A Field Trial to Study the Effects of Stenella attenuata Deterrent on the Catch of a Light-Falling Net Fishery

Abstract

The incidental capture (bycatch) of protected cetaceans in fishing gear has become a serious problem worldwide. Bycatch has also had many serious consequences for pantropical spotted dolphins (Stenella attenuata) in the light-falling net fishery within this research area. We provided a self-developed acoustic deterrent device and conducted trials to investigate the long-term deterrent effects on Stenella attenuata and assess the influence of external factors on fishing catch. In 2022, 50 deterrence trials were conducted, of which 38 were effective and 12 were ineffective. In 2023, 30 deterrence trials were conducted: 24 effective and 6 ineffective. No dolphin bycatch occurred within a two-year period. Overall, the catch per unit effort (CPUE) of the effective deterrence nights was 4.96 ± 3.06 kg/min, while the CPUE of the ineffective nights was 3.78 ± 3.10 kg/min. There was a significant difference (p = 0.041, >0.05) between the two conditions. On the nights when dolphins did not appear, the average CPUE was 5.04 ± 4.44 kg/min. This CPUE was not different from the night on which deterrence was successful (p = 0.981, >0.05), but was considerably higher than night on which deterrence was unsuccessful (p = 0.028, <0.05). General additive model results indicate that month, longitude, flow direction, latitude, wind direction, dolphin number, and wind speed affect fish catch. The ADD may be improved by manually controlling the pulse frequency, transmission interval, and type to improve deter efficiency.

1. Introduction

The incidental capture (bycatch) of protected cetaceans in fishing gear has become a serious problem that is widespread in the South Pacific, Indian, North Pacific, and Atlantic Oceans, involving dozens of cetaceans [1,2,3,4,5,6,7]. Cetaceans often engage in predatory behavior around fish nets that increases their risk of being caught, such as herding fish along nets, consuming discarded catches, and removing fish from the net (depredation) [8]. Interactions with fishing nets may also result in serious injury to cetaceans, reducing their swimming and foraging abilities and increasing their risk of being caught by predators [9]. Although many cetaceans have an exceptional biosonar system, sometimes they cannot detect fishing nets that are a long distance away [10]. Harbor porpoises (Phocoena phocoena) cannot detect monofilament nets from distances of more than 3–6 m, which may lead to their being entangled within a net when preying on fish [9]. Herr et al., (2009) reported harbor porpoise (Phocoena phocoena) strandings along the Baltic Sea coast of Germany to have increased annually, with 47–86% of deaths of this species being the result of accidental capture by fishing vessels [11]. The International Union for Conservation of Nature (IUCN) also recognizes 11 species or populations of small cetaceans as “Critically Endangered”, and their populations are severely threatened by fisheries, especially as bycatch [12].

Cetacean bycatch has also been reported in the South China Sea, which is an important fishing ground for coastal fishermen from neighboring China, Vietnam, and the Philippines [13]. The high-intensity fishing activities in this area have threatened the survival of cetaceans in the South China Sea [13,14,15,16]. Among the cetaceans in this area, the pantropical spotted dolphin (Stenella attenuata) has been recognized as one of the most affected by fishing activities [17]. The South China Sea is an important habitat for S. attenuata [15,18]. Though listed as a second-level nationally protected mammal, bycatch of this species has still occurred in the South China Sea. According to some unpublished data, researchers have estimated that 20 deaths were caused by bycatch in 2015 and 18 in 2017 [19]. These bycatch data may underestimate the real situation, as they were collected from a few vessels, and some fishermen might discard the dolphin directly into the sea and not report this to the fishery’s management.

Regarding the bycatch of S. attenuata, light-falling net fisheries have contributed the most. This type of fishery concentrates on squids in the area at night using lights. However, when vessels operate at night, S. attenuata often surround the vessel and prey on the target catch. After the net is raised from the water, dolphins that did not escape the area covered by it may be caught and possibly die. Additionally, because dolphins drive away squids and therefore reduce the catch, effective methods to deter their presence around vessels and nets are desirable.

