A Study on the Catch Losses and Mesh Selectivity Related to the Attachment of Marine Mammal Bycatch Reduction Devices on Midwater Trawl Gear

Abstract

The National Institute of Fisheries Science in Korea is developing marine mammal bycatch reduction devices (BRDs) for high-risk fishing gear, such as trawls. We experimented with two BRD types (guiding nets) attached in front of codend at 30° and 45° angles, and catch losses and mesh size selectivity were investigated. Experimental fishing operations were conducted along the East Coast of South Korea where whales and dolphins are commonly sighted. The catch was classified according to fishing location, BRD type, codend, and covernet, with measurements recorded for body length, maximum girth, and weight. The average selectivity for each haul was analyzed using the ‘selfisher’ package. The catch loss rates with the BRD attached at tilt angles of 30° and 45° were 11% and 29% for common flying squid, 6% and 28% for sailfin sandfish, and 5% and 8% for pearlside. While the mesh selectivity rates for common flying squid and pearlside remained at 0.2–0.5 across all lengths and tilt angles, the mesh selectivity curve for sailfin sandfish was estimated. There were significant differences in catch loss between 30° and 45° angles, with the 30° angle being more effective in catch loss. We observed a masking effect in the codend.

1. Introduction

The bycatch and mortality of whales as well as dolphins due to fishing activities is a persistently raised issue [1,2,3]. In the United States, measures have been implemented under the Marine Mammal Protection Act to prohibit the import of catches using fishing gear with high rates of marine mammal bycatch [4]. For gillnets and pot gears, which have high marine mammal bycatch rates, the use of acoustic pingers has proven effective in reducing bycatch [5,6,7]. Additionally, ropeless pot gear, which does not use buoy lines, has been proposed to reduce marine mammal bycatch [8,9]. As exemplified by these studies, there are continuous efforts to reduce marine mammal bycatch.

South Korea has also been developing bycatch reduction devices (BRDs) for five types of fishing gear with high potential for marine mammal bycatch (gillnets, traps, trawls, stow nets, and set nets), in accordance with the comparability finding of the U.S. National Oceanic and Atmospheric Administration (NOAA). A BRD for stow nets, developed by Lee et al. (2021) [10], was attached to 30 stow net fishing vessels between April 2021 and December 2023, during which 16,000 operations were monitored. As a result, while 52 finless porpoises (Neophocaena phocaenoidess) were caught in stow nets without the device, those equipped with the device had zero bycatch. Since then, the record of no bycatch of finless porpoises has been maintained in stow nets equipped with the BRD. Based on these monitoring results, the attachment of BRDs to stow nets has been recommended in South Korea since 2021. However, the use of BRDs in stow nets has led to a catch loss of approximately 4% [11], indicating that mesh selectivity also occurs with the device. This highlights the need to determine the optimal mesh size of the BRD to minimize catch losses while allowing the escape of marine mammals.

In this study, we modified the principle of a turtle excluder device (TED), a BRD for non-target species in midwater trawl gear [12,13,14], to adapt it for use with midwater trawl nets as experimental gear. Trawl fishing, including the use of trawl gear, is known for capturing a diverse range of species [15,16]. In the United States, there have been studies in which different acoustic devices, such as pingers and BRDs with grids, were attached to trawl gears with positive results [17]. However, BRDs with grids have possibility of safety risks for fishers, and BRDs made of netting panel have not been experienced.

From these studies, our objectives are to attach BRD with netting panels instead of grids during experiments and to calculate catch losses due to the BRD for both target and non-target species. Additionally, this study aims to estimate the BRD mesh selectivity for target species, and develop methods to reduce these catch losses.

