Dual-beam echo integration method for precise acoustic surveys

ICES Journal of Marine Science, 53: 351 358. 1996 Dual-beam echo integration method for precise acoustic surveys Yoshimi Takao and Masahiko Furusawa Takao, Y. and Furusawa, M. 1996. Dual-beam echo integration method for precise acoustic surveys. ICES Journal of Marine Science, 53: 351 358. The echo integration method has been widely and effectively used for acoustic surveys of fisheries resources. In recent quantitative echo-sounders the dual-beam or splitbeam method has been adopted, which improves the accuracy of in situ target strength (TS) measurement. The dual-beam method utilises a coaxial narrow and wide-beam pair. In ordinary dual-beam systems, echo integration is performed only for the narrow-beam signal, not for the wide-beam signal. If echo integration is also performed for the wide-beam signal, more information and better accuracy and precision of the acoustic survey of fish abundance can be gained. Comparing the integrator outputs of both beams, an index of avoidance behaviour of fish towards the surveying vessel can be compiled and unreliable measurements caused by, for example, noise contamination and failure in bottom detection can be found. Wide-beam integration is of better use when distributions of fish are shallow and sparse, because the sampling volume is greater. Furthermore, the error caused by transducer motion in bad weather conditions is smaller for the wide beam. By combining both beam echoes, the volume backscattering strength within individual fish schools can be measured precisely. The method is discussed theoretically. Experimental investigations were carried out on walleye pollock (Theragra chalcogramma (Pallas)), in the Bering Sea by our quantitative echo-sounding system, which provides independent echo integration results for both narrow and wide beams. 1996 International Council for the Exploration of the Sea Key words: acoustic survey, avoidance behaviour, dual-beam method, echo integration. Y. Takao and M. Furusawa: National Research Institute of Fisheries Engineering, Ebidai Hasaki, Kashima Ibaraki, 314-4 Japan. Correspondence to Takao [tel: +81 479 44 5949, fax: +81 479 44 6221]. Introduction Echo-integration, the most important acoustic method for fisheries resource surveys, has been used widely and effectively (Johannesson and Mitson, 1983). In recent quantitative echo-sounders the dual-beam or split-beam method has been adopted to facilitate accurate observation of in situ fish target strength (TS) (Ehrenberg, 1983). The dual-beam method utilises coaxial narrow and wide beams. Pulses are transmitted on narrow beams and echoes are received on both narrow and wide beams. This method corrects for the directivity pattern by using both beam echoes from the same fish, thus achieving an accurate TS value. In ordinary dual-beam systems, echo integration is performed only for the narrow beam and not for the wide beam. We believe that if echo integration is also performed for the wide-beam signal and the results compared with the narrow beam signal, more information can be gathered and the accuracy and precision of the acoustic survey of fish abundance improved. We first discuss the theory of this method and then examine the data obtained in acoustic surveys of walleye pollock (Theragra chalcogramma (Pallas)), focusing on detecting situations where the fish avoid the research vessel. Theory Echo intensity The echo pressure P r of multiple echoes by N fish at the transducer surface is at the time corresponding to a range r, where P is the source pressure, r is the reference range 1 m, α is the 154 3139/96/2351+8 $18./ 1996 International Council for the Exploration of the Sea on 11 February 218

