Mortality and injuries of haddock, cod and saithe escaping through trawl meshes and sorting grids

Transcription

1 Mortality and injuries of haddock, cod and saithe escaping through trawl meshes and sorting grids Thesis submitted in partial fulfilment of the requirements for the Master of Science degree in Fisheries Science By Ólafur Arnar Ingólfsson Faculty of Natural Resource Science University of Akureyri Iceland May 2002

2

3 Acknowledgement I am in great debt to my supervisors Aud Vold Soldal and Anders Fernö for their unselfish and priceless contribution. They frequently gave helpful comments on my progress and my thesis. To Aud, I am very grateful for allowing me to participate in these experiments and not least, for housing me in the so much giving environment of the Fish Capture Division, Institute of Marine Research in Bergen. Irene Huse gets special thanks for her contribution: Precise fish recordings, valuable help and comments concerning statistics and the writing of the thesis. I also thank Terje Jørgensen for statistical help and Norman Graham for his comments on the thesis. All the research participants from the Fish Capture Division and the crewmembers of the participating vessels I thank for their contribution and collaboration. The staff and students at the Fish Capture Division I thank very much for the goodfellowship and priceless environment during my studies. Jón Þórðarson, manager of the Faculty of Natural Resource Science gets special thanks for support and time saving arrangement of my studies. My girlfriend and coming wife, Elín Halldóra Friðriksdóttir I thank for moving with me to Bergen and her support during the studies. My contribution to the family live as been at minimum level the past months, while her share has increased without complains or impatience. Last but not least, I want to thank the Marine Research Institute in Iceland for their financial support during my studies.

4

5 Table of contents Table of contents Abstract...1 Introduction...2 Materials and methods...5 Trial Vessels and gear...6 Methods...8 Trial Vessels and gear...12 Methods...12 Data analysis...14 Results...16 Mortality and catch composition...16 Trial Trial Fish injuries...26 Trial Trial Fish traps...41 Overview of results...42 Discussion...44 Experimental design...44 Experimental area...44 Experiment procedure...44 Cage construction...46 Trawls and codend...48 Cod and saithe mortality and injuries...49 Haddock mortality and injuries...49 Differences between mesh, grid and control cages...49 Length-dependent mortality and injuries...51 Mesh escape mortality...52 Further experiments...52 Conclusions...54 References...55 Appendixes...59

6

7 Abstract Abstract Mortalities and injuries of the gadoids haddock (Melanogrammus aeglefinus L.), cod (Gadus morhua L.) and saithe (Pollachius virens L.) were studied after escapes from a bottom trawl through 135 mm cod-end meshes and a Sort-X sorting grid with 55 mm bar space. Two full-scale trials in the Barents Sea were carried out in 2000 and 2001, using commercial trawlers and bottom trawls. The fish that escaped were sampled using small-meshed cages attached to a cod-end cover and a cover enclosing the sorting grid. Trawl-caught controls were sampled by removing the cod end and attaching the cage directly to the cod-end extension. In the 2001 trial, control fish were also sampled in fish traps. The timing of sampling during trawling was determined by acoustic closing and releasing devices. Survival rates of cod and saithe that escaped through the cod end and sorting grid were 100%. Mortality rates of haddock were 26.2 to 50.4% (cod-end escapees), 1.6 to 20% (grid escapees), 4.1 to 26.5% (trawl-caught controls) and 0% (trap-caught controls). The haddock mortality decreased with increasing fish length in all groups, with a peak in the mortality k of the mesh escapees whose girth was approximately the mesh size circumference. Cod and saithe had significantly fewer skin and fin injuries than haddock, and the frequency of skin injuries increased generally towards the tail. Scale loss was also related to size, decreasing with increasing fish length in all groups. Fish that escaped from grids had fewer skin and fin injuries than the mesh and control groups. Causes of mortality and injuries are discussed and potential improvements in the experimental procedures are suggested. 1

8 Introduction Introduction For the efficient management of any fishery, the overall mortality associated with the exploited fish population needs to be considered. If this is not done, stock size estimates will be biased, with the degree of inaccuracy depending on the scale of the unknown mortality (Alverson et al., 2000). In Icelandic waters, the recruitment of haddock (Melanogrammus aeglefinus L.) to the catchable stock has not been as expected in recent years, resulting in lower stocks and catches than predicted. A partial explanation of this has been suggested to be discard mortality (Pálsson, in press). Discard mortality differs between species and fishing methods (Alverson et al., 2000) and is reckoned to be high in gadoids (Hokenson & Ross, 1993, Suuronen, 1995; Soldal, 1996), and reducing catches of unwanted fish is therefore a preferable to discarding. In trawl fisheries, most fish selection takes place through the meshes of the cod end (Wileman et al., 1996) or, where appropriate, the bars of sorting grids (Larsen and Isaksen, 1993). The selective devices select out small fish, which are usually of less value or illegal to catch, and it is of vital importance for the fish stock that juvenile fish recruit to the spawning stock. Mortality after escapes of fish from fishing gear is one of several possible sources of unaccounted mortality in trawl fisheries. It is of great importance for sustainable fisheries that fish which escape from fishing gear should survive. Survival rates of fish that escape through meshes differ between species. Mesh size regulations in Baltic herring (Clupea harengus L.) trawl fisheries, for instance, have not been recommended, due to the low survival of escapees (Suuronen, 1995). Studies of cod (Gadus morhua L.) and saithe (Pollachius virens L.) escaping through cod-end meshes and sorting grids have shown little or no mortality (Vinogradov, 1960; Engås et al., 1989; Engås et al., 1990; Soldal et al., 1991; Main and Sangster, 1991; Jacobsen et al., 1992; DeAlteris and Reifsteck, 1993; Jacobsen, 1994; Suuronen et al., 1995). Haddock tend to be more vulnerable, but experimental results have been highly variable, with survival rates ranging from 0 to 100%. Zaferman and Serebrov (1989) questioned the regulation of cod and haddock catches by changing mesh size on the basis of lack of proof of survival of mesh escapees. Main and Sangster (1990) investigated injuries and survival of haddock that escaped from the cod end of a 2

9 Introduction demersal trawl, with divers collecting escaped haddock. Survival rates ranged from 0 to 67% and the fish suffered substantial damage to their scales. These results led to further experiments with improved sampling methods, in which cod-end escapees were collected by a cover surrounding the cod end. Most other studies have found mortality rates in haddock that range from 0 to 30% (Main & Sangster, 1991; Sangster and Lehmann, 1993 and 1994; Lowry and Sangster, 1996; Sangster et al., 1996). Less variable mortality was found by Soldal et al. (1991) and Wileman et al. (1999), who found mortality of mesh and grid haddock escapees to be less than 10%, while Jacobsen (1994) observed 15% mortality. Survival rates increase with fish length, and increasing mesh size and square-mesh cod ends therefore increase the potential survival of haddock and whiting (Main & Sangster, 1990 and 1991; Lowry et al., 1996; Sangster et al, 1996). Smaller fish were found to have more scale damage than bigger fish and the scale damage was greater towards the tail. Most mortality occurred during the first 24 hours after escape, and declined with time (Lowry et al., 1996; Main & Sangster, 1991; Sangster et al., 1996). Sangster et al. (1996) also observed that the smaller escapees died sooner than the larger individuals. Escape mortalities are probably the result of injuries and/or physiological trauma experienced as a result of the fishing process (Breen & Sangster, 1997). Physical injuries such as scale loss and sores can lead to stress and disturbed osmotic balance (Eddy, 1981, Sangster & Lehmann, 1993; Smith, 1993), and exhausted fish in the trawl mouth are likely to be hit and injured by a trawl. The skin and scales constitute an osmotic barrier between the fish s outer and inner environment (McWilliams, 1992), and substantial scale loss can lead to disturbance in osmotic balance and be a cause of death (Engås et al., 1989; Engås et al., 1990; Smith, 1993). Stress can be a vital factor, as described by Jonsson (1994), who observed 40% mortality of the haddock that remained in a tank after he had captured haddock from the tank with a dip net. Intensive exercise, leading to exhaustion, can cause mortality (Wood et al., 1983), probably by intracellular acidosis. Having escaped from a trawl, fish may be more vulnerable to predation if they are injured and/or exhausted. However, one experiment could not demonstrate that cod escaping from a trawl suffered a greater risk of predation (Løkkeborg & Soldal, 1995). 3

10 Introduction Breen et al. (1998) and Wileman et al. (1999) showed that the survival of individual fish may be affected by the length of time spent in the cod-end cover during sampling, exposed to continous waterflow. In previous experiments that used cod-end covers to sample escaped haddock, mortality was probably overestimated. Knowledge of cover-induced mortality is important in further experiments, and stresses the importance of refining future experimental designs. Marteinsson (1991) observed fish resting against the rear netting of the sampling cage, and since swimming performance correlates with length (Wardle, 1977; He, 1993), the smallest individuals may not be able to maintain their position in the cover. Gadoid survival in bottom trawl fisheries was studied by Norwegian marine scientists in the late 80s and early 90s. Ten years later, fishermen and their organisations criticised the experiments for a lack of realism in the setup. The experiments were not carried out on fishing grounds where several vessels were present, and it could therefore be asked if the mortalities of fish that passed through sorting grids several times might have been higher. It was also claimed that the experimental hauls were not of commercial length and with normal catch sizes. Furthermore, in previous experiments, except for the experiments of Wileman et al. (1999), fish were sampled only at the beginning of the trawl hauls, which could have biased mortality estimates. For this reason, a new set of experiments was carried out in 2000 and With the technique used in this study, the sampling period can be controlled by closing and releasing the sampling cages from the trawl by means of acoustic releases. Sampling periods lasted from 5 to 15 minutes after one hour of towing. Data were collected and analysed with the aim of studying: Mortality and injuries of gadoid that had escaped through the cod-end meshes and bars of a sorting grid in the Barents Sea bottom trawl fisheries. Relationships between fish size, mortality and injuries. 4

11 Materials and methods Materials and methods Survival and injury rates of cod, haddock and saithe escaping through cod-end meshes and bars of a sorting grid have been estimated from data from two full-scale fishing experiments. Earlier survival experiments have been criticised for not having been done in a fishing area where trawlers had been actively fishing. In the experiments discussed here, commercial trawlers were chartered to trawl within a specified area before the experiments started. Around 70 hauls were made each year. Trial 1 was carried out from 3 rd to 23 rd of August 2000 and trial 2 from 9 th to 31 st of August Both trials took place in the Barents Sea off the Varanger Peninsula (Figure 1). Figure 1?) in the upper right corner. 5

12 Materials and methods Trial 1 Figure 2?? Five vessels were chartered for this experiment (Table 1). On August 3 rd, six days before the survival trials themselves commenced, three chartered stern trawlers began fishing within the test area (Figure 2), where they towed until August 9 th. Within the area, the skippers themselves decided where to trawl and the length of tows, as long as fishing was carried out under normal commercial conditions. After six days of controlled fishing the survival experiments started. Vessels and gear The three commercial trawlers were all rigged with their own commercial single trawl gears and Sort-X sorting grids. The nominal mesh size in the cod-ends was 135 mm and bar spacing in the grids was 55 mm. Myrefisk II, a commercial stern trawler, was chartered for the survival experiments. Sænes, a commercial Danish seiner, was hired to tow and anchor the cages on the seabed for observations and registration of injured fish. (Table 1). 6

