Target Strength Measurements of Free-Swimming Sandeel Species, Ammodytes spp., in a Large Indoor Experimental Aquarium

Abstract

The sandeel species (Ammodytes spp.) occupy a critical ecological position in connecting lower trophic levels to higher ones. However, they are strongly affected by the marine environment and their catch rates are trending downward. In this study, the target strength (TS) of sandeel species was measured in free-swimming specimens using a split-beam quantitative echo sounder with 38 and 120 kHz frequencies in a physically controlled environment. Parameters a and b used in the estimated TS_mean–fork length (FL: 13.5–21.5 cm) equation were 53.7 and −124.3 dB at 38 kHz, and 71.3 and −153.2 dB at 120 kHz, respectively. The TS of the sandeel species were not proportional to the square of the FL but increased relatively rapidly with increasing body size. In addition, the mean and standard deviation of the swimming angle for the sandeel species from the acoustic data at 120 kHz were −2.2° and 7.7°, respectively, and most were in the −30° to 30°, range. Furthermore, TS was stronger at 38 kHz than at 120 kHz for all FL classes. The use of such frequency characteristics could facilitate the discrimination of fish species in the field and the sustainable assessment of sandeel species stocks.

1. Introduction

Three species of sandeels (Ammodytes japonicas; Ammodytes hexapterus; Ammodytes heian) inhabiting the area around the Japanese archipelago have overlapping distribution areas [1] and external morphologies [2]. Therefore, it is difficult to identify individual species from catches [2], and Japan manages these three species of sandeels (Ammodytes spp.) as sandeel species (hereafter referred to as sandeel) in some areas, such as Hokkaido. In many sea areas, the sandeel occupies a critical ecological position in connecting lower trophic levels to higher ones, because it feeds on zooplankton and is a major prey item of marine mammals [3]. Therefore, sandeel is an important mid-trophic-level species in the marine ecosystem [4]. In Japan, sandeel is widely distributed in coastal areas and is a vital target resource for coastal fisheries nationwide [5]. However, the catch fluctuation of sandeel has been on a downward trend; this may be due to a decrease in the recruitment of sandeel owing a decrease in the number of adult sandeel, which is strongly affected by marine environmental factors [6,7]. Hence, there is a critical need for the adaptive management of the current sandeel stock.

Sandeel has a peculiar ecology as it aestivates in summer and autumn, during seasonal changes such as increasing water temperature and scarcity of plankton [8]. Survival of the aestivation period depends entirely on previously accumulated energy, and spawning occurs immediately after aestivation [8]. Consequently, it is possible to artificially control the number of parental fish in the subsequent fishing season using management methods such as defining a last day of fishing before aestivation [9,10,11]. Accordingly, for the proper management of sandeel stock, the number of individuals should be measured regularly, and long-term monitoring of stock trends is necessary.

In recent years, one of the standard approaches for fish abundance estimation has been acoustic measurement using quantitative echo sounders (hereafter referred to as an echo sounder), which can rapidly and accurately estimate target organism abundance and distribution structure over a broad area [12,13,14,15]. The method calculates population density by measuring volume backscattering strength (Sv, dB) or area backscattering strength (Sa, dB) using a quantitative echo sounder. This is then divided by target strength (TS, dB), which reflects strength per individual of the target organism [16,17]. Therefore, the abundance of the target organism can be estimated using acoustic measurements if TS is known.

