Scottish Industry Science Partnership Report Report no. 01/10 SISP project 004/09 Trials to reduce cod by-catches in Shetland mixed demersal whitefish trawl fisheries by putting large meshes (300 and 600mm) in the front portion of a commercial trawl R J Kynoch, F G O’Neill and R J Fryer February 2010 Trials to reduce cod by-catches in Shetland mixed demersal whitefish trawl fisheries by putting large meshes (300 and 600mm) in the front portion of a commercial trawl R J Kynoch, F G O’Neill and R J Fryer Marine Laboratory, Marine Scotland – Science, Aberdeen. SUMMARY During 2009 experimental trials were conducted to assess the extent to which cod by-catches in the Shetland mixed whitefish fishery would be reduced by replacing the 120mm netting in the front sections of a trawl with 300 and 600mm netting. A further aim was to assess the effect of these modified trawls on the catches of monkfish and megrim, which are economically important to the Shetland fleet. The results showed that both test gears caught significantly fewer cod, hake, megrim and ling than the control trawl across all length classes and significantly fewer monkfish below 76 and 83cm respectively than the control trawl. The 300mm trawl caught significantly more haddock than the other trawls and catches of saithe greater than 53cm were reduced in both test gears. INTRODUCTION During 2008 the Orkney Fisheries Association (OFA) conducted catch comparison trials using a modified whitefish trawl designed to reduce the capture (and consequently potential discarding) of cod (Campbell et al., 2010). The ‘Orkney Gear’ (as the trawl became known) replaced the 160mm diamond mesh netting in the front portion of the trawl (wings, cover and first belly section) with 300mm diamond mesh netting. The Orkney Gear retained fewer cod: the relative catch rate was length dependent with 45% fewer cod caught at 35cm but no significant reduction for lengths >78cm. The Orkney Gear also caught fewer monkfish and megrim: the relative catch rate of monkfish was length dependant with 37% fewer monkfish caught at 37cm, but no significant reduction for lengths >55cm; the relative catch rate of megrim did not depend on length, with about 43% fewer megrim caught at all lengths. Under the 2009 Conservation Credit scheme a ‘buy back’ was offered to whitefish vessels using the Orkney Gear. The incentive offered was either 20 days or 12% depending if the vessel was on flat rate or track record allocations. Owing to the importance of monkfish and megrim to the Shetland mixed whitefish trawl fleet, concerns were raised by local skippers about the viability of fishing the Orkney Gear in their fishery. To assess the impact of this new trawl the Shetland Fishermen's Association (SFA) submitted a Scottish Industry/Science Partnership (SISP) proposal to carry out a catch comparison study similar to the Orkney trials but using a vessel and on fishing grounds typical of the Shetland fishery. The vessel used for the Orkney trial was a 2600hp twin-rig trawler, whereas those typical of the Shetland fleet are in the range 800 -1200hp. Furthermore, two different mesh sizes were proposed, one of 300mm netting and another of 600mm. There was provision under the 2009 Conservation Credit scheme to fish trawls with 600mm diamond mesh netting in their front portion, which would receive an additional 24 days or 15% allocation. This report presents the results of these trials, carried out in the summer of 2009. VESSEL AND FISHING GROUNDS The Resilient (LK195), a 593kW twin rig whitefish trawler (Figure 1), was chartered for the trials. The charter ran for 22 days from 25 June to 16 July 2009. Two Shetland fishing grounds were targeted, one about 15 nautical miles east of Unst and the other 20 to 25 nautical miles North of Muckle Flugga. The species mix was of cod, megrim, monkfish, ling, hake, haddock and saithe, with small numbers of whiting, lemon sole and plaice. Figure 1: Fishing vessel Resilient (LK195). FISHING GEAR – CONTROL TRAWL The vessels own commercial whitefish trawl was used as the control/standard trawl. The design had a 468 meshes x 120mm (nominal) fishing circle with 160mm (nominal) diamond meshes in the upper wings, lower wings and top-square (Figure 