ICES Journal of Marine Science: Journal du Conseil Advance Access originally published online on July 11, 2007
ICES Journal of Marine Science: Journal du Conseil 2007 64(8):1517-1524; doi:10.1093/icesjms/fsm097
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Evaluating the impact of gillnet ghost fishing using a computational analysis of the geometry of fishing gear
1 School of Agriculture, Kinki University, 3327-204 Nakamachi, Nara 631-8505, Japan
2 Graduate School of Fisheries Science, Hokkaido University, Minatocho, Hakodate 041-8611, Japan
3 University of Rostock, 18119 Warnemunde, Germany
Correspondence to T. Takagi: tel: +81 742 43 6169; fax: +81 742 43 6169; e-mail: tutakagi{at}nara.kindai.ac.jp
Takagi, T., Shimizu, T., and Korte, H. 2007. Evaluating the impact of gillnet ghost fishing using a computational analysis of the geometry of fishing gear. – ICES Journal of Marine Science, 64: 1517–1524.We developed a net shape and load analysis system (NaLA) that can estimate the three-dimensional shape of fishing gear underwater computationally. This paper introduces the latest version of the numerical model of the NaLA. Previously, NaLA was used to estimate the net geometries and internal forces of some fishing gear, demonstrating its general versatility. However, the ultimate goal of our study has been to learn about the impact of fishing and the capture process from a physical perspective, not simply to develop elemental technologies for gear design. Accurate, quantitative evaluation of fishing gear performance from a physical perspective can be used to estimate the potentialities of the ghost fishing to gillnet gears. Although the applications are not limited to geometries and internal forces, the paper describes how computer-aided simulations of fishing gear should be applied to investigations of the impact of ghost fishing caused by lost drift and bottom gillnets. The computational results showed that a driftnet with homogenous net panels was deformed slightly and bent only at the two ends of the net. Accumulation of periphyton on a bottom gillnet after 25 d of immersion caused it to settle to the sea bottom.
Keywords: fishing impact, gear performance, ghost fishing, net geometry, numerical simulation
Received 31 August 2006; accepted 26 May 2007; advance access publication 11 July 2007.
 |
Introduction |
|---|
 Top Introduction MethodsResultsDiscussionReferences |
|---|
We developed a net shape and load analysis system (NaLA) that can estimate the three-dimensional shape of fishing gear underwater (Takagi et al., 2002, 2004; Shimizu et al., 2004). Many fishing gear manufacturers require a computational system to evaluate net shape and the internal forces acting on the net underwater, without the need of direct measurement devices, such as electrical data loggers. Because most designers and fishers are unable to provide accurate estimates of these shapes and internal forces, designs and improvements are normally based on empirical approaches. It is difficult to evaluate the dynamic effects of fishing gear in developing new components, such as grid panels and attached separators for trawls and funnels.
Within this context, some numerical models have attempted to simulate fishing net shape. For example, O'Neill (1999) derived the equations governing the geometry of axisymmetric trawl codends made from netting with a mesh of a particular generalized structure. From this, the geometry of the codend can be computed by setting the initial mesh bar to suitable lengths. Bessonneau and Marichal (1998) and Niedzwiedz and Hopp (1998) simulated an entire trawlnet numerically. In Europe, most research on fishing net simulation has dealt with trawlnet shape. The behaviour of the net cage system and wave load (Lee and Pei-Wen, 2001; Lee et al., 2005) was examined using a numerical model proposed by Lee et al. (2002). However, the ultimate goal of our research has not been simply to develop elemental technology on which to build computational aids for net builders and fishers. Accurate, quantitative evaluation of fishing gear performance from a physical perspective is required to determine fishing impact, such as ghost fishing and bycatch problems. Even the same fishing gear used under the same operating conditions behaves differently depending on the environmental conditions. Therefore, a scientific understanding of how the shape and internal forces of fishing gear change under various physical conditions should be the basis for resolving such evaluations. To accomplish this, computer simulations must give users relevant, accurate information in response to a wide range of temporal and spatial phenomena.