13 Materials and methods Table 1 Vessels Role in experiment Length (Loa) m Gross Tonnage (GT) Main engine (HP) Sletnes Fishing Bliki Fishing Glomfjord Fishing Myrefisk II Experiment hauls Sænes Auxiliary vessel Myrefisk II was rigged with an Alfredo no. 3 trawl with an extension between the trawl and cod end of 14 metres. The trawl was rigged with rockhopper footgear. The cod-end used for survival experiments was made of 2 5mm braided Magnet-PE twine. Nominal mesh size was 135 mm (inside measurement). 20 meshes in a row were measured with a mesh gauge. Measurements ranged from 133 mm to 145 mm with an average of 138 mm (SE = 0.7 mm). The overall length of the cod-end was 9.4 m and the width 31 meshes (selvedges included). A standard Sort-X sorting grid, mandatory in Norwegian waters north of 62 N, was used in experiments in which grid escapees were sampled. Two sets of cover nets were used. One was attached to the front of the cod-end, covering the entire cod-end used in trials where cod-end escapees and control fish were collected (Figure 3 B and C), while another enclosed the sorting grid (Figure 3 A). Two acoustic releases (AR 661 B2S from Oceano technologies) were mounted on the cover net in front of the cage. The first one was triggered to close the aft end of the cage and the second to release the cage from the trawl and close the cage in front. A total of nine cages were used to collect fish during the research period. They were cylindrical; 5 m long, 2 m in diameter and constructed of knotted square mesh net with mesh size 50 mm (stretched mesh), and twine diameter 1.8 mm Polyethylene (PE). The rear end of the cage (the closing net) consisted of 19.6 mm (stretched mesh) Polyamide (PA) net. 7

14 Materials and methods A B C Figure 3 The FOCUS - a towed underwater vehicle fitted with an underwater SIT camera and a pan and tilt unit was used to observe the trawl and cages during towing. I order to observe the cages after release, an underwater camera connected to a video recorder was rigged on the lower end of a metal bar and lowered to the cage. More detailed drawings are found in Appendix I. Methods The survival trials started on 9 th August, six days after the three trawlers began fishing in the area. The nine cages were systematically rigged in three sets. Each set consisted of three hauls. Fish were sampled from grid and mesh escapees plus a control haul, for which the cod-end was removed. In order to minimize possible dependent variation in fish density, all cage-groups were spread over the observation period (Table 2). Table 2 Haul number Sampling group Fish Fish sorted out sorted out by grid by mesh Control fish Control fish Fish sorted out by mesh Fish sorted out by grid Fish sorted out by grid Fish sorted out by mesh Control fish After rigging the cover net and cage the trawler started trawling and towed for approximately one hour with the cage open at the rear, letting all fish pass through. Then a signal was sent to the first release unit, which released a sea anchor connected to the rear end of the cage and closed it. The cage was then monitored by FOCUS 8

15 Materials and methods and when a sufficient number of fish ( individuals) was estimated to have entered the cage (after five to ten minutes), a signal was sent to the second acoustic release, which released floats and the cage from the cover net. The floats rose to the surface, maintaining tension on the front end of the cage and keeping it closed (see Figure 3 for chronological order for grid-cage release). A B C Figure 4. Numbered buoy Radar reflector Battery pack Buoy 3 x 8" floats 2 x 8" floats 50 fathoms 2 x deep 50 m G - hook Stopper 2 x 10 Kg weights Weight: Kg chain on lowest hoop 18 Kg anchor + 3 m chain Figure 5 9

16 Materials and methods Figure 6 The auxiliary vessel then raised the cages to a depth of m and anchored them on the fishing grounds. An active radar sonde was mounted on the marking buoy in order to track the cages during the next few days (Figure 5). Immediately after the cages were anchored they were inspected by the underwater camera in order to estimate the quantity of fish and check that they were properly closed. The cages were subsequently inspected every second day until they were recovered after seven days in the sea. Numbers of live and dead fish were then registered. The fish were measured (total length) to the nearest cm, and the extent of injuries of living individuals was recorded. Fish were declared to be alive if they moved when touched. When the number of live fish exceeded 150, one third of them were taken as a sample, otherwise all individuals were measured. All dead fish were also measured. A random sample of 40 live haddock was observed, with the surface of the fish being divided into 38 areas for recording injuries. Each flank was divided into 14 areas and 10 fins were also defined as registration units (Figure 6). Fin injuries (cleft in fins), fin damage (bare fin rays), bleeding (blood visible but epidermis not ruptured), wounds (epidermis ruptured) and infections (purulence visible) were recorded. Fin injuries of all the cod and saithe were noted and skin injuries recorded. 10

17 Materials and methods Trial 2 Figure 7? The experiment procedure was the same as in 2000, except for cage construction and observation technique. The three chartered stern trawlers fished within the defined test area from August 8 th to 15 th. A likely explanation of the high mortality in trial 1 was thought to be strong currents in the water. It was therefore decided to release the cages close to a fjord, tow them further into a sheltered area and anchor them on the bottom. 11

18 Materials and methods Vessels and gear Table 3 Vessels Role in experiment Length (Loa) m Gross Tonnage (GT) Main engine (HP) Nordfjordtrål Fishing Comet Fishing Glomfjord Fishing Anny Kræmer Experimental hauls Kvernsund Auxiliary vessel The chartered vessels involved in trial 2 are presented in Table 3. Anny Kræmer was equipped with Cotesi no. 3 and Euronet/Alfredo no. 3 trawls, both rigged with 19.2 m rockhopper footrope gear, 2200 Kg, 8 m 2 polyvalent trawl doors and 70 m sweeps. Cod-end, sorting grid, observation devices, acoustic releases and cover nets were the same as used in trial 1, except that in order to observe the cages after release, a remote operated vehicle (ROV) was used (Super C cat). Nine square cages, 5 m long, 2 m wide and 2 m high, constructed of 70 mm aluminium tubing, were used to collect fish,. The cage net was square mesh PE net with mesh size 50 mm (stretched length), and twine diameter 1.8 mm. The rear end of the cage (the closing net) consisted of 19.6 mm (stretched length) PA net. Three fish traps (1.8 m 1.8 m 2.15 m) made of steel frame with plastic coated steel net were set out and baited with mackerel. After observing with the ROV that sufficient fish had entered, a 1.8 mm PE twine which held a net hatch from the opening was pulled. The hatch then fell down and closed the entrance to the trap. A current meter that measured current speed in cm/s every 5 min was anchored close to the cages. More detailed drawings are found in Appendix I. Methods The survival experiments started on 14 th of August, five days after the three trawlers began fishing in the area. The nine cages were systematically rigged in three sets in the same way as in trial 1 (Table 2). 12

19 Materials and methods All hauls were performed north of Makkaursandfjord at depths of m. Before setting out the cages, a steel shackle weighing approximately one kilogram was tied to the sea anchor in order to prevent the tension rope from ending up in the ROV s propeller. The trawls were towed for approximately one hour before the rear ends of the cages were closed, then towed for five to ten minutes to sample fish before the cages were released close to the mouth of Makkaursandfjord. After the cages were released from the trawl, they sank to the bottom. The auxiliary vessel then fetched the buoys, slipped a ten kg weight down the rope to prevent the cages from lifting from the bottom and to ensure that the front opening had closed. The cages were towed slowly (= 1 knot) for 50 to 85 min to the site where they were anchored (Figure 7). The cages were then observed with the ROV to check the quantity of fish, their condition and cage closing. Since high mortality of haddock had been observed in the trawl-caught control cages in trial 1, a second control group of trap-caught fish was included in the experiment. Three fish traps were set out and baited with mackerel. After the traps were closed, they were left on the sea-bed for six to seven days before they were taken up for observations. Because of bad weather, the experiments had to be stopped for three days during the experimental period. The sea state and currents also made observations of the cages with ROV difficult after release. In some cases several days passed before it was possible to observe the quantity of fish taken and the condition of the cages. After one week in the sea, the cages were hauled on board the auxiliary vessel, dead and live fish were counted, and species, length and any injuries (Table 1) were recorded. One box of live haddock ( individuals) from each cage was selected for registration of injuries and measurement of length. A further sample of fish was also measured (1-2 boxes). In most cases, the contents of two boxes of fish were measured, and a third was added if the total number of fish in the cage was estimated to exceed Otoliths were taken from the 25 first live haddock from each cage that were recorded. As in trial 1, the surface of the fish was divided into 38 areas; each flank into 14 registration areas plus 10 fins (Figure 5). Bleeding (blood visible but epidermis not ruptured), wounds (epidermis ruptured) and scale loss were classified as small (<1 cm 2 ), medium (1-4 cm 2 ) and large (>4 cm 2 ). Clefts in fins, bleeding and bare fin-rays were registrated. After length measurement and injury 13

20 Materials and methods registration all the fish were counted. All fish from the traps were measured and any injuries noted. Data analysis A database containing all raw data was made using Microsoft Excel 2000, which was also used for mortality analysis. SAS statistical software (version 8) was used for analysis of injuries. The level of significance (α) in statistical tests was In trial one, three of the cages were insufficiently closed. A Mann-Whitney test was used to test for whether open cages affected the number of fish. This was done in order to test our suspicion that some fish escaped from partially open cages. For length-dependent analyses, fish were divided into 5 cm length intervals. To test for differences in the length distributions of live and dead fish and for the General Linear Models analysis (GLM) (see below), true length was used at 1 cm intervals. Total mortality was calculated as the percentages of each species found dead in each cage, and length-dependent mortality was calculated as the percentages of dead fish in each length interval. When length-dependent mortality from more than one cage was calculated, the mean value was weighted by the total number of fish in each length class. To test for differences in length distribution between live and dead fish, the Kolmogorov-Smirnov test for goodness of fit was used. When sample size was smaller than 20, a δ-corrected Kolmogorov-Smirnov test was used (Zar, 1996). The Wilcoxon rank test was used to test for differences in length-dependent mortality between cages. In addition to length treatment of data in the same way as in trial 1, survival probability by length was calculated with the solver in Microsoft Excel 2000 by means of Logistic-curve (Wileman et al., 1996) fitted to measured survival rates by minimizing the residual sum of squares, weighted with the number of measurements. Proportions of dead and living fish have a binomial sampling distribution and the 95% zα / 2 confidence interval was calculated for the observed proportions by pˆ ± where 2 n 14

21 Materials and methods pˆ is the observed proportion, z α/2 is the quantile of the standard normal distribution, and n is the total number of live and dead fish in the length group (Millar, 1995). The curves were found not to fit the data from trial 1. To test the number of skin injuries in each category, the General Linear Model (GLM) of the SAS software was used. Small, medium and large skin and fin injuries were tested against cage groups (mesh, grid, control), fish length (not grouped), flank areas (Figure 8, Figure 8 from snout to tail, 7 areas divided vertically), fin placement (dorsal, ventral and caudal) and between left and right sides. In graphic presentations, numbers of small, medium and large injuries have been merged. On figures showing average wounds divided by area and length group, mean values are chosen as representative. Binomial standard error (SE) of mean values of fin injuries were estimated as pˆ qˆ, where pˆ is the proportion of the special n characteristics, qˆ is 1- pˆ and n is the number of fish. Error bars show approximate 95.4% error limits, equivalent to ±2 SE (Bhattacharyya and Johnson, 1977). 15

22 Results Results Mortality and catch composition Trial 1 The total number of fish caught in trial 1 was 864. Of these, 685 were haddock, 94 cod and 83 saithe. No mortality was found among cod and saithe, but haddock mortality ranged from 1.6 to 50.4% (Table 4). Cages no. 2 (mesh), 7 (grid) and 8 (mesh) contained the smallest numbers of haddock; 45, 40 and 33 respectively (Table 4). These cages were badly closed and the numbers of fish in open and closed cages were significantly different (Mann-Whitney U-test, P<0.02). The badly closed cages were therefore excluded from the survival analysis. The remaining valid mesh cage had the highest mortality rate of all cages and rejection of the other cages leads to high degree of uncertainty due to the lack of sufficient comparative data. If the rejected cages were only open at the anchoring site, dead fish probably sank through the hole, leading to lower observed than actual mortality. If, on the other hand, the cages were open during trawling, the proportion of live and moribund fish that escaped through the hole cannot be estimated. Table 4 Haddock Cod Saithe Cage Days Number Mortality Number Number No. Type in sea Total Dead % Total Dead Total Dead Remarks 2 Mesh Excluded 5 Mesh Mesh Excluded 1 Grid Grid Grid Excluded 3 Control Control Control The sampling time (time from closing the cage to releasing it from the trawl) of cage no. 1 was 15 min, of cage no. 3, 8 min and cage no. 9, 5 min. Sampling time of the other cages ranged from five to ten minutes. Lack of information on accurate sampling times means that estimates of the effect of sampling time on mortality 16