Dolphin bycatch in fisheries can be reduced through the following measures: (1) establishing nature reserves and restricting fishing activities [20]; (2) improving net design or alternative fishing gear [21,22]; and (3) deterrence signals, such as the acoustic deterrent device (ADD) used in gillnet fishery [23]. Emitting a certain acoustic frequency at a certain intensity can ensure that dolphins are “forewarned” and repelled from vessels or nets [24,25,26,27]. Acoustic deterrence is a feasible and cost-efficient method for fishermen [23]. In this study, we test a self-developed ADD on a commercial fishing vessel to (1) investigate the long-term deterrent effects on Stenella attenuata and (2) assess the influence of external factors on fishing catch.

2. Materials and Methods

2.1. Survey Area

The ADD was tested aboard the commercial light-falling net vessel “Yuedian 42212” (44.4 m length, 7.8 m width, 4.3 m draft, and 816.9 t, with four 40 m struts). This vessel has fished in the South China Sea for many years and has often experienced dolphin depredation. The fishermen have collaborated with us many times on marine fishery resource survey programs, and are familiar with the standards for marine resources and experimental data collection. They were willing to test the ADD. The vessel turned on its lights at 7 p.m. every night to start fishing and continued until dawn the next day. The crew fished 8–10 times per night, in 1 h periods. They were asked to conduct at least one trial on a night when dolphins appeared.

The crew received an advance copy of the study protocol that was written specifically for them. It included explanations of the trial’s purpose, with illustrated instructions on ADD use and data-reporting requirements. The trials with ADD were carried out on board by the fishermen after they had received professional training. They could ask any questions and were encouraged to maintain frequent contact. Trial data were recorded on a specific form, for which data-entry instructions were also provided. For familiarization purposes, this form and the ADD were sent to the fishers in advance. There were no prior requirements for fishing locations, and it was up to the fishermen themselves to choose their preferred location. The fishing locations used in 2022 and 2023 are shown in Figure 1 and Figure 2, respectively.

2.2. Acoustic Deterrent Device (ADD)

The ADD is as reported by Fu et al. [17] and includes a high-capacity lithium battery power module (LRS-75-12, MeanWell Enterprises Co., Ltd., Taiwan, China) providing power to the entire system, a signal generator, a digital-to-analog signal transform module, a power-amplifier module, and a transducer (Figure 3). Signal generation was completed by feeding LabView (Version 2017, National Instruments, Austin, TX, USA)-based codes into the system to generate the source signal on demand, which was further transformed into an analog signal through a digital-to-analog signal transform module (DAQ Card USB-6216, National Instruments, Austin, TX, USA) at a sampling rate of 400 kS/s. The output was then fed into the customized power amplifier module (ATA-M112, Aigtek Co., Ltd., Xi’an, China) to boost the signal energy, reaching 120 W. The signal was transmitted into the water through a non-directional underwater transducer (WBT 30, China Shipbuilding Industry Group Co., Ltd., Hangzhou, China) with a centroid frequency of 30 kHz.

After dolphins appeared at night, the ADD was placed into the water and suspended at a depth of 3 m from a rope weighted with an iron block and then the deterrent signals were transmitted. The duration of each trial was 1 h, and the behavioral changes of the dolphins were recorded. After the dolphins disappeared, the ADD was turned off. If the dolphins reappeared, the trial was conducted again. Visual observations of the surface positions and movements of dolphins and their total number were recorded before and after each ADD use. Dolphins’ disappearance from the fishermen’s visibility was considered indicative of an effective test. Otherwise, the test was deemed ineffective.

2.3. Model Construction

In 2022 and 2023, fishing logs were completed nightly, with recorded data including month, day, latitude (°), longitude (°), total catch (kg), total lighting time (min), dolphin number (num), flow direction (°), flow velocity (knots), wind direction (°), and wind speed (m/s). As the total lighting time was different each day, we used catch per unit effort (CPUE) to standardize the daily catch and compare the fishing efficiency [28]:

$$CPUE={\displaystyle \frac{c}{f}}$$

(1)

where c is the total number of squids caught per night (kg) and f is the total lighting time (min). The CPUE of fishing nights on which ADD was and was not used was calculated separately. Student’s t-tests were used to compare differences in the mean CPUE values for nights when the ADD was and was not used.