Here, we introduce the latest version of NaLA and present several applications. These applications are not limited to evaluating net geometry and internal forces. This paper describes how computer simulations of fishing gear can be used to investigate the impact on fishing of ghost fishing by lost drift and bottom gillnets, and presents examples to illustrate potential uses of NaLA.
We examined how a free net drifts and deforms in the open sea over a prolonged period, how a bottom gillnet deforms after prolonged immersion, and estimated the potential for ghost fishing from a dynamic perspective using numerical simulation. To evaluate small local areas of the geometry of a large fishing net, such as the shape and tension of individual mesh bars, we used the numerical procedures and algorithms required for fine-scale estimation. By clarifying these points, we can obtain important basic scientific knowledge to understand the impact of ghost fishing from a physical perspective.
|
Methods |
|---|
|
TopIntroductionMethodsResultsDiscussionReferences |
|---|
A numerical model using an inertia transformation matrix
The calculation model in NaLA assumes that the net consists of lumped point masses that are connected by massless springs. The equations of motion of these point masses can be expressed in a local coordinate system, which simplifies the treatment of the hydrodynamic forces acting on each mass.
Although the formulation of the numerical model used for the simulation was proposed in a previous paper, we present the latest model here, together with the formula of the inertia transformation matrix algorithm. The algorithm facilitates the treatment of the fluid forces acting on a mass point without necessitating a complex procedure for coordinate transformation from a local to a global system (Korte and Takagi, 2005).
Let the motion of equation of a mass point be expressed by Equation (1) in local coordinates applying D'Alembert's principle:
|
| (1) |
To solve the unknown acceleration vector 
numerically, each term in the equation has to be transformed from the local coordinate system to the global one. First, let the acceleration term be transformed into global coordinates using the following transformation matrix:
|
| (2) |
is the external force, m1 the inertia mass matrix, 
the acceleration of the mass in the global system, and Cgl the matrix for transforming the coordinates from the global to the local system. Then, perform matrix multiplication by Clg on both sides of Equation (2) to obtain the equation for the motion of the mass point in the global coordinate system as the following equation: |
| (3) |
where C1g is the matrix for the transformation from the local to the global system. As Equation (3) is equal to 
in the global coordinate system, the first and second terms can be expressed as the following equations:
|
| (4) |
|
| (5) |
This algorithm was proposed as the inertia transformation algorithm by Korte and Takagi (2005).
The force vector acting on an arbitrary point is easily broken down into components based on a rational local coordinate system. Gravity and the tension and fluid forces on both sides of the mesh bar act on mass point i (Figure 1). The inertia transformation algorithm can easily insert fluid forces into the equation of motion, such as an added mass or drag force. The equation of motion for mass point i can be expressed using Equations (4) and (5) as
|
| (6) |
|
where Mi is the inertia matrix;
, 
, 
, and 
the tension force, drag and friction force, submerged weight, and buoyancy of i, respectively; 
the acceleration of mass point i; and 
Mik the added mass matrix of the bar between points i and k expressed as the following equation:
|
| (7) |
|
| (8) |
where (e
x, ey, ez) is the unit vector of the axis. Each mass point can be computed numerically using Equation (6) with the inertia transformation algorithm using a numerical integration method, such as the Runge–Kutta method. Therefore, the dynamic geometry and deformation can be estimated without any complex coding for the computing.
Estimating the deformation of a drift gillnet
A concern exists that lost drift gillnets have a negative impact on the local ecosystem, a process known as ghost fishing. To evaluate the effect of ghost fishing by a drift gillnet, the change in net geometry in the horizontal plane depending on the spatio-temporal profile of the current velocity and net specifications during its drift must be estimated. However, when the net has drifted for a long time and the spatial scale is greater than half a kilometre, it is difficult to measure the net deformation in situ in the horizontal plane. Given this background for evaluating ghost fishing by a driftnet, we evaluated the two-dimensional geometry of a driftnet using a numerical simulation for different current fields and net specifications. The numerical simulation can estimate the temporal change in geometry for more than 1 week.