2). The rockhopper ground gear had 350mm discs throughout, all rigged onto 28mm wire with an overall length of 45.73m. The wire rig had 36.58m single wire sweep, 36.58m of single rubber leg sweep and 27.44m double bridles (top – wire and lower – chain). The gear was fished with a three-warp towing system with a 1500kg roller clump centre weight and spread using 1000kg Morgere Ovalfoil otterboards. During the trials all trawls, the standard and the two test trawls, were rigged with a tickler chain made from 16mm mid-link chain and a flip-up rope assembly, as used by Resilient when fishing commercially. Figure 2: Resilient standard trawl. FISHING GEAR – EXPERIMENTAL TRAWLS Two test trawls were made for these trials, one with a maximum mesh size of 300mm (nominal) diamond mesh (Figure 3) and the other with 600mm (nominal) diamond mesh (Figure 4). The test trawls had the same frame rope lengths (e.g. headline, footrope and fish tails), trawl stretched length and positioning of guard/strengthening meshes as the standard trawl. Both trawls were rigged with similar ground gear and flotation specifications as the standard trawl. Figure 3: 300mm test trawl. Figure 4: 600mm test trawl. To simplify rigging of the 600mm meshes to the headline and footrope combination wire, a 5 mesh deep x 300mm mesh DBL twine guard was used which was intended to minimise slippage of the large meshes and to prevent the 200mm floats fouling on the large meshes (Figure 5). To improve strengthening further, the same 300mm mesh guard was also used in the upper and lower wing tips. To enable the rearmost meshes of the 600mm first belly section to connect to the following 120mm (nominal) mesh panel a 4 mesh deep x 300mm DBL bating section was used (Figure 6). Figure 5: Guard meshes to connect 600mm meshes to headline/footrope. Figure 6: 300mm batings section connecting 120mm to 600mm meshes. During the trials all trawls (standard and test) were rigged with new 120mm (nominal) codends. The codends were made from 5mm DBL high tenacity PE twine with 94 open meshes in the circumference and 100 meshes deep. The 120mm extensions were made from 5mm single high tenacity PE twine with 94 open meshes in the circumference and 100 meshes deep. The codend mesh sizes were measured, whilst the netting was wet, using an EU standard wedge gauge (Table 1). TABLE 1 Mesh measurements of 120mm codends using EU wedge gauge Codend Control trawl 300mm trawl 600mm trawl Mesh size (mm) 126 126 126 FISHING TRIALS PROCEDURE Tow duration was 4 hours except for hauls 11 and 34 which were 3.1 and 2.2 hours respectively due to the tickler chains starting to dig into the seabed. Towing speed over the ground ranged between 2.7 to 3.1 knots, the usual towing speed for Reliant when fishing commercially for monkfish and megrim. Simrad ITI net monitoring acoustic instrumentation was used to monitor otterboard spread during most hauls. Only one headline height sensor was available so height measurements were recorded for each trawl on alternate days. A full haul summary can be found in Tables 2 and 3. The twin trawl technique (Wileman et al., 1996) was used to assess the relative catch rates of the test gears. This involves towing two similar trawls in parallel, with the test gear fished on one side of the twin trawl system and the unmodified control trawl fished on the other side. To reduce any port/starboard bias the test trawls were fished on both sides of the twin trawl system during the trials. As standard practice aboard Resilient, the starboard trawl’s codend was always emptied into the vessel’s fish hopper first. To prevent washout the port codend was hauled up and left suspended on the stone trap above the fish hopper until the starboard catch had been processed. At the end of each haul, catches from the test and control trawls were kept separate with the target species separated into baskets, weighed and then all measured. A breakdown of weights for the target species is given in Tables 4 and 5. The amounts of non-target species caught during the trials were very small and estimated to be no more than 60kg per haul in either test or control codends. Most of the non-target species caught were skate, rays, gurnards and lesser spotted