Target Strength Estimates Research Articles

Inverse problem solution techniques as applied to indirect in situ estimation of fish target strength

In situ indirect methods of fish target strength (TS) estimation are analyzed in terms of the inverse techniques recently applied to the problem in question. The solution of this problem requires finding the unknown probability density function (pdf) of fish target strength from acoustic echoes, which can be estimated by solving the integral equation, relating pdf's of echo variable, target strength, and beam pattern of the echosounder transducer. In the first part of the paper the review of existing indirect in situ TS-estimation methods is presented. The second part introduces the novel TS-estimation methods, viz.: Expectation, Maximization, and Smoothing (EMS), Windowed Singular Value Decomposition (WSVD), Regularization and Wavelet Decomposition, which are compared using simulations as well as actual data from acoustic surveys. The survey data, acquired by the dual-beam digital echosounder, were thoroughly analyzed by numerical algorithms and the target strength and acoustical backscattering length pdf's estimates were calculated from fish echoes received in the narrow beam channel of the echosounder. Simultaneously, the estimates obtained directly from the dual-beam system were used as a reference for comparison of the estimates calculated by the newly introduced inverse techniques. The TS estimates analyzed in the paper are superior to those obtained from deconvolution or other conventional techniques, as the newly introduced methods partly avoid the problem of ill-conditioned equations and matrix inversion.

Target strength of an oily deep-water fish, orange roughy (Hoplostethus atlanticus) I. Experiments

Target strengths (TS) of orange roughy (Hoplostethus atlanticus) are difficult to measure in situ because this species lives at 800–1200 m, has low TS at 38 kHz, shows strong avoidance behavior, and sometimes co-occurs with species having much higher TS. Orange roughy have wax-invested swimbladders and fish are ∼20% wax esters by weight. Estimates of orange roughy near-dorsal aspect tilt-averaged target strength (〈TS〉) are controversial, varying from [−39, −35] dB to [−52, −48] dB for a 35-cm fish in different studies. 〈TS〉 is estimated using experiments on both live and dead fish, and new swimming orientation data. Target strength of previously frozen and thawed fish decline rapidly, stabilizing after ∼12 h as air and oil diffuse out of tissues during rehydration. A 〈TS〉 of −46.3 dB is predicted for a 35-cm live orange roughy (bootstrapped 95% confidence intervals=−48.0 to −45.5 dB). This value is 3.7 dB higher than Australian estimates of ≈−50 dB for 〈TS〉 of a 35-cm live orange roughy. It is shown that the magnitude of the difference between in situ estimates of 〈TS〉 and experimental measurements of 〈TS〉 on live fish can be explained by the effect of strong avoidance reactions to a towed transducer on the tilt angle of the fish. Previously much higher estimates of dead orange roughy 〈TS〉 were an artifact of air pockets caused by freezing. Comparison of orange roughy TSmax measured in our experiments with a large dataset for nonswimbladder and swimbladder fish shows that orange roughy have an unusually low target strength at 38 kHz.

Target localization for a three-dimensional multibeam sonar imaging system

One advantage of three-dimensional multibeam imaging systems is that they have the ability to image and track many targets concurrently in their field of view. In this regard, the accurate estimation of both position and target strength are important data which can be derived from the echo responses of an active multibeam imaging system. This article considers an algorithm which can be used to yield more accurate positional estimates for targets than available from the preformed beams of a given multibeam system: FishTV [, Deep-Sea Res. 42, 1495 (1995)]. The method refines a least squares residual of the predicted data versus observed data by adjusting the parameters of a model. The parameters are the position and target strength of the target. It is demonstrated that this procedure is a maximum likelihood estimator for position and target strength in high signal-to-noise. In addition, the shape of the least squares surfaces indicates that the algorithm is somewhat robust with respect to noise.