2. Materials and Methods

2.1. Test Fishing Gear with Bycatch Reduction Device

The fishing gear used in this study is classified as midwater trawl gear according to the Food and Agriculture Organization (FAO) classification, primarily designed to catch species inhabiting surface or midwater layers. Its design and schematic diagrams are illustrated in Figure 1 and Figure 2. According to the design in Figure 1, the total length of the net, excluding the trawl rope, is 140 m, with the codend being 11.2 m long. The mesh size is 60–10,000 mm. In the schematic diagram in Figure 2, the codend is composed of three layers of nets (A, B, and C from the outermost layer inward), with their mesh sizes measured 20 times as 120 ± 2.4, 62 ± 4.0, and 28 ± 0.4 mm, respectively. Here, the covernet functions as both codend nets A and B. Additionally, as shown in Figure 2, a BRD was attached to reduce marine mammal bycatch, and the guiding net of the BRD was installed just in front of the codend of the midwater trawl gear. The guiding net was made of the escape opening size of the BRD is specified in 2 × 1.5 m for considering the marine mammals escape from BRD, and a BRDs covernet (PE 46 ± 1.4 mm) was attached to the escape opening to investigate species escaping through the BRD. The guiding net of the BRD was made of netting panel with a square mesh (1 bar: PE 200 mm, mesh perimeter: 800 mm), and two types of BRD guiding nets with tilt angles of 30° (2 × 2.23 m) and 45° (2 × 1.4 m) were prepared for use when the midwater trawl net was deployed in a parallel position.

2.2. Fishing Area

This study was conducted in the coastal area of Pohang, South Korea (35°30.000′–36°30.000′ N, 129°0.000′–129°30.000′ E, Figure 3), an area known for frequent whale sightings [18,19].

2.3. Fishing Experiment with Test Fishing Gear with Bycatch Reduction Device

The fishing operations were conducted over 12 days (10–21 June 2024) aboard the research vessel Tamgu-20 (885 tons, 2600 horsepower) of the National Institute of Fisheries Science (NIFS). The operations were carried out during the daytime, with each trawl lasting 1 h, at a speed of 4.5–5 knots. The average depth of the fishing grounds ranged from 200 to 300 m, with the trawl net deployed at the surface (20–50 m) and midwater (100–150 m) layers a total of 16 times. After hauling, the catch was classified by BRD, codend, and covernet. The body length (measured as fork length, total length, mantle length, or anal length, depending on species), body weight, and maximum girth of the classified catch were measured to the nearest millimeter and gram. A full count was conducted for species with larger individuals. For smaller fish species, including juveniles, quantitative sampling involved placing individuals of the same species in uniform boxes and counting the number from a randomly selected box [20]. For catches classified by BRD, codend, and covernet, the length, weight, and maximum girth of each individual within a sampled box were measured. These measurements were then multiplied by the number of boxes sorted by species or fishing location to estimate the length composition of the caught species.

2.4. Data Analysis

Until recently, most studies used the methods of estimating the selectivity by pooling the total catch from all hauls [21,22,23]. However, mesh selectivity can vary even within the same species depending on the haul [24]. For example, in Mediterranean trawl fisheries, selectivity curves in the codend change significantly due to variations in swimming ability dependent upon water temperature or increased girth during the spawning season [25,26]. In such cases, it is considered necessary to assess the changes in mesh selectivity for each haul. Currently, studies have been reported to compare the method of estimating the selectivity by pooling the total catch from all hauls or the average selectivity for each haul [27,28,29]. Thus, we estimated and compared the BRD and codend mesh selectivity by analyzing both the catches per haul and the total catch of all hauls in this study. Additionally, selectivity curves were calculated separately and overall for the BRD at tilt angles of 30° and 45° (hereinafter referred to as 30°, 45°, and total). Equation (1) describes the probability $r_{BRD}\left(l\right)$ that a species of length class $l$ passes through the BRD and is caught in the codend, and is expressed as follows:

$$r_{BRD}\left(l\right)={\displaystyle \frac{n_{l2}+n_{l3}}{n_{l1}+n_{l2}+n_{l3}}}$$

(1)

where $n_{l1}$, $n_{l2}$, and $n_{l3}$ are the numbers of individuals of length class $l$ caught in the BRD, codend, and covernet, respectively.

$r_c\left(l\right)$ is expressed as a function of the length class l and is approximated using a logistic equation according to the analytical method described as follows:

$$r_c\left(l\right)={\displaystyle \frac{\exp \left(a+bl\right)}{1+\exp \left(a+bl\right)}}$$

(2)

where $a$ and $b$ are parameters of the logistic equation.

