1 Introduction

As primary prey of the complex ecosystem of the Antarctic Ocean, Antarctic krill (Euphausia superba) is a keystone species in the energy and material flows throughout the Southern Ocean ecosystem (Sun and Liu, 2009). This organism is rich in protein and fat and is the largest renewable source of animal protein, implying its potential for further development and utilization (Wang and Sun, 1995). The lysine content of dried Antarctic krill is higher than that of beef, tuna, or tiger shrimps and can compensate for the low lysine content of daily grain crops (Wang et al., 2020). Moreover, Antarctic krill is one of the largest single biological resources on the earth, and its large biomass and potential to support a large fishery have received increasing attention from many countries (Xu et al., 2015).

A mid-water trawl net is the main fishing gear for slowmoving pelagic species. Given the increasing impact of greenhouse gases on the environment, the trawl net must be designed for less energy consumption. Owing to the lack of detailed quantitative information about resource distribution and the poor skills for trawl design, devising a midwater trawl with target species is a difficult task (Wan et al., 2015). Ensuring that the trawl position is the same as the depth of fish schools is necessary to improve the fishing efficiency. Lee et al. (2001) described the model of a simplified trawl system and the design of a control system that uses a fuzzy algorithm to control the depth of a mid-water trawl net. To decrease the energy consumption and increase the trawling range, Lee et al. (2017) used ultra-high molecular weight polyethylene and achieved a 44% decrease in the resistance force compared with that of currently used trawls at 4 knots towing speed. Balash et al. (2016) designed a new trawl with a low energy consumption and a high spread ratio. Kinzey et al. (2015) and Breen et al. (2009) used the data from marine fisheries to assess the distribution of Antarctic krill and rock lobster (Jasus edwardsii) resources to provide theoretical guidance for accurately measuring the amount of resources and production. Previous studies mainly operated on the assumption that the trawl is stable with average properties, such as trawl tension, spread, and depth performance, which are dependent on the towing speed and gear parameter (Robins and McGilvray, 1999; Graham and Fryer, 2006); however, the oscillation characteristics of the trawl were ignored.

The trawl oscillates during towing and changes its position in the depth of the fish schools due to turbulent flow, current flow, fishing vessel motion by wave or wind, and natural underwater flows (current and tides) (Thierry et al., 2021a, 2021b). These trawl oscillations can cause the relative motions of the outer and liner of the trawl (O'Neill and Neilson, 2008), which in turn can reduce resistance, increase the roughness of the twine bar, and consequently amplify the drag of the trawl (Thierry et al., 2020a, 2020b). In addition, trawl oscillation can disrupt the trajectory of fish or mammals that accidentally enter inside the trawl and cannot escape (Lomeli and Wakefield, 2019). In Antarctic krill fisheries, trawl oscillation can substantially increase the chances of krill epidermal damage and especially fluoride exposure from the krill shell into the flesh leading to krill spoilage (Ade Lung et al., 1987; Jung et al., 2013). Another reason to study trawl oscillations is that they negatively affect fish production and cause errors in the assessment and distribution of fish resources. Therefore, avoiding catch storage in the trawl for a long time and improving the trawl stability are necessary to improve the fishing efficiency, reduce the energy consumption, and ensure the fish quality.