The specifications of the net are shown in Table 1. Driftnets made of nylon monofilament are used to catch salmon and trout in the Pacific Ocean near Hokkaido. In this case, the driftnet consisted of ten panels. We computed the driftnet geometry in the two-dimensional horizontal plane for various current velocities. We assumed that the sea was sufficiently deep to neglect the bottom effect and that the current velocity was uniform in the vertical plane. A flowfield for the numerical simulation was produced, based on current velocity and direction data measured near south Hokkaido (42°10'21''N 140°49'29''E) every hour for 1 month by the Japan Oceanographic Data Centre (JODC). The JODC data were used as reference data for the numerical simulation. For comparison, we also used a uniform spatio-temporal flow. The uniform flowfield had the same average velocity as the JODC data.
|
A driftnet is generally made of multiple panels with the same mesh size. In a uniform current, the fluid force acts on each panel homogenously. However, when fish are captured in the net, the apparent area and mass of the net panels increase, and the inertia of the driftnet changes, as do the fluid forces acting on it. In this study, we estimated the deformation of a driftnet in such flowfields using changes in the projected area and inertial mass of the driftnet. Figure 2 shows schematic views of the changes in the projected area and inertial mass for three configurations (Types A–C) of the net used in the simulation. In all cases, in the computation, the initial flow direction was perpendicular to the net panels.
|
Estimating the deformation of a bottom gillnet
When a bottom gillnet is fixed to the seabed, the net is deployed vertically, like a screen, by the buoyancy of floats on the float line. For a lost bottom gillnet, it is postulated that prolonged immersion increases the sinking force caused by the accumulation of periphyton and detritus on the net. Consequently, net height would decrease. To evaluate how a bottom gillnet would be deformed after prolonged immersion, we used a numerical simulation to estimate the three-dimensional deformation of a net subject to a change in density caused by the accumulation of organisms. Yamaguchi and Nishinokubi (1998) and Yamaguchi (1999) quantified the accumulation of organisms and detritus on immersed net panels over time using image analysis. A net panel was attached to a 0.4 x 0.4-m rectangular frame. Three panels were immersed at depths of 5, 10, and 15 m off the coast of Nagasaki in summer. The measured projected area and weight allow us to estimate the apparent density of organisms and detritus as 1.33 g cm–3. Figure 3 shows the relationship between immersion time and R, the ratio of the apparent projected area of the immersed net panel to that of the net panel before immersion, based on the results of Yamaguchi and Nishinokubi (1998) and Yamaguchi (1999). The projected area was 2.5 times greater than that of the initial net after 30 d. We estimated the apparent diameter and density of a net bar covered with organisms to compute the three-dimensional deformation of a net covered with organisms using these properties. The bottom gillnet that we simulated is used to catch walleye pollock in Hokkaido, and has a stretched mesh size of 83 mm and a length of 29 m (Table 1). The geometry was estimated for changes in the projected area based on the characteristics of accumulation shown in Figure 3, and for changes in current speed.
|
The catch by a bottom gillnet depends on its mesh size, geometry, and internal forces. If the current flow and biomass accumulation on a net differ, then the geometry of the entire net and the mesh must also differ. Therefore, to evaluate fishing efficiency and the impact of a gillnet, the mesh geometry in a portion of the net must be investigated on a real scale. A computer-aided net shape simulator requires much greater computational effort to calculate the values associated with the increased number of mass points needed to estimate the geometry (Takagi et al., 2004). Therefore, to reduce the computational effort, adjacent meshes are grouped into one large mesh, reducing the number of mass points for which computations are required. However, for the portion on which we want to focus, the mesh-grouping algorithm is not needed to investigate the true mesh geometry. In this study, we used a hybrid method that includes both mesh grouping and actual mesh size computations to investigate how an entire gillnet and its mesh geometry deform with changes in the current flow and the bioaccumulation of organisms.
|
Results |
|---|
|
TopIntroductionMethodsResultsDiscussionReferences |
|---|
Drift gillnet
Figure 4 shows the trajectory of a lost drift gillnet over 10 d, computed using the numerical simulation model for the actual current conditions measured by the JODC. In all cases, the gillnet drifted with the current in a southwesterly direction before encountering a counter flow current and moving back towards its initial position. Integrating the trajectory lines shown in Figure 4, the total distance travelled by the net was 79 km. This distance is shorter than the cumulative mean current speed (11.5 cm s–1) over 10 d. With increases in the apparent projected area and mass of the net panels (Types B and C), both trajectories were similar to that of a net with homogenous net panels (Type A).
