Catch Losses and Reduction of Bycatch for Jellyfish Using Marine Mammal Bycatch Reduction Devices in Midwater Trawl Gear

Abstract

The National Institute of Fisheries Science in Korea is developing marine mammal bycatch reduction devices (BRDs) for midwater trawl gear. In this study, we tested two BRD-type guide nets (inclined net panel) with 30° and 45° tilt angles to prevent marine mammals from reaching the codend and facilitating their escape from the net. Fishing operations were conducted along the east and south coasts of South Korea, and cameras were installed in front of the BRDs to monitor their performance. The catch loss of herring with the 30° guide net was 13% and 11% in number and weight. The catch loss of hairtail was 53% and 51% in number and weight with the 30° guide net. Mackerel showed a 97% catch loss in number and weight with the 45° guide net. The 30° guide net resulted in lower catch loss for rudderfish and jack mackerel compared to the 45° guide net. The jellyfish discard rate of the BRD was 5% and 7% in number and weight with the 30° guide net and 12% and 11% with the 45° guide net, indicating that the 30° guide net was more effective at discarding jellyfish. Mesh selectivity was not strongly related to target species body length.

1. Introduction

Cetaceans, which are among the most well-known marine mammals, have historically been considered as competitors to fisheries owing to their perceived role in declining fish stocks [1]. However, subsequent studies have reported a limited impact on fisheries [2]. In the past, cetaceans were perceived as competitor for fisheries resources [3], but now, cetaceans are increasingly recognized as an important resource for carbon problems, since carbon emissions have been designated as a major cause of climate change [4]. These findings underscore the importance of protecting cetaceans and other marine mammals to maintain ecological balance and support fisheries.

The United States implemented the Marine Mammal Protection Act (MMPA), with numerous studies conducted to reduce marine mammal bycatch across various fisheries [5]. In Korea, the National Institute of Fisheries Science (NIFS) is developing bycatch reduction devices (BRDs) for trawls that comply with the equivalency assessment standards of the National Oceanic and Atmospheric Administration (NOAA) and the MMPA. Additionally, a BRD for trawl gear was recently developed and tested [6].

The development of BRDs using acoustic technologies, such as pingers, in gillnets, traps, and trawls has been actively investigated [7,8,9,10,11,12]. However, marine mammals may initially avoid pingers, they may eventually become habituated, which could result in a resurgence in bycatch [10,12,13]. Moreover, the continuous use of pingers could deter marine mammals from accessing coastal waters, thereby leading to substantial disruptions to marine ecosystems. Marine mammals, including whales and dolphins, contribute to the vertical mixing of ocean waters through foraging [14,15], supplying essential mechanical energy to the ocean [16]. Cetaceans support primary production by transporting nutrients to the surface via respiration, digestion, and fecal release, and facilitate carbon sequestration by transporting carbon to deeper ocean layers [17,18,19]. Furthermore, the migration of humpback whales across oceans is considered to contribute to both primary production and carbon sequestration [20]. We designed BRDs that enable marine mammals to escape from fishing gear without the use of pingers, considering coexistence with them in the sea.

In 2024, trawl gears equipped with marine mammal BRDs at 30° and 45° tilt angles were operated off the Korean East Coast in June to assess catch losses for three fish species: common squid, sandfish, and pearlside. For species exhibiting high catch losses, the causes were inferred based on their body and behavioral characteristics [6]. However, the East Sea has seasonal inflows of warm and cold currents, and various fish species are distributed by seasonal change in Korea [21,22]. Additionally, an influx of jellyfish populations from July to late August in Korean coastal areas means that these highly toxic organisms has damaged catch [23]. The jellyfish BRD was developed in the same shape as the BRD marine mammals in a previous study [6,23,24]. It is the same in terms of reducing the bycatch marine mammals and jellyfish, and the ability to escape jellyfish suggests that marine mammals can also be escaped from BRD. Therefore, further studies are necessary to verify the estimated catch loss rates for different fish species in the East Sea and evaluate the effectiveness of BRDs with different tilt angles in reducing jellyfish bycatch.

Previous studies on BRDs in the East Sea have reported a catch loss rate of over 10% for trawl gear equipped with marine mammal BRDs in June [6]. In contrast, stow nets equipped with finless porpoise excluder devices in the West Sea exhibited a low catch loss rate of 4%, resulting in relatively wider adoption by fishers. Therefore, this study aimed to estimate the mesh selectivity and catch loss of gear with marine mammal BRDs with understanding the escaping process and assess the effectiveness of BRDs in reducing jellyfish bycatch, particularly during the summer.