Three methods are commonly used for TS estimation, ex situ methods and modeling methods for live fish in cages and immobile fish, and in situ methods for wild fish [17]. The ex situ methods and modeling methods control the tilt of the experimental species and cannot easily obtain TS data similar to those obtained in the natural state [18]. The in situ method can obtain TS in the natural state; however, the inclusion of other fish species introduces large TS errors because TS often cannot directly reveal the target fish measured [17,18,19]. Generally, the TS values of fish are influenced significantly by having or not having a swim bladder, and measurements of swim-bladderless fish are more affected by body length and swimming behavior [16,20,21]. Thus, to make good and valid TS measurements on the sandeel, which is a swim-bladderless fish, a method of measuring TS while allowing the sandeel to swim in near-natural conditions is required. However, it is extremely difficult to measure swimming sandeel because it has schooling behavior and presents a weak acoustic target. Previous studies of swimming sandeel measured multiple individuals in cages in a tank or in cages under natural conditions [22,23]. Since the TS was measured in a fixed space and from multiple individuals, it was difficult to detect individual targets and the fish behavior was restricted. Therefore, to accurately estimate the abundance of sandeel, it is necessary to estimate TS in conditions close to their natural swimming conditions, while considering various swimming behaviors. Accordingly, a method of measuring TS while allowing the sandeel to swim freely in a relatively large, physically controlled environment is needed.

The purpose of the present study was to obtain the practical TS of the sandeel using an echo sounder in a large indoor aquarium, where the physical environment is stable, and to estimate the relationship between TS and body length. Swimming angles were also estimated by analysing the acoustic reflection passing through the acoustic beam.

2. Materials and Methods

2.1. Experimental Fish

Sandeel was caught in a small set net in Osatsube, Hokkaido, Japan. Subsequently, the fish were acclimated in a 2500 L plastic tank (diameter × height: 2 × 0.8 m). Since bubbles and air could significantly influence TS [17], the fish were released into the acclimated tank and experimental aquarium, exercising care to avoid contact with air and bubbles adhering to their body surfaces or gills. Furthermore, as sandeel is a schooling organism [23], an individual will often settle at the bottom of a tank when it is a single fish, and it rarely swims. Therefore, in the present study, multiple specimens in the same length class were used to achieve general swimming behavior in natural conditions. The fork lengths (FLs) of the specimens ranged from 13 to 22 cm and a total of ten experiments were conducted from 29 June to 10 July 2017, and from 15 to 23 June 2021 (

Table 1. Experiment date and fork length composition of experimental fish.

Start Date	End Date	Number of Specimens	Mean Body Size of Specimens (Lmean, cm)
26 June 2017	27 June 2017	20	13.5
28 June 2017	29 June 2017	19	20.0
29 June 2017	30 June 2017	22	17.0
30 June 2017	3 July 2017	20	15.0
3 July 2017	4 July 2017	14	19.0
15 June 2021	16 June 2021	11	21.5
17 June 2021	18 June 2021	20	19.5
19 June 2021	20 June 2021	20	16.5
20 June 2021	21 June 2021	20	17.5
22 June 2021	23 June 2021	20	18.5

).

2.2. Experimental Conditions

The experiments were carried out in a large indoor aquarium (W × L × H: 5 × 10 × 6 m) at the Hakodate Research Center for Fisheries and Oceans, Hakodate, Japan. A split-beam echo sounder (38 and 120 kHz; KSE-300, Sonic Co., Ltd., Tokyo, Japan) was used for the TS measurements (

Table 2. Settings of quantitative echo sounder KSE-300 in all experiments.

Transducer	T–178	T–182
Frequency (kHz)	38	120
−3 dB beam width (degree)	8.4	8.2
Transducer type	Split beam
Pulse length (ms)	0.3
Pulse width (ms)	0.6
Ping interval (s)	0.2
Recording range (m)	10.0

). During the experiments, the transducer was moved to the center of the experimental compartment and positioned with the surface of the transducer approximately 50 cm below the water surface (Figure 1). As the sandeel aestivates when the water temperature exceeds 20 °C [8,24], the water temperature in the experimental aquarium was maintained in the 13–14 °C range, which does not affect its swimming behavior.