|
Figure 5 shows the deformation of the net in the horizontal plane while drifting. In all cases, after 1 d, the shape of the net had not changed from that on release. Three days after release, both ends of the net were bent. In the net with homogenous panels (Type A), the net was bent only near its ends, whereas in the other types, the net panels were bent markedly at the boundary between the panel with a large projected area and that with a normal area.
|
The relationship between the elapsed time after release and the distance between the ends of the net is shown in Figure 6. In all cases, the distance decreased with time. For the homogenous net (Type A), the distance was longer than in the other cases, which means that the deformation of the homogenous net was less. For the nets in which the projected area and inertia increased, little difference was observed in the decrease in distance over time, regardless of the position of the panel with increases in the apparent projected area and mass of the net. Ten days after release, in Types B and C, the distances were shorter than 70% of the initial length, whereas in Type A, the distance was greater than 80% of the initial length. For comparison, the net deformation in all three types was also estimated under a spatio-temporally uniform current field. For the homogenous net panels, the net shape did not deform after drifting for 10 d. For the other types, inflection points were found at the boundaries between the panel with a large projected area and the area of normal netting during drifting, but the bend was not as great as the actual flow.
|
Bottom gillnet
We estimated the geometry of a clean bottom gillnet without any accumulated organisms or flow, as shown in Figure 7a. The net was kept vertical to the sea bottom by the buoyancy of the floats and float line. Between buoys, the net was shaped like a catenary curve. The internal force acting on the netting was not homogenous over the entire surface. Discontinuities and strong localized internal forces developed on the net surface. Generally, this is a physical characteristic of bottom gillnets. The tension force on the net bar was
400 gramme-force (gf) in the area subjected to the strong internal force. The force on the net was stronger close to where the float or sinker lines were attached.
|
The geometry when a current flowing at 10 cm s–1 encountered the net perpendicularly is shown in Figure 7b. The current caused the net to incline and bulge. The internal force acting on the net was stronger than when there was no flow caused by fluid-dynamic forces, such as the drag and friction forces. We estimated the geometry of a fouled net in which the projected area and mass increased after 15 d of immersion (Figure 7c) using the relationship between projected area and the accumulation of periphyton with immersion time, as shown in Figure 3. The incline and bulge of the net caused by the current were greater, and the loads on the float lines increased. Because Figure 3 shows that the accumulation periphyton follows a sigmoid growth curve, the following logistic growth curve function was identified using the non-linear least-squares method to model the relationship between the projected area because of the accumulated organisms and immersion time:
|
| (9) |
|
|
After estimating the accumulated mass and projected area, we simulated the geometry of a net fouled with periphyton. Figure 8 shows the relationship between immersion time and net height calculated as dimensionless values divided by the initial height without current flow.
|
At a current speed of 25 cm s–1, after 20 d immersion, the net height was less than 25% of the initial height without flow. In contrast, at current speeds of 5 and 10 cm s–1, the decrease was less than 30% of the initial net height. Consequently, under such conditions, the nets will continue to catch fish for at least 20 d. After 25 d, regardless of current speed, the buoyancy of the floats and float lines could not counter the weight of the accumulated organisms, and the net settled to the sea bottom.
If we need to focus on the actual shape of part of the net mesh, we can first compute the shape of the entire net using grouped meshes to reduce the computational effort, then estimate the focal real-scale meshes from the result for the entire net geometry. Figure 9 shows the real-scale mesh of a gillnet immersed for 15 d. The mesh shape differed with current speed. In the marked mesh subject to strong internal forces, the horizontal diagonal at a current speed of 25 cm s–1 was 17% shorter than when there was no current flow, and the area was 15% smaller owing to stream-wise stretching. The orthogonal area projected on the vertical plane perpendicular to the flow at 25 cm s–1 was 19% of that in the initial conditions with no current. Table 2 shows the length of the diagonal line and the orthogonal projected area of the marked mesh shown in Figure 9.
|
|
|
Discussion |
|---|
|
TopIntroductionMethodsResultsDiscussionReferences |
|---|
Drift gillnet
Because drift gillnets are extremely long, some panels might catch many more fish than others. Consequently, the mass and projected area of such a panel increases, and the net tends to bend at the boundary between a panel that has caught many fish and one with a small catch. In contrast, if the distribution of the catch on the net is even, then no remarkable deformation or bending occurs, and the net with homogenous panels takes much longer to become deformed.