2. Materials and Methods

2.1. Test Fishing Gear with Bycatch Reduction Device

This study used a midwater trawl, as classified by the Food and Agriculture Organization (FAO), with same fishing gear as that used in a previous study [6]. Schematic diagrams of the midwater trawl and BRD installation without pinger are shown in Figure 1 and Figure 2. The net had a total length of 140 m (excluding the towing warp), including a head rope and ground rope measuring 59.3 m, with the codend (hereinafter COD) measuring as 11.2 m. As shown in Figure 1, the COD comprised three layers of netting with varying mesh sizes: 120 (Codend A), 62 (Codend B), and 28 mm (Codend C), from the outermost to the innermost layer. For the marine mammal bycatch, a guide net equipped with a net panel as a BRD was installed directly in front of the COD of the midwater trawl gear (Figure 2). The net panel was made using square mesh polyethylene (PE) netting with a mesh size of 200 mm (mesh perimeter 800 mm), and two guide nets with tilt angles of 30° and 45° were prepared for use. Both nets had the same width (2.0 m), but their length differed depending on the tilt angle: 2.23 m for 30° and 1.4 m for 45°. The position where the body net was attached varied accordingly. The escape opening measured 2 × 1.5 m (vertical × horizontal), and a covernet with a mesh size of PE 46 ± 1.4 (D) mm was attached to the escape opening to capture fish species that escaped into the BRD via the guide net.

2.2. Fishing Area and Sea Trial

As shown in Figure 3, this study conducted fishing operations in the southern [24] and eastern [25,26] coastal regions of Korea, where marine mammals such as whales, dolphins, and finless porpoises are frequently sighted.

Fishing operations were conducted for 12 days, from 2 to 13 September 2024, using the NIFS research vessel Tamgu-20 (comprising 885 tons and 2600 horsepower). All operations were conducted during the daytime, with each trawl lasting 1 h at a trawling speed of 4.5–5 knots. Seventeen trawls were conducted at 200–300 m depth, targeting the surface (20–50 m) and midwater (100–150 m) layers by changing towing depth from living different depth in fish finder and sonar. After hauling, the catch was sorted to species levels and classified into BRD and COD based on capture location. A complete survey of all captured fish was conducted, measuring body length, as fork length (FL), total length (TL), mantle length (ML), anal length (AL), maximum girth (G) to the nearest millimeter, and weight (W) in grams. From these data, we calculated the relationship between body length and maximum girth for comparing with their body size and mesh size of BRDs’ guide net.

2.3. Camera Observation

In a previous study, we discussed the possibility of escape because of long times of hauling on net or species behavior characteristics [6]. Thus, we need to record to how target species move in the net. A camera (GoPro Hero 6) was installed facing the front of the guide net (Figure 4) to monitor conditions inside the fishing gear during operation. The camera was positioned 1.5 m in front of the starting point of the net leading to the COD. The angle was calibrated to shoot a photo of both the catch entering the COD through the guide net and escaping through the escape opening. Camera shots were taken every second. The recorded photo was analyzed and compared with catch data to determine the pathways that fish used to pass through the guide net or escape via the escape opening (

Table 1. Dates and positions of codend mesh selectivity experiments by Tamgu-20.

Haul No.	Date	Start Position of the Haul		Underwater Camera Record
		Latitude	Longitude
#101	2 September 2024	34°56.058′ N	129°03.446′ E	-
#102	3 September 2024	34°31.000′ N	128°24.128′ E	-
#103	4 September 2024	34°46.076′ N	128°37.967′ E	-
#104	4 September 2024	36°50.784′ N	129°39.087′ E	-
#105	4 September 2024	35°56.072′ N	129°40.945′ E	-
#106	5 September 2024	35°58.026′ N	129°40.785′ E	-
#107	5 September 2024	35°05.693′ N	129°42.921′ E	-
#108	6 September 2024	36°09.762′ N	129°43.325′ E	✓
#109	6 September 2024	36°16.434′ N	129°34.745′ E	✓
#110	8 September 2024	36°03.900′ N	129°41.283′ E	✓
#111	9 September 2024	36°00.205′ N	129°41.414′ E	✓
#112	9 September 2024	36°00.205′ N	129°41.706′ E	✓
#113	10 September 2024	35°42.863′ N	129°41.067′ E	✓
#114	10 September 2024	35°43.841′ N	129°43.267′ E	✓
#115	11 September 2024	35°43.059′ N	129°44.584′ E	✓
#116	11 September 2024	35°43.577′ N	129°41.777′ E	✓
#117	12 September 2024	35°31.000′ N	129°34.210′ E	✓

“✓” designates a tow which the underwater camera recorded.

).