dogfish. TABLE 2 Haul summary table for the 300mm test trawl Haul number R09/01 R09/02 R09/03 R09/04 R09/05 R09/06 R09/07 R09/08 R09/09 R09/10 R09/11 R09/12 R09/13 R09/14 R09/15 R09/16 R09/17 R09/18 R09/19 R09/20 R09/21 R09/22 R09/23 R09/24 R09/25 R09/26 R09/27 R09/28 R09/29 R09/30 Test trawl 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm 300mm Side test Fished (P or S) P P P P P P P P S S S S S S S S S S S S S S S S P P P P P P Area fished Haul Duration (Hrs) Unst Unst Unst Unst Unst N Flugga N Flugga N Flugga N Flugga N Flugga N Flugga Unst Unst Unst Unst Unst Unst Unst Unst Unst N Flugga N Flugga N Flugga N Flugga N Flugga N Flugga N Flugga N Flugga N Flugga N Flugga 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 3.1 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 Mean speed Made good (kts) 3.0 3.0 3.1 3.0 2.8 2.9 3.0 2.8 2.7 2.7 2.8 3.1 2.9 2.9 2.8 2.8 2.7 2.8 2.9 2.7 3.0 2.9 2.8 2.7 2.7 2.8 3.1 3.0 2.8 3.0 Mean headline height test (m) 5.0 4.9 N/R N/R N/R 4.2 4.2 4.3 N/R N/R N/R 4.7 4.9 4.9 N/R N/R N/R N/R N/R 4.6 N/R N/R N/R 4.9 4.5 4.6 N/R N/R N/R 4.4 Mean headline height control (m) N/R N/R 3.3 3.2 3.4 N/R N/R N/R 3.0 3.0 3.0 N/R N/R N/R 3.0 2.9 2.9 N/R N/R N/R 2.9 3.5 3.1 N/R N/R N/R 3.0 3.3 2.7 N/R Mean speed Made good (kts) 2.8 2.9 2.9 2.7 2.8 2.9 2.9 2.9 3.1 3.0 2.8 2.8 2.9 3.0 Mean headline height test (m) 5.2 4.9 5.0 4.9 4.7 4.8 4.7 N/R N/R N/R N/R N/R N/R N/R Mean headline height control (m) N/R N/R N/R N/R N/R N/R 3.3 3.1 3.4 3.0 3.2 2.9 2.9 3.1 TABLE 3 Haul summary table for the 600mm test trawl Haul number R09/31 R09/32 R09/33 R09/34 R09/35 R09/36 R09/37 R09/38 R09/39 R09/40 R09/41 R09/42 R09/43 R09/44 Test trawl 600mm 600mm 600mm 600mm 600mm 600mm 600mm 600mm 600mm 600mm 600mm 600mm 600mm 600mm Side test Fished (P or S) P P P P P P S S S S S S S S Area fished Unst Unst Unst Unst Unst Unst Unst Unst Unst Unst Unst Unst Unst Unst Haul Duration (Hrs) 4.0 4.0 4.0 2.2 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0 TABLE 4 Summary of target species catch composition for 300mm test and control trawl codends Haul No R09/01 R09/02 R09/03 R09/04 R09/05 R09/06 R09/07 R09/08 R09/09 R09/10 R09/11 R09/12 R09/13 R09/14 R09/15 R09/16 R09/17 R09/18 R09/19 R09/20 R09/21 R09/22 R09/23 R09/24 R09/25 R09/26 R09/27 R09/28 R00/29 R09/30 300mm test trawl codend catch (kg) HAD 60 89 62 55 41 24 16 23 8 11 8 74 82 68 45 60 34 45 38 19 23 19 9 23 18 26 31 20 27 32 WHI 24 43 29 11 19 0 0 0 0 0 0 35 23 34 23 30 15 27 5 11 0 0 0 0 0 0 3 0 3 0 COD 134 92 131 124 242 42 23 17 7 11 30 74 128 183 168 120 185 118 152 172 5 15 14 40 21 46 61 72 68 57 SAI 48 15 89 36 25 47 83 38 36 30 57 93 33 69 58 19 29 26 33 60 55 22 42 45 45 40 127 66 60 107 HAK 30 75 50 53 83 5 9 5 1 2 5 63 93 53 71 56 22 41 30 18 21 5 6 8 7 7 25 10 22 20 LIN 30 45 93 104 89 38 63 13 32 43 34 33 78 63 38 22 33 33 45 45 23 36 24 44 46 31 21 21 9 19 Control trawl codend catch (kg) ANG 80 58 78 124 94 89 78 69 128 82 106 77 55 72 121 65 70 41 76 79 63 97 102 77 78 90 73 84 92 94 MEG 1 3 2 4 1 11 16 10 20 12 9 1 1 1 0 2 3 2 6 2 10 17 14 16 15 14 5 4 4 2 PLA 4 5 6 4 9 0 0 0 0 0 0 8 12 1 6 5 2 2 5 4 0 0 0 0 0 0 0 0 0 0 LSO 2 1 1 2 1 0 0 0 0 0 0 0 2 0 1 1 0 2 2 1 1 0 0 0 0 0 0 0 0 0 Total 413 426 541 517 605 256 288 175 232 191 264 458 507 544 531 380 393 337 392 311 201 121 211 253 210 254 346 277 285 331 HAD 44 52 30 27 45 18 13 7 10 4 7 47 51 29 31 44 31 30 14 10 9 10 7 21 16 16 20 27 10 24 WHI 40 33 18 10 18 0 0 0 0 0 0 33 30 16 32 20 10 17 9 11 0 0 0 0 0 0 1 5 7 2 COD 171 118 289 215 168 26 36 24 27 53 50 168 107 310 311 280 316 179 298 254 50 42 25 42 76 57 181 208 92 182 SAI 53 23 72 87 50 106 60 33 86 50 86 134 33 108 120 27 32 38 33 27 103 40 54 58 91 56 211 35 103 178 HAK 92 61 105 104 74 11 19 9 8 4 6 71 70 58 68 78 49 59 66 18 20 1 13 20 19 7 41 19 25 34 LIN 60 52 117 158 122 60 53 57 63 101 61 48 60 103 53 52 66 51 111 57 68 52 36 32 43 50 49 29 14 17 ANG 107 98 85 120 107 108 103 103 77 143 130 81 103 86 110 99 82 89 102 141 141 152 97 116 112 123 142 160 148 104 MEG 8 9 7 6 4 59 74 59 75 76 61 3 6 8 9 11 8 11 16 9 68 71 46 64 75 66 16 13 2 16 PLA 7 12 14 14 15 0 1 0 0 0 0 19 11 10 19 7 5 6 5 5 0 0 0 0 0 0 0 0 0 0 LSO 2 2 2 3 3 1 0 0 0 0 0 3 2 1 2 3 1 4 1 1 1 1 0 0 0 1 1 0 0 0 Total 