To examine any changes in CPUE while controlling for other variables, a generalized additive model (GAM) was built to investigate the relationships between CPUE and variables. The general expression is as follows [27,29,30]:

$$Y=s\left(x_1\right)+\cdots s\left(x_m\right)+\epsilon$$

(2)

where Y is a connection function (we selected a natural logarithm for this study); x is the explanatory variable; m represents the number of variables; s(x) is the natural cubic spline smoothing function; and ε represents the residual.

We counted the ln of CPUE as the response variable Y. As some CPUE data were less than 1, a constant “1” was added to all CPUE values to prevent the occurrence of a logarithm being 0. A Q-Q normal analysis was performed to ensure the ln (CPUE + 1) followed a normal distribution prior to model construction [29,31]. The test results verify this, and the ln (CPUE + 1) data can be used to construct a GAM. Before model construction, the multicollinearity of all variables was investigated by calculating the variance inflation factor (VIF). Explanatory variables with the highest VIF value ≥ 4 were individually removed from the model [30,32], and multicollinearity was re-checked to verify that the remaining variables were not correlated. The results showed that the VIF values of all variables were less than 4, indicating a low degree of multicollinearity among them, and all variables could be used for model construction. The model can be expressed as shown below:

ln (CPUE + 1) = s(month) + s(day) + s(latitude) + s(longitude) + s(dolphin number) + s(flow direction)
+ s(flow velocity) + s(wind direction) + s(wind speed) + ε

(3)

According to the Akaike Information Criterion (AIC), coefficient of determination (R²), and deviation explained (DE), we gradually removed or added variables to obtain the optimal model (with the minimum AIC) [29,30,32]. The AIC formula is as follows:

$$AIC=D+2d_f\phi$$

(4)

where D is the deviation (sum of squared residuals); φ is the deviation parameter (variance); and d_f is the effective degree of freedom. All analyses were conducted in R (version 4.0.2) using the “stats” and “mgcv” packages.

3. Results

3.1. Acoustic Deterrent Device (ADD) Trial Results

In 2022, the vessel was conducted 168 night-time fishing activities from January to October, of which dolphins appeared on 50, while 118 nights were free from dolphins, with a frequency of 29.76%. Of the 50 deterrence trials conducted, 38 (76.00%) were effective (● in Figure 4) and 12 (24.00%) were ineffective (▲ in Figure 4). The deterring effect in two areas was poor (Figure 4: area A and B). Of the 12 trials in area A (116°24–116°55′ E, 11°27–12°05′ N), 7 trials were effective, with an efficiency of 58.33%; of the 9 trials in area B (113°20–113°36′ E, 10°37–11°04′ N), 5 trials were effective, with an efficiency of 55.55%. The efficiency in these two areas was lower than the annual average deterring efficiency.

In 2023, the vessel was conducted 85 night-time fishing activities from January to May, of which dolphins appeared on 30 and did not appear on 55, with a frequency of 35.29%. Of the 30 deterrence trials, 24 (80.00%) were effective (● in Figure 5) and 6 (20.00%) were ineffective (▲ in Figure 5). No dolphin bycatch occurred aboard FV “Yuedian 42212” during any of the ADD trials.

3.2. Catch per Unit Effort (CPUE)

In 2022, the vessel used 1252 fishing nets, with a total catch of 486,250 kg and an average CPUE of 4.71 ± 4.34 kg/min. On the nights when dolphins appeared, 148,629 kg of squid was caught, with an average CPUE of 4.78 ± 3.54 kg/min. On the nights when dolphins did not appear, 337,261 kg of squid was caught, with an average CPUE of 4.72 ± 4.67 kg/min. There was no significant difference (p = 0.894, >0.05) between the two conditions (Figure 6). In addition, the CPUE of the effective deterrence nights (5.55 ± 3.15 kg/min) differed significantly (p = 0.018, <0.05) from the CPUE of the nights when it was ineffective (3.54 ± 3.86 kg/min) (Figure 6).