2.4. Data Analysis

Mesh selectivity can vary across different fishing operations, even for the same species [27]. For example, factors such as water temperature, which affects swimming ability, and increased girth during spawning seasons have notably influenced the mesh selectivity of COD in Mediterranean trawl fisheries [28,29]. Consequently, the method of averaging mesh selectivity for each haul has become widely adopted [30,31,32]. In this study, we analyzed the results separately for BRDs with tilt angles of 30° and 45°, as well as for the combined case. The probability of a species with length class $l$ passing through the BRD and being caught in the COD ($r_c\left(l\right)$) is expressed as follows:

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

(1)

where $n_{l1}$ and $n_{l2}$ represent the number of length class $l$ fish caught in the BRD and COD, respectively. A representative $r_c\left(l\right)$ value, $r\left(l\right)$, is expressed as a function of length class $l$ and is approximated using a logistic equation according to the analytical method described below:

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

(2)

where $a$ and $b$ are parameters of the logistic equation. The mesh selectivity of species for the BRD with a 400 mm mesh guide net described earlier was estimated using the R package selfisher (selectivity of fisheries gear, version 4.3.3) [33]. Selfisher is a statistical package that uses the SELECT model based on maximum likelihood estimation [34,35,36,37] to estimate the selectivity of fishing gear using generalized linear mixed models (GLMMs). It supports the fixed effects model, which aggregates data across all hauls by length class to estimate mesh selectivity for each length class, and the random effects model, which assumes that each haul is influenced by random effects such as fishing conditions or environmental factors. Bootstrapping can produce multiple selectivity curves from haul data, which can be used to calculate a mean curve. For this study, we estimated mesh selectivity for the 30° and 45° guide nets, overall hauls, and the COD. Catch loss for each species, classified into BRD and COD, was calculated using the method shown in [6]:

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

(3)

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

3. Results

3.1. Body Length Composition

Table 2. Total catch results from 17 fishing operations of midwater trawl in the fishing ground.

Common Name	Scientific Name	Number Caught	Weight of Catch
Hairtail	Trichiurus lepturus	597	30.77 kg
Rudderfish	Psenopsis anomala	55	4.07 kg
Jack mackerel	Trachurus japonicus	61	1.43 kg
Mackerel	Scomber japonicus	417	58.25 kg
Herring	Clupea pallasii	264	8.42 kg
Firefly squid	Watasenia scintillans	265	2.01 kg
Pearlsides	Maurolicus muelleri	1,144,551	1230.20 kg
Nomura’s jellyfish	Nemopilema nomurai	47	807.47 kg
Total		1,146,257	2142.62 kg

shows the results of 17 fishing operations conducted over 12 days. A total of 1659 fish were caught, comprising 264 herring Clupea pallasii, 597 hairtail Trichiurus lepturus, 417 mackerel Scomber japonicus, 55 rudderfish Psenopsis anomala, 61 jack mackerel Trachurus japonicus, and 265 firefly squid Watasenia scintillans. Additionally, 47 jellyfish, including Nomura’s jellyfish Nemopilema nomurai, were caught. But marine mammals were not caught in this study. The total weight of the fish catch was 107.8 kg, comprising herring (8.5 kg), hairtail (30.8 kg), mackerel (60.1 kg), rudderfish (4.1 kg), jack mackerel (1.5 kg), and firefly squid (2.8 kg). The jellyfish weighed 807.5 kg. These six fish species, excluding jellyfish, accounted for 97% of the catch by number and 12% by weight. Figure 5 shows the size distribution of the six species, excluding jellyfish. Herring captured in the COD had FLs of 12–18 cm, whereas those in the BRD had FLs of 15–18 cm. Hairtails were captured with ALs of 11–20 cm in both the COD and BRD. Mackerel in both sections had FLs of 20–25 cm. Rudderfish in the COD had FLs of 9–14 cm, whereas those in the BRD had FLs of 10–14 cm. Jack mackerel had FLs of 9–16 cm in the COD and 9 cm or 12–16 cm in the BRD, with no fish in the 10–11 cm range.

3.2. Relationship Between Body Length and Maximum Girth (And Weight)

Based on a complete survey of the catch, relationships between body length (FL for herring and mackerel, AL for hairtail) and G were estimated for hairtail, herring, and mackerel, with each species having a sample size from the number caught greater than 200.

The relationships between the body length (AL, FL) and the maximum girth (G) of hairtail, herring and mackerel were expressed as follows:

• Hairtail: G = 0.4716 × AL + 0.7431 (3) ($r$ = 0.899, $p$ < 0.05, $n$ = 597, Figure 6)
• Herring: G = 0.3671 × FL + 1.5031 (4) ($r$ = 0.611, $p$ < 0.05, $n$ = 264, Figure 6)
• Mackerel: G = 0.4909 × FL + 0.0813 (5) ($r$ = 0.771, $p$ < 0.05, $n$ = 417, Figure 6)

The results indicated a significant positive linear relationship between body length (AL or FL) and both G and W for all three species ($p$ < 0.05).

3.3. Bycatch Reduction and Catch Loss Inside the Gear

Table 3. Fish behavior monitoring inside the gear with BRD.