2.3. TS Data Collection and Calculation

Before the start of each experiment, the echo sounder was calibrated using a tungstencarbide sphere (φ38.1 mm), which has a standard acoustic reflection with a known TS [25]. The calibration sphere was suspended approximately 2–3 m below the transducer, and the transmission/reception coefficient (K_TR: product of voltage sensitivity and gain) was calibrated by comparing the measured TS to the theoretical TS. The seawater sound velocity and absorption attenuation used in the TS calculation were obtained by measuring the water temperature and salinity values during the experiment [26]. Each experiment was recorded for more than 12 h, and acoustic data, when fish passed within the beam, were recorded. The TS data identified by the single-target detection algorithm in the echo sounder were detected from the recorded acoustic data, and standard parameters, including the voltage phase difference of single echo in front-back and left-right directions of the transducer, were exported along with the TS data as a CSV file [25]. The single-target detection algorithm uses a method that detects only those echoes that can ensure a signal-to-noise ratio (SNR) of 20 dB. Results were recorded only for single echoes recorded within a cutoff angle of 5.0° considering a −3 dB beam width.

Since bladderless fish such as sandeel have weaker acoustic reflections than fish with swim bladders [20,27], a lower threshold of TS is needed for analysis in consideration of background noise. In this study, the maximum TS value of background noise was determined from the raw data of each experiment. The maximum value of background noise was in the range of −83 to −73 dB at 38 kHz and in the range of −87 to −84 dB at 120 kHz (

Table 3. Analysis range, mean TS, and max TS for each body length class.

−	Analysis Range (Lower Threshold, dB)		Analysis Range (Upper Threshold, dB)	TSmean (dB)		TSmax (dB)
	38 kHz	120 kHz	38 kHz & 120 kHz	38 kHz	120 kHz	38 kHz	120 kHz
13.5	−76.0	−87.0	−45.0	−62.9	−74.4	−61.1	−56.7
15.0	−79.0	−85.0	−45.0	−64.8	−66.1	−60.2	−55.5
16.5	−79.0	−87.0	−45.0	−56.7	−61.5	−52.7	−50.6
17.0	−79.0	−87.0	−45.0	−60.4	−72.7	−56.9	−54.5
17.5	−77.0	−85.0	−45.0	−54.2	−65.1	−45.4	−49.6
18.5	−73.0	−86.0	−45.0	−54.5	−62.4	−49.2	−50.3
19.0	−77.0	−85.0	−45.0	−53.6	−60.8	−45.1	−50.5
19.5	−83.0	−84.0	−45.0	−56.5	−58.9	−48.7	−48.6
20.0	−78.0	−84.0	−45.0	−55.3	−63.3	−46.9	−45.2
21.5	−78.0	−86.0	−45.0	−54.2	−57.6	−47.1	−47.4

). Since the SNR of 20 dB is ensured by a single-target detection algorithm, the lower threshold of the analysis was set to the maximum background noise for each experiment. In addition, when multiple individuals are measured simultaneously, the distance between individuals may become too small and multiple individuals may be erroneously detected as a single target, resulting in an overestimation of TS [28]. Since previous studies have shown that the maximum TS of the largest sample used in this study, a 22 cm sandeel, is approximately −50 dB [27], the upper threshold of TS used in the present study analysis was set at −45 dB. Secondly, if an object is close to the transducer (i.e., within the near-sound field where the sound pressure changes rapidly), stable and accurate TS cannot be obtained. Therefore, an acoustical far-field boundary (Rend, m) was set, where the sound pressure was stable and accurate measurement was possible, as the standard, and the data below it were excluded [29]. Rend was 1.56 m at 38 kHz and 0.63 m at 120 kHz. To measure TS with high accuracy, the 1/2 FL of the experimental fish should be smaller than the diameter of the first Fresnel zone of the beam [30]. The measurable FLs at each Rend were calculated to be 35 and 7 cm at 38 and 120 kHz, respectively. Since the maximum FL of the experimental fish was 22 cm, the analysis range was determined by the Rend to be 38 kHz; however, at 120 kHz, the analysis range needs to be calculated based on each fish’s FL. The depth at which the diameter of the first Fresnel zone of the maximum FL in each experimental class could be ensured was calculated, and data shallower than that were removed. Finally, because of the performance of the echo sounder, an acoustic blind zone occurs near the bottom [31,32]. Data obtained within 0.24 m from the bottom of the tank were also excluded.