23 Results cannot be made, but the three documented sampling times and mortalities indicate no such effects. Haddock The observed mortality of haddock disregarding cages 2, 7 and 8, and irrespective of length, was 50.4% of mesh-selected fish, 12.1% of grid-selected fish and 9.2% of the controls. Percentages are weighted by the total number of haddock in cages. 50 Dead Alive Mean length (cm) Mesh (5) Grid (1) Grid (6) Control (3) Control (4) Control (9) Cages Figure 9 The mean length of dead and live haddock in valid cages is shown in Figure 9 and the length distributions of dead and live haddock are shown in Figure 10. The length distributions in each cage differed significantly in all cases except cage 1 (grid) and 9 (Control). The mean length of haddock in control cages was lower than in grid (P<0.05) and mesh cages (P<0.01). 17

24 Results Number Grid - Cage no. 1 Alive n=62 Dead n=1 Number Grid - Cage no. 6 Alive n=68 Dead n= Length (cm) Length (cm) Number Mesh - Cage no. 5 Alive n=69 Dead n=70 Number Control - Cage no. 3 Alive n=171 Dead n= Length (cm) Control - Cage no Length (cm) Number Control - Cage no. 4 Alive n=71 Dead n=3 Number Alive n=122 Dead n= Length (cm) Length (cm) Figure 10 Mortality in all control cages was length dependent and mortality rates are merged in the figures (Figure 11). In cage 1 (grid), there was only one dead fish and mortality in that cage shows therefore significant difference from all others. In neither of the grid cages (cages 6 and 7) was haddock mortality length related, and these two cages were therefore also merged (Figure 11). Mortality (%) n=1 n=0 n=9 n=2 n=7 n=45 n=32 n=49 n=108 n=16 n=23 n=44 Mesh n=139 Grid n=148 Control n= Length class (cm) n=47 n=2 n=37 n=29 n=28 n=33 n=10 n=14 n=4 n=2 n=1 n=0 Figure 11 18

25 Results Mesh-selected fish suffered higher mortality than the other groups (Wilkoxon rank test, P<0.01). The mortality in the control groups was not significantly different from that of the grid-selected fish (P>0.1). Mortality rate was inversely proportional to fish length in the mesh group (r 2 =0.48, P<0.05) and the control group (r 2 =0.81, P<0.005). In the grid group, there was no correlation between length and mortality (r 2 =0.025, P>0.25). A tendency to a peak in mortality could be observed in mesh and grid groups in the length interval 40 to 54 cm. Cod The length of the cod ranged from 30 to 59 cm in all valid cage groups (n = 61) (Table 5). Mean length in all groups combined was 44.1 cm, (SD = 5.9). Saithe In all cage groups combined, length ranged from 32 to 50 cm (n = 54), with a mean length of 38.6 cm, (SD = 4.9). Table 5 Cod Saithe Mesh Grid Control Mesh Grid Control Cage 5 Cage 1 Cage 6 Cage 3 Cage 4 Cage 9 Cage 5 Cage 1 Cage 6 Cage 3 Cage 9 Mean n Min Max

26 Results Trial 2 A total of 8663 fish were caught in trial 2 (excluding fish from fish traps). Of these, 7442 were haddock, 999 cod, and 222 saithe. One cod and one saithe were found dead in control cages. There was no mortality of these species among mesh and grid escapees. The mortality of haddock in valid cages varied from 3.8 to 32.1%. Cages no. 1 (mesh), 4 (control), 5 (mesh) and 6 (grid) were inadequately closed. With the aid of a diver, cages no. 5 and 6 were closed before they were taken up, but when cage no. 5 came to the surface it appeared to be full of holes and fish slipped out of it. Cages no. 1, 4 and 5 were excluded from analysis since many fish were observed slipping out of it, and so was cage no. 7 (grid) because it did not contain sufficient fish. Cage no. 6 was the only remaining cage containing grid escapees, and although the diver closed it, it had been open for part of the sampling period, with an outer diameter of the opening estimated to be about 30 cm. The net lay partly under the cage at the anchoring site, preventing escapes. A total of 62 haddock, 25 saithe and 6 cod were caught in the fish traps. There was no mortality among the haddock and cod, but one saithe was found dead. The dead saithe had an approximately 2 cm diameter sore on its left flank, penetrating to the vertebrae. The sore was thought to be caused by a parasite. Table 6 Haddock Cod Saithe Cage Cage Sampling Mortality Total Dead Total Dead Total Dead Remarks no. group time (min) (%) 2 Mesh Mesh 10 Excluded 8 Mesh Grid 10 Excluded 6 Grid Grid Excluded 3 Control Control 10 Excluded 9 Control Mortality rates of haddock, disregarding cages 1, 4, 5 and 7, and irrespective of length, was 28.6% of mesh-selected fish, 3.8% of grid-selected fish and 25.2% of controls. Percentages are weighted by the total number of fish in the cages. Cod and saithe in the mesh and grid groups suffered no mortality. The mortality rates of cod 20

27 Results and saithe in the control group were 0.02 and 1.0% respectively. Sampling times are shown in Table 6. No relation was found between sampling time and mortality. Haddock Mean length (cm) Dead Alive Mesh (2) Mesh (8) Grid (6) Control (3) Control (9) Cages Figure 12 The mean length of dead haddock was significantly lower than that of live fish in all cages (Figure 12). Total average lengths did not differ between cage groups (P>0.05). The distribution of live and dead haddock in valid cages is shown in Figure 13. There were significant differences between the length distributions of live and dead haddock in all cages (Kolmogorov-Smirnov test for goodness of fit, P<0.01). The length of haddock in the fish traps ranged from 25 to 49 cm, with an average length of 33.1 cm (SD = 5.1). An overview of length distribution of alive and dead haddock is shown in Appendix II. 21

28 Results Number Mesh - Cage no. 2 Alive n=1154 Dead n=546 Number Mesh - Cage no. 8 Alive n=1824 Dead n= Length (cm) Length (cm) Number Control - Cage no. 3 Alive n=472 Dead n=129 Number Control - Cage no. 9 Alive n=1316 Dead n= Length (cm) Length (cm) Number Grid - Cage no. 6 Alive n=853 Dead n= Length (cm) Figure 13 Length distribution of two- and three-year-old survivors 14 2 years (n=28) 3 years (n=68) Number Length (Cm) Figure 14 Most fish smaller than 34 cm were less than three years old (Figure 14). There were four measurements of 4- and 5-year-old haddock, cm, not shown in the figure. 22

29 Results Length-dependent mortality Mortality (%) 100 % 80 % 60 % 40 % 20 % 0 % n=0 n=16 n=284 n=50 n=53 n=31 Mesh - Cage no. 2 n=43 n=13 n= Length (cm) Mortality (%) 100 % 80 % 60 % 40 % 20 % 0 % n=13 n=62 n=53 n=71 n=65 n=46 Mesh - Cage no. 8 n=22 n=6 n= Length (cm) Mortality (%) 100 % 80 % 60 % 40 % 20 % 0 % n=0 n=10 n=72 n=12 n=20 n=15 Control - Cage no. 3 n=26 n=8 n= Length (cm) Mortality (%) 100 % 80 % 60 % 40 % 20 % 0 % n=4 n=97 n=288 n=32 n=55 n=43 Control - Cage no. 9 n=67 n=22 n= Length (cm) Figure 15 Mortality (%) 100 % 80 % 60 % 40 % 20 % 0 % n=0 n=11 n=8 n=26 n=31 n=19 Grid - Cage no. 6 n=15 n=5 n= Length (cm) Length-related mortality was registered in all cages (Figure 15). For cages 2, 8, 3, 9 and 6, r 2 was 0.856, 0.784, 0.834, and respectively, P<0.005 in all cases. Table 7 α Experiment type Mesh Mesh Grid Control Cage no <0.05 <0.01 <0.05 < < <0.01 NS NS Grid 6 <0.01 < <0.05 <0.01 Control 3 <0.05 NS < NS 9 <0.05 NS <0.01 NS 1 Cage no. 2 (mesh) had significantly higher mortality than any other cage and cage no. 6 (grid) had significantly lower mortality than the other cages. There was no significant difference in size-related mortality between control groups nor between control groups and cage no. 8 (mesh). P-values of the Wilcoxon test are shown in Table 7, where mortalities in length groups (Figure 15) were tested pairwise. 23

30 Results Survival curves Survival is the opposite of mortality rate, and the survival probability of length classes in each cage group can be estimated by fitting logistic curves (Figure 16 and Figure 17). As in Figure 15, the length related mortality is shown. In addition, survival probability of a given length can be estimated. Survival probability 1,00 0,75 0,50 0,25 Control cage no. 3 Control cage no. 9 Survival probability 1,00 0,75 0,50 0,25 Mesh cage no. 2 Mesh cage no. 8 0,00 Figure Lengt class (cm) 0, Length class (cm) Best-fit logistic curves plotted against length in mesh and control cages gave a relatively good fit (Figure 16). The figure shows a drop in survival rate in the cm length class in the mesh group. A best-fit logistic curve plotted against length from the grid haul (cage 6) is shown in Figure 17. There is a relationship between length and mortality, but the data are limited. Survival probability 1,00 0,75 0,50 0,25 Grid cage no. 6 Table 8 Cage group L 25 L 50 L 75 S.R. Mesh Grid Control ,00 Figure Length class (cm) Table 8 shows 25, 50 and 75% survival probabilities (L 25, L 50 and L 75 ) from all sample groups as well as survival range (SR), reflecting the slope of the curve, defined as L 75 - L 25. The lowest values of L 25, L 50 and L 75 in the grid group indicate a better survival probability of smaller fish in grid cages than the mesh and control cages. The range of survival in the grid cage was also the lowest, and there was a difference of less than 4 cm between fish that had 25 and 75% possibility of survival. 24

31 Results Cod The length of the cod ranged from 18 to 63 cm in all cage groups (n = 360), (Table 9). Mean length was 30.7 cm, (SD = 7.2) and median length 40.5 cm. The mean length of the 6 cod caught in the fish traps was 34.7 cm, ranging from 26 to 52 cm (SD = 9.4). Saithe In all cage groups combined (n = 168), the length of the saithe ranged from 21 to 45 cm (Table 9), with a mean length of 31.5 cm and SD = 4.4 and median length 33.0 cm. In the fish traps, the length ranged from 18 to 38 cm, with a mean length of 26.9 cm, (n = 26, SD = 5.1). Table 9 Cod Saithe Mesh Grid Control Mesh Grid Control Cage 2 Cage 8 Cage 6 Cage 3 Cage 9 Cage 2 Cage 8 Cage 6 Cage 3 Cage 9 n Mean Min Max