Codend is the end and main part of the trawl for storage and selective fish catch and has the strongest trawl oscillation (Wileman et al., 1996). Many recent researches investigated the fluttering motion of codend using numerical simulation, flume tank experiments, and sea trials. Madsen et al. (2015) evaluated the behavior of six different codends at a full scale in a flume tank of Norwegian Institute of Science, Technology and Industry (SINTEF) and found that each codend oscillates considerably when loaded with fish, and the standard codend is the most stable among those six codends. Bouhoubeiny et al. (2011) evaluated the flow around rigid and oscillating codend models. These preliminary analyses showed that a symmetrical vortex exists behind the rigid codend. However, the possible turbulent flow interactions with the fluttering codend structure cannot be determined without a complete study of the Particle Image Velocimetry (PIV) database. Druault and Germain (2016) used the PIV method to characterize the flow in the unsteady wake developing behind the moving codend structure and analyze the frequency content of codend oscillations, and determine the physical mechanism of the whole oscillating process. They found that the local hydrodynamic effects (vortex shedding developing on the wake zone around codend) and the motion of the flexible codend structure can cause codend oscillation. Liu et al. (2021) analyzed the effect of cutting ratio and catch on the drag characteristics and codend fluttering motions and found that the unfolding degree of codend netting and the height of cod-ends are directly proportional to the current speed and inversely proportional to the cutting ratio. Blevins and Saunders (1977) believed that the codend structure can oscillate intensely due to local hydrodynamic effects (such as fluctuating velocities and vortex shedding wake), warp tension variations, motions, and the deformable structure itself. Each excitation mechanism can simultaneously occur and interact with each other, thereby making the fluid structure interaction difficult to be characterized (Blevins and Saunders, 1977). Most studies used the Fast Fourier Transform (FFT) and Morlet wavelet transform method as powerful tools to objectively identify the flexible structures in turbulent flows and their possible characteristics. Kim(2012, 2013) investigated the interaction between water flow velocity and codend tilt through FFT and Morlet wavelet methods and found that the vortex shedding period varies from 2 s to 8 s. Thierry et al. (2022) used FFT and wavelet transform methods to analyze the time-periodicity of instantaneous flow velocity fields and fluctuating part inside and around the trawl net, and found the low frequency activities of the dominant frequencies of the interaction between the unsteady turbulent flows and fluttering trawl motions. In addition, the wavelet coefficient transforms from smallscale to large-scale motions under unsteady shear turbulence.

Existing studies and available data are mostly based on numerical simulation and flume tank experiments, which are quite different from the real sea trials. Thus, the present work aimed to discuss the effect of codend oscillations on the hydrodynamic analysis of codend to reliably improve the engineering performance and understand the catch process and fish selectivity under real conditions. The relation-ship of some factors, such as towing speed, warp length, warp tension, and catch size, with the fluttering motions (oscillation periodicity and amplitude) of Antarctic krill trawl codend was investigated by Morlet wavelet analysis using the measured data from sea trials. The findings will provide a theoretical basis for the improvement and optimization of mid-water trawl stability.

2 Materials and Methods 2.1 Area and Time of Sea Trials

The fishing operation was conducted from February to March of 2020 in waters off the South Orkney Islands of the Antarctic Peninsula (61˚S, 44˚W), and the South Shetland Islands (63˚S, 59˚W) (Fig.1).

2.2 Vessel and Full-Scale Trawl

The Antarctic krill trawl is designed for the target China commercial vessel, which has a main engine power of 2 × 2650 kW. The main characteristic parameters of the vessel are listed in Table 1. A four-panel mid-water trawl commonly used in the Antarctic fisheries was selected as the trawl design for this fishing vessel (Fig.2). The circumference of the trawl net at the mouth, the trawl length, and the headline length were 300 m, 132.8 m, and 55.38 m, respectively, and the fishing line length was 54.88 m. Both trawl wings and sections 1st to 7th of trawl body were made of polyethylene twine with a diameter of 6.0 mm, the mesh size of wing and first section of trawl body is 400 mm, and the mesh size of the 2nd to 7th section of the net body is 200 mm. The sections 8th to 11th of trawl body and codend were constructed from polyethylene netting with diameter of 4.0 mm and mesh size of 144 mm. The liner was constructed from 2.0 mm diameter polyamide twine and consisted of a 16 mm mesh on the second section to the 11th section of the trawl body and a 12 mm mesh on the codend.

2.3 Data Collection

Data were collected during 18 fishing operations and included the towing speed, warp length, warp tension, catch size, and codend depth. The towing speeds were recorded according to the data displayed by the GPS radar of the vessel and varied from 2 to 3 kn. The warp length and tension were measured using the NAUST MARINE ATW towed system controller (manufactured by Iceland) with a P ≤ 1% accuracy. During fishing operation, the warp length was maintained in the range of 150 – 550 m, and the warp tension varied within 8 – 12 t. The codend depth (the distance from the sea surface) was measured and stored automatically using a DR (Dr-1050 bathymetric meter) (produced in Canada). This instrument has a measuring range of 0 – 750 m and a specified accuracy of ± 0.05%, as stated in the manufacturer's specifications. The measuring position was the upper middle point of the codend. According to the Antarctic krill swarm size, the total operating time was between 4254 s and 7254 s, which was convenient for accurate data recording and analysis. The data were assessed every 6 s. During the operation, a camera was used to capture the towing speed, warp tension, and warp length recorded by a GPS radar and a towing system controller. The towing speed, warp tension, and warp length were matched to the depth of codend by the DR.