2.4. Relationship between TS and FL

Using the TS data extracted using the above procedures, we obtained the mean TS (TS_mean) and maximum TS (TS_max) under each experimental condition. The procedure for calculating the TS_mean was to convert the TS of each ping to a linear value, calculate their mean, and then reconvert this to dB. In addition, the following equation was used to estimate the TS–FL relationship, which is necessary for estimating fish abundance:

$$TS=a lo{g}_{10}FL+b$$

(1)

where a and b are coefficients and FL is the fork length of the fish (cm). Since fish in their natural environment may experience caudal fin damage, the body length of TS equation used for future abundance estimates was the fork length, which is unaffected by caudal fin damage. In general, the linear value of TS is proportional to the square of the FL [33,34], in which case it is expressed by the following equation:

$$TS=20 lo{g}_{10}FL+T{S}_{cm}$$

(2)

In the present study, parameters a, b, and TS_cm were optimized by the non-linear least-squares method using the ‘nls’ function in R ver. 4.1.0 (R Core Team, 2021. A Language and Environment for Statistical Computing, Vienna, Austria, https://www.R-project.org/, accessed on 1 June 2021).

2.5. Swimming Angle

In the present study, whether the TS_mean estimated in the experimental aquarium was similar to what would be measured in a natural marine environment was verified based on swimming behavior. TS varies greatly depending on the angle of inclination during swimming, and it is important to obtain information on the swimming tilt, swimming direction, etc. of fish in their natural state [25]. A split-beam type echo sounder was used to obtain three-dimensional coordinates (X, Y, Z) of a target organism simultaneously with the TS. Based on these coordinates, the swimming angle (i.e., the traveling direction for the dorsal aspect of the fish) was estimated using the echo trace method [35,36]. Unlike methods that determine the swimming behavior of fish from video cameras or other images, the echo trace method has the advantage of providing a clear correspondence between TS and swimming behavior. In addition, to estimate the swimming angle more accurately, we used data from fish that only swam linearly and obtained single echoes for more than five consecutive pings [18]. Additionally, tortuosity (i.e., the ratio of the path between start and end coordinates of the consecutive echoes to the direct distance) was considered in accordance with Henderson et al. [37], and among the range of echoes with tortuosity, more than 1.2 were removed from the analyses. The above processes were carried out using R ver. 4.1.0. Sandeel swimming behavior was verified based only on the 120 kHz results because the 120 kHz data were more continuous than the 38 kHz data.

3. Results

3.1. TS_mean and TS_max

The numbers of ping data obtained for each length class were 14–186 and 157–1769 pings at 38 and 120 kHz, respectively; more TS data were recorded at higher frequencies (Figure 2). The TS_mean and TS_max ranges calculated using the data were −64.8 to −54.2 and −61.1 to −45.1 dB, respectively, at 38 kHz, and −74.4 to −57.6 and −56.7 to −45.2 dB, respectively, at 120 kHz (Table 3). At both frequencies, TS_mean and TS_max tended to increase with an increase in FL. Additionally, TS_mean was stronger at 38 kHz than at 120 kHz in all length classes.

3.2. TS–FL Relationship

Using the TS values of each length class, the relationships between FL and TS_mean, and FL and TS_max, at the two frequencies were determined (Figure 3). With respect to the relationship between FL and TS_mean, in the case of no fixed coefficients, a and b were 53.7 and −124.3 dB (r² = 0.67), respectively, at 38 kHz, and 71.3 and −153.2 dB (r² = 0.62), respectively, at 120 kHz.

$$38 \mathrm{kHz}: T{S}_{mean}=53.7 lo{g}_{10}FL-124.3$$

(3)

$$120 \mathrm{kHz}: T{S}_{mean}=71.3 lo{g}_{10}FL-153.2$$

(4)

When coefficient a was fixed at 20, the TS_cm values were −82.8 dB (r² = 0.40) and −89.2 dB (r² = 0.30) at 38 kHz and 120 kHz, respectively.