32 Results Fish injuries Trial 1 Haddock Skin injuries were tested for differences between cage groups, for relationship to length, between right and left sides and between flank areas and fins (Table 10 to Table 12). Figures 18 to 21 show frequency of injuries, location and length relationship. The length distribution of fish on which injuries were registered is shown in Appendix II. Table 10 α Test variables Skin injuries Fin injuries Bleedings Wounds Infections Fin damage Bleedings Infections Cage group NS <0.05 NS <0.01 <0.01 <0.01 Length NS NS NS NS NS NS Cage*Length NS <0.05 NS NS NS NS All fin injuries and skin wounds differed between cage groups, among which meshselected fish showed the highest values and grid selected the lowest (Table 10 and Table 11). No relationship was found between length and injuries when cage groups were pooled. There were intercage differences in skin wounds, but no length relationship was found when individual cage groups were tested. The left and right sides of the haddock showed no differences when tested for bleeding, wounds and infections. Table 11 Cage group Mesh n=41 Grid n=103 Control n=123 Skin injuries Fin injuries Bleedings Wounds Infections Fin damage Bleedings Infections n % n % n %

33 Results Table 12 Injuries Skin area/fins Cage * Area Skin bleedings <0.01 NS Skin wounds <0.01 <0.05 Skin infections <0.05 NS Fin damages NS <0.05 Fin bleedings <0.01 NS Fin infections <0.01 <0.01 cage groups (see pages 29 and 31). Skin injuries were not homogeneously distributed between all flank areas and fins (Figure 8), except for fin damage (Table 12). There were different patterns in the location of skin wounds, fin damage and fin infections between 27

34 Results Skin injuries There was no difference in number of the most common injury, skin bleeding, between cages (Table 10, Figure 18). Approximately 40% of the haddock suffered skin bleeding. A 100 Mesh Grid Control Frequency (%) Number of bleedings B 100 Frequency (%) Number of wounds C 100 Frequency (%) Number of infections Figure 18 Mesh-selected haddock had the highest mean number of wounds ((P<0.05), Figure 18B and Figure 19B). 14.6% of mesh-selected haddock had one or more wounds, compared to 5.8% of grid-selected haddock and 7.3% of the haddock from control cages. No fish from mesh cages had any skin infections, but 4.9% of the grid cage fish and 4.1% from the control cages were infected. Differences in skin infection between cages were not significant (Table 10). 28

35 Results A Mean number of bleedings B Mean number of wounds C Mean number of infections 1,0 0,8 0,6 0,4 0,2 0,0 1,0 0,8 0,6 0,4 0,2 0,0 1,0 0,8 0,6 0,4 0,2 0,0 Mesh Grid Control Flank area Flank area Flank area Figure 19 The number of bleeding sites increased from snout to tail flank areas in all cage groups, with the highest mean number in flank areas 5 and 6. Bleeding did not differ between groups (Table 12, Figure 19A). The average number of wounds in all cage groups increased by flank area from snout to caudal fin with a peak in flank area 6. A GLM test showed difference in wounds by flank area of the fish (Table 12) and cage group (Table 10). Skin infection differed significantly by flank area (Table 12). Infections in grid-selected haddock were found only in flank areas 6 and 7 (five fish infected), but in haddock from control cages, the number of infections rose from flank area 3 to flank area 6 (five fish infected). Despite having the highest number of wounds, haddock from the mesh cage had no infections. 29

36 Results Fin injuries A B C Frequency (%) Frequency (%) Frequency (%) Mesh Grid Control Number of fin bleedings Number of fin infections Number of fin damages Figure 20 All mesh-selected fish suffered more than two fin damages. Grid-selected haddock had the lowest frequency (65%) of fin damage, and 94.9% of the control haddock had one or more fin damages. The difference between cage groups was significant (Table 10). Mesh-selected haddock displayed the highest frequency of fin bleeding, and gridselected the lowest (Figure 20A). 63% of mesh-selected haddock had more than one fin bleeding, while 20% of control and 16% of grid-selected haddock had more than one fin bleeding. There was a significant difference between cage groups (Table 10). Grid-selected haddock had the lowest rate of fin infections. 78.6% of grid-selected haddock had no infections at all, while 51.2% of mesh-selected and 53.7% of the control haddock had no infections. There was a significant difference between cage groups (Table 10). 30

37 Results A Mean number of fin bleedings B Mean number of fin infections C Mean number of fin damages 1,0 0,8 0,6 0,4 0,2 0,0 1,0 0,8 0,6 0,4 0,2 0,0 1,0 0,8 0,6 0,4 0,2 0,0 Mesh Grid Control Dorsal fin 1 Dorsal fin 2 Dorsal fin 3 Ventral fin 1 Ventral fin 2 Caudal fin Fins Dorsal fin 1 Dorsal fin 2 Dorsal fin 3 Ventral fin 1 Ventral fin 2 Caudal fin Fins Dorsal fin 1 Dorsal fin 2 Dorsal fin 3 Ventral fin 1 Ventral fin 2 Caudal fin Fins Figure 21 Mean numbers of fin injuries are shown in Figure 21. The three dorsal fins are numbered in ascending order from the foremost fin to the rearmost. The two ventral fins are numbered in the same order. The dorsal and caudal fins in all cage categories were the most liable to suffer fin damage, while ventral fins had the lowest frequency in all cases. The frequency of damage in all fins but the rear ventral fin was highest in mesh-selected haddock, and lowest in all fins in grid selected fish (Figure 21C). The caudal fin was most exposed to fin bleedings in all categories, with mesh-selected fish the most exposed and grid-selected the least. Approximately half of the meshselected (47%) and control (46%) haddock had infections in the caudal fin, which was the most exposed, compared to only 21.4% of grid selected haddock. 24% of the control haddock had infections in the foremost dorsal fin and 10% in the second dorsal fin, while these frequencies were below 4% for the other groups. The differences in fin infections between fins and cage groups were significant (Table 12). 31

38 Results Trial 2 Haddock Test results are shown in Table 13 to 15. Figures are presented in Figure 22 to 27. The length distribution of fish in which injuries were recorded is shown in Appendix II. Table 13 α Skin injuries Fin injuries Test variables Bleeding Wounds Scaleloss Bleeding Wounds Fin split Fin rot S M L S M L S M L S M L S M L Cage group NS NS NS NS NS <0.01 NS NS NS <0.05 NS NS <0.05 NS NS NS <0.05 Length NS NS NS <0.05 NS <0.01 NS <0.01 NS NS NS NS NS NS <0.01 <0.05 NS Cage*Length NS NS NS <0.01 NS <0.01 NS NS NS NS NS NS NS NS NS NS NS Inter-cage variations were found for skin wounds, fin bleedings, fin wounds and fin rot, where grid-selected fish had the lowest rates of injuries in all cases (Table 13). Skin bleedings, skin wounds, scale loss, large fin wounds and fin split were related to the length of the fish for some of the sizes of injuries. Skin wounds showed significant effect of length interaction by cage group. The wounds of control haddock were not related to size. The prevalence of injuries is presented in Table 14. Table 14 Cage group Mesh n=185 Grid n=94 Control n=161 Skin injuries Fin injuries Bleedings Wounds Scale loss Bleedings Wounds S M L S M L S M L S M L S M L Fin split Fin rot n % n % n % No difference in injuries were found between the left and right flanks of the fish, Table 15 Injuries Skin area/fins Cage*Area Skin bleedings <0.01 NS Skin wounds NS NS Scale loss <0.01 NS Fin bleedings <0.01 NS Fin wounds <0.01 <0.01 Fin split <0.01 NS Fin rot <0.01 NS neither when all fish were compared (n=440), nor cage interaction. All skin and fin injuries except skin wounds were non-homogeneously distributed between flank areas and fins (Table 15). Fin wounds showed interaction between cage groups and fin location (Figure 26). 32

39 Results Skin injuries A Frequency (%) Mesh Grid Control Number of bleedings B Frequency (%) Number of wounds C Frequency (%) Number of scale losses Figure 22 Frequencies of skin injuries are shown on Figure 22. No significant relationship between cage groups was found in skin bleedings (Table 7) but grid-selected haddock had the lowest proportion of individuals with one or more bleedings, and meshselected the highest. The wounds, which were mostly small (<1 cm 2 ), did not differ between cage groups. Less than 5% of the haddock in grid cages had skin wounds, compared to 14% of mesh-selected fish and 12% of the controls. Scale losses were quite common; 86.5 to 90.4% of haddock in the various groups had lost scales. No difference was found between cage groups. 33

40 Results A Mean number of bleedings B Mean number of wounds C Mean number of scale losses 1,6 1,4 1,2 1,0 0,8 0,6 0,4 0,2 0,0 1,6 1,4 1,2 1,0 0,8 0,6 0,4 0,2 0,0 1,6 1,4 1,2 1,0 0,8 0,6 0,4 0,2 0,0 Mesh Grid Control Area 1 Area 2 Area 3 Area 4 Area 5 Area 6 Area 7 Flank area Area 1 Area 2 Area 3 Area 4 Area 5 Area 6 Area 7 Flank area Area 1 Area 2 Area 3 Area 4 Area 5 Area 6 Area 7 Flank area Figure 23 The mean numbers of skin injuries in individual flank areas are shown in Figure 23. The rate of bleedings by flank area differed (Table 15). The highest rate of bleedings was in flank area 6 in all cage groups; control haddock had the highest rate and gridselected fish the lowest. The number of wounds did not differ between flank area and there was no interaction between flank areas and cage groups, i.e. the distribution of wounds in terms of flank areas did not differ between cage groups. Since the haddock s head has no scales, and it is the largest part of flank area 1, a low number of scale losses from that flank area is to be expected. In all three cage groups, flank areas 4, 5 and 6 had most scale losses. Difference between flank areas were significant but there were no differences between cage groups (Table 15). 34

41 Results A B C Mean number of bleedings Mean number of wounds Mean number of scale losses Mesh Grid Control Length class Length class Length class Figure 24 Mean numbers of fin injuries in each length class of the cage groups are shown in Figure 24. No difference in bleedings between length classes was found. Nor was there any length-related difference between cage groups (Table 13). The number of wounds differed significantly between length classes. High values in the cm length class (grid) and cm length class (mesh) caused this length-class effect. There was also a significant interaction between length and cage groups. Mesh and grid groups showed a length relationship, while the controls did not (Table 13). However, only four haddock from the control group had wounds. Numbers of scale losses fell significantly with increasing length. No differences were found in scale losses between mesh, grid and control cages. 35

42 Results Fin injuries A Frequency (%) B Frequency (%) C Frequency (%) D Frequency (%) Mesh Grid Control Number of fin bleedings Number of fin wounds Number of fin split 0 Figure Number of fin rot All fin injuries except fin split differed among the cage groups, and grid-selected haddock had the lowest rate in all cases (Figure 25, see Table 14 for segmentation). Fin bleedings were mostly small; 90% of mesh-selected fish, 89% of the controls and 81% of the grid-selected haddock had small fin bleedings. The total frequency of occurrence of one or more wounds was 25.9% in the mesh-selected fish, 32.9% in the controls and 16% in the grid-selected fish. Fin wounds on mesh selected and control haddock were mostly small, but they occurred at an equal rate of small and medium- 36

43 Results sized wounds on grid selected fish. 42 % of mesh-selected haddock were suffering from fin rot, compared to 43% of the controls and 33% of the grid-selected fish. A Mean number of fin bleedings B Mean number of fin wounds C Mean number of fin splits D Mean number of fin rot 1,0 0,8 0,6 0,4 0,2 0,0 1,0 0,8 0,6 0,4 0,2 0,0 1,0 0,8 0,6 0,4 0,2 0,0 1 0,8 0,6 0,4 0,2 0 Mesh Grid Control Dorsal fin 1 Dorsal fin 2 Dorsal fin 3 Ventral fin 1 Ventral fin 2 Caudal fin Fins Dorsal fin 1 Dorsal fin 2 Dorsal fin 3 Ventral fin 1 Ventral fin 2 Caudal fin Fins Dorsal fin 1 Dorsal fin 2 Dorsal fin 3 Ventral fin 1 Ventral fin 2 Caudal fin Fins Dorsal fin 1 Dorsal fin 2 Dorsal fin 3 Ventral fin 1 Ventral fin 2 Caudal fin Fins Figure 26 Mean numbers of fin injuries are shown in Figure 26. Cases of zero fin injuries were homogeneously distributed over fins. Only fin wounds showed a different pattern of distribution between cages. Grid-selected fish had the fewest wounds, and then only on the caudal fin. No wounds were found on ventral fins. The highest frequencies of fin bleedings were on the foremost dorsal fin and caudal fin in all cage groups, and mesh-selected haddock had the highest frequency of bleedings in all fins. The 37

