A wideband echo sounder: measurements on cod, saithe, herring, and mackerel from 27 to 54 khz

Transcription

1 Rapp. P.-v. Réun. Cons. int. Explor. Mer, 189: A wideband echo sounder: measurements on cod, saithe, herring, and mackerel from 27 to 54 khz E. J. Simmonds and F. Armstrong Simmonds, E. J., and Armstrong, F A wideband echo sounder: measurements on cod, saithe, herring, and mackerel from 27 to 54 khz. - Rapp. P.-v. Réun. Cons, int. Explor. Mer, 189: The paper describes a wideband constant beamwidth echo-sounder system and presents spectral responses obtained from caged aggregations of cod (Gadus morhua), saithe (Pollachius virens), herring (Clupea harengus harengus), and mackerel (Scomber scombrus). The transmitter generates swept-frequency pulses with uniformfrequency spectrum output, using stored reference levels and a digital-to-analogue converter to provide amplitude and frequency stability. The transducer is a 217- element partial sphere with frequency-independent shading achieved using an autotransformer on transmission and a series of separate amplifier stages on reception. The receiver uses a series of low-noise pre-amplifiers in an underwater housing behind the transducer with a summing point amplifier to define the beam. The received signal is processed through an eight-channel bandpass filter and a 100 khz sampling module with a 12.5 khz sample rate per channel. Control of the transmitter and receiver is computerized. Measurements on caged aggregations of fish are reported. The fish were placed in a 2-m-diameter cage and observed over several days with both the wideband sounder and stereo 35-mm still cameras to determine spatial distributions and tilt angles. Data for different species are presented, showing frequency-dependent differences in the reflectivity in the range 27 to 54 khz. These data are related to the behaviour and information obtained from the photographic system. The differences among species in spectral reflectivity are discussed and the implications for species identification considered. E. J. Simmonds and F. Armstrong: DAFS Marine Laboratory, P. O. Box 101, Victoria Road, Aberdeen A B 98D B, Scotland. Introduction Using echo integration to estimate the abundance of pelagic fish stocks can provide estimates of biomass or the numbers of fish. This method assumes that the average fish backscattering strength or target strength (TS) is known. Present procedures require that the TS does not depend significantly upon fish behaviour and that echoes from species of interest can easily be distinguished from those of lesser importance. The identification is usually done either by visual inspection of echo traces or by partitioning of the echo-integrator data in proportion to the composition of trawl catches. The target strength is usually assumed to be constant, or perhaps dependent on fish length but independent of fish behaviour. Unfortunately, this is probably too simple a model. Fish are known to have highly variable backscattering strengths which depend upon tilt angle (Nakken and Olsen, 1977; Foote, 1985). Moreover, the size of the gas-filled swimbladder and its target strength will change with the depth of the fish (Sand and Hawkins, 1974; Blaxter et al., 1979). At normal survey echosounder frequencies the acoustic wavelength is generally of the same order as some of the dimensions of the fish. It is therefore expected that there will be some frequency dependence in backscattering strength resulting from changes in both tilt angle and swimbladder size. By increasing the system bandwidth, it might be possible to use frequency-response information to help identify species (i.e., swimbladder size), or smooth out differences in backscattering strength due to fish-orientation changes. A wideband echo sounder was produced to investigate these possibilities. The main measure of a wideband system is the extent to which the transducer beamshape both on transmission and on reception remains constant and independent of frequency, thus ensuring that the sampled volume is not frequency dependent. In addition, there is a need to keep the sample volume size similar to that used in present surveys so that data collected from such a system can be used directly with our present knowledge of 381