Haul No.	Date *1	Recorded Duration	Remark
			COD *2					BRD *2
			H	R	JM	M	J	H	R	JM	M	J
#108	6 September 2024	00 h 59 min 50 s	1	-	2	-	1	-	-	36	-	6
#109	6 September 2024	00 h 59 min 20 s	2	-	15	-	1	-	-	5	-	13
#110	8 September 2024	00 h 34 min 19 s	-	3	-	-	-	-	1	-	-	-
#111	9 September 2024	00 h 59 min 01 s	4	2	-	-	1	12	1	-	-	5
#112	9 September 2024	01 h 00 min 34 s	28	-	-	-	-	39	-	-	-	-
#113	10 September 2024	00 h 50 min 25 s	-	-	-	10	-	-	-	-	30	-
#114	10 September 2024	01 h 02 min 22 s	-	-	-	-	1	-	-	1	-	5
#115	11 September 2024	01 h 00 min 44 s	-	-	-	-	-	10	1	-	-	2
#116	11 September 2024	00 h 44 min 17 s	12	-	-	-	-	53	-	-	-	-
#117	12 September 2024	00 h 27 min 27 s	10	-	-	-	-	11	-	-	-	3

*1, all observations were carried out in the daytime. *2, number of observed species (“H”, hairtail, “R”, rudderfish, “JM”, jack mackerel, “M”, mackerel, “J”, Nomura’s jellyfish) were categorized by observed position (COD or BRD).

presents the monitoring results inside the fishing gear. During two hauls using a 45° BRD, the camera recorded 18 jack mackerel, 2 hairtails, 1 common squid, and 22 Nomura’s jellyfish. Three jack mackerel and one common squid entered the COD, whereas fifteen jack mackerel escaped through the BRD. However, only 2 Nomura’s jellyfish entered the COD, with 20 escaping through the BRD.

During seven hauls using a 30° BRD, the camera recorded 1 common squid, 4 firefly squid, 2 rudderfish, 87 hairtails, 1 jack mackerel, 3 blue crabs, and 15 Nomura’s jellyfish. Of these, 1 common squid, 3 firefly squids, 32 hairtails, and 3 blue crabs were caught in the COD, whereas 1 firefly squid, 2 rudderfish, 55 hairtails, and 1 jack mackerel escaped through the BRD. One Nomura’s jellyfish entered the COD, whereas fourteen escaped through the BRD. Additionally, a significant number of pearlside and mackerel were filmed inside the fishing gear. Most pearlside were caught in the COD; however, most mackerel, presumably caught while hunting pearlside, escaped through the BRD.

Camera observation analysis revealed that fish entering the gear from below typically passed through the BRD and were captured in the COD, whereas those entering from above bypassed the BRD and escaped. In contrast to the fish, Nomura’s jellyfish frequently collided with the BRD and were deflected upward, escaping through the opening even when they entered from below. Jellyfish that were eventually found in the COD did not pass through the BRD but were trapped in the narrow spaces between the guide and body nets. Many of these jellyfish adhered to the sides of the BRD during towing and subsequently entered the COD during hauling or were torn apart by the guide net and strong currents, with the resulting fragments entering the COD.

3.4. Mesh Selectivity of BRD

The mesh selectivity of BRDs was analyzed by fish species and guide net tilt angle. Mesh selectivity for the 30° and 45° BRDs, as well as for the combined data, is summarized in Figure 7 and

Table 4. Estimated parameters for sorting selectivity curves after capture (hairtail).