352 Y. Takao and M. Furusawa pressure based absorption coefficient in nepers per unit distance, D i =D(θ i, i) is the directional response in the direction of the i-th fish which is located at spherical coordinates (r, θ i, i), and σ bsi is the backscattering cross-section of the i-th fish (Clay and Medwin, 1977). Equation (1) can be rewritten as: where n, Ω, c, and τrepresent volume density, distribution range of fish in solid angle, sound speed, and pulse duration, respectively. If fish distribute homogeneously in space (Assumption 1), we have where D N 4 for the narrow channel and by D N 2 D W 2 for the wide channel where the suffixes N and W mean narrow and wide, respectively. The avoidance may cause a reduction in density (n in Equation (9) changes) and/or an extreme orientation distribution for fish near the beam axis in near range (σ bs in Equation (9) changes). Therefore, SV for the narrow channel will be sensitive to the avoidance reaction, because the weight of echoes near the beam axis is larger for the narrow channel. Thus, the existence or magnitude of the avoidance effect can be measured by the ratio of measured SV values by narrow and wide channels. Noise Here, we discuss noise received on the transducer, excluding reverberation. The noise pressure at the transducer surface P n is amplified and integrated to give so-called noise SV (Takao and Furusawa, 1995) expressed as: and n is the average volume density. The angle Ψ is called the equivalent beam angle and <σ bs > is the apparent average backscattering cross-section (Furusawa et al., 1986). In Equation (5), σ bs (θ, ) is generally a function of θ and and is an averaged value with respect to orientation distribution. If the orientation distribution is not dependent on fish position with respect to beam axis (Assumption 2) and beam width is small, then we can separate the directivity and backscattering cross-section terms as where σ bs is the average backscattering cross-section in Ω and η is the contribution factor (Furusawa et al., 1986). If the size of the fish school is large compared with the beam width (Assumption 3) or the integration is done for many pings, Ω will be equal to 2π and we get Avoidance effect In the general situation where Assumptions 1 and 2 are not valid, measured volume backscattering strength (SV) at range r from Equations (2) and (3) is: where r w is the width of the integration layer and P n is assumed constant in r w. This equation shows that noise SV increases with the range r, the error caused by noise thus becoming more serious when fish are deep. If we assume that (1) the receiver bandwith is small and the noise spectrum is flat in the band, and (2) the noise is received omni-directionally, the P n are shown as: where N P is the noise spectrum level, Δf is the receiver bandwidth, and D I is the directivity index of the receiving transducer. Then the ratio of the noise SV of narrow to wide channels is expressed as: where Ψ C is the equivalent beam angle for the composite narrow and wide beam. The parameters on the right side of Equation (12) give a constant value for the noise SV ratio. If noise is directional, for example propeller noise, the value of the ratio changes, thus giving information on the noise characteristics. Sampling volume Bodholt (1977) showed the following expression for the variance of the normalized integrator outputs: where Sv is the linear value of SV. In the dual-beam echo integration method, D 4 of Equation (9) is replaced by on 11 February 218

Dual-beam echo integration 353 Range (m) 1 1 2 224 223 Time Figure 1. An example echogram made on 16 September 1989. The upper and lower windows show the narrow and wide 2 log r TVG outputs, respectively, original in colour. where m is the ping numbers for averaging, and <óbs2 >/ <óbs > 2 is the moment ratio of backscattering crosssection which is generally about 2.7. The moment ratio for the transducer directivity Ø4/Ø is about.5 where Ø4 =8 D8 dù. The second term in the bracket is larger for the narrow channel than for the wide channel by a factor of ØC/ØN. If the depth r is shallow and the weight of the second term becomes large, and if a large on 11 February 218

354 Y. Takao and M. Furusawa integration period is not applied (m is small), then the wide channel will give more precise results. Transducer motion error The dual-beam method needs a narrow beam in order to resolve single echoes. This produces large errors in echo integration when the transducer is unstable. Stanton (1982) analysed the transducer motion error in the echo integration method. The transducer motion gives a negative bias due to reduction of the apparent equivalent beam angle. Therefore, the wide-beam echo integration is preferable in bad weather conditions, particularly for vessel mounted transducers. Narrow SA (db) 3 4 5 6 7 8 Observation of volume backscattering strength within schools The volume backscattering strength within a school can sometimes be important information; for example, for fisheries and species identification purposes. It is not easy to measure SV in schools accurately, because (1) the beam must be sharp compared with school size (Assumption 3; see Equation (7)) and (2) the echo must be formed by a sufficient number of fish echoes if precise SV measurements are to be made (see Equation (13)). The first point suggests that the narrow beam is preferable in lessening the bias (1-η), while the second point suggests that the wide-beam is better in lessening variance (V 2 ). One practical way of alleviating the above contradiction is to measure SV in the wide channel, where the difference between SV in the two channels is small. Experiment Methods We examined several simultaneous narrow and wide SV measurements using our quantitative echo-sounding system (Furusawa et al., 1993). The acoustic frequency is 38 khz. The equivalent beam angle of the narrow and wide channels is.72 sr and.12 sr, respectively. Thus, the reverberation volume of the wide channel is about 1.6 times that of the narrow channel. This difference is not very large, but any abnormal situation, such as strong avoidance, will be detectable. The transducer is mounted on a towed body to reduce the error caused by transducer motion, bubbles, and noise. The echointegrator has two independent channels to perform integration on both narrow and wide channels. The CEIP (colour echo image processor) displays absolute echo levels, and also special functions combining narrow and wide TVG outputs. For example, SVB mode shows the differences between the narrow and wide SV. We can use this information effectively in the survey. 9 We have been conducting acoustic surveys for walleye pollock in the Aleutian Basin and the eastern Bering Sea shelf since 1988. We calibrated our system using a copper sphere (Foote, 1982) several times during each survey we measured the noise level received to determine the characteristics of the noise and to set appropriate threshold parameters on the echo-integrator. The dependence of the noise on engine speed or ship speed was also measured before and after each survey (Takao and Furusawa, 1995). Results and discussion 8 7 6 5 4 Wide SA (db) 3 Figure 2. Comparison of narrow and wide SA values. Black and white dots are SA values observed on 16 September 1989 and 17 September 1989 respectively. We can assume that the shallower the fish depth, the greater the fish avoidance reaction against the research vessel (Olsen et al., 1983). Therefore, we selected data sets collected on the shelf area for the analysis of avoidance effect. Figure 1 presents an example of echograms of walleye pollock observed on the eastern Bering Sea shelf from 2221 h to 2245 h on 16 September 1989. The water depth was about 15 m and the noise level was low. We observed a patchy distribution of fish between the surface and 6 m depth, and a layer-like distribution between 8 m and the bottom. Frequently we observed fish distributions similar to those in Figure 1 on the same day. The maximum range of the echo-sounder was 2 m and the integration period was 4 min (the ping rate was about 1 min 1 ). Narrow and wide SA values summed over the depth range of 14 m to the bottom are indicated in Figure 2 by black dots. These values were observed from 228 h to 2253 h. The total number of SA measurements was 37. The ratio of narrow to wide on 11 February 218