44 Results haddock s caudal fin was most liable to suffer splitting, where 91.5% of grid-selected fish, 82.6% of the controls and 76.2% of the mesh-selected haddock showed clefts. The second most exposed fin was the rear dorsal fin. The frequency of fin rot was highest in the caudal fin. Less fin rot was found in grid-selected haddock. A Mean number of fin bleedings B Mean number of fin wounds C Mean number of fin split D Mean number of fin rot Mesh Grid Control Length class Length class Length class Length class Figure 27 Mean numbers of fin injuries in each length class of the cage groups are shown in Figure 27. Large fin wounds and the frequency of fin splits were length-related. The combined sizes of fin wounds were also length-related. The numbers of wounds show relatively high values in the cm (one fish) and cm (one out of three fish) 38

45 Results length classes. A significant, decreasing trend in the number of fin splits occurred with increasing fish length (Table 13). The average number of fin splits in the cm length class was 3.5 (all cage groups), falling to 2.6 in the cm length class. Cod Numbers of skin and fin bleedings differed among cage groups. The lowest values were found in the grid cage (Table 16 and Table 17). No length dependence was found in skin bleeding independent of cage groups, but the number of bleedings decreased with in line with fish length in both grid and control groups (P<0.01). Skin wounds were length-related, but only two fish from the grid group (35 and 36 cm) and one from the mesh group (42 cm) displayed wounds. A length-related relationship was found in small and total scale losses in all cage groups taken together. The only group that did not show a length relationship in scale losses was the mesh group (P>0.1). Small fin bleedings differed among cage groups. 67% of the control cod suffered small fin bleeding, compared to 65% of mesh-selected and 32% of the gridselected fish. Medium-sized fin bleedings were length related. Four cod from the control group in the size interval cm had medium-sized fin bleedings. No fin wounds were found in the cod. The frequency of fin splits was inversely related to fish length when tested independent of cage. The length-dependent relationship of fin split differed between cage groups. Control and mesh-selected fish showed a lengthdependent relationship (P<0.01) while grid-selected fish did not (P>0.5). Fin rot differed neither between cage groups nor length groups. Table 16 α - Skin injuries Fin injuries Test Bleedings Wounds Scale loss Bleedings Wounds Fin split Fin rot variables S M L S M L S M L S M L S M L Cage group <0.01 NS - NS - - NS NS - <0.05 NS NS NS Length NS NS - < <0.01 NS - NS < <0.01 NS Cage*Length <0.01 NS - < <0.05 < NS NS <0.05 NS Table 17 Skin injuries Fin injuries Cage group Bleedings Wounds Scale loss Bleedings Wounds Fin split Fin rot S M L S M L S M L S M L S M L Mesh n=71 Grid n=31 Control n=104 n % n % n %

46 Results The distribution of skin and fin bleedings and skin wounds differed between cage groups. Most skin bleedings in all cage groups were found in flank areas 1 and 2 (Table 19). Two fish from the grid group and one from the mesh group all had skin Table 18 α Injuries Area Cage*Area Skin bleedings <0.01 <0.01 Skin wounds <0.01 <0.05 Scale loss <0.01 NS Fin bleedings NS <0.05 Fin wounds - - Fin split <0.01 NS Fin rot NS NS wounds on the head. The most frequent sites of fin bleeding in mesh-selected and control-group cod were on the foremost dorsal fin, but on the foremost ventral fin in grid-selected fish (10 fish). No fin wounds were found in the cod. The distribution of fin bleedings and fin rot did not differ among fins when this was tested with all cage groups pooled (Table 20). The distributions of scale loss, fin split and fin rot over the flank area and fins did not differ among cage groups. Fin splits differed among fin and were most frequently found on the caudal fin and the rearmost dorsal and ventral fins. There was no trend in the distribution of fin rot by fins or cage groups. Table 19 Flank area Injuries Cage group Σ Mesh Bleedings Grid Control Wounds Scale loss Mesh Grid Control Mesh Grid Control Table 20 Injuries Bleedings Fin split Fin rot Fins Cage group Dorsal fin 1 Dorsal fin 2 Dorsal fin 3 Ventral fin 1 Ventral fin 2 Caudal fin Σ Mesh Grid Control Mesh Grid Control Mesh Grid Control

47 Results Saithe Injuries were registered on a total of 17 saithe from mesh cages and 20 from grid cages, but on none from the control cages Only fin bleedings showed differences between cage categories (α = 0.05). Twelve saithe (71%) from the mesh group had small fin bleedings, while eight (40%) from the grid cages had this type of injury. Injuries were not length-related. Table 21 Cage group Mesh n=17 Grid n=20 Skin injuries Fin injuries Bleedings Wounds Scale loss Bleedings Wounds Fin split Fin rot S M L S M L S M L S M L S M L n % n % Skin bleedings, scale losses and fin split differed significantly between flank area and fin location. The four mesh-selected saithe showed bleedings in areas 1 and 2, and the only saithe from the grid group had skin bleeding in area 6. Flank areas 3-5 showed the highest frequencies of scale losses in both categories, with mesh-selected fish having 83% and grid-selected fish 74% of scale losses in these areas. No fin splits were found in the caudal fins of mesh-selected saithe and only in one fish out of 20 in saithe from the grid group. These data are limited and definite conclusions cannot be drawn. Fish traps Thirty haddock from two fish traps were examined in order to quantify injuries. Fish length ranged from 26 cm to 49 cm (mean length 33 cm). Fewer scale losses, fin bleedings and fin wounds were found in haddock taken in traps than in trawl hauls (pooled data, P<0.01). No length relationship was found among registered injuries. Mean numbers of injuries per fish were: Bleedings 1.97 Skin wounds 0.3 Scale loss 2.03 Fin bleedings 1.6 Fin split 2.86 Fin rot 0.7 No fin wounds were observed and injuries were predominantly small. Skin bleedings and skin wounds were predominantly on the snout (flank area 1). Scale losses did not 41

48 Results differ among areas (GLM test, P=0.67). Fin splits were most common on the tail, while fin bleeding and fin rot showed less variation between fins. Overview of results The mean length of haddock in the control cages in trial 1 was lower than in mesh and grid cages in the same trial. Only in the control cages were the dead fish significantly smaller than live fish. Total average lengths did not differ among cage groups in trial 2, but the dead haddock were significantly smaller than living fish in all the cages. The haddock in trial 2 were mostly two or three years old. The age group barrier lies at approximately 34 cm. In trial 1, irrespective of length, mortality was highest in the mesh cage, but no difference between grid and control cages was observed. In trial 2, the highest mortality found was in a mesh cage and the lowest in the grid cage. No difference was found between mortality in the mesh cage and the control cages. Mortality was inversely related to fish length in all cage groups except the grid cages in trial 1. From survival curves fitting data from trial 2, it was observed that the smallest individuals of the three-year-old cohort had more than a 75% chance of survival. All haddock and cod caught in fish traps in trial 2 survived. One saithe out of 25 was dead, and the cause of death is believed to be a parasite. In trial 1, the rates of skin wounds, fin damage, fin bleedings and fin infections differed between cage groups. Mesh-selected fish had the highest values and gridselected fish the lowest. In general, injuries were not homogeneously distributed over the flank areas and fins, while skin wounds, fin damage and fin infections showed different pattern of distribution among cage groups. Most flank injuries were found on the rear part of the fish, with the highest values recorded in areas 5 and 6 (counting from 1 to 7 from snout to tail). The caudal fin tended to have the highest frequency of fin injuries in all cage groups. Mesh-selected haddock also had a high frequency of bleeding in the foremost dorsal fin. All cage groups suffered a relatively high frequency of fin damage in their dorsal fins compared to ventral fins. In trial 2, large skin wounds, small fin bleedings, small fin wounds and fin rot differed among cage groups. Grid-selected haddock had the lowest frequency in all cases while the injuries in mesh selected and control haddock were similar. Small and large skin wounds, medium scale loss, large fin wounds and fin split were less frequent as 42

49 Results fish length increased when pooled data were tested. Skin wounds of control haddock were not size-related. Except for skin wounds, fish injuries were not homogeneously distributed over the flank areas and fins. Only fin wounds showed different distribution patterns among the cage groups. The highest rates of skin bleedings and scale loss were found in the rear parts of the fish (areas 4-6). The foremost dorsal fin and the caudal fin showed the highest rates of fin bleedings in all cage groups. Fin wounds were almost entirely found on the caudal fin. Fin split was most frequent on the caudal fin, but the rearmost dorsal fin also had a high rate of fin splits. Ventral fins had the lowest rate of fin rot and the caudal fin the highest. The haddock from the fish traps suffered less scale loss, fin bleedings and fin wounds than haddock from pooled trawl hauls. Skin bleedings and wounds were mostly on the snout, and scale loss did not differ among flank areas. No mortality of cod and saithe was observed in the mesh and grid cages, while one cod and one saithe were found dead in the control cages in trial 2, giving mortality rates of 0-1%. Differences in size distribution between cages showed no trends. In cod, only small skin and fin bleedings differed between cage groups. The lowest values were found in the grid cage. Small skin wounds, small scale loss, medium fin bleedings and fin split were length-related irrespective of cage. Skin wounds and fin bleedings were uncommon (three and four fish injured, respectively). The mesh group did not show length dependency in scale loss, and fin split was not lengthdependent in the grid cage. Skin bleedings and wounds were found on the foremost flank areas of the cod, but the highest scale losses were towards the tail (areas 5 and 6). The highest rates of fin split were found in the caudal fins and rearmost dorsal and ventral fins in all groups. Of saithe in the mesh and grid groups, only fin bleedings showed intercage differences, with lower values in the grid cage. Scale losses, which were the most frequent injuries, were mostly small and predominantly in flank areas % of the saithe suffered fin splits, but only one fish out of 37 had a fin split in the caudal fin. No saithe from the control group were examined to quantify injuries. 43

50 Discussion Discussion Experimental design Experimental area The experimental catch area was chosen on the basis of species availability since the preferred catch was a combination of haddock, cod and saithe. Prefered anchoring depth of cages in trial 1 for observation was 50 m. Since rapid pressure reduction of about 70% results in ruptured swim bladders in cod, haddock and saithe (Tytler and Blaxter, 1973), towing depth was kept below 100 m in trial 1. In trial 2, the fish cages were anchored in the mouth of Makkaursandfjord. The cages were released from the trawler in the fjord mouth. Since towing speed was knots and towing time approximately one hour, the start position of the trawl hauls was limited to a radius of four nautical miles from the fjord mouth. These considerations limited the choice of study area. Experiment procedure Cage number 6, the only valid grid cage in trial 2, was partly open after anchoring and was therefore closed by a diver before being taken up for observation. The inner diameter of the opening was approximately cm. If the cage was open while towing, fish may have leaked out, affecting the observed survival rate. The control cages in trial 1 were closed before the trawl was shot and towed for approximately 10 minutes. This may have caused a bias in the size distribution of fish. Smaller fish tend to swim more slowly than larger individuals of the same species (Wardle, 1977; Isaksen and Valdemarsen, 1994). There is a marked change in tactics, when maintaining station with the gear gives way to turning or dropping back (Wardle, 1993). The smallest individuals tire first and will be the first to enter the codend. If sampling only takes place during the first few minutes of towing, the proportion of small fish may be higher than if sampled after one hour of towing. In trial 1, a difference in length distribution was observed between haddock in control cages and the other groups. However, samples from most length classes were 44