Acoustic characterization of Mysis relicta

It is possible to increase the sampling resolution and the available information on zooplankton populations by applying acoustic sampling techniques. Knowledge of an organism's acoustic target strength (TS) is critical for translating acoustic data into meaningful biological information, such as numerical abundance. We have taken several approaches to evaluate the acoustic TS of Mysis relicta, a key benthic‐pelagic invertebrate in the Laurentian Great Lakes. This included in situ TS measurements with a dual‐beam 420‐kHz system and comparison among net‐based, optical plankton counter (OPC)‐based, and acoustic‐based estimates of numerical abundance of mysids. In addition, based on the in situ TS data, we modified a model previously applied primarily to marine zooplankton taxa. The resulting predicted TS was compared to other TS estimates. The estimated mean TSs from the in situ TS measurements and the OPC‐acoustic comparison were within a narrow range (–76 to –74.8 dB). Analysis of the net‐acoustic data comparison resulted in an estimated average TS of –73.1 dB. We suggest that the net‐acoustic discrepancy relative to other methods is due to net avoidance and low filtering efficiency. The model predicted similar mean TSs for the size distribution present when R (reflectivity coefficient) was changed to 0.0675. We hypothesized that the modification of the model parameter values was necessary because of the differences in the medium surrounding marine vs. freshwater zooplankton.

In situacoustic target-strength measurement of bigeye (Thunnus obesus) and yellowfin tuna (Thunnus albacares) by coupling split-beam echosounder observations and sonic tracking

A research programme was carried out in French Polynesia to study tuna behaviour using acoustics and fishing experiments. Acoustics are of great importance for the study of tuna behaviour and estimation of abundance but estimates of individual target strength that are available are particularly inaccurate. In this study, four yellowfin tuna (Thunnus albacares) and two bigeye tuna (Thunnus obesus) of weight 4‐50 kg were individually caught, identified, and equipped with ultrasonic tags for telemetry experiments. While tracking the fish, simultaneous underwater acoustic data were recorded with a split-beam echosounder in order to estimate their in situ acoustic target strength. It was observed that target strength was stronger when fish were diving than when they were ascending toward the surface. This can be explained by the tilt angle orientation of the swimbladder. A target strength bias according to depth was also observed. ? 1999 International Council for the Exploration of the Sea

Comparison of Ayu target strength estimates determined from tank measurement and calculations

Ayu, Plecoglossus altivelis, is one of the most popular freshwater fish which can be seen in many rivers in Japan. Especially in the North Basin of Lake Biwa, Ayu is the highest predator. It is important to know the spatial and temporal variations of Ayu biomass to understand the predator’s effect to the food chain in the lake. Acoustic survey has been conducted since 1995. There is little information about the target strength (TS) of Ayu. Since Ayu has an open swimbladder, it is necessary to measure TS before the swimbladder was deflated by the suspension method in a tank. Target strength of live Ayu was measured in a tank and a three-dimensional image of swimbladder was obtained by an x-ray computer tomography (CT) system. In addition, a side and a dorsal image of the swimbladder were obtained by a soft x-ray system. All measurements were done in live condition. Kirchhoff approximation was used for the calculation on the basis of the three-dimensional image of swimbladder. A deformed cylinder model and a prolate spheroid model were also used for the calculation on the basis of side and dorsal image of the swimbladder. Calculation and measurement are compared and discussed in this paper.

Wavelet decompositions versus singular value decomposition (SVD) in a fish target strength estimation

The fish target strength estimation when using acoustic echoes from a single-beam echosounder is possible after solving ill-conditioned equations where probability density functions (PDFs) of echo level, target strength, and beam pattern are involved. The modified singular value decomposition (SVD) algorithms are typically used in such cases. Unfortunately, this technique often leads to artifactual modes in solution due to Fourier nature of underlying eigenfunctions. The expectation maximization smoothing (EMS) method is used to improve smoothness of solution but by the cost of increasing the time of computation of estimates. D. L. Donoho (1995) proposed the wavelet-vaguelette decomposition (WVD) as being well suited for solving ill-conditioned equations with inhomogeneous functions. This concept was extended to fish target strength estimation and obtained results were compared to SVD performances.

