Development and trial of a demersal otter trawl cod escape panel (EMFF project SCO1711): technical summary

(EMFF project SCO1711): technical summary Shaun Fraser, Chevonne H. Angus, Arthur Johnson, Davie Riley April 2018

Development and trial of a demersal otter trawl cod escape panel (EMFF project SCO1711): technical summary Authors: Dr Shaun Fraser, Dr Chevonne H. Angus, Arthur Johnson, Davie Riley Corresponding author: Dr Shaun Fraser shaun.fraser@uhi.ac.uk NAFC Marine Centre Port Arthur Scalloway Shetland ZE1 0UN email: info@uhi.ac.uk web: www.nafc.ac.uk Suggested citation: Fraser, S., Angus, C.H., Johnson, A., and Riley, D. (2018). Development and trial of a demersal otter trawl cod escape panel (EMFF project SCO1711): technical summary. NAFC Marine Centre report. pp 15. Copyright NAFC Marine Centre UHI 2018. All rights reserved. NAFC Marine Centre is the trading name of the Shetland Fisheries Centre Trust, Scottish Charity Number SC003715. i

Contents Executive summary... iii 1 Background and aims... 1 2 Materials and methods... 2 2.1 Gear development... 2 2.2 Sea trials and data recording... 2 2.3 Data analysis... 5 2.4 Underwater camera system... 5 3 Results... 6 3.1 Catch weights... 6 3.2 Length distributions... 8 3.3 Underwater video observations... 10 4 Discussion... 12 5 Acknowledgements... 14 6 References... 15 ii

Executive summary This project was initiated in response to recent increases in cod (Gadus morhua) abundance in the North Sea, especially in the waters around Shetland. The recent increases have proved challenging because cod are mostly caught in a mixed demersal fishery and cod quota has been a limiting factor. The landings obligation has exacerbated the problem of fishermen trying to avoid unwanted catches of cod, while still catching other target species. The escape panel that was designed utilises known, or reported, behavioural patterns of cod in the net. The concept that was developed, constructed and trialled consisted of a horizontally split extension section with an inclined panel to guide fish upwards. The lower sheet of the extension beyond the inclined panel was constructed of large mesh to form an escape panel. The hypothesis was that cod, being powerful swimmers that are inclined to swim downwards would, after passing over the horizontal panel, swim forwards and downwards in the extension and then escape through the large mesh escape panel, while other species would fall back into the cod-end. The trials were undertaken using the NAFC Marine Centre s fishing vessel MFV Atlantia II. Trials were undertaken using a standard comparative paired trawls methodology. Thirteen valid comparative pairs were completed and an initial assessment of the results has been undertaken. In total, 1525 kg of cod was caught with the control net compared to 429 kg with the experimental net. Average cod catch per unit effort (CPUE) was 78 kg hr 1 in the control net. CPUE was significantly lower (p < 0.001) in the experimental net at 18 kg hr 1. The results suggest that overall over two-thirds of cod were escaping from the experimental net, compared to the control net. No consistent differences in catches of other species were recorded between the control and experimental nets suggesting that other commercially important species were not finding the escape panel. Overall the length distributions of cod were significantly different between the control and experimental nets (p < 0.001). In total, more small cod (< 35 cm) were recorded from the experimental net, however two-thirds of these were caught during one haul. Length distributions of plaice were significantly different between the nets (p = 0.001). Length distributions of haddock were not significantly different between the nets (p > 0.05). Based on the results of this trial it can be concluded that the design of escape panel has significant potential as a technical measure to significantly reduce cod catches without negatively impacting on catches of other commercially important species. Whilst these initial results are extremely promising, further trials are recommended to determine if the panel is as effective during longer tows, within different mixes of fish, and onboard larger vessels. iii