51 Discussion collected in the control cages. However, there could be effects of premature closing when sampling in areas where the range of sizes of fish is great. After the cages had been released from the trawl in trial 2, they sank to the bottom before being towed at barely 1 knot (0.5 m/s) over the sea bed, disregarding the current speed in the area (0 0.3 m/s, see Appendix II). Even though towing speed was lower than the presumed maximum sustainable swimming speed of the fish, it is not reasonable to assume that the fish, after having been exhausted during trawling, would have been able to remain swimming for an hour. The cages could not be monitored during towing, and they might have risen vertically from the bottom, causing stress and buoyancy disruptions. Furthermore, towing the cages might have raised bottom sediments (sandclouds), causing stress and gill irritations. In trial 1, the mesh cages were released and anchored at depths of less than 100 m, except for cage 5, the only valid mesh cage which had 50% haddock mortality. This cage was released at a depth of between 100 and 130 m and then floated up to 50 m depth and anchored. Assuming a similar ability in haddock and cod to resorb gas from the swimbladder, this ascent is close to the tolerance limits of the fish and may have affected the mortality rate. Cod is estimated to need three hours to adapt to a 50% reduction in depth (Harden Jones and Scholes, 1985), and a rapid 70% depth reduction results in ruptured swim bladders (Tytler and Blaxter, 1973). The results obtained by Tytler and Blaxter indicate in fact that haddock have a lower tolerance to pressure reduction, and this may explain the high haddock mortality in this cage. No mortality of cod or saithe was observed. High fish density in the cages might affect mortality through increased stress, transmittion of infections and oxygen insufficiency. The number of fish in valid cages in trial 1 ranged from 76 to 217 and in trial 2 from 972 to 2623, but no correlation between mortality and fish density was found. The registrations of injuries allowed comparisons of the injuries of trawl and mesh escapees and the control group to be made. The differences in injury level among cage groups may explain differences in mortality rates. Injuries of dead fish were not examined, but have been shown to be of the same kind, but occurring more frequently than in live fish (Breen & Sangster, 1997). 45

52 Discussion Due to high turbidity in the sea (visibility approximately 1 m), no observations of behaviour in the cages or of the functionality of the cage mechanism could be made in trial 2. Nor could the number of fish in the cages or technical failures such as ripped net, net cage twisting, shackle tangled in sea anchor, premature closing or opening of cages be observed. Cage construction The cages in trials 1 and 2 were constructed of knotted net. The knots lead to rough texture of the net. During towing, exhausted fish were observed resting against the small-meshed closing net at the rear end of the cages. There is reason to believe that this caused external injuries such as scale loss, bleedings and wounds. The injuries could have been reduced by use of knotless netting. Underwater observations did not show fish touching the 50 mm square netting of the cage sides, either during trawling or when the cages were anchored. When the cages were brought aboard the auxiliary vessel at the end of the experiments, fish were left lying on the netting for some minutes. In trial 1, the cages were in a horizontal position when they were brought out of the sea and all fish had collected in the rear end of the cage (the small-mesh closing net). In trial 2, the cages were raised horizontally and fish rested on the knotted square-meshed bottom, sliding around on it as the cage tilted. The effect of this treatment on scale loss and other skin and fin injuries was not recorded, but there is reason to believe that it added significantly to the observed level of injuries in all experimental groups. The towing resistance of the cover and cages may have changed the opening of the rearmost trawl panels and the opening of the cover net over the Sort-X grid. During trawling, the control and mesh cages hung behind the trawl, while the grid cages floated above the cod-end. The grid cover was kept open by a float of only kg buoyancy, and a strong pull caused by the attached cage may well have affected cover functionality and grid angle, and closed the passage from grid to cage. In trial 2, the captain of M/TR Anny Kræmer noted that the pull on the trawl wires was significantly less when the cages had been released. At a speed of 4 knots (2.1 m/s), the calculated resistance (Fridman, 1986) of the aluminium frame and floats would be approximately 850 kg. Some changes due to water turbulence may occur, and 46

53 Discussion resistance because of netting construction, i.e. cover nets and cage nets, comes in addition. The mesh and control cages were probably closer to the bottom than the grid cage during trawling. The risk of the cages being covered by sand clouds from the trawl gear is therefore higher in these cages and thus also the risk of stress and gill irritation. This could have affected both injury rates and mortality. Breen et al. (1998) found that exposure of fish in a cover, when they are forced to swim at speeds above sustainable speed limits, after swimming to exhaustion during trawling, induces mortality in fish survival experiments. Since fish in all cage groups in trial 1 were observed resting against the closing net during towing, it is assumed that they were exhausted and could not keep up with the water flow. When fish are exhausted even before entering the cage, the waterflow is even more crucial. A small mesh size was needed both to retain the fish and to reduce waterflow in the cages so that fish could keep swimming. Maximum sustainable swimming speed is related to fish length and temperature (He, 1993; He & Wardle, 1988). A 25 to 35 cm gadoid can sustain a maximum 0.9 m/s swimming speed at 9 C, the temperature observed at the anchoring sites in trial 2. Sea temperature was not recorded in trial 1. Engås et al. (1990) found that most haddock and cod could maintain a swimming speed of 1.35 m/s (35 to 50 cm fish) for 15 min. As water speed rose, more and more fish gave up swimming. The water flow rate in the cages was not measured, but it has been shown that the flow through a trawl is 30% lower than the trawling speed (Thiele et al., 1997). As mentioned by Breen et al. (1998), future trials should attempt to minimise water flow within the cover. However, the towing speed of the trawls in survival experiments must be similar to that used during standard fishing operations in order to obtain results comparable with commercial conditions. It can be assumed that the mesh sizes in the cages, and thus the water velocity inside the cages in our experiments, could have been substantially reduced. When investigating mortality of cod and herring, Surronen et al. (1995; 1996) used cages with 6 16 mm mesh sizes (stretched full mesh length) while towing at speeds of knots. Decreased mesh size increases towing resistance (Hansen & Wileman, 1984), increasing the pull on the trawl and thereby the likelihood of changing the opening of the trawl funnel. All such effects need to be taken into account. Engås et al. (1999) observed that by 47

54 Discussion inserting a cone of webbing in front of a cod-end, waterflow may be altered. Fish that could not sustain a swimming speed equal to towing speed soon dropped back in the trawl body, but responded to the turbulent flow within the webbing cone by taking up position in the low flow area behind the cone. It may therefore not be necessary to construct the whole cage with tight material to reduce waterflow. A panel or cone inside the cage that provides shelter from the waterflow might be worth considering. To close the rear end of the cages, sea anchors, pulling constraining ropes through climb locks were used. Closing frequently failed in both trials because the sea anchors lacked sufficient pull, unsuitable rope material or rope thickness or failure of the locks themselves. The closure system will have to be improved in future experiments. Trawls and codend In trial 2, two trawls were used (Cotesi no. 3 and Alfredo no. 3, see Appendix I). The trawls were of similar size, headline and fishing line length. The same sweep lengths and footrope gears were also used. However, the cutting rate of the bellies was different, and considering that exhausted fish that do not manage to swim with the trawl have been observed slinging from side to side between the net walls (Marteinsson, 1991), the cutting rate may affect the rate of collision and therefore of injuries. Since haddock mortality was size-related, all variables that affect size selection will affect haddock mortality. The single cod- end used in these experiments was constructed of Magnet, a tightly-braided PE twine. It is known that the circumference of cod ends may affect selectivity (Broadhurst & Kennelly, 1996; Reeves et al., 1992). The width of the codend was 31 meshes, and it is expected to have the same selectivity properties as the standard double codend used in Norwegian waters (Valdemarsen, personal communication). Magnet has recently been adopted by Norwegian trawlers. In a selection experiment carried out in 2001 and 2002, it was noted that cod-ends of Magnet had similar selection range (SR) for gadoids as the Sort-X sorting grid and therefore a lower SR than traditional PA codends used in Norwegian waters (Isaksen, personal communication). 48

55 Discussion Cod and saithe mortality and injuries No mortality of cod and saithe was recorded in these experimentsin either mesh or grid escapees. These results are in agreement with previous experiments (Vinogradov, 1960; Engås et al., 1989; Engås et al., 1990; Soldal et al., 1991; Main and Sangster, 1991; Jacobsen et al., 1992; DeAlteris and Reifsteck, 1993; Jacobsen, 1994; Suuronen et al., 1995), and show the high tolerance of these species. The only dead cod and saithe were found in control cages in trial 2. The most likely explanation is that they had been left in the trawl from previous hauls and ended up in the cages during towing of the control cages. The injuries sustained by cod and saithe tended to be less than in haddock, and in all cases, the fewest injuries among the cage groups were seen in the grid group. Many saithe suffered small scale losses, but this did not lead to mortality. Haddock mortality and injuries Differences between mesh, grid and control cages In both trials, mortality in mesh selected haddock was higher than in grid-selected fish. Overall mortality, disregarding fish length and cage groups, ranged from 1.6 to 50.4%. In trial 1, the mesh-selected fish had the significantly highest mortality rate, with no difference between grid and control cages. In trial 2, there was no statistical difference between mortality in mesh and control cages, but the mortality in the grid cage was significantly lower. The lowest and highest mortalities in our experiments were for grid (1.6%) and mesh (50.4%) escapees in trial 1. Similarly, the lowest and highest mortalities in trial 2 were 3.8% in grid and 32.1% in mesh cages. However, the mortality in the mesh cage in trial 1 may have been caused by rapid pressure reduction (see page 45, discussion of Experiment procedure), and the survival rate in the grid cage in trial 2 may have been affected by the opening in the cage. It was evident that injuries were significantly less frequent in grid cages than in control and mesh cages. These results are in agreement with previous experiments (Marteinsson, 1991). One likely explanation is that the grid escapees did not have to pass through the narrow extension between the trawl belly and the cod end, since the grid was mounted in front of the extension. Another possible explanation lies in the 49

56 Discussion fact that the sampling cage for the grid escapees was floating above the cod end during towing, attached to the cover net above the sorting grid, while the mesh and control cages were towed behind the trawl, with resulting differences in water flow within the cages. The cages attached to the cod-end cover were also closer to the bottom than the grid cages, and possibly covered by sand and mud clouds from the trawl gear. These clouds may have caused stress and gill irritations. Most fin injuries were found on the caudal and dorsal fins. In both trials, the injuries were significantly lowest in grid-escaped haddock; only fin split showed no inter-cage variation. In trial 2, rates of fin bleedings and fin wounds were significantly lower in trap-caught controls than trawl-caught haddock, which suggests that the trawling process and/or the sampling technique are likely to be the main causes. Fin injuries were usually small and not length-related, and were thus unlikely as an explanation of the mortality. If we assume that a certain proportion of the fish in our experiments sampled in all experiment groups and traps had already passed through a sorting grid of the commercial trawlers that had been engaged to fish in the experimental area, the grid selection caused no mortality since all fish caught in the traps survived. If this is true, the mortality and injuries originated mainly from the research procedure. If, on the other hand, the grid escapes led to rapid death, the trapped control fish would scarcely have been in contact with the grid. These considerations remains can only be speculative and are brought up in order to highlight the importance of acquiring knowledge of the effects of repeated selection. There was no mortality of trapped control fish, except for a single dead saithe whose cause of death was probably unrelated to capture or cage captivity. Although the trapcaught haddock had some small injuries, some of them were significantly less serious than those of the trawl-caught groups. These injuries (scale loss, fin bleeding and fin wounds) should therefore be focused on when explaining mortality. Furthermore, since mortality of all trawl-caught haddock in trial 2 was size-related, the injuries most likely to affect mortality are also expected to be size-related. Of the injuries that differed from those of trap- and trawl-caught haddock, only skin injuries were lengthrelated and therefore the most likely physical explanatory factor of mortality. It has nevertheless to be borne in mind that only live haddock, and not dead fish, were examined for injuries. 50