In 2023, the vessel used 712 fishing nets, with a total catch of 261,031 kg and an average CPUE of 5.03 ± 3.67 kg/min. On the nights when dolphins appeared, 51,772 kg of squid was caught, with an average CPUE of 3.41 ± 2.63 kg/min. On the nights when dolphins did not appear, 209,259 kg of squid was caught, with an average CPUE of 5.82 ± 3.84 kg/min. There was a significant difference (p = 0.002, <0.01) between the two conditions (Figure 6). Furthermore, the CPUE of the nights on which deterrence was effective (3.93 ± 2.69 kg/min) differed significantly (p = 0.001, <0.01) from the CPUE of the nights when it was ineffective (1.33 ± 0.55 kg/min) (Figure 6).

Over a two-year period, the CPUE of the nights on which deterrence was effective was 4.96 ± 3.06 kg/min, while the CPUE of the nights on which it was ineffective was 3.78 ± 3.10 kg/min. There was a significant difference (p = 0.041, >0.05) between the two conditions. On the nights when dolphins did not appear, the average CPUE was 5.04 ± 4.44 kg/min, which was not significantly different (p = 0.981, >0.05) from the CPUE when the dolphins were successfully deterred, but was significantly higher (p = 0.028, <0.05) than when the dolphins were not deterred.

3.3. General Additive Model (GAM) Results

The correlation test results between the CPUE and variables in GAM are shown in

Table 1. Pearson correlation test results between CPUE and variables in GAM.

Variables	p Value	Explanatory Variance
s(month)	1.29 × 10−5 ***	11.70%
s(flow direction)	0.0125 *	4.40%
s(latitude)	0.0040 **	4.60%
s(wind speed)	0.0065 **	1.00%
s(wind direction)	0.0086 **	3.30%
s(dolphin number)	0.0238 *	2.40%
s(longitude)	0.0680	6.10%
s(day)	0.3842	2.20%

* represents significant differences of 0.05; ** represents significant differences of 0.01; *** represents significant differences of 0.001.

. The variables with p > 0.05 (low correlation) are the longitude, day, and flow velocity. Although the p values of the longitude and day are greater than 0.05, the AIC value of the model that includes them is the smallest. Therefore, the final variables included in the model are the month, flow direction, latitude, wind speed, wind direction, dolphin number, longitude, and day. The explanatory power of these variables on the CPUE variance was 35.7%, with a coefficient of determination for model fitting of 0.271. According to the AIC, the optimal model expression for the GAM is as follows:

ln (CPUE + 1) = s(month) + s(day) + s(latitude) + s(longitude) + s(dolphin number) + s(flow direction)
+ s(wind direction) + s(wind speed) + ε

(5)

The temporal trend of the CPUE changes with the month was obvious. From February to April, the CPUE gradually increased; from April to July, it gradually decreased; from July to mid-August, it again gradually increased, peaking in mid-August; and from mid-August, it decreased significantly (Figure 7). The longitude and latitude also affected the CPUE (Figure 7). The CPUE slowly increased with increased longitude west of 112.5° E; as longitude increased from 112.5° E to 115° E, the CPUE gradually decreased; and east of 115° E, the CPUE increased with increased longitude. For latitude, the CPUE took on a “hump” shape, centered on 13° N. The CPUE also increased with increasing flow direction and decreased with increasing wind speed, wind direction, dolphin number, and day (Figure 7).

4. Discussion

4.1. Study Justification

Research on dolphins linked to fishing activities in the South China Sea is relatively scarce, mainly focusing on statistical descriptions of bycatch information. Wang et al. [33] studied bycatch data concerning marine mammals in Chinese mainland waters from 2000 to 2006, during which time 66 bycatch events attributed to five species were reported. Liu et al. [4] performed a large-scale interview of local fishermen around Hainan Island, China, and reported 150 bycatch events involving more than 600 marine mammals from 2000 to 2013. Both studies focused on nearshore waters, and dolphin bycatch in offshore South China Sea waters remained poorly studied. Bycatch of cetaceans in the offshore area is also serious, and one representative species is Stenella attenuata, for which the majority of deaths were attributed to fishing activities in this area [33], and light-falling net fisheries have contributed the most to S. attenuata bycatch. The ADD described in the current study has been validated for its effectiveness in deterring dolphins previously [17]. However, the impact of this deterrent on the squid catch still needs to be studied to support its widespread use in fisheries.