In all simulations, very little deformation occurred after the net had drifted for 1 or 2 d, whereas marked deformation was observed after 5 or 6 d. Therefore, a driftnet may function normally in terms of fishing for 2 or 3 d from a physical perspective. However, because the numerical model of the driftnet was restricted to the two-dimensional horizontal plane, the vertical movement was not considered, and the changes in fluid dynamics and friction with the entanglement and accumulation of net panels could not be considered. Such non-linear effects might accelerate net deformation. Nevertheless, the simulation shows that, although the fishing ability of a lost drift gillnet would be relatively poor, the net can still catch fish for many days, if it is not damaged.
In the ocean, currents are not homogenous in time or space, so an actual net would likely deform. The result in Figure 5 shows that the temporal changes in the currents and inertia of captured fish strongly affect net geometry. Therefore, to reduce ghost fishing by driftnets, a change in the mass and projected area of the net panels can be used to accelerate deformation and entanglement. Our numerical simulation should allow us to develop measures to prevent ghost fishing by driftnets.
Bottom gillnet
Some experimental studies based on observations by divers have examined the prolonged deployment of bottom gill or trammelnets (Kaiser et al., 1996; Erzini et al., 1997; Matsushita et al., 2004). These studies reported that, after deployment for 1 month, the net height had decreased markedly, which concurs with our results. Erzini et al. (1997) reported that by 20 d after deployment, the height of the gillnet was half the initial height, whereas Kaiser et al. (1996) reported that it was less than 20% of the initial height. In both cases, the catches decreased drastically. Matsushita et al. (2004) reported that the bottom gillnet settled down completely to the sea bottom within 2 weeks. Because the apparent projected area and surface area of the net bars increase with the accumulation of periphyton with prolonged immersion, drag and friction also increase, increasing the internal forces acting on the net. Prolonged immersion may make the net more likely to break. In particular, as the internal forces at the boundaries between the float and sinker lines increase, the net may break more easily.
Nevertheless, a bottom gillnet with an accumulation of periphyton can retain an effective net area to catch fish for 2 to 3 weeks. Subsequently, however, the fishing ability decreases markedly with the decrease in net height in response to the exponential growth of periphyton, in a logistic curve, according to the simulation. Because the ability to resist the underwater physical environment depends on the periphyton growth rate and the properties of the gear, such as the material used to make the net and the buoyancy of the floats, it is impossible to conduct experimental studies for all net types and situations. As the bottom gillnet was used in northern Japan, the periphyton growth rate may be relatively lower than in the results of Yamaguchi and Nishinokubi (1998) and Yamaguchi (1999). However, if the rate of periphyton growth in a certain area can be checked seasonally, our simulation approach should easily predict the impact of fishing in general terms from a physical perspective.
Because an increase in periphyton accumulation and current speed increase the inclination of a bottom gillnet, the orthogonal projection area of the mesh in the vertical plane decreases. Therefore, an inclined bottom gillnet may encounter fish able to pass through the net at a lower rate. We found that the passage rate of walleye pollock decreased for a vertically stretched mesh because of the decrease in porosity (Shimizu et al., 2005). The inclination of the net decreases the apparent porosity of the net for fish that swim horizontally. Furthermore, fish can detect the net more easily because the accumulated periphyton increases its visibility. Therefore, the number of enmeshed fish attempting to pass through the net would decrease. The real-scale mesh shape and internal force in each mesh estimated in the simulation are useful for elucidating the capture process, because mesh geometry determines the probability that a fish will contact the net bar. Once contact is made, the internal force is the most important parameter determining whether a fish is caught.
