UNCERTAINTY IN REVERBERATION MODELLING AND A RELATED EXPERIMENT C.H. HARRISON, M. PRIOR AND A. BALDACCI SACLANT Undersea Research Centre, Viale S. Bartolomeo 400, 19138 La Spezia, Italy. E-mail: [email protected] A serious stumbling block in modelling reverberation is the uncertainty in bottom scattering strength, and in the absence of detailed surveys we are still faced with many possible physical mechanisms. For instance, bottom penetration at low sound speeds allows scattering from buried layers and volume scattering which distorts the angle dependence. Typically in shallow water the propagation fall-off with range is controlled by ‘mode-stripping’, and the joint scattering and propagation angle dependence has a profound effect on the reverberation. Some predictions of the reverberation and signalto-noise are made using very simple analytical calculations, and it is shown that there is a regime where signal-to-reverberation is a constant, independent of range, if we assume mode-stripping and Lambert’s law! To check these ideas an experiment is proposed to investigate simultaneously the reverberation intensity and one-way path intensity as functions of range and vertical angle. By comparison of the two trends vs range it is possible to estimate the angle dependence of the scattering. The relative importance of position-dependent scattering strength, bottom shape, scattering law, and propagation are considered. These are backed up by calculations from a new multistatic sonar performance model SUPREMO. 1 Introduction Reverberation is strongly influenced by a number of distinguishable scattering and propagation mechanisms. Amongst the scattering phenomena are: angle-dependence of the scattering law, geographic variation of the scattering strength, and local tilt of bottom facets. Propagation phenomena include range-dependent propagation intensity and range-dependent arrival angle at the scatterer. So, can we predict diffuse reverberation when so many things are going on? Some of these quantities are more susceptible to surveying than others. Some can already be deduced from existing known data. From the opposite point of view, given a (non-local) reverberation measurement, is it possible to tell what mechanism is what? Can we uniquely determine scattering properties? On the face of it there is enormous scope for mathematical complication. But in this paper we try to gain some insight into the relative importance of the mechanisms by first taking an analytical approach. The importance of a mechanism can be judged by its impact on the signal-to-reverberation-ratio (SNR) directly. Finally we return to a numerical model to look at some more detailed effects. 361 N.G. Pace and F.B. Jensen (eds.), Impact of Littoral Environmental Variability on Acoustic Predictions and Sonar Performance, 361-368. © 2002 Kluwer Academic Publishers. Printed in the Netherlands. 362 2 C.H. HARRISON ET AL. Anlytical (closed-form) reverberation and implications for SNR Propagation in shallow water tends to follow a mode-stripping law, culminating in a single mode at long ranges [1]. The reason for this is that the bottom loss (in dB) almost always obeys a linear law with angle until the critical angle θc (if it exists) R = αdBθ. Then the number of reflections is determined by the ray cycle distance rc which for isovelocity is given in terms of the water depth H by rc = 2H/tanθ. This results in an exponential decay in range but a gaussian angle distribution exp(−αθ 2r / 2 H ) . By considering either a sum of eigenrays or a sum of modes [2] the final result for isovelocity can be shown to be a soluble integral over ray angles I= 2 rH θc ∫0 exp( − αθ 2 r ) dθ = 2H 2π Hα r 3 erf( αr / 2 H θ c ) (1) When the square root term in the error function is large (i.e. the gaussian is narrow compared with the critical angle) the intensity reduces to [1] I= 2π Hα r 3 = 20π log(e) (2) Hα dB r 3 which is the familiar three-halves law or 15 log(r). Closed-form solutions also exist [2] with smooth transition to single mode propagation (an exponential decay in range), and these effects can be seen in some of the later plots here. For a monostatic sonar the result of two-way propagation with a point target of target strength ST this becomes I= 2 π ST