4.2. Deterrence Results

The CPUE of the nights on which deterrence was effective differed from that of the nights on which it was ineffective, indicating that after the dolphins were repelled from the vessel, fish were again attracted to the net lights, and the CPUE subsequently increased. On the nights when dolphins did not appear, the average CPUE was not significantly different from that when dolphins were successfully deterred, but was considerably higher than that when the dolphins were not deterred. This also indicated the impact of the dolphins’ appearance on the fishing catches. On the nights when dolphins were present, the CPUE was significantly lower; after the dolphins were deterred, the CPUE increased remarkably, approaching the values achieved when no dolphins were present, and the adverse effects of the dolphins’ appearance on the fishing activity were eliminated. The results of this study indicate that dolphin acoustic deterrents do not affect the fishing catch, which is consistent with Moan and Bjørge’s [27] research: the pingers they used had no significant effect on the catch rates of fish in three Norwegian commercial gillnet fisheries.

During the 2022 trials, the deterring effect in the two areas was poor (Figure 4: area A and B). These two areas are traditional fishing grounds, and reports of dolphins in this area are frequent. Vessels have used various deterrence techniques (e.g., underwater firecrackers, alarms, horns, and tapping on the ship’s side), but dolphins exposed to long-term and high-frequency sound stimulation may gradually become used to such noises, thereby rendering the ADD less effective. Some studies have demonstrated that dolphins may adapt to some types of ADD over time. Gearin et al. [34] performed field tests in a salmon gillnet fishery in Washington, USA. No porpoises were taken in the first 18 d of the 1997 season, and of the 12 porpoises eventually taken, 11 were caught in the last two weeks of the fishing activities, indicating that their adaptation may have increased the entanglement rate. Cox et al. [35] used a Netmark 1000 pinger to continuously deter harbor porpoises and reported the deterrence distance to have reduced by half on day 4, and that by day 11, the dolphins had fully adapted to the signal. Carlström et al. [36] conducted a similar study and reported that with prolonged stimulation, the frequency of harbor porpoises appearing near a dolphin-deterring device increased, indicating their increased tolerance of the deterrent signal. Kindt-Larsen et al. [37] studied the vocalization behavior of porpoises in response to two different pingers, AQUAmark100 (20–160 kHz) and AQUAmark300 (10 kHz). The results of AQUAmark300 revealed a significant effect at a 0 m distance and either no effect or a 17% reduction at a 300 m distance. At one station, habituation effects were found, indicated by an increase in clicks over time. However, some field studies have shown no evidence of habituation. Moan and Bjørge [27] found that as there were no changes in the bycatch rates over two years, there was no evidence suggesting that harbor porpoises habituated to the pingers. Paitach et al. [38] developed an experiment to test the deterrent effect of Banana pingers. Their data indicated that franciscanas avoided ensonified areas and no habituation effects were recorded during 65 days of testing.

This adaptation has previously been considered to be a form of habituation, but this may not be accurate. After the initial period of ADD use, the dolphins were deterred away from the nets; they may have later returned to the net, ignoring the sound, because the food was there. This is a strategic decision for feeding purposes. Therefore, it is uncertain if dolphins always adapt to acoustic deterrence signals. In future studies, this ADD could be improved to achieve better deterring effects, such as expanding the frequency range, the transmission interval, and signal type. The tests could also be conducted in more vessels to collect more data.