Common Name (Angle of BRD)	Estimation Method *1	Logistic Parameters *2			Selectivity Curve Parameters *3			Model Fit
		a	b	L50	C.I. of L50	S.R.	C.I. of S.R.	Deviance	df	p-Value
Hairtail (30°)	Average of each haul	−0.110 (0.841)	0.005 (0.072)	20.8	Error	416.3	Error	0	19	0.98
	Sum of hauls	−0.032 (0.732)	0.001 (0.052)	28.7	-	1997.2	-	0	19	0.94
Hairtail (45°)	Average of each haul	−0.161 (0.845)	0.003 (0.072)	52.3	Error	715.5	Error	0	19	0.97
	Sum of hauls	−0.019 (2.82)	0.002 (0.191)	9.2	-	1046.3	-	0	19	0.99
Hairtail (total)	Average of each haul	−0.096 (0.843)	0.004 (0.072)	21.4	Error	491.3	Error	0	19	0.95
	Sum of hauls	−0.003 (0.071)	0.001 (0.005)	299.0	-	21974.9	-	0	19	0.99
Rudderfish (30°)	Average of each haul	−9.16 (5.26)	0.55 (0.21)	16.6	15.2–17.9	4.0 (2.7)	Error	0	15	0.28
	Sum of hauls	−11.20 (7.68)	0.67 (0.53)	16.5	-	3.2 (0.5)	-	0.1	15	0.19
Rudderfish (45°)	Average of each haul	−2.63 (1.55)	0.12 (0.01)	22.8	Error	19.0 (12.3)	Error	0	15	0.42
	Sum of hauls	−2.60 (1.77)	0.11 (0.03)	20.4	-	20.4 (16.3)	-	0.1	15	0.75
Rudderfish (total)	Average of each haul	−4.86 (2.95)	0.26 (0.13)	18.7	Error	8.4 (4.6)	Error	0	15	0.27
	Sum of hauls	−4.97 (2.72)	0.27 (0.16)	18.6	-	8.2 (4.2)	-	0.1	15	0.32
Jack mackerel (30°)	Average of each haul	−11.31 (7.79)	0.72 (0.31)	15.6	15.0–16.2	3.0 (1.6)	0.1–4.8	0	15	>0.05
	Sum of hauls	−7.90 (4.47)	0.47 (0.21)	16.7	-	4.6 (2.1)	-	0	15	<0.05
Jack mackerel (45°)	Average of each haul	−3.63. (1.94)	0.22 (0.16)	16.6	14.9–18.3	10.0 (6.63)	Error	0	15	0.18
	Sum of hauls	−3.37 (1.95)	0.20 (0.15)	16.8	-	11.0 (7.51)	-	0	15	0.19
Jack mackerel (total)	Average of each haul	−5.95 (3.57)	0.37 (0.27)	16.1	15.5–16.7	5.9 (4.1)	Error	0	15	>0.05
	Sum of hauls	−6.00 (2.02)	0.37 (0.15)	16.4	-	6.0 (2.5)	-	0	15	0.02
Mackerel (30°)	Average of each haul	−35.97 (26.74)	1.81 (1.34)	19.9	19.8–20.0	2.2 (0.9)	2.0–2.4	0	24	<0.05
	Sum of hauls	−35.00 (4.38)	1.75 (0.21)	20.0	-	2.3 (0.1)	-	0	24	<0.01
Herring (45°)	Average of each haul	−20.61 (16.13)	1.20 (0.84)	17.1	17.0–17.2	2.1 (1.3)	2.0–2.2	0	17	0.12
	Sum of hauls	−18.08 (3.11)	1.05 (0.19)	17.1	-	2.1 (0.4)	-	0	17	<0.01

*1, average of each haul; parameters were estimated by a random effects model for each haul and average values are shown. Sum of hauls parameters were estimated from catch results summed from all hauls by a fixed model. *2, parameters for Equation (2) in text. Values in parentheses are standard errors. *3, C.I., the 95% confidence interval, error; the value of C.I. (95% confidential interval) of L₅₀ and C.I. could not be estimated.

.

In species of hairtail, no significant mesh selectivity was found in all cases (p > 0.05). For the 30° BRD, the selectivity rate remained below 0.5 for all body lengths. For the 45° BRD, the selectivity rate for most measured body lengths was estimated to be between 0.4 and 0.5. When data from both angles were combined, the selectivity rate was approximately 0.5 in the sum of the haul, while the selectivity rate rose with larger body length in the average of each haul. The L₅₀ (body length at 50% retention probability) was estimated as follows:

• 30° BRD: 28.7 cm (sum of haul) and 20.8 cm (average of each haul)
• 45° BRD: 9.2 cm (sum of haul) and 52.3 cm (average of each haul)
• Combined data: 299.0 cm (sum of haul) and 21.4 cm (average of each haul).

The estimated selective range (SR) varied in all cases. The 95% confidence interval (95% CI) could also not be estimated for the L₅₀ and SR in all cases.

In rudderfish mesh selectivity for the 30° and 45° BRDs, as well as for the combined data, no significant mesh selectivity was observed in all cases (p > 0.05). For the 30° BRD, the selectivity rate rose with larger body length, but remained below 0.5 for all body lengths. For the 45° BRD, the selectivity rate for most measured body lengths was estimated between 0.1 and 0.3. When data from both angles were combined, the selectivity rate rose with larger body length, but remained below 0.3 for all body lengths. The L₅₀ (body length at 50% retention probability) was estimated as follows:

• 30° BRD: 16.5 cm (sum of haul) and 16.6 cm (average of each haul)
• 45° BRD: 20.4 cm (sum of haul) and 22.8 cm (average of each haul)
• Combined data: 18.6 cm (sum of haul) and 18.7 cm (average of each haul).

The estimated SR was different between the 30° and 45° BRDs. The 95% CI could not be estimated for the L₅₀ and SR in all models except for the 30° BRD, for which the 95% CI for L₅₀ was 15.2–17.9 cm.