The interpretation of the mesh selectivity for species based on the 400 mm guiding net mesh of the BRD described earlier was estimated using the statistical software package R (Version 4.3.3), specifically the selfisher package (selectivity of fisheries gear [30]). The selfisher package utilizes the SELECT model based on maximum likelihood estimation [31,32,33,34]. Its key feature is the use of a generalized linear mixed model (GLMM) specialized for estimating the selectivity of fishing gear. It is designed to concurrently analyze the fixed effects model, which calculates selectivity curves by summing the catches of all hauls by length class, and the random effects model, which assumes that catches per haul are influenced by random effects due to variations in fishing conditions or resource environmental factors. In the analysis using the random effects model, the average mesh selectivity was estimated from each haul’s selectivity using the bootstrap method. Using these analytical methods, we estimated the BRD mesh selectivity for each species under the three conditions of 30°, 45°, and total (both 30° and 45°).

Catch losses due to the BRD were estimated from the catches classified as BRD, codend, and covernet as follows:

$$\mathrm{C}\mathrm{a}\mathrm{t}\mathrm{c}\mathrm{h}\mathrm{l}\mathrm{o}\mathrm{s}\mathrm{s}\mathrm{e}\mathrm{s}={\displaystyle \frac{\mathrm{T}\mathrm{h}\mathrm{e}\mathrm{n}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r}\mathrm{o}\mathrm{f}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{c}\mathrm{h}\mathrm{i}\mathrm{n}\mathrm{B}\mathrm{R}\mathrm{D}}{\mathrm{T}\mathrm{h}\mathrm{e}\mathrm{n}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r}\mathrm{o}\mathrm{f}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{c}\mathrm{h}(\mathrm{B}\mathrm{R}\mathrm{D}+\mathrm{c}\mathrm{o}\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{d}+\mathrm{c}\mathrm{o}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{n}\mathrm{e}\mathrm{t})}}$$

(3)

Catch losses were calculated for each haul and averaged by species.

3. Results

3.1. The Composition of Body Length from Catch

During the 12-day test fishing operation, 16 hauls were conducted, resulting in a total catch of 1,093,070 individuals across 14 species, weighing a total of 1312.1 kg. Among these, 2120 sailfin sandfish (Arctoscopus japonicus) individuals (77.0 kg), 5458 common flying squid (Todarodes pacificus) individuals (55.9 kg), and 1,085,193 pearlside (Maurolicus muelleri) individuals (1166.4 kg) were caught, comprising 99% of the total catch by number and weight (

Table 1. Total catch results of 14 species from 16 hauls of midwater trawl in fishing ground.

Common Name	Scientific Name	The Number of Catch	Weight of Catch
Common flying squid	Todarodes pacificus	5458	55.9 kg
Sailfin sandfish	Arctoscopus japonicus	2120	77.0 kg
Pearlsides	Maurolicus muelleri	1,085,193	1166.4 kg
Pacific cod	Gadus macrocephalus	2	1.7 kg
Green rough-backed puffer	Lagocephalus lunaris	4	0.8 kg
Striped puffer	Takifugu xanthopterus	5	2.2 kg
Schoolmaster gonate squid	Berryteuthis magister	4	1.4 kg
Japanese Spanish mackerel	Scomberomorus niphonius	1	0.1 kg
Blackmouth angler	Lophiomus setigerus	1	1.3 kg
John dory	Zeus faber	20	1.1 kg
Largehead hairtail	Trichiurus lepturus	17	0.3 kg
Firefly squid	Watasenia scintillans	240	3.3 kg
Pacific herring	Clupea pallasii	3	0.1 kg
Silver pomfret	Pampus argenteus	2	0.5 kg
Total		1,093,070	1312.1 kg

). This allowed for the collection of sufficient data over a wide range of body lengths. These three species were selected for analyzing the mesh selectivity of the BRD and codend. Figure 4 illustrates the total number of individuals caught with the BRD by species (common flying squid, sailfin sandfish, and pearlside, abbreviated as CFS, SS, and PS, respectively, in Section 3 and Section 4) and tilt angle (30° and 45°). The mantle length (ML) of the captured CFS ranged from 4.0 to 22.1 cm, with individuals caught in the BRD ranging from 4.0 to 22.1 cm, in the codend from 4.0 to 19.3 cm, and in the covernet from 4.1 to 5.4 cm. For SS, the fork length (FL) ranged from 11.2 to 24.3 cm, with individuals caught in the BRD ranging from 11.4 to 24.3 cm, in the codend from 6.1 to 23.7 cm, and in the covernet from 5.2 to 13.1 cm. For PS, the total length (TL) showed a narrow range of 3.5 to 6.7 cm, with individuals caught in the BRD ranging from 4.0 to 6.7 cm, in the codend from 4.0 to 6.5 cm, and in the covernet from 3.5 to 5.7 cm. In terms of tilt angle, more individuals were caught at a 30° angle than at a 45° angle across all three species.