2.4 Parameter Definition and Calculation

The operation was divided into three stages: net shooting, towing, and hauling (Fig.3). The net shooting stage is the period when the codend is shot into the water until its depth is basically stable; the hauling stage covers the last significant change in the codend depth to the codend out of the water; and the towing stage is between the net shooting and hauling stages.

The speed of net shooting and hauling was calculated using the following formula:

$ S = \frac{{\left| {{L_2} - {L_1}} \right|}}{T}, $

(1)

where S is the speed obtained during the net shooting or hauling stage (m s⁻¹), L₂ is the warp length after the change (m), L₁ is the warp length before the change (m), and T is the time series (s).

The mean towing speed, warp tension, and warp length at different operation stages were calculated using the following formula:

$ M = \frac{{\sum\nolimits_{i = 1}^n {{M_i}} }}{n}, $

(2)

where M is the mean towing speed, warp tension, or warp length (kn, t, m), M_i is the towing speed, warp tension, and warp length at each 6 s (kn, t, m), and n is the number of data.

2.5 Wavelet Analysis

Wavelet transform is a powerful tool to analyze nonperiodic signals, especially when they present complex transient behaviors or even singularities. The only requirement of this approach is that the signal must have finite energy, which is important in codend motion analysis and particularly in detecting coherent structures emerging from the fluttering motions and the associated unsteady turbulent flow (Farge, 1992; Sung, 2011; Shaw et al., 2015). This method effectively analyzes the local feature of a signal and is a useful time frequency analytical tool for non-stationary signals (Zheng and Akira, 2017). In this study, wavelet analysis was performed to obtain the time-periodicity of the codend oscillation using the Matlab 2019B software package of MathWorks and its wavelet toolbox.

In wavelet analysis, the function or signal sequence can be analyzed at multiple scales by the operation functions, such as stretching and translation, which can reflect the overall characteristics of the signal in the time and frequency domain and can provide localized information for the time and frequency or periodicity domain. The analysis process is as follows:

Let ψ(t) ∈ L²(R) and M (∈) be a Fourier transform. When M (ω) satisfies the condition $C = \int_R {\frac{{\left| {M(\omega)} \right|}}{{\left| \omega \right|}}} {\text{d}}\omega < \infty $, ψ(t) is called a fundamental or mother wavelet and can be expanded and migrated into:

$ {\omega _{a, b}}(t) = \frac{1}{{\sqrt {\left| a \right|} }}\psi (\frac{{t - b}}{a}) a, b \in R; a \ne 0, $

(3)

where a is the scaling factor, and b is the translation factor. In Morlet mother wavelet, a is approximately equal to 1/f or period in numerical value.

For any function F(t) ∈ L² (R), the continuous wavelet transform can be expressed as follows:

$ {W_f}(a, b) = < f, {\psi _{a, b}} > = {\left| a \right|^{ - 1/2}}\int_R {f(t)} \psi (\frac{{t - b}}{a}){\text{d}}t. $

(4)

For a discrete signal F(t) ∈ L² (R), t = (1, 2, ···, N − 1), Eq. (3) can be discretized as:

$ {W_f}(a, b) = {\left| a \right|^{ - 1/2}}\Delta t\sum\nolimits_{i = 0}^{N - 1} {f(i\Delta t)} \psi *(\frac{{i\Delta t}}{a}), $

(5)

where (*) represents the complex conjugate.

The continuous wavelet energy spectrum of a given signal (Farge, 1992; Rioul and Flandrin, 1992; Jiang and Mahadevan, 2011) can be defined as:

$ E = {\left| {{W_f}(a, b)} \right|^2}. $

(6)

The wavelet variance is defined as:

$ {\rm{var}} (a) = {\sum\nolimits_{i = 0}^{N = 1} {\left| {{W_f}(a, b)} \right|} ^2}. $

(7)

Two time series, X(t) and Y(t), can be obtained by the wavelet variation $W_f^x(a, b)$ and $W_f^y(a, b)$ as shown in Eq. (5), respectively, and the cross wavelet spectrum is defined as:

$ W_f^{xy}(a, b) = \left| {W_f^x(a, b)W_f^{y*}(a, b)} \right| . $

(8)

In this study, the Morlet wavelet was chosen as the complex wavelet, and the phase difference between its real and imaginary parts was π/2. This phase difference eliminates the oscillation of the real wavelet transform system modulus and separates the modulus and phase of the wavelet transform coefficient. The modulus represents the number of a certain scale component, and the phase can be used to study the singularity and real-time periodicity of the signal.