2 fish reflectivity. Thus a wideband system operating around the most commonly used frequency of 38 khz is desirable. Several approaches have been suggested and tried using non-linear acoustics for transmission and/or reception and twisted arrays for reception (Berktay et al., 1968; Berktay and Learly. 1974). The technique we have chosen is suggested by Rodgers and van Buren (1978) and uses a partial-sphere transducer which provides constant beamwidth on transmission and reception in a single unit. The transducer is constructed from elements arranged in concentric rings on the surface of a partial sphere. The rings are amplitude shaded on transmission and reception to form the required beamshape. The system has been designed with a multifrequency and swept-frequency transmitter and multi-channel variable-gain receiver. The equipment is described in this paper along with measurements on cod, saithe, herring, and mackerel. Methods W id eb an d echo so u n d er The transducer (Fig. 1) was constructed from 217 elements arranged in eight concentric rings around a central element, on the surface of a partial sphere. The individual transducer elements were of a pre-stressed sandwich resonator type with 20-mm-diameter aluminium head and 10-mm steel tail mass. The case was machined from a solid block of the nylon derivative, nylatron GSM. The active face of the transducer subtends an angle of 28.3 and has a radius of curvature of 86 cm, giving a nominal beamwidth of An underwater housing containing transmit and receive shading compo nents is situated on the back of the transducer. The constant beamwidth was formed using frequency-independent amplitude shading on both transmission and reception. The shading functions for transmission and reception are the same (Fig. 2). On transmission, the voltage shading was produced from an auto-transformer with nine taps, one for each ring of the transducer. On reception the same shading function, which depends upon the number of elements and was therefore different from ring to ring, was achieved by using a single low-noise operational amplifier per ring with a summing point amplifier to form the beam. Both the receive and transmit shading components were contained in the housing on the back of the transducer, along with mercury reed relays which are used for transmit receive switching. The transducer thus had a single transmit drive, and a single receive line along with transmit receive switching and receiver power supplies. The transmitter consisted of a modulator providing V2, 1, 2, 4, and 8 millisecond pulses. For the present experiments a 1-ms pulse was used with swept frequencies from 27 to 54 khz. The power stage was a 1 kw mos-fet amplifier which operated up to 100 khz (Fig. 2). After the beam forming the receive had a wideband filter and a 50-ohm line driver to match the cable. The signal was then connected to eight frequency channels, each with a five-resonator filter giving 3.3 khz bandwidth between 3 db points, followed by two gain stages with computer control of gain to match the differences in transducer sensitivity with frequency. Each frequency channel had a separate detector, and the outputs were multiplexed and sampled at 100 khz with 12-bit binary precision giving a 12.5 khz sample rate per channel (80 [as interval). The transmission rate was set at 300 ms, giving 1000 transmissions in 5 minutes, with one minute for data analysis per block. Figure 1. Transducer, 217 elements on the surface of a partial sphere. E x p erim en tal rig Measurements of the transducer beamshape and the spectral reflectivity of fish were carried out at the M a rine Laboratory s Loch Duich field station, on the west coast of Scotland. Fish were placed in a cage 2 m in diameter and 1 m deep. The cage was placed between two aluminium frames supporting stereo 35-mm still cameras and low light TV cameras below the cage. The complete rig (Fig. 3) was suspended from a raft, the transducer 15 m below the surface in a motorized gimbal table and a 38.1-mm-diameter tungsten carbide reference target positioned 10 m deeper. The cage support frame was 12 m from the transducer, which placed the fish at a range of 14 m from the transducer. This arrangement allowed the ball to be used as a reference target, with transducer position adjusted for maximum echo strength and the fish located in the region of the centre of the beam. Photographs were taken throughout the experiments 382

3 Q UJ ZüE < D m Q u j UJ «o. o z o ï<2 > o n id s to X CL «5 u j o K g K -J < < o x cc > w in co ^ ^ ( O N N r - r - S O O l C O N S l f l ^ N O cc ut UJ u. 8 ; Ol Û* z > CO Z UJ < o X uj O E Figure 2. System block diagrai»- O < Œ -J < 3 O D