3.2. The Relationship between Body Length and Maximum Girth

Counting the common flying squid (CFS) and sailfin sandfish (SS) catches yielded the following results:

• The relationship between mantle length (ML) and maximum girth (G) for CFS:
  G = 0.5287 × ML + 0.8663 (4) (r = 0.76, p < 0.03, n = 266, Figure 5a).
• The relationship between fork length (FL) and maximum girth (G) for SS:
  G = 0.459 × FL + 1.0491 (6) (r = 0.79, p < 0.02, n = 446, Figure 5b).

In CFS and SS, body length and maximum girth were found to be correlated, with both maximum girth and body weight increasing as body length increased (p < 0.05, Figure 5a,b). However, in PS, the relationship between body length and maximum girth did not show a significant correlation (p > 0.05, Figure 5c).

3.3. The Mesh Selectivity of Bycatch Reduction Device

BRD selectivity was analyzed by species and tilt angle. The BRD selectivity curves for CFS were estimated under the three conditions of 30°, 45°, and total (Figure 6,

Table 2. Estimated parameters for sorting selectivity curves after capture on the East Coast of South Korea during June, 2024 (common flying squid, CFS).

Scientific Name	Estimation Method (30°, 45°, Total)	Logistic Parameters		Selectivity Curve Parameters				Model Fit
		a	b	${\mathbf{L}}_{50}$	C.I. of ${\mathbf{L}}_{50}$	S.R.	C.I. of S.R.	Deviance	df	p-Value
Todarodes pacificus	Average of each haul	−0.43 (0.82)	0.02 (0.06)	28.4 (78.2)	Error	146 (381.3)	Error	0	21	0.813
	Sum of hauls	−0.41 (0.18)	0.02 (0.04)	27.0 (58.3)	-	146 (615.5)	-	0	21	0.701
	Average of each haul	−0.33 (0.03)	0.02 (0.06)	13.1 (18.7)	Error	93.6 (252.6)	Error	0	21	0.711
	Sum of hauls	−0.31 (0.81)	0.02 (0.03)	17.4 (17.6)	-	115.2 (186.3)	-	0	21	0.536
	Average of each haul	−0.16 (0.81)	0.01 (0.06)	12.3 (32.5)	12.1–12.5	167.8 (806.8)	Error	0	21	0.408
	Sum of hauls	−0.16 (0.07)	0.01 (0.02)	12.4 (9.4)	-	168.5 (203.5)	-	0	21	0.835

). The estimation indicated no significant selectivity in either model (p > 0.05). At 30°, the selectivity rate did not exceed 0.5 across all body lengths. At 45°, the selectivity rate ranged from 0.4 to 0.5 for the measured body lengths. For the total, the selectivity rate was distributed around 0.5, showing a smoother curve compared to the estimation at 45°. Additionally, the 50% selectivity length (${\mathrm{L}}_{50}$) was 17.0 cm (fixed effects model) and 28.4 cm (random effects model) at the 30° angle, 17.4 cm and 13.1 cm, respectively, at the 45° angle, and 12.4 cm and 12.3 cm, respectively, without considering the angle factor. The selection range fell outside the measured length range, and the 95% CI was 12.1–12.5 cm only in the ${\mathrm{L}}_{50}$ of the total model; estimation of 95% CI of ${\mathrm{L}}_{50}$ and S.R. was not possible for other models.

The BRD selectivity curves for SS were estimated under the three conditions of 30°, 45°, and total (Figure 7,

Table 3. Estimated parameters for sorting selectivity curves after capture on the East Coast of South Korea during June, 2024. (sailfin sandfish, SS).