3 Results 3.1 Oscillation Characteristics of the Antarctic Krill Trawl Codend Based on Morlet Wavelet Transform

The Morlet wavelet transform was used to obtain the corresponding wavelet transform diagram and curve (Fig.4). The maximum codend depth of 18 fishing operations varied from 20 m to 150 m depending on the fish schools' position. As examples, only six of the fishing operations are shown in Fig.4. For all fishing operations, the codend depth quickly increased with time during the net shooting stage. During the towing stage, the codend depth increased with the time series for fishing operations A and F. In this case, the maximum codend depths of about 80.39 m and 99.33 m for fishing operations A and F, respectively, were attained at the beginning of the towing stage t < 20 min (Figs.4A(a) and 4F(a)). For fishing operations B, C, D, and E, the codend depth slightly decreased with the increasing time series at the towing stage. In this case, the maximum codend depth was obtained at the end of the towing stage (100 < t < 120 min) for all the fishing operations except B (Figs. 4B(a), 4C(a), 4D(a), and 4E(a)). The maximum codend depth was approximately 77.70, 174.14, 95.25, and 91.49 m for fishing operations B, C, D, and E, respectively. Finally, during the hauling stage, the codend depth quickly increased with the increasing time series for all the six fishing operations.

The Morlet wavelet transform showed the oscillation of spatial wavelets with time and periodicity domains. In addition, the amplitude of the wavelet oscillation can accurately reflect the oscillation changes of the trawl codend. Fig.4 shows that for all the six fishing operations, two large events were observed during codend oscillation. For the first event, high wavelet energy coherent structure was observed at the periodicity range of 4 – 5.6 s and time range of 0 – 17.6 min during all the net shooting stages and at the beginning of the towing stage. For the second event, the high wavelet energy coherent structure was observed in the periodicity range of 2 – 6 s at the time greater than 90 – 106, 64 – 71, 80 – 96, 111 – 121, 100 – 111, and 85 – 91 min during fishing operations A, B, C, D, E, and F, respectively, at the end of the towing stage and all the hauling stages. These results implied the effect of codend oscillation on the trawl performance and indicated that this oscillation has evolved into a high-periodicity structure and contains more energy than the small-scale structures. The unsteady codend motion comprised large (high periodicity) and small (low periodicity) scales during all the six fishing operations. At the net shooting and hauling stages, the wavelet coefficient changed with a large amplitude. Meanwhile, the amplitude of the wavelet coefficient was stable at the towing stage (Figs.4A(c), 4B(c), 4C(c), 4D(c), 4E(c), and 4F(c)).

3.2 Relation Between the Warp Tension and the Codend Oscillation of Antarctic Krill Trawl

The amplitude and time series of the wavelet coefficients at 0.75 s were selected by taking the average value of 18 wavelet coefficients for 18 fishing operations to reflect codend oscillation. The results were linked to the warp tension to explore the relationship between the trawl codend oscillation and the warp tension. The warp tension of the trawl varied at 1.25 – 8.40, 7.95 – 11.95, and 7.60 – 12.0 t for the net shooting, towing, and hauling stages, respectively. The period of codend oscillation varied from 50 s to 90 s and increased with the warp tension during the net shooting and hauling stages but decreased when the warp tension increased during the towing stage. The amplitude of the codend oscillation coefficient varied from 0 to 4 and increased with the warp tension during all the stages of the fishing operation. During the towing stage, the amplitude was relatively stable (between 0.2 and 0.6) and was about 80.00% and 82.85% lower than those obtained during the net shooting and hauling stages, respectively (Fig.5).

3.3 Effect of Towing Speed and Warp Length on the Antarctic Krill Trawl Codend Period and Amplitude

Fig.6 shows the relationship between the towing speed and the oscillation of the codend depth. The towing speed of the trawl was maintained in the ranges of 2 – 4.5, 1.5 – 3.5, and 0 – 3 kn at the net shooting, towing, and hauling stages, respectively. The codend oscillation period increased with the towing speed during the net shooting and hauling stages but decreased when the towing speed increased during the towing stage. The period and amplitude of codend depth oscillation were positively correlated with the towing speed during the net shooting and hauling stages. The amplitude of codend depth increased with the towing speed during all the fishing operations. These results indicated that the variation in codend amplitude between different towing speeds was approximately 33.78%, 12.56%, and 30.91% during the net shooting, towing, and hauling stages, respectively.

