpretation of this relationship, a linearized inversion method can be developed relating modal eigenvalues, given by discrete horizontal wavenumber values, to sediment geoacoustic properties >1H. A key aspect of this approach is the accurate estimation of the individual modal eigenvalues 385 N.G. Pace and F.B. Jensen (eds.), Impact of Littoral Environmental Variability on Acoustic Predictions and Sonar Performance, 385-392. © 2002 Kluwer Academic Publishers. Printed in the Netherlands. 386 K.M. BECKER AND G.V. FRISK contributing to the propagating field. This problem becomes more complicated when spatial dependence enters into the problem, through variation in sediment properties with range [3], or variability in the water column due to sound speed fluctuations or changes in bathymetry. In the cases of changes in sediment properties or bathymetry that occur slowly over range, horizontal wavenumber estimates can be obtained by assuming regions of local range-independence in the waveguide and applying techniques analogous to short-time Fourier transform methods [4]. However, although the effects of sound speed fluctuations due to internal waves on modal propagation, as well as their corresponding effects on broadband geoacoustic inversion, have been treated in the literature recently (see, for example, [5] and [6]), their effects on horizontal wavenumber estimation have received less attention. Two general assumptions are usually made for wavenumber analysis based on the Hankel transform. The first assumption is the adiabatic mode approximation, which states that no energy is transferred amongst the different modes and that the number of propagating modes remains constant. The second assumption is that the effects of sound speed fluctuations in the water column are minimized due to the averaging effect that occurs when transforming data over sufficiently large apertures. In the case of internal waves, it is well known that the adiabatic assumption does not apply and that a complex competition between mode coupling and mode stripping occurs [7]. In this paper, a study is presented addressing the wavenumber estimation problem for a range-dependent acoustic field where mode coupling is known to occur. Synthetic complex pressure fields are generated using a range-dependent parabolic equation (PE) model [8] for both a range-independent shallow-water waveguide, and a waveguide with a deterministic internal wave (IW) field that introduces range-dependence into the waveguide through perturbations to a background sound speed profile. Applying short-window methods based on the Hankel transform to the resulting fields, modal content is plotted as a function of range and compared for the different models. In addition, using a fully coupled mode code [9], reference values of modal amplitudes and horizontal wavenumbers are determined as a function of range for both the background and internal wave models. Comparisons are made between estimated wavenumbers and reference values for both acoustic fields. The influence of the internal wave field on wavenumber estimates is examined in terms of the mean and variance of the estimates compared to their reference values. A comparison of modal amplitudes between the different models is also made yielding information about the detectability of particular modes. 2 Methods For a cw point source, and assuming horizontal stratification, the equations governing sound propagation in a shallow-water waveguide can be reduced to a Hankel transform relationship between the complex-pressure field p(r; z, zo ) and the depth-dependent Green’s function g(kr ; z, zo ), p(r; z, zo ) = g(kr ; z, zo ) = ! ∞ !0 ∞ 0 g(kr ; z, zo )J0 (kr r)kr dkr , (1) p(r; z, zo )J0 (kr r)rdr. IW EFFECTS ON WAVENUMBER ESTIMATION 387 In the above, pressure is measured as a function of range r, and the Green’s function is a function of horizontal wavenumber kr and satisfies the inhomogeneous depth-dependent wave equation, along with impedance boundary conditions at the surface (z = 0) and the bottom (z = h) [10], " % # $ d 1 d d2 2 2 + ρ(z) + k (z) − kr g(kr ; z, zo ) = −2δ(z − zo ), dz 2 dz ρ(z) dz iZT (kr ) ∂g T (kr ; 0) g T (kr ; 0) + = 0, ωρ(0) ∂z iZB (kr ) ∂g B (kr ; h) g B (kr ; h) + = 0. ωρ(h) ∂z (2) In Eq. (2), the source and receiver depths, zo and z, are treated as parameters in the problem, ρ(z) is density, and k(z) = ω/c(z) is the total wavenumber for the angular frequency ω. The solution to the depth-dependent Eq. (2) is obtained by incorporating into the total solution, the independent solutions gT (kr , z) and gB (kr , z) which satisfy impedance boundary conditions at the top and bottom, represented by ZT (kr ) and ZB (kr ), respectively. The general form of the depth-dependent Green’s function satisfying both boundary conditions can then be written, g(kr ; z, zo ) = −2 gT (kr , z< )gB (kr ; z> ) , W (zo ) 0 ≤ z ≤ h, (3) where z< = min(z, zo ), z> = max(z, zo ), and W (zo ) is the Wronskian evaluated at the source location and given by, & & ∂gB && ∂gT && − gB (zo ). (4) W (zo ) = gT (zo ) ∂z &zo ∂z &zo For a horizontally stratified ocean, and assuming the top surface satisfies a pressurerelease boundary condition, g(kr ; z, zo ) becomes a function of the impedance boundary condition at the water-bottom interface alone. Through the impedance relationship, the Green’s function is a complete characterization of the waveguide as needed for the solution of Eq. (1). For shallow-water waveguides, the Green’s function is characterized by sharp peaks, or resonances, that occur when the Wronskian in Eq. (3) goes to zero. These peaks occur at distinct values of horizontal wavenumber that correspond to the modal eigenvalues kn in a modal expansion of the field. The mode sum can be obtained by inserting Eq. (3) into Eq. (1) and applying complex contour integration methods, to obtain p(r, z) = nmax iπ ' (1) φn (zo )φn (z)H0 (kn r) + I(r), ρ(zo ) n=1 (5) where φn are mode functions evaluated at the source and receiver depths, nmax is the maximum number of propagating modes, and I(r) represents a continuum contribution that can often be ignored for ranges greater than a few water depths from the source. For analysis purposes, by applying the large argument asymptotic approximation to the Bessel function, J0 (kr r) in Eq. (1), the Hankel transform operation can be replaced 388 K.M. BECKER AND G.V. FRISK with a Fourier Transform [10], eiπ/4 g(kr ; z, zo ) ∼ √ 2πkr ! ∞ −∞ √ p(r; z, zo ) re−ikr r dr, kr r % 1, kr > 0. (6) The asymptotic form allows the Green’s function to be obtained through a Fast Fourier √ Transform (FFT) of the product p(r; z, zo ) r. This form is particularly useful in that various spectral estimation tools can be applied, including high-resolution methods, to determine wavenumber content. Using these methods, horizontal wavenumbers corresponding to modal eigenvalues are determined from the peak locations in the Green’s function estimate. The relationship between the depth-dependent Green’s function and complex pressure field described above is strictly valid for range-independent environments. For the rangedependent case, a convenient representation of the pressure field is given by the adiabatic mode sum expressed as, (r √ nmax i k (r! )dr! e 0 n 2πeiπ/4 ' φn (0, zo )φn (r, z) )( . (7) p(r, z) ∼ r ρ(0, zo ) n=1 k (r# )dr# 0 n In this representation, mode functions and their corresponding eigenvalues adapt themselves to the local waveguide properties as they change in range. By performing the integral in (6) over a finite range interval given by −L/2 ≤ r ≤ L/2, where L is the aperture length, a local estimate of the Green’s function can be obtained, where it is assumed that the waveguide environment is approximately range-independent over L. For a truly adiabatic environment, moving the aperture along in range using short range steps gives a picture of the modal content of the waveguide as it evolves with range [4]. The adiabatic representation provides an intuitive way to understand how modal content evolves with range, but is not a necessary requirement for extracting range-dependent modal information using the Hankel transform. For the synthetic pressure fields generated for this study, the introduction of the internal wave field to the otherwise range-independent waveguide causes energy in the propagating modes to be transferred amongst each other violating the adiabatic assumption. However, estimates of the local depth-dependent Green’s function can still be obtained by integrating the field over a finite aperture as described above. 3 Numerical examples and wavenumber estimation The waveguide environment for the numerical studies was one of several designed as a test bed for benchmarking range-dependent acoustic propagation models [11]. The waveguide for this study was a fluid model with range-independent sediment properties and a constant water/sediment interface depth of 200 m. Density in the water column was 1 g/cm3 with no attenuation. The bottom had a compressional wave speed of 1700 m/s, a density of 1.5 g/cm3 , and an attenuation of 0.1 dB/λ. The background sound speed profile was generally downward refracting with a slight duct occurring above 26 m depth and was given by c(z) = 1515 + 0.016z, z < 26 m c(z) = co [1 + a(e−b + b − 1)], z ≥ 26 m, (8) 389 IW EFFECTS ON WAVENUMBER ESTIMATION a) TL vs Range (100 Hz) b) 0−5 km (100 Hz) −30 −30 background internal wave TL (dB) −40 −40 −50 −50 −60 −60 −70 −70 −80 −80 −90 −90 −100 0 −100 0 5 10 15 20 1 TL (dB) c) 10−15 km (100 Hz) −50 −60 −60 −70 −70 −80 −80 −90 −90 11 12 13 14 3 4 5 d) 15−20 km (100 Hz) −50 −100 10 2 15 −100 15 16 Range (km) 17 18 19 20 Range (km) Figure 1. 