4 - 1 6 E quivalent Beam Angle E.B.A. against Theory 54kH z 12 m wires wires reference target 27 54kH z experimental (2m dia. 1m /» m wires 1 E.B.A. through the C age -20 ; ^ ~ 54kH z benthos. cameras TV camera Figure 4. Transducer equivalent beam angle (db) from 27 to 54 khz. measured values replotted against the theoretical value and for the part of the beam that passes through the fish cage. experiments the fish were removed from surface holding pens and placed in the experimental cage, lowered to a depth of 29 m, and left for a period of days. Table 1. Numbers of fish, mean length, standard deviation, and mean weight for eight experiments. Figure 3. Experimental rig. Species Number L (cm) LSD (cm) W (g) to record the fish distribution and behaviour within the cage. Results were collected in 6 min or 1000 transmission blocks. The beam patterns were measured by removing the cage and camera frames and recording the echo from the reference target only. The transducer was steered through a range of angular locations using the motorized gimbal table. The method of obtaining the equivalent beam angle is described in Simmonds (1984). B eam -angle m e asurem ents The measured beam angles are shown in Figure 4. This has been replotted showing deviation from the theoretically predicted sample volume, and for the section of the beam which passes through the experimental fish cage. Fish frequency-response m easu rem en ts Eight experiments were carried out, two each with cod, saithe, herring, and mackerel (Table 1). For all the Cod Cod Saithe Saithe Herring Herring Mackerel Mackerel D a ta analysis The sampled data from the received echoes were summed to give integrals for the reference target and for the fish for each frequency for each transmission. First the individual integrals, from the fish, for each frequency from each transmission were examined. For correlation between frequencies for 1000 transmissions: 2 (Iti Ii)(Itj Ij)/ X V'[(Iti- ï i)2(iij Ij)2] i t where 1 and Itj are the echo integrals from transmission t for frequencies i and j respectively, and ij and Ij are the mean values for 1000 transmissions. 384

5 For the correlation between adjacent transmissions for the same frequency: Herring where I(1_u)i is the transmission integral shifted by u transmissions from transmission t. To examine the frequency-response data all the experiments were treated in a similar manner. The small contribution produced by the cage was subtracted from the integrals. The eight frequency channels were corrected for different sensitivity by dividing the fish integrals by the integral from the reference target and multiplying by its calculated acoustic cross-section (Fig. 5) and then dividing by the equivalent beam angle through the cage shown in Figure 4. The resulting eight average integrals from 1000 transmissions were then divided by their geometric mean to provide an eightpoint spectrum. This procedure provides a spectrum for each six-minute period which was then used to define average spectra (Fig. 6). To establish the accuracy of the individual independent sample spectra as a method of determining species, groups of 1000 and 100 samples were compared with the four average spectra using a least-squares fitting procedure, weighted inversely by the standard deviation of each point in the average spectra. Examination of the cod and saithe data showed changes of spectra during the experiments. To improve the discrimination between these species, the data from the four experiments on gadoids were used to provide separate average spectra for unacclimatized and acclimatized fish. These two stages were defined as the first 24 hours and the period after 48 hours of the experiment respectively; the four spectra are shown in Figure 7. The least-squares fitting procedure used was (for four species): minimum 2 ( x i ~ aji)2/sji where x, is the mean sample value for frequency i, ajj is the mean value for species j frequency i, Sji is the standard deviation of a^. *-2 Saithe Mackerel Figure 6. Average spectra ±1 standard deviation for four species: herring, saithe, cod, and mackerel. Results C o rrelation With the exception of the cod experiment with only 17 fish, there was no significant correlation between adjacent frequency samples that would have confirmed the possibility of phase relationships between multiple targets dominating the single independent sample results. In addition, the transmission-to-transmission correlation at the same frequency was not significant in any experiment. This allows each integral from each transmission to be regarded as an independent sample. C alculated T arget Response Figure 5. Frequency response (TS db) of 38.1-mm tungsten carbide ball from 27 to 54 khz. M easured targ et strength The mackerel showed no significant change in target strength over time, and the herring showed a small decrease in target strength, combined with a diurnal variation of about 6 db, over the period of observation. Both these species were measured over a period of four days. The saithe and cod slowly adapted to depth over the first two days. During this time the target strength rose 25 R apports et Procès-Verbaux 385