In jack mackerel mesh selectivity for the 30° and 45° BRDs, as well as for the combined data, no significant mesh selectivity was observed in all cases (p > 0.05) except for the 30° BRD and combined data with fixed effects model (sum of haul, p < 0.05). The selectivity rate was rising with larger body length in all cases. The L₅₀ was estimated as follows:

• 30° BRD: 16.7 cm (sum of haul) and 15.6 cm (average of each haul)
• 45° BRD: 16.8 cm (sum of haul) and 16.6 cm (average of each haul)
• Combined data: 16.4 cm (sum of haul) and 16.1 cm (average of each haul).

The estimated SR was different between the 30° and 45° BRDs. The 95% CI could not be estimated for the L₅₀ and SR in all models except for the 30° BRD, for which the 95% CI for L₅₀ was 15.0–16.2 cm, and 95% CI for SR was 0.1–4.8 cm.

In mesh selectivity for mackerel (30° BRD) and herring (45° BRD), significant mesh selectivity was observed in either model for both species (p < 0.05). The selectivity curve for mackerel with the 30° BRD showed a 1.0 selectivity rate in larger body length, while the selectivity curve for herring with the 45° BRD was not reached. The selectivity curve was sharper for mackerel with the 30° BRD than for herring with the 45° BRD. The L₅₀ was estimated as follows:

• Mackerel (30° BRD): 20.0 cm (sum of haul) and 19.9 cm (average of each haul)
• Herring (45° BRD): 17.1 cm (sum of haul) and 17.1 cm (average of each haul).

The estimated SR was 2.2–2.3 cm for mackerel and 2.1 cm for herring. The 95% CI estimated for the L₅₀ and SR was 19.8–20.0 cm and 2.0–2.4 cm for mackerel and 17.0–17.2 cm and 2.0–2.2 cm for herring.

3.5. Catch Loss Results by BRD and Bycatch Results for Jellyfish by BRD

Table 5. Catch loss rate for target species by using BRD.

Name	Scientific Name	Catch Loss Rate (%) *1
		30°		45°		Total (Average)
		N	W	N	W	N	W
Hairtail	Trichiurus lepturus	53	52	52	58	53	54
Rudderfish	Psenopsis anomala	18	19	24	26	22	23
Jack mackerel	Trachurus japonicus	14	16	20	19	18	18
Mackerel	Scomber japonicus	97	97	-	-	97	97
Herring	Clupea pallasii	-	-	13	11	13	11
Nomura’s jellyfish *2	Nemopilema nomurai	93	93	88	89	89	90

*1, Catch loss rate (“N”, the number caught, “W”, the weight of the catch) was categorized for each species. *2, catch loss rate of jellyfish is the opposite concept of bycatch rate, which means that the higher rate, the more escape.

presents the catch loss rates for each fish species at different tilt angles of the BRD. Catch loss rates, both in number caught and weight, varied by species and tilt angle: hairtail: 30° BRD: 53.3% (number caught) and 51.8% (weight) and 45° BRD: 51.9% (number caught) 58.4% (weight); rudderfish: 30° BRD: 18.2% (number caught) and 19.1% (weight) and 45° BRD: 24.3% (number caught) and 25.9% (weight); jack mackerel: 30° BRD: 14.1% (number caught) and 15.9% (weight) and 45° BRD: 20.1% (number caught) and 19.4% (weight).

Hairtail showed slightly lower catch loss rates (number caught) with the 45° BRD compared to the 30° BRD, whereas rudderfish and jack mackerel showed the opposite trend. These differences in catch loss rates between the two tilt angles were statistically significant for both number caught and weight (p < 0.05).

For mackerel and herring, data were only available for the 30° and 45° BRDs, respectively. The observed catch loss rates were: 96.9% (number caught) and 96.9% (weight) for mackerel (30° BRD) and 20.1% (number caught) and 19.4% (weight) for herring (45° BRD).

The escape mechanism of jellyfish and their bycatch rates following BRD installation are presented in Figure 8 and Table 5. All jellyfish which remained at COD were cut off. Jellyfish catch loss rates were 93.4% (number caught) and 92.6% (weight) with the 30° BRD and 87.9% and 89.1%, respectively, with the 45° BRD. It means that jellyfish bycatch rate is approximately 7% in 30° BRD and 11~12% in 40° BRD. Statistical analysis revealed a significant effect of BRD tilt angle on bycatch rates for number caught and weight (p < 0.05), with the 30° BRD demonstrating superior performance in reducing jellyfish bycatch.

4. Discussion

The fishing gear used in this study was designed for resource survey to support the total allowable catch (TAC) system [38,39]. The small COD mesh size enabled accurate resource estimation and increased the potential for capturing a wide range of fish species and sizes. Installing a developing marine mammal BRD potentially reduced bycatch, but also presented a risk of target species catch loss. Therefore, we investigated methods to minimize marine mammal bycatch while reducing target species catch loss by adjusting BRD and COD selectivity.