$$38 \mathrm{kHz}: T{S}_{mean}=20 lo{g}_{10}FL-82.8$$

(5)

$$120 \mathrm{kHz}: T{S}_{mean}=20 lo{g}_{10}FL-89.2$$

(6)

With respect to the relationship between FL and TS_max, when the coefficients were not fixed, the values of a and b were 85.1 and −157.43 dB (r² = 0.88), respectively, at 38 kHz, and 53.5 and −117.5 dB (r² = 0.94), respectively, at 120 kHz.

$$38 \mathrm{kHz}: T{S}_{max}=85.1 lo{g}_{10}FL-157.4$$

(7)

$$120 \mathrm{kHz}: T{S}_{max}=53.5 lo{g}_{10}FL-117.5$$

(8)

When the coefficient a was fixed at 20, the TS_cm values were −76.3 dB (r² = 0.68) and −75.8 dB (r² = 0.86) at 38 kHz and 120 kHz, respectively.

$$38 \mathrm{kHz}: T{S}_{max}=20 lo{g}_{10}FL-76.3$$

(9)

$$120 \mathrm{kHz}: T{S}_{max}=20 lo{g}_{10}FL-75.8$$

(10)

3.3. Swimming Angle

The mean and standard deviation of the swimming angles detected at 120 kHz were −2.2° and 7.7°, respectively. All the swimming angles were in the −30° to 30° range (Figure 4).

4. Discussion

4.1. Swimming Behavior of Sandeel

In this study, more than half of the swimming angles obtained were positive, indicating that sandeel often swim in an upward direction. Sandeel lack a swim bladder, which means that they swim continuously except when they remain on the seafloor, and their relatively small caudal fins are characteristic of swimming with their body axis tilted backward [38]. Additionally, previous studies have suggested that sandeel have negative buoyancy, which requires them to swim with their bodies tilted upward to some extent in order to maintain altitude [23]. On this basis, the swimming angles obtained in this study reflect the fact that sandeel, which are bladderless fish, move more easily in the upward direction.

In addition, it has been reported that the average tilt angle of free-swimming sandeel is approximately 20° and that they always swim with their heads up [23,27]. However, the mean swimming angle of the sandeel in this study was −2.2°, indicating that in some cases the direction of movement of the sandeel was negative. This is thought to be due to swimming and diving behavior with the aim of conserving energy used for swimming. This phenomenon is called glide and swim, a behavior commonly observed in negatively buoyant fish [39,40,41]. From these results we concluded that the swimming angles obtained in this study reflect swimming behavior similar to the natural state, including the glide and swimming behavior of the sandeel. It can be determined that there was no abnormal behavior due to the aquarium experiment.

The swimming angle can be viewed as an indicator of postural changes that affect TS; however, it is different from rigorous tilt angle. The orientation of fish changes in response to the direction in which they are moving, but at lower speeds, many fish can have very different tilt angles even during straight swimming. Therefore, it is desirable in the future to conduct TS measurements under conditions close to natural seawater to verify the difference between the tilt angle and swimming angle.

4.2. Measured TS

The TS frequency distributions obtained in this study varied more widely at 120 kHz than at 38 kHz for all length classes. In general, the directivity of TS to fish tilt is sharper at higher frequencies than at lower frequencies [42]. Notably, when TS is measured randomly, the higher the frequency used in the measurement, the greater the variation in the obtained TS [43]. Therefore, in this study, the large variation in TS at 120 kHz is considered to be due to the strong directivity.