6 Saithe before acclimatization Table 2. Percentage species recognition rates for four species categories using average spectra from all experiments: (a) for 1000 samples; (b) for 100 samples; and (c) for 100 samples with six categories, with cod and saithe separated into acclimatized and unacclimatized states. Species recognized Actual species Herring Saithe Cod Mackerel Cod before acclimatization (a) 1000 samples Herring Saithe Cod Mackerel Saithe after aclimatization (b) 100 samples Herring Saithe Cod Mackerel (c) 100 samples with six categories Herring Saithe Cod Mackerel Cod after acclimatization L-2 Figure 7. Average spectra ± 1 standard deviation for cod and saithe unacclimatized and acclimatized to depth. steadily and maintained a more constant value after acclimatization. These species were observed for sixday periods. For herring and to some extent for cod and saithe, differences in the absolute level of acoustic reflectivity were observed between day and night. However, no real differences were found in the spectra at different light levels, and the data for day and night were combined. In all respects the absolute levels of backscattering strength followed the pattern of those reported in Forbes et al.. (1983). Species discrim ination The results for the four species from all eight experiments are shown in Table 2. These show that it is possible to distinguish herring from mackerel and both of these from both gadoid species to better than 95% confidence using 1000 independent samples. However, discrimination between saithe and cod is only at the 90% level. By reducing the number of samples to 100 the discrimination rate drops to 90 % for herring, mackerel, and the gadoids, with the differentiation between cod and saithe dropping to 72%. To improve the discrimination between the gadoids, the saithe and cod spectra were replaced by four new average spectra, two for each species: the first for the initial 24 h of data (unacclimatized) and the second for all data after 48 h (acclimatized). This increase in possible categories has only a very slight effect; it improves the separation of cod and saithe data to 74 % and makes no significant difference to the discrimination of the gadoids from herring and mackerel. Conclusions We have demonstrated that it is possible to use acoustic methods to identify caged herring and mackerel and the gadoid species cod and saithe with 90% confidence, using only 100 independent samples from an eight-channel 27 to 54 khz swept-frequency sounder. This confidence level exceeds 95 % if 1000 samples are taken. The discrimination is independent of day or night, and in the case of the gadoids, state of acclimatization. The repeatability between experiments and over time, despite substantial changes in backscattering strength, indicates a high probability that fish in the wild will also exhibit distinct and repeatable spectra. These may not, however, correspond to the spectra shown here. The requirement for only 1000 samples, equivalent to a 10-mhigh by 45-m-wide shoal observed from a vessel travelling at 10 knots, for 90 % discrimination means that this could be of value as a method of species identification where fish shoals of different species occur within a small area. 386

7 Acknowledgements We would like to thank D. N. MaeLennan for providing the frequency response of the tungsten carbide ball, S.T. Forbes for his assistance in developing the wideband system, and P. J. Copland, I. B. Petrie, and T. V. Taylor for assistance with the experimental work and construction of the system. References Berktay, H.O., Dunn, J.R.. and Grazey, B.K Constant beamwidth transducers for use in sonars with very wide frequency bandwidths. Appl. Acoust., 1: Berktay, H. O., and Learly, D. J Far field performance of parametric transducers. J. acoust. Soc. Am., 55(3): Blaxter, J. H. S., Denton, E. J., and Gray, J. A. B The herring swimbladder as gas reservoir for the acoustics-iateralis system. J. mar. Biol. Ass. UK, 59: Foote, K. G Rather-high-frequency sound scattering by swimbladdered fish. J. acoust. Soc. Am., 78(2): Forbes, S. T., Simmonds, E. 1., and Edwards, J. I Target strength of gadoids. ICES/FAO Symposium on Fisheries Acoustics, Bergen, Norway, June Nakken, O., and Olsen, K Target strength measurements of fish. Rapp. P.-v. Réun. Cons. int. Explor. Mer, 170: Rodgers, P. H., and van Buren, A. L A new approach to constant beamwidth transducers. J. acoust. Soc. Am., 64(1): Sand. O., and Hawkins, A. D Measurements of swimbladder volume and pressure in cod. Nor. J. Zool., 22: Simmonds, E.J A comparison between measured and theoretical equivalent beam angles for seven similar transducers. J. Sound Vib., 97(1): Simmonds, E.J., and Copland. P. J A wide band constant beam width echo sounder for fish abundance estimation. In Proceedings of the Institute of Acoustics, Salford, 1986, pp * 387