1 Background and aims Increases in cod (Gadus morhua) abundance in the North Sea over recent years, especially in the waters around Shetland, have proved challenging for Shetland vessels undertaking the mixed demersal fishery in the area. Vessels, often with limited cod quota, can often encounter high catches of cod when targeting other species which can lead to rapid uptake of available quota. This project involved the development and trial of a modified trawl design that could allow some of the cod catch to escape from the net while retaining the remainder of the catch. Two identical trawls were constructed, one acting as a control, to which catches from an experimental net were compared. The experimental net was modified whereby there was a horizontally split extension section behind an inclined panel that guided fish upward into the top section of the extension. The lower side of the extension, beyond the inclined panel, incorporated a large mesh panel for fish to escape. The proposed design was considerably different from previous trawl modification trials aimed at minimising catches of unwanted species (for a useful review see O Neill and Mutch, 2017) and aimed to utilise the differing behaviours of target species. Cod are known to be powerful swimmers (Krag et al., 2009) that have a preference for the lower part of the trawl (Rosen et al., 2012). In comparison, other gadoids such as haddock (Melanogrammus aeglefinus), saithe (Pollachius virens), and whiting (Merlangius merlangus) are known to have a tendency to swim upwards within the net (Holst et al., 2009). It was hypothesised that during towing or when the vessel slows to haul, cod will swim downwards and forwards going below the horizontal panel and towards the large mesh panel in the bottom section of the extension, allowing them to escape from the trawl. Species with a tendency to swim upwards would not locate the escape panel. The aim of this project was to undertake preliminary trials to assess the effectiveness of the proposed design on the catches of cod and other species, and to determine whether the experimental trawl has potential for further development and modification to optimise escapes of unwanted species. 1

2 Materials and methods 2.1 Gear development An experimental and control trawl were constructed at the NAFC Marine Centre following the evaluation of scale models (Figure 1). The nets were based on a standard Jackson demersal otter trawl with a fishing circle of 352 (160 mm) meshes (Figure 2). Figure 1. Construction of the experimental and control trawls at the NAFC Marine Centre. The experimental and control nets were identical except for modifications in the net extension near the cod-end of the experimental net. These modifications consisted of an inclined panel and a horizontal middle panel in the extension (Figure 2) made using 120 mm mesh net and single 4 mm twine. Under the horizontal middle panel, a bottom sheet escape panel was constructed of 300 mm square mesh using double 5 mm twine. Both the experimental and control trawls were rigged and fished in the same way. The ground gear on both trawls consisted of 23 m of 150 mm and 200 mm diameter rubber disks. The headline was fitted with 30 floats of 200 mm diameter. The trawl doors were Bison No. 6½ (Edwin Ashworth Ltd., York). The sweeps comprised of 46 m of doubles and 46 m of singles. 2.2 Sea trials and data recording Sea trials were undertaken between the 30 th of November 2017 and the 1 st of March 2018 using the NAFC Marine Centre s fishing vessel the MFV Atlantia II (LK 502). The Atlantia II is 12 m in length, has a 131 kw (172 HP) main engine, and is fitted with matching twin net drums. Fishing was undertaken at approximately 2.5 knots with trawl performance (headline height and door spread) routinely assessed using a Notus trawl monitoring system. An alternate haul method was used, with a standardised pattern of haul pairs consisting of experimental and control tows in the same areas. Haul pairs were fished along the same track during daylight conditions as close together in time as possible. The first tow of each pair was swapped between the experimental and control trawls after each valid pair. Depending on the conditions and available time, haul pairs were either fished in the same or opposite directions. Where tide was 2

determined to be factor in haul pairs, efforts were made so that the following pair matched the previous tidal conditions but with opposite control/experimental trawl order. Figure 2. Trawl design schematics. Top schematics show the overall design of the bottom and top panels based on a standard Jackson demersal otter trawl. Bottom schematics show details of the net extension and codend, with coloured arrows demonstrating the movement of retained and escaped fish. Experimental and control nets were identical except for the escape panel, inclined panel, and middle panel in the experimental net. 3