Laboratory target strength measurements of free-swimming Antarctic krill (Euphausia superba)

Target strength measurements of free-swimming krill at 120 kHz were made using a single-beam monostatic system in a 10-m3 laboratory tank. Krill (grouped according to length classes) swam freely in the tank triggering a data acquisition system when generating a backscattered signal larger than a threshold, determined by the system noise level. Dorsal and ventral target strength estimates were calculated indirectly by deconvolution of the cumulative probability function of echo ensembles of single-animal insonifications. For mean length classes in the range [29.6 to 36.2] mm the median single-animal target strengths are in the range [−76.7 to −71.8] dB. Monte Carlo computer simulations were used to evaluate the effects of varying the ratio of largest to smallest echo amplitudes for a given ensemble, thus enabling the estimation of threshold induced bias in the target strength estimates. The threshold induced bias was then determined for each ensemble of experimental data and used to determine corrections which were in the range of [−0.84 to −0.33] dB. An error analysis of the target strength estimates detailing the components due to measurement accuracy and precision, and the indirect signal processing techniques used is also presented.

Target strength of southern blue whiting (Micromesistius australis) using swimbladder modelling, split beam and deconvolution

The tilt-averaged target strength, , of southern blue whiting (Micromesistius australis) at 38 kHz was measured using swimbladder modelling and two in situ methods. Split beam estimates of for southern blue whiting were carefully screened for multiple echoes, but a few useful values of were obtained from the periphery of shoals even at 150-300 m depths. We compare of southern blue whiting derived from swimbladder modelling to split beam and deconvolution estimates of . The -length regression for M. australis has similar slope but lower intercept than the published regression for blue whiting Micromesistius poutassou. Predicted for M. australis are 2.9 to 4.3 dB lower than for M. poutassou of the same size. In contrast to the swimbladder results, split beam and deconvolution estimates of for M. australis were in line with the -length relationship for M. poutassou. The magnitude of the difference between modelling in situ results could arise from the assumed tilt distribution of fish used in the modelling calculations and the actual, but unknown tilt distribution of the fish in situ. Acquiring information on the tilt distributions of southern blue whiting is essential to resolve the measured difference between in situ and modelled estimates of target strength.

Estimation of the target strength of juvenile salmonids at any aspect

Many tasks, such as assessment of the state of stocks and observation of migrant behavior, elemental for recovery of Columbia River basin salmonid stocks, are currently performed using active acoustic methods. Detectability models are required to configure acquisition systems, design elements of studies, and process and analyze data. Accurate and precise estimates of juvenile salmonid target strength are necessary input to detectability models. The target strength at 420 kHz of 87 juvenile salmonids was measured within the length range of 45 to 300 mm. The fish were rotated through pitch, roll, and yaw planes. Forty target strength measurements were obtained for each degree of rotation in each plane for each fish. Mean target strengths were computed for 1° and 5° increments of rotation. The mean target strength estimates for each 5° increment of rotation for all fish were fit to the equation σ/λ2=a(L/λ)b. Estimates of mean target strength for regions of the sphere encompassing the fish between the planes of rotation were estimated by interpolation to complete a description of the target strength of juvenile salmon at any aspect.

Acoustic monitoring of salmonid density, target strength, and trajectories at two dams on the Columbia River, using a split-beam scanning system

Horizontal acoustic scanning was used at two hydropower dams on the Columbia River to monitor downstream migrating juvenile salmonids. Two split-beam systems with computer-controlled rotating capabilities monitored the fish movement patterns in the forebay areas of the Rocky Reach and Rock Island dams. Scanning consisted of 15 or 17 horizontal aiming angles, separated in 10° increments each monitored for 4 min, and tilted downwards from the horizontal at either 10° or 15°. Scanning was carried out for 24 h/day for periods of five days in May 1995 and was part of a larger study which lasted from 4–6 weeks. Estimates of mean fish density, target strength and trajectory were made for 5-m range cells within each horizontal scanning angle. Results show that fish movement patterns differed in the two dams. At Rocky Reach Dam, fish followed a circular trajectory pattern which was similar to the observed pattern of river flow and fish target strengths were higher in areas of low water velocity. At Rock Island Dam, fish behavior differed according to whether turbine units were operating or not. When turbines were in operation, fish moved towards the flow which contrasted with their milling behavior when the turbines were not operating. The fish target strengths were also greater in the latter situation. The difference in body aspect being insonified may account for this random acoustic aspect of milling fish compared with primarily head/dorsal aspect of fish oriented to the operational turbines.