A significant relationship was found between the codend oscillation period and warp length during the towing stage but not during the net shooting and hauling stages (T-test: P > 0.05). In this case, the period of codend oscillation increased with the warp length (145 – 500 m), and its amplitude decreased with the increasing warp length during the towing stage (Fig.7). Instead of the warp length, warp shooting and hauling speeds were used to reasonably compare the relationship between the codend depth oscillation and the warp length. The amplitude of the codend was inversely proportional to the warp shooting speed (0.05 – 0.96 m s⁻¹) and directly proportional to the hauling speed (0.31 – 0.80 m s⁻¹) at the net shooting and hauling stages, respectively.

3.4 Effect of Catch Size on the Amplitude of Antarctic Krill Trawl Codend

No oscillation was observed at the net shooting stage because the trawl had no catch. However, during the towing and hauling stages, the codend amplitude increased with the catch size (Fig.8). The codend motions were generally affected by the catch size (9.29 – 40.29 t) and characterized by quasi-periodic oscillations during the towing and hauling stages (Fig.4).

4 Discussion 4.1 Fluttering Motion Characteristics of Mid-Water Trawl Codend

Trawl motions during sea trials are essential for understanding the fluid structure interaction of this fishing gear in terms of bycatch selectivity, as well as improving profitability and ensuring resource sustainability. During fishing operations, the factors affecting the codend oscillations are the unsteady turbulent flow developing inside and around it, towing speed, warp length, current flow, catches, fishing vessel motion, and natural underwater flow (Kim, 2012, 2013; Bi et al., 2014). Liu et al. (2021) demonstrated that an empty codend does not oscillate significantly. The results of the present work and previous studies by Druault and Germain (2016), Bouhoubeiny et al. (2011), and Madsen et al. (2015) indicated that the codend motions are caused by the presence of the vortex shedding generated by the codend tail and the blockage of the flow passage through the codend caused by the catches. This vortex shedding will cause vertical pressure on the codend, resulting in the codend oscillations and making the hydrodynamic force act on the codend that causes the codend unstable (O'Neill et al., 2005). The present results showed that the oscillation of the Antarctic krill trawl codend only occurred in the vertical direction. This was mainly because the DR instrument could only record the depth data (z direction). This finding differed from that obtained by Madsen et al. (2015), who found that the codend oscillate freely in y and z coordinates by installing a camera behind the codend model to record oscillations of the codend, and it was deduced that the blocking effect of catch is not limited to a particular direction. In fact, codend oscillation is the synthesis of multiple degrees of freedom, which can be verified by multidirectional monitoring means in the future.

4.2 Effect of Towing Speed and Warp Length on the Codend Oscillation of Mid-Water Trawl

The results of the present work and previous studies of Kim(2012, 2013) indicated that the main factors affecting the codend motions are gear geometry by towing and natural underwater flows, such as tides or currents. During the towing stage, the amplitude of codend oscillation increases, and the period decreases with the increasing towing speed. When the towing speed increases, the water motions inside and around the trawl net become intense and unstable, thus leading to energy exchange. Bi et al. (2017) reported that the vortex would be generated behind the twines, and Zhao et al. (2013a, 2013b) found that vortexgenerated speed increases with the flow velocity. This energy exchange creates the vortex shedding around the trawl codend, which considerably increases the water pressure on the trawl. The creation of this vortex shedding and the increase in the water pressure can severely deform the trawl nets and increase the trawl motion, thus strongly modifying their shape and increasing the hydrodynamic forces acting on them. This trend was confirmed previously by Kim(2012, 2013), Druault and Germain (2016) and Liu et al. (2021) demonstrated that the period and amplitude of flow velocity are linked to trawl fluttering motions and the periodicity variation of codend oscillations is mainly due to vortex shedding periodicity. According to the present work and the study by Liu et al. (2021), the oscillations of codend motion and drag force are synchronous due to the catches blocking the flow directly through the codend mesh, resulting in reflux and creating Karman vortices at the endpart of the codend. In the real operation, the towing speed should increase as soon as possible during the net shooting stage, appropriately decrease and maintain stability during the towing stage, and finally decrease at the hauling stage to reduce the trawl amplitude. Following this protocol, the fishing operation will decrease the towing speed during the hauling stage and eventually decrease the oscillation amplitude of the trawl motions.