In total, 31 hauls were carried out during the trials of which 26 hauls in 13 pairs provided valid data to compare the experimental and control nets. One haul was undertaken for recording video footage of the experimental net performance, and the remaining four hauls were invalid due to either damage to the gear or entanglement in either the first or second haul of the pair. The locations of valid haul pairs are shown in Figure 3. Fishing was undertaken on grounds where cod were expected to be present in sufficient numbers at the time of year. Fishing activity was restricted to locations that were operationally manageable in the winter conditions. Valid hauls were carried out in depths ranging from approximately 60 to 130 m. Tow durations ranged from approximately 45 mins to 2.5 hours. Figure 3. Locations of haul pairs by the Atlantia II around Shetland. Each location represents the average midpoint of each valid haul pair (numbered 1 to 13) calculated between shooting and hauling coordinates. Contains Ordnance Survey and SeaZone data Crown copyright/hr Wallingford Ltd 2018. The catch from each haul was sorted and the total live weights were recorded. For commercial species, the catch was counted and individual fish total lengths were recorded (cm below). For non-marketable species (those not typically landed by boats into Shetland markets) the combined total live weight was recorded. During the 13 th pair of valid hauls, only the cod weight could be recorded due to operational limitations. In the first 12 pairs of valid hauls, complete weight and length data was recorded. In some hauls, subsampling for length measurements was required for large numbers of haddock and plaice. No subsampling was required for cod. 4

2.3 Data analysis All data visualisation and statistical analysis was carried out using R (R Core Team, 2018). Preliminary analysis has focussed on live weight data and the length distributions of cod and the other largest contributing species by weight. To compare catch rates, catch data were expressed as catch per unit effort (CPUE), the catch per towing time with units kg hr 1. Following F-tests for variance, CPUE for the control and experimental nets was compared using the non-parametric Wilcoxon signed rank test for matched pairs. Raised length distributions were used for the analysis of length data where any subsampling had been necessary. The non-parametric two-sample Kolmogorov-Smirnov test was used to statistically compare length distributions. The proportion of fish retained by the experimental net per length class (1 cm bins) was defined P(L) = N L,trial /(N L,trial + N L,control ) where N L,trial is the total count of fish of length L from the experimental net and N L,control is the total count of fish of length L from the control gear. The resulting proportions were fitted with a smoothing spline regression with four degrees of freedom as in Vogel et al. (2017). 2.4 Underwater camera system A battery-powered trawl camera system was developed using NAFC equipment to provide additional information on the performance of the experimental trawl and observations of fish behaviour in the cod-end and around the escape panel. A Veho Muvi K2 NPNG was used to record high definition video footage in depths to 100 m. Artificial lighting was required at depth for illumination and two BlueFire 1200 Lumen XM-L2 LED diving lights were used. The camera and lights were mounted in a custombuilt protective stainless-steel housing which provided a clear field of view and prevented any damage or snagging. The housing was fitted to the net by means of bolts attaching a separate float plate. The combined camera, lights, housing, and float plate were neutrally buoyant in sea water and did not affect normal fishing operations or gear rigging. The trawl camera system was attached in the cod-end on the top panel facing forwards towards the escape and middle panels (Figure 4). The camera was set to record continuously from when the gear was shot. The footage was downloaded from an SD card once the gear was hauled and the camera system recovered. Figure 4. Fitting and deployment of the trawl camera system in the experimental net on the Atlantia II. 5

3 Results In the 26 valid hauls, a total catch of 5657.9 kg was recorded of which 5332.5 kg was of marketable species. In the control trawl, a total weight of 3365.9 kg was recorded from approximately 25.5 hours of towing time in which 3209.2 kg was of marketable species; and in the experimental trawl 2292.0 kg was recorded from approximately 25.5 hours of towing time of which 2123.3 kg was of marketable species. In total, 4190 individual fish lengths were measured. Non-marketable species present were predominately: common dab (Limanda limanda), flounder (Platichthys flesus), grey gurnard (Eutrigla gurnardus), red gurnard (Chelidonichthys cuculus), and lesser-spotted dogfish (Scyliorhinus canicular). An example catch from the control net is shown in Figure 5. Figure 5. A typical catch from a single haul of the control net before sorting. 3.1 Catch weights The overall measured weights for different species and summed over all valid hauls is shown in Table 1 and Figure 6. Total towing time of each net was approximately 25.5 hours. These results show that by far the biggest difference between the control and experimental catch is in the catches of cod, with 1524.7 kg of cod in the control hauls compared to 428.9 kg in experimental hauls, corresponding to an overall 71.9% reduction in cod catch in the experimental net. Except for cod, the weight and catch proportions in the experimental net catches were mostly similar to those of the control net. The combined total weight of all other species in the control hauls was 1841.2 kg and 1863.1 kg in the experimental hauls. Cod is the biggest contributor by weight in the control catch, followed by haddock and plaice which together with cod accounted for 72.7% of the total control catch. The higher catch of ling in the experimental net results was due to a single haul in which 123.5 kg of large (up to 116 cm length) ling were caught. 6