100 Hz TL plots for internal wave study. 5 km sub-plots show departure of acoustic field for internal waves from range-independent case. where z is the depth in meters, co = 1490 m/s, a = 0.25, and b = (z − 200)/500 [12]. To the background sound speed was added a perturbation as a function of depth and range to represent the internal wave field. The sound speed perturbation, δc, was given by, δc(z, r) = 4(z/B)e−z/B 5 ' cos(Ki r), (9) i=1 where Ki = 2π[2000 − 300(i − 1)]−1 m−1 and B = 25 m [12]. Using this model, the maximum sound speed perturbation is about 7.5 m/s. Using the above models, PECan [8], was used to generate the acoustic field at 100 Hz for both the background and perturbed sound speed environments from 0 to 20 km on a 5 meter range grid. Transmission loss data are shown in Fig. 1 for the full aperture and for several 5 km sub-apertures for a source depth of 45 m and receiver depth of 60 m. The modal interference patterns in the figure result from the constructive and destructive interference of different propagating modes. The differences in interference patterns for the two fields are most evident at long ranges and indicate the influence of the internal waves on the modal interactions. Using the complex pressure fields, wavenumber estimates were obtained using a high resolution sliding widow estimator based on an autoregressive spectral estimator [13]. The estimator is based on a parametric representation of the signal and requires the model order as an input parameter. For the 20 km apertures of the synthetic data fields, wavenumber estimates were made for 3000 m sliding window sub-apertures with a step size of 200 m and a model order of 1/3 the number of data points in each sub-aperture. Additionally, reference values of horizontal wavenumbers for the range-independent waveguide environment with the background sound speed profile were determined using a normal mode code [9]. The resulting wavenumber estimates with the reference results overlayed are shown in Figs. 2 and 3. In the figures, modes identified using the Hankel transform method correspond with the most energetic modes for the acoustic fields. Where there is a match between an estimated wavenumber and a reference value, the reference value 390 K.M. BECKER AND G.V. FRISK Figure 2. Sliding window wavenumber estimates compared to reference results (dashed) plotted for unperturbed case. The dark lines indicate the locations of the peaks of the Green’s function. Figure 3. Sliding window wavenumber estimates compared to reference results (dashed) plotted for IW case. The dark lines indicate the locations of the peaks of the Green’s function. is plotted as a dashed line. The plots show that for the unperturbed case, 6 modes are estimated that can be correlated with the reference results. For the IW case, 8 modes are identified that correlate with the reference results. In both cases, the wavenumber estimates are stable with range, which is especially evident for the higher-order modes. This is an indication that the waveguide boundary conditions are range-independent. For the low-order modes, there are small undulations about the background wavenumbers that occur with range. However, there is no apparent wavenumber shift observed for the modes obtained from the two fields. Together, the observations suggest that for the given source/receiver geometry, certain modes are not excited or contain relatively little energy. In order to verify this claim and gain a better understanding of the wavenumber estimates shown, it is necessary to examine both modal amplitudes and horizontal wavenumber content as a function of range. 391 IW EFFECTS ON WAVENUMBER ESTIMATION mode 1 dB 40 20 0 0 5 10 dB 15 20 mode 3 20 5 10 15 20 mode 4 20 0 0 5 10 15 20 mode 5 40 0 5 10 20 15 20 mode 6 40 20 0 0 0 5 10 15 20 mode 7 40 dB 0 40 0 dB Modal Amplitude (dB) 0 40 0 5 10 20 15 20 mode 8 40 20 0 0 0 5 10 15 20 mode 9 40 dB mode 2 40 20 0 5 10 20 15 20 mode 10 40 20 0 0 0 5 10 Range (km) 15 20 0 5 10 Range (km) 15 20 Figure 4. Mode amplitude difference (internal wave - background) vs. range. To examine mode amplitudes and wavenumbers, the COUPLE [9] range-dependent normal mode code was used. Horizontal wavenumbers and modal amplitudes were output at a receiver depth of 60 m for each range in the problem. The horizontal wavenumbers corresponding to the propagating modes for the background problem are shown as the straight lines in Figs. 2 and 3. For the IW case, discrete values of horizontal wavenumbers were selected from the peak positions of the Green’s function at each range. Means and variances over range were calculated for the individual eigenvalue estimates and compared to the background, or reference, eigenvalues. The variances of the individual estimates were small, consistent with the variance expected for the