αHr 3 (erf( α r / 2H θ c ) ) 2 (3) The equivalent reverberation with angle-independent scattering strength SB and elementary scattering area (determined by the spatial pulse length p=ctp/2 and the spatial horizontal beam width rΦ ) is I= 2 π S B Φp αHr 2 (erf( α r / 2H θ c ) ) 2 (4) It inevitably falls off more slowly than the target echo because of the widening of the scattering area with range. Eventually at long range reverberation will win, and at the transition there is a “reverberation limit”. Reverberation with Lambert’s Law scatterers behaves differently. Assuming that the scattering strength is separable in the incoming (θ1) and outgoing (θ2) ray angles S = µ sin(θ 1 ) sin(θ 2 ) the intensity integrals also separate (5) UNCERTAINTY IN REVERBERATION MODELLING I= 4 r2H 2 θc θ c ∫0 ∫0 sin θ 1 sin θ 2 exp( − αθ 1 2 2H r ) exp( − αθ 2 2 2H r )dθ 1 dθ 2 µ rΦ p 363 (6) to give a closed form solution for the reverberation 2 2 θc αθ 2 I = r ) dθ µ rΦp θ exp( − 0 2H rH 2 4µ = 2 3 Φp 1 − exp( −α rθ c 2 / 2 H ) α r ∫ ( ) (7) In Eq. (3) we saw that the mode-stripping three-halves range law resulted in r −3 in two-way propagation when the gaussian was much narrower than the critical angle. Interestingly the Lambert’s law reverberation in Eq. (7) shows exactly the same range dependence in this case; the usual proportionality of scattering area to range that alters the range-dependence in Eq. (4) is exactly counteracted by mode-stripping in this new integral. At shorter ranges where the critical angle truncates the gaussian the error function of Eqs. (3) and (4) is replaced by a (1−exponential) term. (a) (b) Figure 1. Target (two-way propagation) and reverberation: flat bottom, 100m depth, (a) 1000 Hz, (b) 100 Hz. Both frames show general formula and HF approximation for comparison. The source term and the µΦp terms have been set to unity in this example. Another interesting point is that the dependence of reverberation on bottom loss α is stronger than that of the target. In fact low bottom loss benefits the reverberation more than the target so the SNR is improved by high bottom loss! Figure 1 shows the behaviour of these formulae for target and reverberation at 1 kHz and 100 Hz. Note that µΦp has been set to unity since we are only interested in shapes of the curves. The regime of constant SNR is seen at long ranges where target and reverberation are parallel. Single mode effects are seen in the right hand frame. Although these formulae assume isovelocity it is easy to see that refraction simply introduces the possibility of ducts at low angles; angles steep enough to hit the boundaries and therefore reverberate are still attenuated in more or less the same way [2]. 364 C.H. HARRISON ET AL. It is shown in [2] that these formulae can be extended to slowly varying rangedependent environments by using ray invariant H sin(θ) = constant [3] to cope with the angle changes. The results are written in terms of water depth at the source Hs and receiver Hr, the shallowest depth Hc, and an intermediate ‘effective’ depth Heff [2,3] defined in terms of the depth profile H(r). r dr ' 0 H 3 (r ' ) H eff = ( H r 2 H s 2 / r ) ∫ (8) (a) (b) Figure 2 Signal-to-reverberation ratios: flat bottom, 100 m depth, (a) 1000 Hz, (b) 100 Hz. Blue line is large critical angle approximation; green line is arbitrary critical angle; red line is no approximation. The target echo becomes I= 2 2π α r 3 H eff erf{ α rH eff θ c H c } S T 2 Hr Hs (9) Reverberation with angle-independent scatterers becomes I= 2 2π α r 2 H eff erf{ α rH eff θ c H c } S Φp B 2 Hr Hs (10) and reverberation with Lambert’s law becomes I= 4H s 2 α 2 r 3 H eff 2 (1 − exp{−θ 2 c ) H c 2α rH eff /(2 H r 2 H s 2 )} 2 µ Φp (11) Surprisingly, propagation is only weakly affected because, for instance, the increased bottom loss up-slope tends to be counteracted by the concentration of flux into a shallower water depth. The main effect on Lambert’s law reverberation is the change of scattering angle rather than incident intensity, which manifests itself in Eq. (11) as the (Hs/Heff)2 term. Up-slope reverberation is therefore generally stronger than down-slope, UNCERTAINTY IN REVERBERATION MODELLING 365 as one might guess. However the difference is small; for instance on a uniform slope halving the depth results in only a 2.5 dB increase, and doubling it results in a 3.5 dB decrease (Heff is just the average of depths at source and receiver for a uniform slope). 