Table 1. Overall summed measured weights for different species from all valid hauls. Species Scientific name Control net total (kg) Experimental net total (kg) Cod Gadus morhua 1524.7 428.9 Haddock Melanogrammus aeglefinus 479.3 506.8 Plaice Pleuronectes platessa 442.2 430.5 Cuckoo ray Raja naevus 228.1 187.1 Monkfish Lophius spp. 185.8 162.9 Thornback ray Raja clavata 89.7 68.0 Spotted ray Raja montagui 65.2 36.3 Lemon sole Microstomus kitt 48.7 46.0 Whiting Merlangius merlangus 45.0 41.2 Squid Loligo spp. 36.5 59.2 Saithe Pollachius virens 24.8 3.3 Ling Molva molva 11.7 126.0 Sandy ray Leucoraja circularis 10.3 8.7 Hake Merluccius merluccius 7.2 0.9 Megrim Lepidorhombus whiffiagonis 5.9 9.8 Witch Glyptocephalus cynoglossus 2.6 4.0 John Dory Zeus faber 1.4 1.1 Mackerel Scomber scombrus 0.1 0.3 Turbot Scophthalamus maximus 0.0 1.4 Herring Clupea harengus 0.0 0.9 Non-marketable --- 156.7 168.7 Figure 6. Overall summed measured weights for different species from all valid hauls. To investigate the catch rates between the control and experimental nets, catch data is presented per haul in Figure 7 and grouped per haul pair. Cod catch results are compared to the combined catch of all other species. In all cases, cod CPUE is greater in the control net than in the experimental net. The results of the Wilcoxon tests show that overall cod catch rates are statistically different (p < 0.001) between the control and experimental nets. However, the reduction in cod catch in the experimental net ranges from as little as 9.3% less in haul pair 5 to as much as 96.6% less in haul pair 4. Average cod 7

CPUE for the control net was 78.2 kg hr 1 and for the experimental net was 17.7 kg hr 1, corresponding to a 77.4% reduction in cod catch rate. Due to the possibility that pair 4 could have been skewing the results, the Wilcoxon test was also applied to the cod CPUE data with pair 4 removed and the difference between the nets was still found to be significant (p < 0.001). As only cod data was recorded from haul pair 13, there were 12 valid pairs available for comparison for other species (Figure 7). Catch rates of other species varied between approximately 30 kg hr -1 in pairs 2 and 12, to approximately 180 kg hr -1 in pair 8. Half of the pairs had a higher catch rate in the control net and half had a higher catch rate in the experimental net. Overall, there was no statistically significant difference in catch rates of other species between control and experimental nets (p > 0.05). Average CPUE for other species was 80.09 kg hr 1 for the control net and 81.93 kg hr 1 for the experimental net. Figure 7. Catch results for cod and for all other species presented per haul pair. 3.2 Length distributions Length distributions were investigated for the three main species (cod, haddock and plaice) caught with the control and experimental nets (Figure 8). Catches of these three species accounted for 73% of catches with the control net and 60% of catches with the experimental net. Kolmogorov-Smirnov tests show that the most significant difference between control and experimental net catches was in the length distributions of cod (p < 0.001). Length distributions of cod ranged from 27 cm to 98 cm. For lengths greater than approximately 35 cm there were consistently less cod in the experimental 8

net when compared to the control. The higher number of small cod (less than around 35 cm) in the experimental net was mainly due to a relatively large catch of small cod in the experimental net of comparative pair 7. Small cod from this haul accounted for 66% of cod 35 cm from all of the experimental net tows. Length distributions of plaice were also significantly different between the control and experimental nets (p = 0.001). Plaice length distributions ranged between 19 to 51 cm in the control and 25 to 55 cm in the experimental net. The length distribution of plaice in the control net peaked at 31 cm while the distribution in the experimental net peaked at 34 cm. There was no significant difference in the length distributions of haddock between the two nets (p > 0.05). The haddock length distributions were bimodal with peaks at approximately 31 cm and 46 cm. Figure 8. Length distribution results for main species caught during trials. 9