Scientific Name	Estimation Method (30°, 45°, Total)	Logistic Parameters		Selectivity Curve Parameters				Model Fit
		a	b	${\mathbf{L}}_{50}$	C.I. of ${\mathbf{L}}_{50}$	S.R.	C.I. of S.R.	Deviance	df	p-Value
Arctoscopus japonicus	Average of each haul	−9.96 (4.38)	0.51 (0.22)	19.4 (1.42)	19.0–20.6	4.3 (1.9)	2.5–6.5	0	20	0.02
	Sum of hauls	−8.75 (0.45)	0.45 (0.03)	19.5 (0.20)	-	4.9 (0.3)	-	0.2	20	<0.01
	Average of each haul	−11.18 (5.20)	0.54 (0.25)	20.9 (1.41)	20.1–21.4	4.1 (1.9)	2.2–6.0	0	20	0.01
	Sum of hauls	−10.85 (1.17)	0.52 (0.07)	20.9 (0.42)	-	4.2 (0.5)	-	0.1	20	<0.01
	Average of each haul	−9.59 (4.42)	0.45 (0.21)	21.5 (1.6)	21.5–21.9	4.9 (2.3)	4.4–5.2	0	20	0.03
	Sum of hauls	−9.58 (0.54)	0.44 (0.03)	21.7 (0.3)	-	5.0 (0.3)	-	0.1	20	<0.01

). SS exhibited significant selectivity in both the random and fixed effects models (p < 0.05). ${\mathrm{L}}_{50}$ estimated at 30° was 19.4 cm for the random effects model and 19.5 cm for the fixed effects model (95% CI, 19.0–20.6). The selectivity range was 4.3 cm and 4.9 cm for the random and fixed effects models, respectively (95% CI, 2.5–6.5). At a 45° tilt angle, ${\mathrm{L}}_{50}$ was 20.9 cm (95% CI, 20.1–21.4) for both the random and fixed effects models, and the selectivity range was 4.1 cm for the random effects model and 4.2 cm for the fixed effects model (95% CI, 2.2–6.0). For the total condition, ${\mathrm{L}}_{50}$ was 21.5 cm for the random effects model and 21.7 cm for the fixed effects model (95% CI, 21.5–21.8), and the selectivity range was 4.9 cm and 5.0 cm, respectively (95% CI, 4.4–5.2).

The BRD mesh selectivity for PS was estimated under the three conditions of 30°, 45°, and total (Figure 8,

Table 4. Estimated parameters for sorting selectivity curves after capture on the East Coast of South Korea during June, 2024. (pearlside, PS).

Scientific Name	Estimation Method (30°, 45°, Total)	Logistic Parameters		Selectivity Curve Parameters				Model Fit
		a	b	${\mathbf{L}}_{50}$	C.I. of ${\mathbf{L}}_{50}$	S.R.	C.I. of S.R.	Deviance	df	p-Value
Maurolicus muelleri	Average of each haul	−1.581 (0.637)	0.001 (0.164)	2106 (45,993)	Error	2928 (64,020)	Error	0	66	0.999
	Sum of hauls	−1.580 (0.032)	0.001 (0.006)	2776 (30,886)	-	3861 (43,031)	-	0	66	0.988
	Average of each haul	−1.499 (0.62)	0.003 (0.159)	541 (30,960)	Error	793 (45,639)	Error	0	66	0.989
	Sum of hauls	−1.501 (0.04)	0.002 (0.008)	502 (1409)	-	736 (2084)	-	0	66	0.985
	Average of each haul	−1.501 (0.62)	0.014 (0.159)	1053 (11,768)	Error	1541 (17,277)	Error	0	66	0.999
	Sum of hauls	−1.501 (0.02)	0.013 (0.004)	1146 (4345)	-	1677 (6390)	-	0	66	0.992

). PS, like CFS, did not show significant selectivity in either model (p > 0.05), indicating a selectivity rate around 0.2 across all measured body lengths. Additionally, with both ${\mathrm{L}}_{50}$ and the selectivity range falling outside the measured length range, the 95% CI could not be estimated.