The Antarctic krill trawl depth is mainly controlled by the warp length (Han, 1995; Su et al., 2017). When the trawl reaches the depth of Antarctic krill swarms, the net oscillates because of the change in flows, warp tensions, and catch sizes. Analysis of codend motions indicated that the codend oscillations at the net shooting and hauling stages were greater than that at the towing stage. When the warp length increased, the traction force was reduced, and the codend motion was accelerated at the shooting stage. Meanwhile, hauling warp would increase the towing speed and drag force, thus increasing the amplitude at the hauling stage. According to the comparison of the trawl position data, the coefficient amplitude of codend at the towing stage was basically less than 0.6, and the period was about 65 – 75 s. However, the amplitude and period of codend suddenly fluctuated greatly mainly because of the changes in towing speed and codend position caused by the adjustment of the warp length. Thus, the warp length should be steady during fishing operations.

4.3 Effect of Catch Size on Mid-Water Trawl Codend Oscillation

This study found that the codend amplitude increased with the catch size because of filterability. The presence of catch inside the codend blocks the mesh opening and bulges the codend shape, thus disrupting the free passage of flow through the codend (O'Neill et al., 2005; Thierry et al., 2021b). This disturbance of the flow passage will create an unstable, turbulent flow; that is, the shedding vortex increases the vertical water pressure on the codend. The increase in this vertical pressure will destabilize the codend motions. The increase in catches inside the codend can lead to a large codend volume, which leads to unsteady turbulent flows inside and around this structure, generates an intense codend, and considerably modifies the mesh opening, thus increasing the drag and affecting the physiology and behavior of fish (Madsen et al., 2015; Druault and Germain, 2016; Bi and Xu, 2018; Thierry et al., 2021b). The results obtained in this study differed from those obtained by Bouhoubeiny et al. (2011), who placed an impermeable hemispherical cap at the bottom of the codend and then fixed the codend symmetrically with the central axis. This hemispherical cap prevented water flow from passing directly through the middle of the net mesh at the end part of the codend, thus causing water backflow around the outer surface of the hemispherical cap and creating vortices around the end part of the codend. In the present study, the trawl attached to the fishing vessel generated more intense oscillation than the rigid codend used by Bouhoubeiny et al. (2011). However, Liu et al. (2021) found that the codend without a catch did not produce oscillation but oscillated immediately when the catch was placed inside the codend. The main reason might be that the presence of the catch reduces the water filtration through the end-part of the codend, leading to the formation of vortex flow and codend oscillations.

Two Antarctic krill vessels, 'Atlantic Navigator' and 'Saga Sea' of Norwegian Aker Company, first adopted continuous fishing technology during the fishing seasons in April and May of 2003 (Li et al., 2010). The stability of the trawl net was significantly improved, as indicated by the enhanced fishing efficiency. The reason for this upgrade could be that the intermediate processes of net shooting and hauling are omitted after the first net shooting, and the warp length is basically unchanged. In addition, the fishing pump can continuously absorb the catches; therefore, the catch in the codend is maintained constant to ensure that the filterability of the codend remains unchanged. Under the less oscillation of the codend depth, continuous fishing production can be carried out for 1 month or even longer time, and the catch quality can be improved to the maximum. This study estimated that in actual operation, when the catch size reaches a certain degree, the trawl net should be hauled as early as possible to prevent the catch quality from deteriorating because of codend oscillation.

5 Conclusions

The effects of towing speed, warp length, warp tension, and catch size on trawl codend motion were investigated by utilizing the Antarctic krill trawl at sea during 18 fishing operations. The non-stationary time series of the oscillatory phenomenon of the undulating codend motion was analyzed by the wavelet transform method, and the results indicated that high-energy coherent structures periodically form at the time-periodicity plane during high-periodicity activity. Therefore, the duration of Antarctic krill trawl operations should not be extremely long, and the net shooting speed should be increased appropriately to reach the depth of the Antarctic krill swarms as soon as possible. However, the towing speed and warp length should be kept constant to decrease the codend amplitude. At the hauling stage, the towing speed and warp length should be reduced, and the catches must be stored in time.

Acknowledgements

This study was supported by the National Natural Science Foundation of China (No. 31902426), the Shanghai Sailing Program (No. 19YF1419800), and the Special Project for Exploitation and Utilization of Antarctic Biological Resources of the Ministry of Agriculture and Rural Affairs (No. D-8002-18-0097).