The differences in length distributions of cod between the nets are looked at in more detail by considering the proportion of fish in the experimental net (Figure 9). The smoothing spline is fitted over the length classes where there is a sufficient count in the control data. A P value equal to 0.5 indicates no difference in catch rates of cod between nets at a given fish length; a P value less than 0.5 indicates less cod of a given length retained in the experimental net when compared to the control net. The fitted trend suggests that there may be an increased probability of escape in the experimental net with increasing cod length. Figure 9. The proportion of cod retained by the experimental net (P) as a function of length. The thick line shows a regressive smoothing spline fitted to the data with four degrees of freedom. The dotted line shows P = 0.5 which represents an equal number of fish in the control and trial results for each length class. 3.3 Underwater video observations Footage from the camera system provided additional insights which were useful in the interpretation of results (Figure 10). The performance of the net and behaviour of fish was clear with consistent good visibility. Some cod were observed swimming quickly underneath the middle panel against the flow to escape through the escape panel. Other species were observed to unsuccessfully interact with the escape panel and middle panel before tiring and falling back to the cod-end. It was noted that thornback rays would occasionally become entangled in the middle panel as they passed down the extension which would force down the middle panel resulting in the temporary blockage of the escape panel (e.g., Figure 10c). 10

TP MP TP MP BP BP a) b) TP MP TP MP BP c) d) BP Figure 10. Screenshots from underwater camera observations. BP = bottom panel, MP = middle panel, and TP = top panel. Footage proved useful in observing: a) the net structure and performance; b) the movement of cod to the escape panel; c) fish interactions with the inclined and middle panel (the white arrow highlights the entanglement of a thornback ray); and d) the behaviour of other species in the net extension. 11

4 Discussion The results and analysis provide a promising indication of the effectiveness of the escape panel trialled in this study. Consistently less cod were retained in the experimental net over the range of depths, conditions, and fishing grounds investigated. This resulted in an overall 71.9% reduction in the cod catch. Of equal importance, the results also showed that catches of other species were not significantly different between the two nets. Consequently, the initial trial results indicate that the experimental design is a highly selective and effective method for reducing cod catch without affecting the remaining catch (under the conditions in which it was used). The sampling method that was used in this study was the alternate haul method. This method uses a single trawl and that is the dominant gear type used to catch demersal whitefish around Shetland. As with any method there are advantages and disadvantages, however for the objectives and circumstances of this study it was deemed appropriate. The method has been used successfully in various gear trials around Shetland in the past (e.g., Bullough et al., 2007) where the method reported relatively low natural variability in catches between successive hauls. The disadvantage of this method is the larger number of hauls required, than for example in gear trials using twin-rig, and the difficulty in maintaining consistent conditions during each haul pair (Wileman et al., 1996). To both maximise the numbers of comparative hauls and minimise the variation in environmental conditions between consecutive tows, the tow duration was kept relatively short, at around two hours. This is a shorter duration than that of normal commercial conditions. However, the aim was to test the concept of the design rather than its performance under normal commercial practice. The other commonly used approach during selectivity trials is the covered cod-end method, or an adaptation of it for escape panel trials where a cover would be put over the panel to catch escaped fish. The covered cod-end method is known to introduce bias (e.g., Masden and Holst, 2002) and for the design trialled in this study, with the panel being on the underside of the extension, it was not a workable option. In this study there was variability between tows and this needs to be taken into consideration when interpreting the results. Examples that have been highlighted are the comparatively high cod catch in the control net of tow pair 4 and the high proportion of smaller cod caught in the experimental net of comparative pair 7. Also, the ling catch was all attributable to one haul using the experimental net. A degree of variability between hauls is inevitable and cod and ling are known to be species where variability between hauls is common. Within each comparative pair the cod CPUE was higher in the control net. However, there was variation in the differences in cod catch rates between the control and experimental between the comparative pairs. Some of this variation will be due to differences in the amounts of cod being encountered during the tows, however there is evidence from the video footage that other species within the catch may reduce the effectiveness of the panel at different times. Skate becoming entangled with the middle panel which caused a temporary reduction in the accessibility of the escape panel was observed on the video footage. Some refinements to the inclined and middle panel to reduce snagging could reduce the likelihood and impact of this effect. Potentially alterations to the experimental gear likely to affect escape probability could include moving the position of escape panel 12