4. Discussion

The fishing gear used in this study was designed with a very small mesh size in the codend to conduct a resource survey aimed at examining the “total allowable catch“ (TAC) system [35,36]. This design increases the likelihood of catching a wide range of sizes and species. Additionally, a marine mammal bycatch reduction device (BRD), currently under development, was attached to the gear, which may help prevent bycatch of marine mammals, although it could also result in catch losses of target species. Therefore, this study explored methods to prevent bycatch of marine mammals while minimizing catch losses of target species by estimating the selectivity curves of the BRD and the codend. It was also deemed necessary to examine the masking effect caused by the covernet, as both the codend and covernet are designed with triple layers.

4.1. The Importance of Caught Fish Species

Among the three species caught and measured in this study, i.e., common flying squid (CFS), sailfin sandfish (SS), and pearlside (PS), CFS and SS are important species widely consumed in Korea and are therefore subject to the “total allowable catch” system [37,38,39,40]. In contrast, PS is considered inedible in Korea, and there is no specific fishing gear or method designated for catching this species. However, in this study, most of the catch, including CFS, SS, and other species, were found with PS in their mouths. Globally, PS are known to be key prey for target species [41,42]. This highlights the need to develop fishing gear to prevent bycatch of PS.

4.2. The Efficiency of Bycatch Reduction Device

Since this study used a research vessel instead of a commercial fishing vessel, the time required for setting and hauling the nets was relatively longer to ensure the safety of the crew. During 16 trial hauls, the average times for setting and hauling the nets were 12 min 32 s and 32 min 7 s, respectively. This suggests that some species could escape from the BRD during the extended hauling time and not be included in the final catch count. A study on a seine boat fishery reported that minimizing the hauling time using a developed hauling system could reduce the catch loss of anchovies [43].

The species-specific selectivity of the BRD was analyzed based on tilt angle selectivity, mesh selectivity, and fish behavior and biological characteristics. In this study, a marine mammal BRD with a mesh size of 400 mm and tilt angles of 30° and 45° was attached directly in front of the codend. Thus, in this study, being selected by the BRD indicates that the species did not enter the codend, resulting in catch loss due to the BRD.

4.2.1. Selectivity According to the Tilt Angle of the BRD

To examine the catch loss rates in terms of tilt angle-dependent BRD selectivity, the catch loss rates for the three species were compared based on the total number and weight of individuals caught at 30° and 45° tilt angles of the BRD. CFS showed a catch loss rate of 11% at 30° and 29% at 45° in terms of the number of individuals, and 7% at 30° and 12% at 45° in terms of weight, showing significant angle-dependent differences (p < 0.05). The catch loss rates estimated for SS at 30° and 45° tilt angles were 6% and 28% by number, and 6% and 14% by weight, also showing significant angle-dependent differences (p < 0.05). Unlike CFS and SS, PS did not show any significant angle-dependent differences (p > 0.05), with catch loss rates of 5% at 30° and 8% at 45° by number, and 5% at 30° and 6% at 45° by weight. These results suggest that, even with the same mesh size, the catch loss rate was lower at a 30° tilt angle compared to a 45° tilt angle, with the loss rate increasing as the tilt angle increased, indicating that a 30° tilt angle is more appropriate than a 45° tilt angle. These results demonstrate that selectivity can vary depending on the tilt angle of the BRD. In a study using a jellyfish separator net as a BRD to reduce jellyfish bycatch, it was confirmed that as the tilt angle increased from 10° to 15° and 20°, the discharge rate of jellyfish decreased to 66%, 41%, and 44%, respectively [44]. Other studies have also reported different selectivity based on the tilt angle of the BRD [45,46,47].

Comparing the catch loss rates between estimations by number and weight, the catch loss rates of CFS and SS at a 45° tilt angle were 29% and 28%, respectively, based on the number of individuals, and 12% and 14%, respectively, based on weight, indicating a difference in catch loss rates between the estimation methods. In this study, the average body lengths of all captured CFS and SS were 4.8 cm and 15.6 cm, respectively, both smaller than their minimum maturation lengths of 15.9–17.2 cm [48] and 13.0–16.7 cm [49]. This suggests that smaller species can be caught in large quantities in the BRD and subsequently escape, which could be beneficial from a resource management perspective. The average rate of BRD-induced catch loss was reported to be 4.86% when the escape opening of the marine mammal BRD was positioned upward and 8.46% when it was positioned downward in stow net gear [11]. The TED-induced catch loss rate in Atlantic trawl gear was reported to be 7–8% [50]. While the catch loss due to the bycatch exclusion device was similar to other catch losses at a 30° tilt angle, it was higher at a 45° tilt angle.