This study demonstrated that using a numerical simulation could clarify the impact of fishing from a physical perspective. Many problems exist related to the impact of fishing, and in most cases, scientific understanding from a physical perspective is limited. We believe that our approach has many uses in solving problems dealing with the impact of fishing.
|
Acknowledgements |
|---|
We thank the staffs of the laboratory of Fisheries Production System of Kinki University. We also thank the editors and anonymous referees for helpful comments. This research was supported in part by a grant-in-aid for the 21st Century COE programme, Japan.
|
References |
|---|
|
TopIntroductionMethodsResultsDiscussionReferences |
|---|
-
Bessonneau J. S., Marichal D. Study of the dynamics of submerged supple nets (applications to trawls). Ocean Engineering (1998) 25:563–583.[CrossRef][Web of Science]
Erzini K., Monteiro C. C., Ribeiro J., Santos M. N., Gasper M., Monteiro P., Borges T. C. An experimental study of gill net and trammel net "ghost fishing" off the Algarve (south Portugal). Marine Ecology Progress Series (1997) 158:257–265.[CrossRef][Web of Science]
Kaiser M. J., Bullimore B., Newman P., Lock K., Gibert S. Catches in ghost fishing set nets. Marine Ecology Progress Series (1996) 145:11–16.[CrossRef][Web of Science]
Korte H., Takagi T. Inertia transformation based motion calculation of fishing nets with Matlab/Simulink. (2005) 9. Proceedings of 4th International Symposium on Automatic Control, 22–23. Germany: Wismar.
Lee C., Lee J., Cha B. Calculation model to describe the fishing gear behavior. In: Contributions on the Theory of Fishing Gears and Related Marine Systems—Paschen M., ed. (2002) 2. Rostock: Neue Hochschulschriften. 15–20.
Lee H. H., Pei-Wen W. Dynamic behavior of tension-leg platform with net-cage system subjected to wave forces. Ocean Engineering (2001) 28:179–200.[CrossRef][Web of Science]
Lee M., Lee C., Cha B., Lee J., Yoon H. Dynamic analysis of a cage subjected to the current. In: Contributions on the Theory of Fishing Gears and Related Marine Systems—Paschen M., ed. (2005) 3. Rostock: Shaker Verlag. 51–59.
Matsushita Y., Honda N., Fujita K., Watanabe T. Time series changes in gillnets abandoned at shallow water. Bulletin of Fisheries Research Agency (2004) 10:15–17.
Niedzwiedz G., Hopp M. Rope and net calculations applied to problems in marine engineering and fisheries research. Archive of Fishery and Marine Research (1998) 46:125–138.[Web of Science]
O'Neill F. G. Axisymmetric trawl codends made from netting of a generalized mesh shape. IMA Journal of Applied Mathematics (1999) 62:245–262.
Shimizu T., Takagi T., Hiraishi T., Yamamoto K. Physical model of capture process of gill net. (2005) 8. In Proceedings of 5th Conference of the Society of Mathematical and Physical Fisheries of Science, 27–28. Japan: Hakadate.
Shimizu T., Takagi T., Suzuki K., Hiraishi T., Yamamoto K. Refined calculation model for NaLA, a fishing net shape simulator, applicable to gill nets. Fisheries Science (2004) 70:401–411.[CrossRef][Web of Science]
Takagi T., Shimizu T., Suzuki K., Hiraishi T., Yamamoto K. Validity and layout of "NaLA": a net configuration and loading analysis system. Fisheries Research (2004) 66:235–243.[CrossRef][Web of Science]
Takagi T., Suzuki K., Hiraishi T. Development of the numerical simulation method of dynamic fishing net shape. Bulletin of the Japanese Society of Scientific Fisheries (2002) 68:320–326.
Yamaguchi Y. Effect of porosity on drag coefficient of plane nettings encrusted with marine foulings. Fisheries Engineering (1999) 35:247–251.
Yamaguchi Y., Nishinokubi H. Changes of drag and attaching porosity of marine foulings on plane nets in a set-net fishing ground for given period of time. Fisheries Engineering (1998) 35:43–50.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||