Incoherent localization of targets for a three-dimensional multibeam sonar imaging system

One advantage that three-dimensional multibeam imaging systems have over more conventional systems is that they have the potential to track many targets that are concurrently in their field of view. As such, the accurate estimation of both position and target strength are important data which can be derived from the echo responses of an active multibeam imaging system. An algorithm has recently been developed which can be used to obtain higher positional estimates for targets than available from the preformed beams of a given multibeam system (Fish TV). The method refines a least-squares residual of the amplitudes of a set of predicted data versus a set of observed data by adjusting the parameters of a model. The parameters are the position and target strength of the target. The least-squares norm results in a maximum likelihood estimator for position and target strength in high signal-to-noise. Two fortunate aspects of the algorithm are that it takes into account the geometry of the sonar system and also uses only the amplitude data. The special collection geometry and acquisition method make these features important.

HOTWIT: An unconstrained shallow-water reverberation estimation algorithm

A new algorithm/method for the inversion of activated towed array reverberation data in shallow water, known as the heading optional transformation of Wagstaff’s iterative technique (HOTWIT), has been developed. By taking advantage of the activation and motion of the towed array, the algorithm resolves the inherent left–right ambiguity of the towed array without the need for multiple headings and is an active analog to the passive 3-D WIT algorithm. The result is an unambiguous estimate of bottom target strength from which the reverberation density at a given receiver location can be inferred. The algorithm lends itself well to a rapid characterization of shallow-water reverberation in shallow-water areas. A description of the algorithm with a discussion of results from the algorithm with measured and modeled towed array data will be discussed.

Potential improvements to current methods of recognizing single targets with a split-beam echo-sounder

The precision of in situ target-strength (TS) estimates generated by TS analysers is directly related to the ability of single-target detectors to reject multiple echoes in dense aggregations. The rejection performance of five single-target detection algorithms, utilizing phase, amplitude, and echo duration information from a split-beam echo-sounder as rejection criteria, were tested by simulation for target pairs randomly generated in a three-dimensional resolution volume. Each algorithm was also tested for selection bias. It was found that all five algorithms accept multiple echoes preferentially when they are in phase. Of the phase methods, that which used the standard deviation in phase rejected the multiples most effectively, while the echo duration method performed relatively poorly. A method which relies on the magnitude of amplitude differences between split-beam elements performed well, particularly for stronger targets. Another advantage of this method is that it also preferentially accepts out-of-phase multiples, thus partly compensating for its bias towards multiples which are in phase. We conclude that implementation of algorithms using either the standard phase deviation or the amplitude difference will lead to a significant improvement in single-target detection. The performance of the amplitude difference technique is particularly encouraging, considering the potential improvements possible by using averaging techniques to minimize the influence of noise.

Target-strength measurements of Pacific hake (Merluccius productus) in Canadian waters using quasi-ideal and conventional beam transducers

A joint Japanese/Canadian research cruise was carried out during May and June 1994 off the west coast of Canada to obtain two independent estimates of the in situ target strength (TS) of Pacific hake (Merluccius productus). Two vessels with independent quantitative echosounder systems were used. One was equipped with 120 and 50 kHz transducers, while the other used 38 kHz. The 120 kHz transducer featured a quasi-ideal beam pattern with a relatively flat response near the beam axis and low side lobes. Different single-beam TS estimation procedures were used to estimate TS from either system. Fish targets were identified by midwater trawling on the same or the next day. On one occasion both vessels were used to fish and measure TS simultaneously. The observed mean night-time TS for adult hake ranged from −36.4 to −30.4 dB kg −1 .