584 460 739 744 606 389 359 292 346 431 401 607 473 729 755 621 600 484 655 533 460 369 278 353 432 376 662 496 401 557 TABLE 5 Summary of target species catch composition for 600mm test and control trawl codends Haul No R09/31 R09/32 R09/33 R09/34 R09/35 R09/36 R09/37 R09/38 R09/39 R09/40 R09/41 R09/42 R09/43 R09/44 600mm test trawl codend catch (kg) HAD 23 14 0 13 11 17 6 9 10 10 14 15 30 19 WHI 1 2 0 4 1 3 1 1 0 0 2 2 4 2 COD 22 48 30 42 41 40 18 29 9 25 21 33 44 50 SAI 11 9 8 10 9 9 10 15 7 27 12 26 40 25 HAK 23 50 24 35 14 46 39 22 24 20 12 14 16 14 LIN 4 16 6 0 25 7 14 4 9 8 2 10 21 6 Control trawl codend catch (kg) ANG 25 20 5 23 25 37 47 51 55 41 35 20 47 35 MEG 1 0 0 1 0 2 2 0 0 0 1 0 2 1 PLA 0 4 1 0 1 0 4 0 2 2 3 1 0 0 LSO 2 0 0 0 2 1 1 1 2 1 1 1 2 1 Total 112 163 74 128 129 162 142 132 118 134 103 122 206 153 HAD 23 0 0 12 14 15 6 2 10 9 23 7 15 15 WHI 4 0 0 3 1 3 0 1 0 0 3 2 5 1 COD 113 118 146 61 127 173 186 76 98 103 92 138 163 159 SAI 26 15 4 4 5 18 17 30 34 29 22 11 58 60 HAK 22 45 35 42 28 85 99 77 84 54 35 41 21 37 LIN 44 15 21 21 27 39 39 20 46 20 15 40 25 17 ANG 64 100 57 38 77 100 93 99 131 104 68 121 100 90 MEG 7 7 5 1 5 16 11 13 10 7 7 7 11 9 PLA 3 11 11 3 11 9 5 11 3 19 15 1 6 0 LSO 1 0 0 0 2 3 3 1 4 2 2 5 4 0 Total 307 311 275 185 297 461 459 330 420 347 284 368 408 388 RESULTS Gear Performance Mean headline heights for the control trawl were about 1.6m less than for both test trawls. The otterboard-to-clump spreads for the side fishing the test trawl were also generally higher, possibly indicating that the larger mesh test trawls had less drag and were easier to tow. There were some problems operating the 600mm trawl. The 600mm netting in the belly and top sheets consistently got caught on the 200mm floats while shooting. On several occasions it took 3-4 attempts to clear the floats, even after a temporary float rope was rigged. A solution was eventually devised using a rope run through the headline centre and then wound back around the powerblock. This kept the headline clear of the deck as the trawl was being shot until the wingends were clear of the stern roller. This technique was used for the rest of the trials but would not be considered practical for commercial fishing. Catch comparison A total of 30 hauls were completed with the 300mm trawl and 14 with the 600mm trawl. The 300mm trawl was fished on the Unst and Flugga grounds but the 600mm trawl only on the Unst grounds. Adequate numbers of cod, hake, monkfish, megrim saithe and ling were encountered during the trials, but the numbers of whiting, plaice and lemon sole were low and insufficient for subsequent analysis. The catches retained in the test and control trawl codends were analysed in two ways: using the smoother based methodology of Fryer et al. (2003); and using a generalised linear mixed model (Holst & Revill, 2009) (see Appendix for details). Both methods estimated and compared the catch rates of the test gears relative to the control gear and only differed in the assumptions they made about the shape of the catch rate curves. The results were virtually identical and the fitted relative catch rates from the generalised linear mixed modelling approach are shown in Figure 7. Cod, hake, megrim, and ling: the relative catch rates did not depend on length. Both test gears caught significantly fewer fish than the control gear. The 600mm gear caught significantly fewer fish than the 300mm gear. Haddock: the relative catch rates did not depend on length. The 300mm gear caught significantly more haddock than the control gear. The catch rates of the 600mm gear and the control gear did not differ significantly. Monkfish: the relative catch rates increased with length. The 300 and 600mm gears caught significantly fewer monkfish than the control gear below 76 and 83 cm respectively. The relative catch rate of the 600mm gear was significantly below that of the 300mm gear below 78 cm. Saithe: the relative catch rates decreased with length. Both test gears caught significantly fewer saithe than the control gear above 53 cm. The relative catch rate of the 600mm gear was significantly below that of the 300mm gear above 61 cm. Figure 7: Estimated catch rates of the test gears relative to the