In this study, the vertical length and the number of meshes of the BRD varied depending on the tilt angle. Moreover, although the BRD used in this study was based on the TED principle for beam trawls, unlike the TED, it was made of ordinary netting material rather than a metal grid [14,51]. If installed over the same area inside the body of the trawl net, the difference in vertical length would lead to different drag coefficients and resistance on the meshes, potentially altering the shape of the BRD depending on the tilt angle. This could explain why the BRD with a 45° tilt angle, which has a different mesh shape due to the changed angle, is more likely to block the escape from the BRD, resulting in a higher discharge rate compared to the BRD installed at a 30° tilt angle, which has more meshes. In a midwater trawl model experiment comparing shapes based on codend mesh size changes, it was reported that a non-tapered codend with more meshes was less likely to deform, despite experiencing greater resistance, compared to a codend with fewer meshes [52].

4.2.2. Selectivity According to the Mesh Size of the BRD

From a mesh selectivity perspective, we examined the ratio of the maximum girth to the internal mesh perimeter of the BRD (800 mm), based on the size of each species. Mesh selectivity is known to occur when the maximum girth of a fish species exceeds the internal mesh perimeter, resulting in its capture [22,23]. In this study, the maximum girth of CFS ranged from 3.0 to 12.5 cm, with the ratio to the mesh perimeter ranging from 0.03 to 0.12 across all size classes. SS showed a similar range of 3.3 to 13.0 cm, resulting in a ratio to the mesh perimeter of 0.03 to 0.13. For PS, the maximum girth ranged from 1.8 to 2.8 cm across all size ranges, with a ratio to the mesh perimeter of 0.02–0.03. Even though all three species had a maximum girth significantly smaller than the mesh perimeter, they were still selected by the BRD, resulting in bycatch. This suggests that the mesh selectivity of different species is influenced more by factors such as the hanging ratio or the projected area of the mesh, depending on the tilt angle, rather than by the mesh size alone. A selectivity study in the Baltic Sea also reported that selectivity varies not only with the size of the mesh but also with the mesh shape and hanging ratio [29,53]. The BRD used in this study had a tilt angle, which suggests that species with a smaller maximum girth could not entirely pass through the BRD when viewed from the hauling direction. That is, the lower ratio of maximum girth to the mesh perimeter of the BRD and the angle-dependent variation in selectivity suggest that if the mesh size is set to allow most species to escape, the size of the escaped catch may vary depending on the tilt angle.

4.2.3. Selectivity According to the Behavioral and Biological Characteristics of Fish Schools

When considered from the perspective of fish behavior and biological characteristics, the BRD selectivity and catch loss for the three species are believed to stem from their specific behavioral and biological traits. For example, CFS have been reported to exhibit diel vertical migration [54], which may influence their interaction with the BRD. Additionally, it has been reported that when encountering obstacles such as trawl gear, CFS tend to move vertically to avoid them [55]. A comparative study of catch efficiency based on the shape of the trawl gear entrance found that the catch efficiency for another squid species nearly doubled when the mesh height at the net entrance was increased [15]. Therefore, the catch loss of CFS due to the BRD is likely attributable to their vertical migration behavior to escape the BRD, regardless of size.

SS primarily inhabit the lower layers and are distributed vertically depending on water temperature, generally residing near the seabed [56]. Therefore, their behavioral characteristics are not considered significantly related to BRD selectivity. Nonetheless, the observed catch loss suggests that selectivity may be influenced by their physical features, such as their hard fins. Another study reported that selectivity could change seasonally due to physical characteristics, even within the same species [57]. In this study, SS, despite being a benthic species, were frequently observed swimming at lower depths for feeding in middle water. This behavior indicates that selectivity could also change due to seasonal variations in water temperature and other factors associated with depth changes [57]. Therefore, it is highly likely that selectivity may vary with changes in water temperature related to towing depth.