57 Discussion The haddock mortality in trawl-caught control cages was higher than in earlier experiments in which controls were sampled by barbless hooks (Main and Sangster, 1990; Sangster, 1992; Sangster and Lehmann, 1994; Sangster et al., 1996). It had been expected that the fish in the control groups would suffer the lowest mortality and injury rates since they avoided the mesh and grid escape process. The control groups were not real controls, since the fish went through the fishing process, except for escaping from the mesh and grid. Nevertheless, they were important, since the comparison to the other groups showed that escaping from the mesh and grid is not the main cause of mortality. Length-dependent mortality and injuries Haddock mortality was size-related, and was significantly inversely related to fish length except in the grid cages in trial 1. The causes of mortality, physical or physiological, must therefore be related to the size of the fish. The most likely reason for the size-dependent mortality is the lower swimming capacity of smaller individuals. The smallest haddock have been observed slinging from side to side between the net walls of the trawl when exhausted (Marteinsson, 1991), probably leading to skin damage. Most skin injuries were found on the most flexible region of the body, the rear part. Beating the tail, probably when pressed against net walls of the trawl or cage during trawling, is therefore the most likely cause of the skin injuries. If the injuries had been caused mainly by squeezing through meshes, they would have been predominantly at the broadest body part, while random collisions with the net panels would have led to injuries being more evenly distributed over the flank area. The cages could also have size-related mortality effects since the smallest individuals may not be able to maintain their position in the cage during towing (see page 47, discussion of Experimental design). In all cages, scale loss and mortality in trial 2 were length-dependent, while there were no length-related injuries in the haddock in trial 1, and length dependency in mortality was not observed in all the cage groups. One possible cause of this difference between trials is the differences in the techniques employed in the two trials, such as the towing of cages to a sheltered area after release from the trawl (see page 45, discussion of Experiment procedure). The degree of scale loss in trawl- 51

58 Discussion caught haddock in trial 2 was size-related, and differed from that of haddock in fish traps, in which no mortality occurred. Scale loss may therefore be regarded as one explanatory variable of mortality. The results regarding length-related scale loss and mortality of mesh-selected haddock are in agreement with previous experiments (Marteinsson, 1991; Soldal et al., 1991; Sangster and Lehmann, 1994; Sangster et al., 1996; Breen and Sangster, 1997). Mesh escape mortality A peak in mortality of mesh-escaped haddock was found in length classes whose girth approximated the mesh circumference. The probable girth of a 50 cm haddock is approximately the same as the circumference of a 135 mm mesh (Marteinsson, 1991). The peak in mortality of mesh escapees between 40 and 54 cm in trial 1, and the fact that mesh escapees in trial 2 had highest mortality rates in cm and length class, indicate that escaping through meshes was a cause of death in individuals whose girth was close to the mesh circumference. The mesh-selected haddock in trial 1 had most skin wounds, and the highest rate of large skin wounds in trial 2. The wounds were size-related, with the highest values recorded in the cm length class. The observation that the skin wounds of haddock in mesh cages were more frequent than in both grid and control cages indicates that the wounds were obtained by the mesh escapement. Further experiments As noted above, several factors other than trawl escapement may have caused injuries and mortality. In future experiments, it will be important to eliminate any possible sources of mortality caused by the experimental procedure. The knowledge acquired in the course of this study may lead to improvements in the experimental design of future studies. The following suggestions are based on the same experimental procedures as we employed in our research, and do not include observations by divers. 52

59 Discussion Water flow velocity in the cover and cage must be minimized so that the fish can swim freely. At the same time, trawling speed must be the same as on commercial vessels. Two possible solutions have been suggested: 1) Create a bucket effect with a small meshed netting or tight material in the rear end of the cage. 2) Use objects such as cone or net panel inside the cage to create a sheltered area. Heavy cage construction can cause deformation of the trawl or grid due to high towing resistance. The hoops or frames used in the cages should be designed with as little projected area as possible in order to minimize water resistance. The lining of the cages must be soft in order to minimize the risk of injuring the fish, especially in the rear end of the cages, where the fish will end up when they are exhausted, and in the part of the cage where fish are lying when the cages are raised at the end of the experiment. The risk of covering the cages with sand clouds from the trawl gear must be minimized. This could be achieved by using lighter ground gears or raising the cod end from the bottom with floats. If this is done, however, the effects of differences from commercial trawl gears will have to be considered. Since selection takes place mostly through the upper part of the cod end (Graham & Kynoch, 2001), a possible solution might be to have the cover only above the cod end, just as the grid cover is above the grid (Figure 3). The closure system must be reliable. If not, the risk of losing valid cages is high, increasing the uncertainty of the mortality estimates. It is important to be able to tow with open cages and to close them when wished, since short tows with prematurely closed cages can lead to biased size distribution and long tows to increased mortality. Ideally, the cages should be left in the sea where they are released, either on the bottom or floating in the water column. Towing cages after release causes stress. Even though divers may be used in shallow waters to move the fish from the cover to an underwater container in which they can be transported in a sheltered environment without water flow, the possibility of leaving the cages where released makes it possible to do the experiments independent of towing depth. If the cages could be left floating in the water column, it would be possible to study the survival of pelagic species in deeper waters. Observation of cages may be made by ROV, or by a camera 53

60 Discussion rigged inside the cages with a cable to the surface. Observations would be easier if the cages were raised from the bottom. This must be done in accordance with known decompression regimes (Tytler & Blaxter, 1973; Blaxter & Tytler, 1978; Harden Jones & Scholes, 1985) in order to avoid rupturing the swim bladder. If two or more trawls are to be used in the experiments, they should be identical unless the objective is to investigate the mortality-inducing effects of different trawl types. Conclusions Our experiments indicated that cod and saithe tolerate selection through meshes and grid without their survival being affected. Haddock mortality was higher and sizerelated, irrespective of the channel of escape and control sampling. Swimming ability is regarded as a critical factor, and the observed mortality could have been caused by fish passing through the trawl or by the sampling procedure. Mesh and grid escapes are not believed to be the main cause of mortality, although mesh escape might have increased mortality in larger haddock whose girth was similar to the circumference of the cod-end meshes. Scale loss was size-related in the same way as mortality, and could therefore be one of the factors explaining mortality. Since there was no significant difference between the scale loss of mesh-, grid- and trawl-caught control cages, escapes are not believed to have been the main cause of scale loss, but either collisions of exhausted fish with the front part of the trawl or the sampling procedure. In general, grid escapees sustained significantly fewer injuries than mesh escapees and trawl-caught haddock, probably because they avoided having to pass through the extension between trawl and cod end. A possible degree of mortality caused by the sampling methods used has been discussed. Before further studies are carried out, it is important to evaluate the possible mortality-inducing effects of the experimental procedure. 54

61 References References Alverson, D.L., Breen, M., Carr, A., Engås, A., Farrington, M., Frechet, A., Graham, N., Kynoch, R. Milliken, H., Soldal, A.V. & Sangster, G.I. (2000). Report of the FTFB Topic Group on Unaccounted Mortality in Fisheries. ICES FTFB Working Group, 74 pp. Bhattacharyya, G.K. & Johnson, R.A. (1977). Statistical concepts and methods. New York: John Wiley & sons. Blaxter, J.H.S., & Tytler, P. (1978). Physiology and function of the swimbladder. In O. Lowenstein (Ed.), Advantages in comparative physiology and biochemistry (pp ). New York: Academic Press Inc. Breen, M. & Sangster, G. (1997). The injuries sustained by haddock (Melanogrammus aeglefinus) escaping from trawl codends and the implications of these to survival. ICES FTFB Working Group, 19pp. Breen, M., Sangster, G. & Soldal, A. (1998) Evidence of cover induced mortality in fish survival experiments A cautionary note. ICES FTFB Working Group. Broadhurst, M.K. & Kennelly, S.J. (1996). Effects of the circumference of codends and a new design of square-mesh panel in reducing unwanted by-catch in the New South Wales oceanic prawn-trawl fishery, Australia. Fisheries Research, 27, 4, DeAlteris, J. T. & Reifsteck, D. M. (1993). Escapment and survival of fish from the codend of a demersal trawl. ICES mar. Sci. Symp., 196, Eddy, F.B. (1981). Effects of stress on osmotic and ionic regulation in fish. In A.D. Pickering (Ed.), Stress and fish (pp ). Bristol: Academic press inc. Engås, A., Fosseidengen, J.A., Isaksen, B. & Soldal, A.V. (1989). Simulerte redskapsskader på sei, Fiskeriteknologisk forskningsinstitutt, rapport (report, in Norwegian), 9pp. Engås, A., Foster, D., Hataway, B.D., Watson, J. W. & Workman, I. (1999). The behavioural Response of Juvenile Red Snapper (Lutjanus campechanus) to Shrimp Trawls that Utilize Water Flow Modifications to Induce Escapment. Marine Technology Society Journal, 33, 2, Engås, A., Isaksen, B. & Soldal, A.V. (1990). Simulated gear injuries on cod and haddock, a tank experiment. ICES Fish Capture Committee, FTFB Working Group Meeting, 9 pp. Fridman, A.L. (1986). Calculations for fishing gear designs. Farnham: Fishing News Books Ltd. Graham, N. & Kynoch, R.J. (2001). Square mesh panels in demersal trawls: some data on haddock selectivity in relation to mesh size and position. Fisheries Research, 27, Halliday, R.G., Cooper, C.G., Fanning, P., Hickey, W.M. & Gagnon, P. (1999). Size selection of Atlantic cod, haddock and pollock (saithe) by otter trawls with square and diamond mesh codends of mm mesh size, Fisheries Research, 41,

62 References Hansen, U.J. & Wileman, D. (1984). Netmotstand. Nyt om oliefisk-projektet (In Danish), 16. 4pp. Harden Jones, F.R. & Scholes, P. (1985). Gas secretion and resorption in the swimbladder of the cod Gadus morhua. J. Comp. Physiol. B, 155, He, P. & Wardle, C.S. (1988). Endurance at intermediate swimming speeds of Atlantic mackerel, Scomber scombrus L., herring, Clupea Harengus L., and saithe, Pollachius virens L. J. Fish Biol. 33, He, P. (1993). Swimming speeds of marine fishes in relation to fishing gears. ICES Marine Science Symposium, 196, Hokenson, S.R. & Ross, M. (1993). Finfish bycatch mortality in the Gulf of Maine northern shrimp fishery. Northwest Atlantic Fisheries Organization, NAFO SCR Doc. 93/124, 4pp. Isaksen, B. & Valdemarsen, J.W. (1994). Bycatch Reduction in Trawls by Utilizing Behaviour Differences. In Anders Fernö and Steinar Olsen (Ed.), Marine fish behaviour in capture and abundance estimation (pp 69-83). London: Fishing News Books Jacobsen, J. A. (1994). Survival experiments of fish escaping from 145 mm diamond cod-end trawl meshes at Faroes in 1992 and ICES FTFB, Fish Capture Committee, 8 pp. Jacobsen, J.A., Thomsen, B. & Isaksen, B. (1992). Survival of saithe (Pollachius virens L) escaping through trawl meshes. ICES Fish Capture Committee, CM 1992/B:29, Ref. G. Jonsson, E. (1994). Scale damage and survival of haddock escaping through cod-end meshes (tank experiment). ICES Fish Capture Committee, CM 1994/B:16. Larsen, R.B. & Isaksen, B. (1993). Size selectivity of rigid sorting grids in bottom trawls for Atlantic cod (Gadus morhua) and haddock (Melanogrammus aeglefinus). ICES marine Science Symposium, 196, Lowry, N. & Sangster, G. (1996). Survival of gadoid fish escaping from the cod-end ef trawls. ICES FTFB Working group meeting,38 pp. Lowry, N., Sangster, G. & Breen, M. (1996). Cod-end selectivity and fishing mortality. DIFTA, study contract no. 1994/005, 80 pp. Løkkeborg, S. & Soldal, A.V. (1995). Vulnerability to predation of small cod (Gadus morhua) that escape from a trawl. ICES Fish Capture Committee, CM 1995/B:15 Ref. G, 7 pp. Main, J. & Sangster, G.I. (1990). An Assessment of the Scale Damage to and Survival Rates of Young Gadoid Fish Escaping from the Cod-end of a Demersal Trawl, Scottish fisheries research report no. 46, 28 pp. Main, J. & Sangster, G.I. (1991). Do fish escaping from cod\ends survive. Scottis Fisheries Working Paper No 18, 15 pp. Marteinsson, J.E. (1991) Skader på fisk som sorteres ut i fiske med bunntrål. Thesis, The University of Tromsø, 82 pp (In Norwegian). McWilliams, P. (1992). Ioneregulering. In K. Døving & E. Reimers (Ed.), Fiskens fysiologi (pp ). Stavanger: John Grieg Forlag AS. 56