control gear from the generalised linear mixed modelling approach. The lines are solid when the relative catch rate is significantly different from unity at the point-wise 5% significance level and dashed otherwise. The catch rates are plotted for lengths that were present in most hauls. DISCUSSION In general, the catch rates of the 300mm trawl were significantly below the control gear and those of the 600mm trawl were significantly below the 300mm trawl. The one exception was for haddock where significantly more were caught in the 300mm gear than the control gear and there was no difference between the 600mm and the control gear. The catch rates of cod, megrim, ling, hake and haddock did not depend on length whereas that of monkfish increased with length while that of saithe decreased. Cod, megrim and monkfish are of prime importance in the Shetland whitefish fishery. The 300 and 600mm trawls caught 49 and 74% less cod respectively than the control trawl at all lengths (although very few cod > 91cm were caught). The respective losses of megrim were estimated to be 80 and 93%, again across all length classes. The relative catch rates for monkfish were length dependent. The 300mm trawl caught ~50% fewer monkfish at 30cm with no significant differences >76cm. The 600mm trawl caught 85% fewer monkfish at 30cm with no significant differences >83cm. The results for the 300mm trawl are broadly similar to those of the ‘Orkney Gear’ (Campbell et al., 2010). The Orkney Gear caught 45% fewer cod at 35cm with no significant differences >78cm, about 43% fewer megrim at all lengths, and 37% fewer monkfish at 37cm with no significant differences >55cm. The 300mm trawl might have caught less megrim than the Orkney Gear relative to the control, because the mesh size of the control trawl during the Orkney trials was 160mm whereas here it was 120mm. Ling, hake and saithe are also important to the Shetland fleet. The 300mm gear caught 28% fewer hake and 36% fewer ling than the control gear while the 600mm gear caught 53% fewer hake and 68% fewer ling. Catches of saithe greater than 53cm were also reduced. The relative catch rate decreased with larger fish and with increasing mesh size, suggesting that saithe are increasingly more able or prepared to penetrate netting the bigger they are and the bigger the mesh size. As found during the Orkney trials, haddock catch rates were significantly higher for all lengths in the 300mm trawl compared to the control trawl. This is possibly due to the higher headline height of the 300mm trawl (~1.6m) compared to the control trawl. However, there was no difference in catch rates between the 600mm and control trawls, which may indicate that haddock are escaping through the 600mm meshes. These results show that the gear modifications made here will result in large catch reductions of some species that are of economic importance to the Shetland whitefish fleet. The extent to which these gears are deemed viable will depend on a range of factors such as target species and species mix, available fishing opportunities and market prices etc. Given, however, that the rationale for these trials was to reduce the by-catch of cod, further research should be directed towards the development of gears that reduce catches of cod but retain species such as monkfish, megrim, ling, saithe and hake and which, presumably, would be more attractive economically. CONCLUSIONS The main aims of these trials were to assess the effect of increasing the mesh size in the front sections of a trawl from 120mm to 300 and 600mm on the ground fish and cod catches of the Shetland mixed whitefish fishery. • The 300 and 600mm trawls caught 49 and 74% fewer cod respectively than the control trawl at all lengths • The 300 and 600mm trawls caught 80 and 93% fewer megrim respectively than the control trawl at all lengths. • The catch rates for monkfish were length dependent. The 300mm trawl caught ~50% fewer monkfish at 30cm with no significant difference > 76cm. The 600mm trawl caught 85% fewer monkfish at 30cm with no significant difference > 83cm. • The 300mm trawl caught significantly more haddock than the 600mm and control trawls which had similar catch rates. • The 300 and 600mm trawls caught 