Dual-beam echo integration 355 Average SV (db) 1 9 8 7 6 5 Narrow SV Wide SV (db) 2 1 1 2 1 2 3 Range (m) 4 5 6 7 8 9 (a) 1 2 3 4 Coef. of variation 5 6 Figure 3. An example of dual-beam echo integration on the eastern Bering Sea shelf. (a) Average SV and its coefficient of variation (CV) as a function of depth. A bold-solid line and a fine-solid line show narrow and wide SV, respectively. A dashed line and a dotted line show narrow and wide CV, respectively. (b) Differences between narrow and wide SV for each integration layer (short vertical bar) as a function of depth with horizontal line showing the range. A bold line represents the average difference. (b) SA was almost unity. The above result suggests that the fish did not avoid the survey vessel. Figure 3a shows the average SV values for each integration layer (4 m width) and their coefficients of variation against range from the transducer obtained for the same data set as the black dots in Figure 2. The average SV curves show that the fish distribution was denser near the bottom than in midwater. There are four peaks of the coefficient of variation (at 1, 32, 58, and 78 m), which seem to coincide with fish layer boundaries. Figure 3b shows the differences between narrow and wide SV plotted against range from the transducer. Horizontal lines show the extent of the differences. Near the bottom, the variability in the SV differences decreases as the average SV value increases. A bold line represents the average difference which was small. Next, we show an example of dual-beam echo integration applied to a shallower fish distribution. Figure 4 is an echogram observed on the eastern Bering Sea shelf from 24 h to 38 h on 17 September 1989. We took a sample by midwater trawl between the observations of Figures 1 and 4 and caught juvenile walleye pollock. The fish signal near the bottom is weaker than that in Figure 1 and the density is less. However, the patchy signal near the surface is bigger and stronger than that indicated in Figure 1. The maximum range of the echo-sounder was 1 m and the integration period was 4 min (the ping rate is 17 min 1 ). The white dots in Figure 2 indicate SA based on the data from both channels from 36 to 441 on the same day. The integration was carried out from 12 m depth to the bottom. The total number of SA data was 62. The difference between narrow and wide SA is larger compared to the black dots. Figure 5a shows the average SV values of each layer (2m width) and their coefficient of variation against range observed for the same data set as the white dots of Figure 2. It can be seen that SV levels decrease and the coefficient of variation increases with range. The coefficient of variation of the narrow SV is larger than that of the wide SV. Figure 5b shows the differences between narrow and wide SV values against range. The wide SV was larger than the narrow SV at shallow depth. Since the contribution of high-frequency noise in echointegrator outputs was low in this case, because of the on 11 February 218

356 Y. Takao and M. Furusawa 5 Range (m) 5 1 25 3 35 Time Figure 4. An example echogram made on 17 September 1989, original in colour. narrow range (see Equation (1)), we can ignore the influence of noise on the difference between narrow and wide SV. The towed body that we used was stable during these surveys, so that transducer motion errors would be small. We therefore consider that the observed difference between wide and narrow SV is caused by fish avoidance, because the fish distribution was shallow and patchy. on 11 February 218