3 A proposed experiment The fact that we expect different range dependences according to the scattering law angle dependence suggests a possible experimental way of determining the scattering law. Two related methods are proposed, and it is hoped that they will be tried this year. There is one experimental arrangement but two sets of measurements. 3.1 Compare Reverberation and Propagation Range-Dependence In principle, referring to Fig. 1 a single shot and a single monostatic hydrophone are sufficient to obtain the entire reverberation curve. To obtain the ‘target’ curve we need to move the source and receiver apart, measuring transmission loss as we go. Doubling the dBs gives a ‘point target’ echo level. The parallel region on the right of the figure shows that we expect the same range dependence for ‘target’ and ‘reverberation’ if there is Lambert’s law. Otherwise we need to try different analytical forms for the law until the trend is explained. Uncertainty in bottom reflection loss is unimportant for this ratio. Given ideal experimental conditions, no numerical modelling is required. However one can think of many practical problems such as finding an environment that is rangeindependent for, say, 50 km radius (in all directions) from the source. In addition there is the problem of ducted propagation. The above analysis is strictly for isovelocity, although as noted, there is not much difference for reverberation with or without refraction, but we need to be careful that, if, for instance, there is downward refraction the source and receiver are either both in the duct or both out. Also we need to remember that we are interested in the bottom scattering law rather than the surface one. These latter problems are controlled and possibly solved by using a vertical array that can separate ducted and completely reflected arrivals. 3.2 Reverberation and Propagation Angle-Dependence If we use a vertical array as receiver in the same experiment we have access to vertical arrival angle for the reverberation as well as for the propagation. This means we have access to the integrands of Eq. (1) and (partially) Eq. (6). In more general terms we can write the angle- and range-dependent propagation and reverberation responses as dI P (θ , r ) = P(θ , r ) dθ (12) dI R (θ 1 , r ) = s (θ 1 ) P(θ 1 , r ) dθ 1 ∫ s(θ 2 ) P (θ 2 , r ) dθ 2 = s (θ 1 ) P(θ 1 , r ) dθ 1 F (r ) (13) where s(θ) is the unknown scattering strength and P(θ,r) and F(r) are initially unknown functions (F(r) is defined by the right hand side of Eq. (13)). The array measurements with moving source (dIP) determine P completely. The single shot monostatic reverberation measurement (dIR) determines s×P×F, but we now know P, and s depends only on θ while F depends only on r. This means that s and F are separable without a 366 C.H. HARRISON ET AL. model! Taking each range in turn, the ratio of the responses always has the angledependence of the scattering strength s but with a range-dependent multiplier F. dI R (θ , r ) = s (θ ) F (r ) dI P (θ , r ) (14) If we take logs of both sides to obtain, say A(θ , r ) = B(θ ) + C ( r ) , then by taking the average over all ranges we get an estimate of the function B(θ) , i.e. s*(θ). Similarly we obtain an estimate of F*(r) by taking the average over all angles, but there is still an unknown numerical factor which could be attributed to either term. Introducing a factor γ such that s=s*γ and F=F*/γ we can use the definition of F implicit in Eq. (13) to solve for γ . γ2 = F * (r ) ∫ s * (θ ) P(θ , r ) dθ (15) Thus we obtain absolute, separated values for s(θ) and F(r). Although γ appears to be a function of r it ought to be flat. Knowing F(r) we can also extract bottom reflection parameters such as α and hence geoacoustic parameters by well-known methods. 