4.1. Importance of Caught Fish Species

The fish species caught in this study included hairtail, rudderfish, jack mackerel, mackerel, and herring. Notably, hairtail, jack mackerel, and mackerel are key commercial species that are widely consumed in Korea and managed under the TAC system. In particular, hairtail is an important target species in trawl fisheries. The rudderfish, which is primarily harvested and consumed in Taiwan [40], is anticipated to be a valuable resource for Korea in the future as climate change alters their habitats. These species commonly prey on pearlside, as evidenced by our catch data and the presence of pearlside in their mouths. Our previous study on marine mammal BRDs in the East Sea also identified pearlside as a key prey species [6], further underscoring the need to develop fishing gear that reduces pearlside bycatch.

4.2. Efficiency of the Bycatch Reduction Device

Species-specific BRD selectivity was analyzed in relation to BRD tilt angle, mesh selectivity, fish school behavior, and biological characteristics. A marine mammal BRD with a 400 mm mesh size was installed directly in front of the COD, and two tilt angles (30° and 40°) were evaluated. For this study, catch loss due to the BRD was defined as the capture of fish within the BRD, as it prevented them from entering the COD.

4.2.1. Selectivity Regarding the Bycatch Reduction Device Tilt Angle

To evaluate the selectivity of BRDs with different tilt angles, this study compared the catch loss rates of hairtail, rudderfish, and jack mackerel by fish number and weight. Hairtail exhibited a lower catch loss rate with the 45° BRD than with the 30° BRD, suggesting the 45° BRD is better suited for this species. However, both BRDs showed catch loss rates exceeding 50%, indicating a need for further refinement. In contrast, rudderfish and jack mackerel had lower catch loss rates with the 30° BRD, indicating that it is more appropriate for these species. This is consistent with our previous finding [6] that a 30° tilt angle was more effective than a 45° angle for common squid and sailfin sandfish, although catch loss rates for rudderfish and jack mackerel still exceeded 10%, indicating a need for further improvement. Only a single tilt angle was used during this study for mackerel and herring, making direct comparisons impossible. Further research is needed to determine the optimal BRD tilt angle for these species.

Regarding jellyfish bycatch reduction, the catch rate was lower with the 30° BRD than with the 45° BRD. This indicates that marine mammal BRDs can effectively reduce jellyfish bycatch during periods of jellyfish influx. The selectivity of BRDs in reducing jellyfish bycatch depends on their tilt angle [41,42,43]. Notably, as the tilt angle increased from 10° to 15° and 20°, jellyfish discard rates decreased by 66%, 41%, and 44%, respectively [44]. However, our findings indicated that bycatch rates were lower at 30° than at 45°, suggesting that the bycatch reduction was not linear with increasing tilt angles, but decreased beyond a specific threshold.

Camera observation analysis showed that the net panel of 30° BRD unfolded more effectively compared to the one in 45° BRD, which was less stable. The net unfolded loosely for the 30° BRD, creating ample internal space. Target fish from below entered the COD, whereas those entering from above either reached the COD or were redirected into the BRD. However, for the 45° BRD, the net failed to unfold properly owing to resistance, causing distortion. This led to the front section of the COD remaining partially collapsed, disrupting water flow and resulting in fish being discarded through the BRD. These findings indicate that a tilt angle of 30° is more suitable for the net panel than a 45° tilt angle. Although BRDs with net panels are commonly used in turtle excluder devices [45,46,47], few studies have examined how the unfolding of guide nets affects net performance. The further study of this will establish the groundwork for future research.

4.2.2. Selectivity Regarding the Bycatch Reduction Device Mesh Size

To evaluate the effect of mesh size on BRD selectivity, we analyzed the ratio of the G of hairtail, rudderfish, jack mackerel, mackerel, and herring to the mesh perimeter (800 mm) of the BRD. The selectivity principles state that fish are retained when their girth exceeds the mesh size [48,49]. The G range of each species across all body lengths was as follows: hairtail: 2.3–8.9 cm, rudderfish: 11.1–16.6 cm, jack mackerel: 6.5–8.8 cm, mackerel: 9.8–12.6 cm, and herring: 6.1–8.4 cm, with a ratio to the mesh perimeter of 0.03–0.21. Despite having a G much smaller than the mesh perimeter, some were selected by the BRD, which resulted in catch loss. This suggests that factors such as the hanging ratio and projected area of the mesh, rather than mesh size alone, influence selectivity. Similar findings have been reported in the Baltic Sea, where selectivity was affected by mesh shape and hanging ratio in addition to mesh size [50,51]. These results indicate that the low ratio of G to the mesh perimeter had a minimal impact on selectivity, while other factors, such as the tilt angle of the net panel or fish school behavior, were more influential. Consistent with our earlier findings [6], this study supports the conclusion that mesh size has a relatively smaller effect on selectivity compared to tilt angle.