36 and 68% fewer ling respectively than the control trawl at all lengths • The 300 and 600mm trawls caught 28 and 53% fewer hake respectively than the control trawl at all lengths • Catches of saithe greater than 53cm were also reduced, with the catch rate decreasing with larger fish ACKNOWLEDGEMENTS The authors would like to thank the skipper and crew of the fishing vessel Resilient for their technical help and expertise, the staff at L.H.D Lerwick who provided shore side facilities, the Shetland Fishermen’s Association and Shetland P. O. for advice and assistance throughout the project, and Iain Penny and Findlay Burns, Marine Scotland – Science and Paul MacDonald, NAFC Marine Centre, Shetland. APPENDIX The catches from the test and control codends were analysed in two ways: using the smoother based methodology of Fryer et al. (2003) and using a generalised linear mixed model (Holst & Revill, 2009). The smoother based analysis was in three stages: 1. a smoother was used to model the log catch rate of the test gear relative to the control gear for each haul; 2. the fitted smoothers were combined over hauls to estimate the mean log relative catch rate for each gear; 3. bootstrap hypothesis tests using the statistic Tmax were used to assess whether the mean log relative catch rates depended on gear, the side that the test gear was fished (port or starboard) or fishing area, and to compare the mean log relative catch rates to zero (or equivalently the mean relative catch rates to unity). The generalised linear mixed model was based on the usual SELECT model of selection (Millar & Fryer, 1999). Let yhl be the number of fish of length l in haul h measured in the test codend, and nhl be the number of fish of length l in haul h measured in both codends. Then, conditional on nhl, yhl was assumed to have a binomial distribution with probability rh(l) where logit rh(l) is a linear function of length logit rh (l ) = α h + β h l and is interpreted as the log catch rate of the test gear relative to the control gear for haul h. There was no sub-sampling to consider. The parameters αh, βh were assumed to vary between hauls as: αh ~ N(α g , σ 2α ) β h ~ N(β g , σ β2 ) Thus, the selection parameters αh, βh are normally distributed about intercepts αg and slopes βg which depend on the fixed effects (typically gear, side and area) with standard deviations σα and σβ respectively. The haul selection parameters αh, βh were assumed to be correlated. The significance of the fixed effects on selection was assessed by Wald tests in a backwards stepwise procedure. The smoother analysis makes fewer assumptions than the mixed model and in particular assumes no specific shape for the relative catch rate curves. Because the smoothers use local weighting, the fitted catch rates track the data, even at lengths with few fish. The mixed model can be more powerful than the smoother analysis and can incorporate haul-specific covariates such as catch size (although these were not considered here). However, care is needed when interpreting the fitted catch rates from a mixed model since their shape is often determined by the lengths where there are many fish and might not be appropriate elsewhere. For example, in catch comparison trials, the two gears might have similar catches once fish get above a certain size. Such behaviour is usually well described by the smoother, but a linear mixed model will force the relative catch rate to continue increasing (or decreasing) at large lengths. For the data presented here, the generalised linear mixed modelling approach seems appropriate. The smoothers found little evidence of nonlinearity in the individual haul or mean relative catch rate curves so linear logistic functions of length can be used. Further, the mean relative catch rates (from both approaches) were often well below unity or showed no length dependence, so inferences about catch rates at large lengths were not an issue. Attempts were made to incorporate quadratic terms in length into the mixed model to account for any nonlinearity, but these were not pursued as there was only one instance (monkfish) where a quadratic term was significant and then only marginally. Both analyses were on the logistic