Target-strength estimates of schooling herring and mackerel using the comparison method

Misund,O.A.andBeltestad,A.K.1996.Target-strengthestimatesofschoolingherringandmackerelusingthecomparisonmethod.–ICESJournalofMarineScienceSymposia,53:281–284.Thetargetstrengths(TS)ofschoolingherringandmackerelhavebeenestimatedusingthecomparisonmethod;purse-seinecapturesofschoolsmeasuredbysonarandechointegration. For herring about 33cm long, the results indicate an average TS of40.8dB,givingb=71.1dBastheconstantintheusualTSequation(TS=20logL+b).Inthecaseofmackerelabout38cmlong,theaverageTSis54.9dB,whichcorrespondstob=86.5dB.Theresultsarediscussedinrelationtoother

Effect of Scuba Divers on Fish Density and Target Strength Estimates from Stationary Dual-Beam Hydroacoustics

The effect of the presence of scuba divers conducting visual point count surveys on the abundance and size distribution of fishes estimated with dual-beam hydroacoustics was examined at a petroleum platform in the northern Gulf of Mexico during November 1991 and February, March, and June 1992. Mean density of fishes declined 41–77% for all sampling periods when scuba divers were present. Target strengths also dropped significantly for all sampling periods when scuba divers were present. Our results show that fishes at petroleum platforms avoid scuba divers. Therefore, visual surveys of fish at petroleum platforms and other similar environments should be supplemented with hydroacoustic assessments.

Evidence of bias in estimates of target strength obtained with a split-beam echo-sounder

In addition to testing the echo duration, the Simrad EK500 split-beam echo-sounder uses the between-sample variation in signal phase to isolate echoes from single fish. The performance of these discriminators was tank-tested on a 38-kHz EK500 sounder, using spheres as targets. It was found that for typical settings of the allowable phase variation and echo duration: (1) there was a pronounced bias against accepting weaker targets; and (2) multiple echoes from targets as far as 700 mm apart were falsely accepted as single echoes, particularly if the individual echoes happened to interfere constructively. These effects will all result in target strengths being over-estimated, by an amount which could be large in many cases and which cannot be corrected for. We conclude that, except in cases where the fish are widely separated and differ comparatively little in target strength, the single fish discriminators used in the EK500 (software Version 4.01) are unreliable, and should be used with great circumspection. Possible improvements to the phase algorithm are suggested, although it is considered unlikely that these will eliminate all biases in the method.

Simple model on the relationship between fish acoustical target strength and aspect for high-frequency sonar in shallow waters

Summary The relationship between fish target strength and the orientation of fish in horizontal (yawing) and vertical (pitching) planes was modelled by the equation TS = m. cos 2 α+ b (also with cos3, cos5 and cos7). The cos3 model gave the best fit of data for three fish species: brown trout, roach and rudd. The model can be used for 1) estimation of the most probable aspect in the horizontal plane; 2) estimation of the mean all-aspect target strength of fish and 3) calculation of probability of recording of a fish of given size when a known noise threshold is applied.

ADCP measurements of the distribution and abundance of euphausiids near the Antarctic Peninsula in winter

Euphausiid populations were observed in Gerlache Strait, Antarctic Peninsula region, during austral winter 1992 (12 July–14 Aug) using a hull-mounted 153-kHz acoustic Doppler current profiler and a multiple opening and closing net and environmental sensing system (MOCNESS). Theoretical estimates of target strength based on a straight cylinder model and applied to the numbers and sizes of euphausiids in net samples yielded estimates of mean volume backscattering strength (MVBS) that compared favorably with observed values at low abundances. However, net samples underestimated biomass by almost two orders of magnitude at high abundances, a feature we attribute to net avoidance. Other taxonomic groups of macrozooplankton did not appear to contribute significantly to acoustic backscatter. Distributions of euphausiids were similar to those observed in summer; we found no evidence for abundant aggregations immediately below surface sea-ice. Euphausia superba dominated the biomass at all stations; individuals of approximately 22 mm were most abundant at 30–70 m depth, whereas 40-mm individuals were most abundant at 90–130 m. Thysanoessa macrura was positioned deeper in the water column, from 150 to 250 m, and Euphausia crystallorophias was centered on 100 m. The first two species were distributed throughout the study area, but E. crystallorophias was concentrated in coastal embayments. Total wet weight biomass ranged from 0.0009 to 9 g m −3, dry weight ranged from 0.002 to 18 g m −3, and protein ranged from 0.0009 to 9 g m −3, mostly concentrated in the 30–70 m depth stratum. Over the upper 200 m of the water column integrated wet weight biomass estimated from ADCP data was in the range from 29 g m −2 to 92 kg m −2.