4.2.3. Selectivity Regarding Fish School Behavior and Biological Characteristics

Although all fish had a G much smaller than the mesh perimeter, catch losses were still observed. This indicates that fish school behavior and biological characteristics considerably influence selectivity compared to mesh selectivity alone. To explore this comprehensively, we examined hairtail behavior. Studies using longlines and gillnets have reported that hairtail schools exhibit diel vertical migration [52]. Additionally, fish finder studies have revealed that the daytime swimming layer of hairtail schools is deeper than their nighttime swimming layer [53]. These findings indicate that the substantial catch loss in hairtail may be attributed to their vertical migration patterns and deeper daytime distribution. Camera observation confirmed that most hairtail entering the net were positioned near the upper section of the gear, making it easier for them to escape through the BRD located at the top. Rudderfish, classified as demersal fish [46], remained attached to the bottom of the gear, resulting in a lower catch loss compared to hairtail. In contrast, jack mackerel and herring, classified as pelagic migratory fish, are characterized by high swimming speeds [53]. Camera observation revealed that these species swam continuously in front of the guide net before eventually passing through the BRD and entering the COD after fatigue. As larger jack mackerel and herring were smaller than the mesh size, they could be guided into the COD through the guide net in the upper section, even when schooling near the top of the gear, resulting in relatively lower catch loss. In contrast, mackerel, another pelagic migratory species with high swimming speeds [54,55,56], exhibited a catch loss rate of 97%, as most escaped through the BRD. Mackerel have larger school sizes, faster swimming speeds, and greater body size than jack mackerel and herring [55,56]. Camera observations captured an event where numerous mackerel collided with the guide net and escaped through the BRD. Unlike jack mackerel and herring, mackerel flipped over when exhausted; however, owing to their larger size, they could not pass through the mesh and ultimately escaped through the BRD. These findings indicate that the current mesh size of the guide net is too small. To address this, the mesh size should be increased, and a covernet should be designed to prevent escape through the escape opening, thereby guiding fish toward the lower section and into the COD.

4.3. Catch Losses and Mitigation Measures

This study examined the catch loss of target species owing to marine mammal BRDs. In our previous study, the catch loss rate exceeded 10% [6]. Similarly, this study also demonstrated a catch loss rate exceeding 10% for the species investigated, with hairtail exceeding 50% and mackerel reaching 97%. To address these high rates, we propose four mitigation measures.

First, the three-layer COD design should be redesigned. The current COD net, intended for resource surveys, has a small mesh size, resulting in significantly slower water flow and higher internal pressure than that of the commercial trawl nets with a single-layer COD. Previous studies on trawl nets have indicated that the selectivity of the COD mesh can change when a kite cover is installed, as it further reduces water flow [57]. These findings suggest that fish may be unable to enter the COD via the guide net owing to the three-layer netting. In future research, we intend to use a single-layer net similar to those used in commercial trawling and conduct further investigations.

Second, the mesh size of the guide net should be adjusted. Our findings showed that herring and jack mackerel could pass through the net panel owing to their smaller size, even when flipping over from exhaustion, whereas larger mackerel could not. To address this, we should focus on redesigning the guide net by increasing its mesh size to an appropriate size that prevents marine mammals from passing through while improving fish retention.

Third, an additional guide device is needed to redirect escaping fish back into the net. To ensure that this device does not obstruct marine mammal escape, we propose designing a covernet with a suitable mesh size that selectively retains target species while allowing marine mammals to escape. This approach is similar to the seal exclusion device (SED) used in Australian fisheries [58,59], which effectively reduces sea lion bycatch without causing mortalities and is an effective solution for mitigating dolphin bycatch [58]. We aim to evaluate the potential application of this method in Korean fisheries.

Fourth, a funnel net can effectively capture target species while allowing marine mammals to escape. This method has been successfully implemented in midwater trawls internationally [60] and in Korean stow nets [4,5], notably reducing the catch loss of target species while enabling finless porpoises to escape. However, its application in trawl gear presents challenges owing to maintenance and repair difficulties. Future research should focus on developing improved maintenance and repair methods to facilitate the use of funnel nets in trawl gear.

5. Conclusions

In this study, no marine mammal was caught in fishing gear. But this study demonstrated that the marine mammal BRD can be thought to be highly effective, achieving near-complete exclusion of marine mammals by excluding jellyfish in experiment. However, its use may result in considerable target species catch losses, increasing concerns among fishers. To adapt this device for commercial midwater trawls, the guide net’s mesh size tested in this study should be increased, and additional covernets, similar to those used in SED systems [57,58], should be installed to mitigate catch losses. Furthermore, enhancing the maintenance and repair methods of funnel nets is necessary, considering their potential in reducing catch losses. Although this study focused on minimizing marine mammal bycatch in midwater trawls, future studies should focus on improving target species retention and exploring the potential of the improved marine mammal BRD to reduce jellyfish bycatch, which is particularly high from August to September.