relative to the cod-end and varying the length or height of the middle panel. Previous studies have shown the importance of the location of separator panels along the length of the trawl for species selectivity (Holst et al., 2009). The results from length measurements suggested that the cod escape probability in the experimental net increased with fish length. The maximum sustained swimming speeds of fish around fishing gear is related to their length (He, 1993), and so the reduction in proportion of retained cod with length is interpreted as being due to the ability of larger individuals to swim at speed forward against the flow and under the middle panel then out the escape panel. Due to the relatively high number of small cod in the experimental net of pair 7 it is suggested that the results in relation to effects on escape probabilities with increasing fish length are taken with caution and further data collection would be recommended. In contrast to the situation for cod, there were no significant differences in length frequency distributions between control and experimental nets for or haddock and this indicates that even the larger individuals are either not inclined to, or not capable of swimming under the net extension and out through the escape panel. This report describes the promising preliminary results that have been obtained during that trials that have been undertaken and indicate that the design concept warrants further investigation as a technical measure that has good potential to reduce unwanted catches of cod, without significantly affecting catches of other commercially important species. Should there be sufficient industry interest and support it is recommended that the design is trialled onboard a larger demersal trawler operating under normal fishing practices. There are also a range of refinements to the panel design which are possible and warrant further investigation. More video footage would be highly informative and would enable a more quantitative video analysis of behaviour. Further trials and increased numbers of haul pairs would improve the statistical power of analysis and could provide information on the effectiveness of the experimental design within different mixes of catch, at different times of the year or in different locations. 13

5 Acknowledgements This project received funding as a Fisheries Local Action Group (FLAG) project through the European Marine Fisheries Fund (EMFF). Arthur Johnson is credited with the design of the escape panel. The authors wish to thank Dr Paul Macdonald for his contribution to the early development of the project, Mr Joseph Fraser for fabrication of the camera housing, and Mr Mark Hamilton for assistance with data collection. 14

6 References Bullough, L.W., Napier, I.R., Laurenson, C.H., Riley, D., Fryer, R.J., Ferro, R.S.T. and Kynoch, R.J., 2007. A year-long trial of a square mesh panel in a commercial demersal trawl. Fisheries research, 83(1), pp.105-112. Krag, L.A., Madsen, N. and Karlsen, J.D., 2009. A study of fish behaviour in the extension of a demersal trawl using a multi-compartment separator frame and SIT camera system. Fisheries Research, 98(1-3), pp.62-66. He, P., 1993. Swimming speeds of marine fish in relation to fishing gears. ICES Marine Science Symposia, 196, pp.183 189. Holst, R., Ferro, R.S., Krag, L.A., Kynoch, R.J. and Madsen, N., 2009. Quantification of species selectivity by using separating devices at different locations in two whitefish demersal trawls. Canadian Journal of Fisheries and Aquatic Sciences, 66(12), pp.2052-2061. Madsen, N. and Holst, R., 2002. Assessment of the cover effect in trawl codend selectivity experiments. Fisheries Research, 56(3), pp.289-301. O Neill, F.G. and Mutch, K., 2017. Selectivity in Trawl Fishing Gears. Scottish Marine and Freshwater Science, 8(1), pp.1-85. R Core Team, 2018. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Rosen, S., Engås, A., Fernö, A. and Jörgensen, T., 2012. The reactions of shoaling adult cod to a pelagic trawl: implications for commercial trawling. ICES journal of marine science, 69(2), pp.303-312. Vogel, C., Kopp, D., Morandeau, F., Morfin, M. and Méhault, S., 2017. Improving gear selectivity of whiting (Merlangius merlangus) on board French demersal trawlers in the English Channel and North Sea. Fisheries Research, 193, pp.207-216. Wileman, D.A., Ferro, R.S.T., Fonteyne, R. and Millar, R.B., 1996. Manual of methods of measuring the selectivity of towed fishing gears. ICES Cooperative Research Report, 215, pp.1-126. 15