high-resolution estimator, as if it were applied to a signal with additive noise and a high signal to noise ratio (SNR greater than 45 dB) [13]. For the 8 modes identified in the IW case, the biases between the estimates and the background values were less than 1 standard deviation for all modes. This suggests that the mean values for the estimated horizontal wavenumbers could be used as input data for a range-independent geoacoustic inversion algorithm. Figure 4 shows the difference in individual modal amplitudes between the internal wave field and background field as a function of range for the first 10 propagating modes. The figure shows the influence of the internal wave field on the individual modes. While it is seen that all of the modes are affected, modes 6 and 10 each have a large amount of additional energy in them due to the internal wave field. This accounts for the enhanced number of modes observed in the wavenumber estimates in figure 3, where the coupled mode code verified that modes not observed were only weakly excited in either case. Further, the smaller perturbations seen in the mode amplitudes can account for some of the scintillations observed in the wavenumber estimates. 4 Conclusions Internal waves greatly influence the propagation of sound in shallow water by redistributing energy amongst different propagating modes. This in turn has an effect on the estimation of local modal content for use in linear inversion algorithms. In the case of the weak internal wave field used as a model for this study, the following conclusions can be 392 K.M. BECKER AND G.V. FRISK drawn. The effect of the internal wave field on local horizontal wavenumber estimates is minimized by the averaging effects controlled by the size of the local aperture used in the processing. A more complete study needs to be done to access an optimal aperture size for a given environment. Finally, the presence of internal waves serves to excite modes that otherwise would not be excited for a particular source receiver geometry and leads to an enhancement of the wavenumber estimates. This enhancement yields additional data which can be used in an inversion algorithm to determine seabed properties. Acknowledgements Richard B. Evans was extremely generous with his time in helping to get the coupled mode code up and running. PECan data was generously provided by Gordon Ebbeson of the Defense Research Establishment of the Atlantic (DREA), Canada. This work was partially supported by an ONR Special Postdoctoral Fellowship Award in Ocean Acoustics [Contract No. N00014-02-1-0334]. References 1. G.V. Frisk and J.F. Lynch, Shallow water waveguide characterization using the Hankel transform, J. Acoust. Soc. Am. 76(1), 205–216 (1984). 2. S.D. Rajan, J.F. Lynch and G.V. Frisk, Perturbative inversion methods for obtaining bottom geoacoustic parameters in shallow water, J. Acoust. Soc. Am. 82(3), 998–1017 (1987). 3. G.V. Frisk, J.F. Lynch and S.D. Rajan, Determination of compressional wave speed profiles using modal inverse techniques in a range-dependent environment in Nantucket Sound, J. Acoust. Soc. Am. 86(5), 1928–1938 (1989). 4. K. Ohta and G.V. Frisk, Modal evolution and inversion for seabed geoacoustic properties in weakly range-dependent shallow-water waveguides, IEEE J. Oceanic Eng. 22, 501–521 (1997). 5. J.C. Preisig and T.F. Duda, Coupled acoustic mode propagation through continental-shelf internal solitary waves, IEEE J. Oceanic Eng. 22, (1997). 6. M. Siderius, P.L. Nielsen, J. Sellschopp, M. Snellen and D. Simons, Experimental study of geo-acoustic inversion uncertainty due to ocean sound-speed fluctuations, J. Acoust. Soc. Am. 110(2), 769–781 (2001). 7. D. Tielbürger, S. Finette and S. Wolf, Acoustic propagation through an internal wave field in a shallow water waveguide, J. Acoust. Soc. Am. 101(2), 789–808 (1996). 8. G.H. Brooke, D.J. Thomson and G.R. Ebbeson, PECAN: A Canadian parabolic equation model for underwater sound propagation, J. Comp. Acoust. 9(1), 69–100 (2001). 9. R.B. Evans, COUPLE, 1997 Version (Nov. 20, 1997). ftp://oalib.saic.com/pub/oalib/couple/ 10. G.V. Frisk, Ocean and Seabed Acoustics: A Theory of Wave Propagation (Prentice Hall, Englewood Cliffs, New Jersey, 1994). 11. Various Authors, Acoustical oceanography and underwater acoustics: Benchmarking rangedependent reference models, chaired by K.B. Smith and A.I. Tolstoy, J. Acoust. Soc. Am. 109 (5 pt. 2), 2332–2335 (2001). 12. K.B. Smith, Benchmarking shallow water range-dependent acoustic propagation modeling, Test Case III: Internal waves (Dec. 20, 2000). http://web.nps.navy.mil/∼ kbsmith/Chicago ASA/iws.html 13. K.M. Becker, Geoacoustic inversion in laterally varying shallow-water environments using high-resolution wavenumber estimation, Ph.D. dissertation, MIT/WHOI, WHOI-02-03 (2002).
© Copyright 2025 Paperzz