Dual-beam echo integration 357 Average SV (db) 1 9 8 7 6 5 Narrow SV Wide SV (db) 2 1 1 2 1 2 3 Range (m) 4 5 6 7 8 9 (a) 1 2 3 4 Coef. of variation 5 6 Figure 5. An example of dual-beam echo integration of a patchy fish distribution. (a) Average SV and its coefficient of variation (CV) as a function of depth. A bold-solid line and a fine-solid line indicate narrow and wide SV, respectively. A dashed line and a dotted line indicate narrow and wide CV, respectively. (b) Differences between narrow and wide SV for each integration layer (short vertical bar) as a function of depth with a horizontal line showing the range. A bold line represents the average difference. (b) Here we discuss the noise SV ratio. The theoretical value of Equation (12) is about 5 db for our system. In most cases, this value concurs with the experimental data when the noise level is low (vessel drifting or sailing at low speed). The narrow noise SV was more sensitive to increasing ship or engine speed. For more details, see Takao and Furusawa (1995). We have used the ratio of narrow to wide SV effectively during survey operations. When sufficient fish echoes under good conditions are observed, this ratio is almost unity. When large differences between narrow and wide SV are noticed, we examine the echograms. If noise appears to be the cause, we measure noise levels. When the noise level is not so high, suitable threshold parameters are set. When the noise level is relatively high or the causes are not solely noise, we sail more slowly. This is an effective way of reducing the problems of fish avoidance, noise, and transducer motion. We are also using this ratio in post-processing of the echo-integration data, where we can readily find unreliable measurements in a huge data set by automatic means. Conclusion We have discussed some of the advantages of the dual-beam echointegration method. More information can be gained and the accuracy and precision of the acoustic estimates of fish abundance can be improved. By comparing the narrow and wide channel outputs of the echo integrator we can: (1) compile an index of avoidance behaviour of fish to a surveying vessel, and (2) readily find unreliable measurements caused by, for example, noise, contamination, and failure in bottom detection. Wide-beam echo integration is advantageous because: (3) the variance of the integrator output of the wide channel can be smaller than that of the narrow channel for shallow and sparse distributions of fish, and (4) the transducer motion error in bad weather conditions is smaller for the wide beam. Comparing echo levels from both beams, we can: (5) provide an accurate measure of volume backscattering strength (SV) within individual fish schools. It is theoretically and experimentally shown that the ratio of narrow to wide SV is unity under good on 11 February 218

358 Y. Takao and M. Furusawa conditions. This SV ratio is an effective index for ensuring the high quality of data during surveys and also for finding unreliable data in post-processing. The dual-beam echo-integration method should be preferred in view of the above-mentioned advantages. The most useful ability is in monitoring the avoidance effect, which could not be accomplished with single-beam instruments. Acknowledgements We thank Kouichi Sawada and Yoichi Miyanohana, National Research Institute of Fisheries Engineering, Tatsuya Hoshou Japan NUS Co., for their support during this study. We are grateful to Neal Williamson, Alaska Fisheries Science Center, NOAA, USA, for his advice and for reviewing the manuscript. Thanks are extended to the staff of Kaijo Co. and the Japan Radio Co., for their design and manufacture of our echosounding system. References Bodholt, H. 1977. Variance error in echo integrator output. Rapports et Procès-Verbaux des Réunions du Conseil International pour l Exploration de la Mer, 17: 196 24. Clay, C. S. and Medwin, H. 1977. Acoustical oceanography. Wiley, New York. 544 pp. Ehrenberg, J. E. 1983. A review of in situ target strength estimation techniques. FAO Fisheries Report, 3: 85 9. Foote, K. G. 1982. Optimizing copper spheres for precision calibration of hydroacoustic equipment. Journal of the Acoustical Society of America, 71: 742 747. Furusawa, M., Ishii, K., and Maniwa, Y. 1986. A theoretical investigation on ultrasonic echo methods to estimate distribution density of fish. Journal of the Acoustical Society of Japan, 42: 2 8 (in Japanese). Furusawa, M., Takao, Y., Sawada, K., Okubo, T., and Yamatani, K. 1993. Versatile echo sounding system using dual beam. Nippon Suisan Gakkaishi, 59: 967 98. Johannesson, K. A. and Mitson, R. B. 1983. Fisheries acoustics: a practical manual for aquatic biomass estimation. FAO Fisheries Technical Paper No. 24. 249 pp. Olsen, K., Angell, J., and Løvik, A. 1983. Quantitative estimations of the influence of fish behaviour on acoustically determined fish abundance. FAO Fisheries Report, 3: 139 149. Stanton, T. K. 1982. Effects of transducer motion on echointegration techniques. Journal of the Acoustical Society of America, 72: 947 949. Takao, Y. and Furusawa, M. 1995. Noise measurement by echo integrator. Fisheries Science, 61: 637 64. on 11 February 218