Unlike CFS and SS, the BRD-induced catch losses of PS were as low as 1% and 4%, indicating that catch loss may not be a significant issue for this species. However, even these low catch losses indicate a potential association between catch loss and the physical characteristics and behavior of the fish schools. PS are characterized by small, slender bodies [11,58,59]. In this study, PS were primarily caught moving from the BRD to the codend but were also frequently caught on the wings or body nets of the trawl gear and in the mesh of the BRD. Additionally, PS exhibit vertical migration behavior and tend to move in large schools [42]. Thus, it is believed that they display behavioral characteristics that allow them to escape from the BRD into the codend or get caught vertically in the covernet of the BRD. However, the catch loss rate of PS at a 30° tilt angle was 99%, which is more effective in reducing non-target species catch loss compared to the 4% loss at a 45° tilt angle, suggesting that a 30° tilt angle is the optimal angle.

The analysis of these three species revealed that catch loss could be influenced by the behavioral and biological characteristics of each fish school, underscoring the need to consider alternative methods to reduce catch loss. For trawl gear targeting shrimp, significant reductions in catch loss have been achieved using auxiliary devices such as metal frames on the BRD or fish eyes installed at escape openings [60,61].

4.3. The Masking Effect between Codend and Covernet

Various codend selectivity studies have also been conducted on trawl gear with the aim of catching mature individuals while conserving juvenile resources [62,63,64,65,66,67,68,69,70]. The International Council for the Exploration of the Sea (ICES) recommends using gear that meets the design requirements of the mesh selectivity manual. For example, the total length of the covernet should be at least twice that of the codend, the width of the covernet at least 1.5 times that of the codend, and hoops should be used to maintain sufficient space between the codend and the covernet [32]. If these requirements are not met, a masking effect can occur, in which small fish that could escape are caught, which significantly impacts selectivity [71]. In other studies, sufficient space was secured by using hoops [16], or a kite covernet [72], or by placing a camera between the codend and covernet to ensure adequate space [73].

In this study, however, fish that passed through the BRD encountered a triple-layered codend with mesh sizes of 122 (A), 62 (B), and 28 (C) mm, with the covernet functioning as codend nets A and B, as illustrated in the schematic diagram in Figure 2. This gear did not meet the design requirements specified in the manual, and the insufficient space between the codend and covernet resulted in a masking effect. To address this issue, further research should implement measures to minimize the masking effect of the codend. Studies examining the masking effect have demonstrated that the 50% selectivity length (${\mathrm{L}}_{50}$) of a codend with a covernet is smaller than that of a codend without a covernet, likely due to the hoop attached to the covernet, indicating a difference in selectivity between the two trawl net types [74]. However, there has been a scarcity of reports on the masking effect of covernets without hoops. Most studies have focused on avoiding the masking effect in selectivity studies of the codend, with little research investigating the impact of the masking effect per se.

5. Conclusions

In summary, this study focused on estimating the selectivity curves for three species based on two mesh selectivity options based on the tilt angles of the marine mammal BRD. However, despite the large mesh size of the BRD, some species that could have passed through the mesh instead escaped from the net, leading to BRD-induced catch loss. When comparing BRD-induced catch losses between the two tilt angles, a 30° angle was found to have lower catch loss rates than a 45° angle. Considering these factors, follow-up research will involve mounting cameras onto the BRD to determine why smaller fish are able to escape the BRD. Camera monitoring will also help confirm whether fish caught during hauling could escape due to the extended time required for hauling. We only conducted a midwater trawl survey on the East Coast of Korea. Future studies will be conducted on the South Coast of Korea, which is known for frequent whale sightings and other target fish species inhabiting it. Additionally, we aim to develop auxiliary devices for the escape openings of the marine mammal BRD to reduce the catch loss of target species. On the other hand, the codend did not provide sufficient space between itself and the covernet, resulting in a masking effect. While a masking effect can be beneficial for stock assessment, it should be avoided in selectivity studies. To address this issue, future research will utilize the covered codend method, following mesh selectivity guidelines, to improve the accuracy of selectivity while avoiding the masking effect. Additionally, studies will be conducted in areas with diverse species to estimate BRD-induced catch losses across different species.