63 References Millar, R.B. (1995). The functional form of hook and gillnet selection curves cannot be determined from comparative catch data alone. Canadian Journal of Fisheries and Aquatic Sciences, 52: Pálsson, Ó.K. (In press). A length-based analysis of haddock discards in Icelandic fisheries. Fisheries Research 1384, 10 pp. Reeves, S.A., Armstrong, D.W., Fryer, R.J. & Coull, K.A. (1992). The effects of mesh size, cod-end extension length and cod-end diameter on the selectivity of Scottish trawls and seines. ICES journal of marine science, 49, 3, Sangster, G. (1992). The survival of fish escaping from fishing gears. ICES CM 1992/B:30 Sangster, G.I. & Lehmann, K. (1993). Assessment of the survival of fish escaping from commercial fishing gears. ICES Fish Capture Committee, CM 1993/B:2. Sangster, G.I. & Lehmann, K. (1994). Commercial fishing experiments to assess the scale damage and survival of haddock and whiting after escape from four sizes of diamond mesh cod-ends. ICES CM 1994/B:38, 69 pp. Sangster, G.I., Lehmann, K. & Breen, M. (1996). Commercial fishing experiments to assess the survival of haddock and whiting after escape from four sizes of diamond mesh cod-ends. Fisheries Research, 25, Smith, L.S. (1993). Trying to explain scale loss mortality: a continuing pussle. Review in Fisheries Science 1(4), Soldal, A.V., Isaksen, B., Marteinsson, J.E. & Engås, A. (1991). Scale damage and survival of cod and haddock escaping from a demersal trawl. ICES Fish Capture Committee, CM 1991/B:44, 12 pp. Soldal, A.V. (1996). Ansvarlig fangststrategi. Fisken og havet nr. 14 (In Norwegian), 82 pp. Suuronen, P. (1995) Conservation of young fish by management of trawl selectivity, Finnish Fish. Res., 15, Suuronen, P., Lehtonen, E., Tschernij, V. & Larsson, P.O. (1995) Skin injury and mortality of baltic cod escaping from trawl codends equipped with exit windows. ICES CM 1995/B:8, 13 pp. Suuronen, P., Perez-Comas, J.A. & Lehtonen, E. (1996). Size-related mortality of herring (Clupea harengus L.) escaping through a rigid sorting grid and trawl codend meshes. ICES Journal of Marine Science, 53, Thiele, W., Kroeger, M., Gau, S., Hopp, M, Klueber, U. (1997). Flow measurements in cod-end. ICES CM 1997/FF:10, 9 pp. Tytler, P. & Blaxter, J.H.S. (1973). Adaptation by cod and saithe to pressure changes. Neth. J. Sea Res. 7: Vinogradov, N.N., (1960). Survival of fish escaping from the codend of trawls. ICES CM Wardle, C.S. (1977). Effects of size on the swimming speeds of fish. In T.J. Pedley (Ed.). Scale effects of size on swimming speeds in fish (pp ). New York: Academic press. 57

64 References Wardle, C.S. (1993). Fish behaviour and fishing gear. In Tony J. Pitcher (ed.), Behaviour of teleost fishes (2. ed). (pp ). 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 No Wileman, D.A., Sangster, G.I., Breen, M., Ulmestrand, M., Soldal, A.V. & Harris, R.R. (1999). Roundfish and nephrops survival after escape from commercial fishing gear. EU Contract Consolidated Progress Report. EC Contract No: FAIR-CT , 37pp. Wood, C.M., Turner, J.D. & Graham, M.S. (1983). Why do fish die after severe exercise? J. Fish Biol. (1983) 22, Zaferman, M.L. & Serebrov, L.I. (1989). On fish injuring when escaping through the trawl mesh, ICES CM 1989/B:18. Zar, J.H. (1996). Biostatical analysis. (3 rd ed.) New Jersey: Prentice-Hall, Inc. 58

65 Appendixes Appendixes Appendix I: Diverse drawings Appendix II: Tables and figures Appendix III: Photographs 59

66 This document was created with Win2PDF available at The unregistered version of Win2PDF is for evaluation or non-commercial use only.

67 Appendix I Diverse drawings

68 Mesh size mm PE o mm N M ,5 15,0 7,30 m 7,30 m Headline 36,5 m Fishing line 18,90 m 18 mm Wire 13 mm LL chain (5 x 7,30 m) 1T2L B 7,30 m Mesh size mm 155 PE N M o mm 2x4 21,5 3,3 2,5 m ,5 67,5 10,5 B ,5 8,8 B 1N4B ,5 7, N8B ,5 7, N4B ,5 7,4 75 1N8B ,5 7,4 75 2N5B 1N3B 1N2B 65,1 m Circumference (Stretched) Name: Cotesi nr. 3 Drawing by: Ólafur Ingólfsson, redrawn from Refa as

69 Mesh size mm PE o mm N M 1,0 m ,5 14,8 7,30 m Headline 36,5 m 18 mm vire 7,30 m 1T2B B Fishing line 19,2 m 13 mm LL chain 7,30 m ,5 10, N17B hm mm PE mm S M 51,5 8,0 6,40 m 6,40 m L 55 3,0 m 38 1N4B ,5 6, N8B ,5 6, N2B ,5 7, x4 14,5 2, B 1N8B ,5 7, x4 14,5 2, B 1N2B 70,2 m Circumference (Stretched) Name: Euronet/Alfredo nr Drawing by: Ólafur Ingólfsson redrawn from Refa a/s

70 2,0m 5,0m 2,0m 3,6m 70 N 20 Plastic Rings 250T 1,8mm PE 50 mm Diamond mesh Front Frame: 70mm aluminium pipes Floats on frame: 6 x 11" plastic floats Lower part of frame wrapped with 16mm PP rope 3 x 11"Floats 200m 14mm PE 50m 16mm PA 5,0m 1,8mm PE 50 mm Square mesh 1,85m 1,85m 1,85m 1,85m Sea anchor Shackle 4,0m 19,6 mm PA 20 Plastic Rings 2,5m Rear Front Rear Drawing of cage and nettings used in trial 2.

71 Cage used in trial 1.

72 2 akustiske utløsere The Sort-X grid with enclosing cover and a sampling cage, not drawn in scale. 2 akustiske utløsere Cover used in both trials and a cage used in trial 2. The diameter of the hoops was 2m, and the mesh size of the PE cover 40 mm. The distance between hoops was 2m.

73 Appendix II Tables and figures

74 Table I Length distribution of haddock in cage categories in trial 1. Length Cage 1 - Grid Cage 6 - Grid Grid - total Cage 5 - Mesh Cage 3 - Control Cage 4 - Control Cage 9 - Control Control - total class Alive Dead Alive Dead Alive Dead Alive Dead Alive Dead Alive Dead Alive Dead Alive Dead Table II Length distribution of haddock in cage categories in trial 2. Length Cage 2 Mesh Cage 8 - Mesh Mesh - Total Cage 6 - Grid Cage 3 - Control Cage 9 - Control Control - Total class Alive Dead Alive Dead Alive Dead Alive Dead Alive Dead Alive Dead Alive Dead Table III Length distribution of injury recorded haddock in trial 1. Length Mesh Grid Control class Cage 5 Cage 1 Cage 6 Cage 3 Cage 4 Cage Table IV Length distribution of injury recorded haddock in trial 2. Length class Mesh Grid Control (cm) Cage 2 Cage 8 Cage 6 Cage 3 Cage Table V Length distribution of cod in trial 1. Grid Mesh Control Length class (cm) Cage 1 Cage 6 Cage 5 Cage 3 Cage 4 Cage Table VI Length distribution of fish from fish traps in trial 2. Length class (cm) Haddock Cod Saithe Table VII Length distribution of cod in trial 2. Grid Mesh Control Length class (cm) Cage 6 Cage 2 Cage 8 Cage 3 Cage

75 Table VIII Length distribution of saithe in trial 1. Length Grid Mesh Control class (cm) Cage 1 Cage 6 Cage 5 Cage 3 Cage 4 Cage Table IX Length distribution of saithe in trial 2. Length Grid Mesh Control class (cm) Cage 6 Cage 2 Cage 8 Cage 3 Cage Table X Results from GLM tests of injuries between haddock s left and right sides in trial 1. Left/Right Cage * Left/Right Bleedings Wounds Infections Table XI Results from GLM tests of injuries between haddock s left and right sides in trial 2. Left/Right Cage * Left/Right Bleeding Wounds Scaleloss Table XII Results from GLM tests, testing homogenity of haddock injuries between cages, flank areas and fins in trial 1. Injuries Skin area/fins Cage * Area Skin bleedings < Skin wounds < Skin infections Fin damages Fin bleedings Fin infections < < Table XIII Results from GLM tests, testing homogenity of haddock injuries between cages, flank areas and fins in trial 2. Injuries Skin area/fins Cage*Area Skin bleedings < Skin wounds Scale loss < Fin bleedings < Fin wounds < Fin split < Fin rot <

76 35 Current measurements at anchoring sites in trial 2 30 Current speed (cm/s) : : : : : : : : : :03 Date and time : : : : : :03 Figure I Current measurements at the anchoring sites in trial 2. 9,4 Temperature measurements at the anchoring sites in trial 2 9,3 9,2 Temperature ( C) 9,1 9 8,9 8,8 8,7 8,6 8, : : : : : : : : : : : :53 Date and time : : : : : : :03 Figure II Temperature measurements at the anchoring sites in trial 2.

77 Appendix III Photographs

78 Figure I Setting out fish trap, baited with mackerel for sampling of control fish. Figure II Heaving a cage onboard the assistant vessel at the termination of the experiment.

79 Figure III Preparing for setting out a cage in trial 1, the collapsible cage is seen with a sea anchor on top. Spherical floats are tied to the acoustic release unit. Figure IV Several haddock from a cage Figure VI Caudal fin of a haddock with a fin rot and bleeding. Figure V A haddock with a cleft in caudal fin, wound and bleeding. Figure VII A wounded haddock, with scale loss, skin bleeding, fin bleedings and fin wound.

80 Figure VIII A cod with scale loss after lying on the meshes of the cage when taken onboard. Figure IX A dorsal fin with a bleeding and a slight cleft. Figure X A haddock with skin bleeding, scale loss and skin wound.