In addition, the TS_mean obtained in this study differed among frequencies for all length classes; this is thought to be due to the relationship between the normalized body length L/λ, where L is normalized by the wavelength λ, and the scattering characteristics. In general, bladderless fish show Rayleigh scattering characteristics when L/λ is less than 1, the transition from Rayleigh scattering to geometric scattering is in the range of 1–10, and geometric scattering where TS depends only on size when L/λ is greater than 10 [42]. In the body length range of this study, L/λ at 120 kHz was in the geometric scattering region, whereas L/λ at 38 kHz was located in the transition region. The difference in TS between the two frequencies is considered to be an inevitable result of the different scattering characteristics of L/λ. This difference in the scattering intensity between the two frequencies may be used to separate sandeel from other organisms on an echogram. Furthermore, the use of 120 kHz data, for which TS_mean is in the geometric scattering region, for the distribution density calculation of sandeel is expected to simplify the population calculation, because the increase or decrease in TS_mean depends only on body length.

4.3. Estimated TS–FL Formula in This Study

The values a obtained in this study were far from 20 for both TS_mean and TS_max. This indicates that the TS of the sandeel is not proportional to the square of the body length, but rather increases relatively rapidly with body growth. In general, the TS of swim bladder fish is proportional to the square of body length because the size of the swim bladder, which accounts for 90% of acoustic reflection, is proportional to body length [44]. However, sandeel does not have a swim bladder, and TS tends to increase in proportion to the 4th–5th power of body length [17,20]. Previous studies have also suggested that the TS of sandeel estimated by theoretical models is proportional to the 4th–5th power of body length [21,45]. Therefore, TS = a ∗ log FL + b is more appropriate for the regression equation between TS and body length in sandeel.

4.4. Theoretical and Measured TS Comparison

Figure 5 shows the relationships between TS and body length in sandeel species reported in previous studies (only A. personatus) and the present study (including related species);

Table 4. Coefficients for TS_mean—body length equations for sandeels in our study and previous research. (Previously, the A. personatus and A. hexapterus have been recognized in the waters sur-rounding Japan. However, a taxonomic reexamination based on genetic and morphological analyses has led to the description of a new species, A. heian, and the scientific name A. japonicus should be applied to the A. personatus).

Reference	Method	Species	FL (cm)	Frequency (kHz)	TSmean
					a	b (TScm)
this study	Free-swimming	Ammodytes spp.	13.5–21.5	38	53.7	−124.3
				120	71.3	−153.2
				38	20.0	−82.2
				120	20.0	−89.2
Safraddin et al.	Theoretical calculation (DWBA model)	A. personatus	16.1–28.7	38	8.2	−74.2
				120	20.9	−92.6
Yasuma et al.	Theoretical calculation (DCM model)	A. personatus	3.4–6.7	38	56.5	−125.1
				120	34.0	−98.2
			7.5–11.5	38	20.0	−89.2
				120	20.7	−92.1
Matsukura et al.	Theoretical calculation (DWBA model)	A. personatus	3.1–7.7	38	46.5	−118.6
				120	34.3	−96.4

shows the estimated coefficients in the equations. First, we compared the theoretical TS estimated from the distorted-wave Born approximation (DWBA) model with the measured TS from this study in similar body length classes [27].

The theoretically estimated TS_mean was −64.9~−63.3 dB at 38 kHz and −69.0~−64.8 dB at 120 kHz, whereas the measured TS_mean was −64.8~−54.2 dB at 38 kHz and −74.4~−57.6 dB at 120 kHz. The measured values were higher than the theoretical values at both frequencies, and the difference between theoretical and measured value is larger at 120 kHz than at 38 kHz. One possible reason for the discrepancy between the theoretically estimated TS_mean and the measured TS_mean is that the parameters used in the DWBA model, such as tilt angle, may have differed from those used in the actual measurements. In addition, the directionality becomes stronger at higher frequencies; even small changes in tilt angle cause large fluctuations in TS values. The tilt angle in the DWBA model was assuming a normal distribution with a mean of −5° and a standard deviation of 15°; however, because this study was free swimming, the tilt angle distribution may differ from the normal distribution, which could bias the model calculation. Therefore, the difference between theoretical and measured values may reflect the difference in tilt angle distribution, and the large difference at 120 kHz may reflect the strong directivity at high frequencies. Possible reasons other than tilt angle distribution are the density ratio g between the fish body and seawater and sound speed ratio h, which are the important parameters for TS estimation by the theoretical model [21]. The DWBA model is used under the assumption that g does not change with fish body tilt; however, because actual fish bodies are composed of flesh and bones, g changes with changing tilt angle. Similarly for h, it can be assumed that the values of h used in this study and in the model are different because the same seawater is not used. Therefore, it is considered that changes in g and h have a large effect on the difference in TS between model measurement and this study [46]. Furthermore, the TS of fish without a swim bladder is affected by the degree of flexion of the fish body at the same tilt angle [47]. Since the sandeel swims with a wavy motion of the whole body [38], TS is expected to change significantly owing to flexion of the fish body caused by the swimming motion. However, the theoretical values calculated by the model cannot predict the behavior of sandeel in natural conditions, and it is quite difficult to calculate the TS considering the bending of the fish body. Therefore, it is considered that the measured TS in this study was the result of considering the parameter changes caused by the swimming behavior of the sandeel under natural conditions.