4 Reverberation examples using the model SUPREMO Returning to the original question of which phenomena need to be modelled we illustrate some of them here using a new multistatic sonar performance model SUPREMO [4]. The message is that there are many mechanisms that we can model, but given experimental measurements it may still not be possible to identify them. Figure 3. Monostatic and bistatic reverberation from a square seamount modelled with SUPREMO. Ambiguous beam returns can be seen on the right of each frame. For example, a black and white photograph of craters on the moon could be interpreted as geographic changes in scattering strength rather than sloping facets! UNCERTAINTY IN REVERBERATION MODELLING 4.1 367 Scattering: Bottom Topography Constructed from Multiple Sloping Facets Figure 3 shows the reverberation response of a monostatic and a bistatic towed array sonar to a square seamount. One can easily see a bright front side, dark back side, and additional ambiguous image. SUPREMO has a “plug-in” propagation section, and this demonstration used only a range-independent model since the path on the way out is indeed range-independent. Figure 4. Lambert’s law (lower line – red), Lommel-Seeliger (middle line – green), and angleindependent (upper line – blue) reverberation with SUPREMO. Figure 5. Monostatic and bistatic reverberation from the Ragusa Ridge (depth contours overlaid). Reverberation is underestimated because SUPREMO included a tilted-facet bottom but used a “plug-in” range-independent propagation model. 368 4.2 C.H. HARRISON ET AL. Scattering Law Angle Dependence Figure 4 shows the effect of three different scattering angle laws (Lambert, LommelSeeliger, and angle-independent) on an otherwise horizontal seabed. 4.3 Realistic Bathymetry Figure 5 shows the effect of tilted-facet scatterers on reverberation from the Ragusa Ridge south of Sicily. Although echoes follow the contours their intensities are underestimated because SUPREMO has been run with a range-independent propagation model. The analytical calculations above suggest that we can probably ignore the additional effects of the increased bottom loss up-slope since there is a counteracting concentration of flux in the shallower water. However, they also suggest that the raw multiplier (Hs/Hr)2 would be an overestimate. This could be confirmed with a rangedependent propagation module at a future date. 5 Conclusions There are a number of scattering and propagation effects that are mathematically distinguishable, for instance, the scattering law angle dependence, local tilt of bottom facets, scattering strength variation, ray arrival angle (as modified by water depth), and propagation intensity. We have used an analytical approach to find first order dependences, then a numerical approach to look at detailed mechanisms. The implications for coping with environmental uncertainty when predicting reverberation are that scattering is more important than propagation effects. Subjectively ranking the effects we have scattering law, local bottom tilt, and absolute scattering strength. Range-dependent propagation intensity effects are probably less important although the accompanying (predictable) arrival angle change is important. An experiment is proposed that could help to solve the scattering law problem by deducing the angle-dependence and absolute value directly from long range VLA measurements. The same ranking of effects also has implications for the design of models such as SUPREMO. For instance, one could spend a lot of computational effort on rangedependent propagation modelling when a range-independent code with a simple correction term for depth profile could conceivably do just as well. The pay-off would be that trials planning and operational research studies using such models could then deal with moving platforms and targets in range-dependent environments with acceptable computation times. References 1. Weston, D.E., Intensity-range relations in oceanographic acoustics, J. Sound and Vib. 18, 271–287 (1971). 2. Harrison, C.H., Reverberation and signal-excess with mode-stripping and Lambert’s law. Rep. SR-356, SACLANT Undersea Research Centre (2002). 3. Weston, D.E., Propagation in water with uniform sound velocity but variable-depth lossy bottom, J. Sound and Vib. 47, 473–483 (1976). 4. Harrison, C.H., Prior, M. and Baldacci, A., Multistatic reverberation and system modelling using SUPREMO. In Proc. 6th European Conference on Underwater Acoustics, Gdansk, Poland (2002).
© Copyright 2025 Paperzz