4.3. General Additive Model (GAM)

The temporal trend of the CPUE changes with the month was obvious (Figure 7). The CPUE increased gradually from February to April and declined from May to August (Figure 7). Due to the fishing ban policy in the South China Sea from 1 May to 15 August, the vessels could only fish in waters south of 12° N, and this limitation may have decreased the CPUE in this period. After August, the CPUE increased again to a peak in September and then significantly decreased after that. As the northeast monsoon creates strong winds and waves after September, the marine environment is no longer suitable for fishing activities at this time [39]. Therefore, the CPUE decreases from September to the January of the following year.

Spatial variables also affect the CPUE, mainly between 9° N and 16° N, and 111° E and 117° E (Figure 7). This is consistent with the research on the distribution of fishing grounds in the South China Sea [40,41,42]. Variation in the CPUE at different latitudes is related to the month [43,44]. In the central and southern South China Sea, between February and July, fishing efforts are mainly concentrated near 12° N to capitalize on the increased water temperatures and nutrient-rich upwellings generated by spring (southeast) and summer (southwest) monsoons that result in abundant food for purpleback flying squids [45]. Purpleback flying squids mainly exist in open-sea waters from a depth of 200 m between 111° E and 117° E, with the shallower waters near the continental shelf west of 111° E being less suitable for them. While the water depth is unsuitable for this species east of 117° E, factors such as the high fishing costs, complex sea conditions, or limited storage capacity of fishing vessels may restrict fishing activities in these waters. Other variables that affect the CPUE include fd, ws, and wd (Figure 7). It is dangerous to deploy nets at high wind speeds, and the CPUE will be decreased significantly. The same conditions apply to wd. The flow direction has the opposite effect; as the flow angle increases, the CPUE gradually trends upward. The number of dolphins also negatively affects the CPUE. On nights when more dolphins appear, the catch will be correspondingly poor.

In the study by Li et al. [28], variables such as the longitude, month, lunar day operation time, and sea surface temperature (SST) affected the CPUE of the squid, with a model explanatory variance of 59%. In the study by Fan et al. [46], the latitude, longitude, sea surface height (SSH), and sea surface salinity (SSS) presented the optimal explanatory model variables, with a total explanatory variance of 54.7%. Compared to these studies, the lack of variables like sea surface height (SSH), sea surface salinity (SSS), and sea surface temperature (SST) may be the reason for the low total explanatory variance of this study. Some studies [43,44,47] have confirmed that the SST is an important environmental factor affecting the distribution of squid habitats. Cephalopods are sensitive to the SST, which not only affects their growth (hatching rate, adult size, and lifecycle) but also the distribution of prey [48]. In addition, squid habitats are mainly distributed in the intersection area of water masses and currents, reflected in the changes in the SSH and SSS [46], and this intersection may have a significant impact on the CPUE [35]. The absence of the mentioned environmental factors may result in the explanatory power of this model being reduced. In future studies, researchers might consider including additional variables and their interaction effects in a revised GAM.

5. Conclusions

We investigated the effect of dolphin deterrence on the catch of fishing vessels by testing a self-developed ADD system. Over a period of two years, the CPUE of the effective deterrence nights was 4.96 ± 3.06 kg/min, while the CPUE of the ineffective nights was 3.78 ± 3.10 kg/min. There was a significant difference (p = 0.041, >0.05) between the two conditions. On the nights when dolphins did not appear, the average CPUE was 5.04 ± 4.44 kg/min, which was not significantly different (p = 0.981, >0.05) from the CPUE when the dolphins were successfully deterred, but was significantly higher (p = 0.028, <0.05) than when the dolphins were not deterred. Overall, this ADD does not affect squid catch. For nights on which deterrence was unsuccessful, this may be because the spotted dolphins had adapted to the signal; however, further study is needed to verify this. We plan to improve the existing ADD to transmit signals with different parameters and achieve better protection in future studies. The influence of external factors on CPUE was studied by constructing a GAM, for which the total explanatory variance was 35.7%. The habitat of purpleback squid is relatively complex and influenced by various factors. While SSH, SSS, and SST also affect CPUE, we did not include these in this study. Further research could examine the influence of more factors and their interaction effects on CPUE. This study could provide technical support for fishery management departments, such as encouraging fishermen to widely use ADD on vessels through government subsidies to better protect dolphins in the South China Sea.