Next, the TS_mean equation for adult sandeel obtained in this study was compared with the results for juvenile sandeel. The a and b for juvenile sandeel (standard length SL: 3.4–6.7 cm) estimated by the Dynamic-Casual-Modelling (DCM) model were 56.5 and −125.1 dB at 38 kHz, showing a small difference from the present study. On the other hand, the values at 120 kHz were 34.0 and −98.2 dB, which were different from those of the present study [45]. In addition, the a and b of juvenile sandeel (SL: 3.1–7.7 cm) estimated by the DWBA model are different from the present study for both frequencies [21]. One of the reasons for this difference may be the difference between the parameters used in the model and those measured, as previously discussed. Another possibility is the effect of different growth stages of fish on different swimming behaviors. Fish swimming abilities change significantly as they grow [48]. To travel the same distance, juveniles move rapidly in small increments, whereas adults swim more slowly [39,48]. Therefore, the probability density distribution of the tilt angle used to estimate TS_mean is likely to be different for different growth stages. However, since the measurements in this study were conducted using free-swimming organisms, we believe that the results consider the swimming behavior of sandeel better than previous studies that assumed a general probability density distribution of tilt angles.

5. Conclusions

The free-swimming measurements conducted in this study provide practical TS values that are similar to those in natural conditions. In addition, since the behavior of the fish during TS measurement was known, we were able to examine the validity of TS based on the swimming behavior of the fish. Sawada et al. [25] stated that it is desirable to detect individual fish from a school of fish in a natural state and to use the data to obtain the standardized TS. Therefore, we consider it significant that practical TS was obtained in this study by using free-swimming schools of sandeel. However, for negatively buoyant fish, such as sandeel, even with a high degree of tilt, the swimming angle from tracking may be linear horizontal. In the future, it will be necessary to determine the difference and relationship between tilt angle and swimming angle.

Conversely, since all sandeel used in this study were adults, it would be desirable to also measure the TS of juvenile sandeel in order to obtain the TS–length relationship for a wider range of body lengths. In addition, it is known that the juvenile Sandeel species has a unique ecology with visual feeding or diving sand behavior, and a vertical swimming habit with increasing swimming ability as they grow [49,50,51]. Therefore, it is expected that the TS–length relationship equation will be estimated more practically in the future by improving the measurement method to take into account the unique ecology and behavior of sandeel. Improving the TS accuracy of bladderless fish such as sandeel, whose TS fluctuates greatly depending on the tilt angle, is expected to enable the estimation of abundance with a high degree of accuracy.

Furthermore, the frequencies used in this study are those carried by many research vessels. By using the frequency characteristics of sandeel revealed by these frequencies, it will be possible to discriminate fish species in the actual sea, and this will facilitate appropriate management of sandeel resources. Since the difference in reflection intensities among frequencies is different among species [52,53], in the future, by clarifying the TS and frequency characteristics of other important fish species, it will be possible to discriminate among many more species. Moreover, it is expected that species identification and abundance estimation for each fish species will be carried out with improved accuracy.