scale, but the results have been back-transformed for presentation. RESULTS The proportions of fish retained in the test codend (of those retained in both codends) are shown below for each species and haul. The black and red lines are the fitted values from the smoother analysis and the mixed model respectively. The grey shaded areas are pointwise 95% confidence bands on the fitted smoothers. The megrim data were sparse in fishing area 1, so the smoothers were only fitted to the data for fishing area 2 and a mean relative catch rate could only be estimated for the 300mm gear. For most hauls, the fitted values from the mixed model are broadly comparable with those from the smoother. The largest discrepancies occur in sparse hauls, where the width of the smoother confidence bands gives due warning about over-interpretation. The estimated mean relative catch rates for each species and gear are shown below: the black lines are the fitted values from the smoother analysis (with pointwise 95% confidence bands indicated by the grey shaded areas) and the red lines are the fitted values from the mixed model. The points are the proportion of fish retained in the test codend (of those retained in both codends) pooled over hauls. There is strong agreement between the results from the smoother analyses and the mixed models. megrim 600mm monkfish saithe 1.0 0.5 0.0 300mm 1.0 0.5 0.0 20 40 60 50 cod 600mm 100 50 haddock 100 hake ling 1.0 0.5 0.0 300mm 1.0 0.5 0.0 50 100 20 40 60 50 100 50 length (cm) There was no strong evidence that the relative catch rate of any species depended on the side that the test gear fished or the fishing area. A couple of terms were marginally significant (at levels between 1 and 5%) but these were higher order interactions and were treated as spurious given the number of tests made in the model selection procedure. Detailed results for each species are given below. 100 Cod: the smoother analysis found that the catch rates of both test gears differed from that of the control gear (p < 0.001 in both cases). Both test gears caught significantly fewer cod than the control gear at all lengths (pointwise 5% level). The relative catch rates of the two gears differed (p < 0.001) with the relative catch rate of the 600mm significantly below that of the 300mm gear at all lengths above 37 cm (pointwise 5% level). The relative catch rates showed no suggestion of length dependence, which was confirmed by the mixed model (p = 0.64). The mixed model also confirmed the significant difference between the relative catch rates of the two gears (p < 0.001), estimating the relative catch rate of the 300mm and 600mm gears to be 51% and 26% respectively (with 95% confidence intervals of 44-59% and 20-32%). Haddock: the smoother analysis found that the catch rate of the 300mm gear differed from that of the control gear (p < 0.001) but that the catch rate of the 600mm gear did not differ from that of the control gear (p = 0.78). The 300mm gear caught significantly more haddock than the control gear at all lengths (pointwise 5% level). The relative catch rates of the two gears differed (p = 0.042) with the relative catch rate of the 600mm gear below that of the 300mm gear for haddock between 40 and 53 cm (pointwise 5% level). The relative catch rates showed no suggestion of length dependence, which was confirmed by the mixed model (p = 0.22). The mixed model also confirmed the significant difference between the relative catch rates of the two gears (p = 0.007), estimating the relative catch rate of the 300mm and 600mm gears to be 142% and 105% respectively (with 95% confidence intervals of 130-157% and 87-128%). Hake: the smoother analysis found that the catch rates of both test gears differed from that of the control gear (p = 0.005, < 0.001 for the 300mm and 600mm gears respectively). The 300mm and 600mm gears caught significantly fewer hake than the control gear between 46 and 72 cm and below 80 cm respectively (pointwise 5% level). The relative catch rates of the two gears differed (p = 0.012) with the relative catch rate of the 600mm gear below that of the 300mm gear at all length below 70 cm (pointwise 5% level). The relative catch rates showed no suggestion of length dependence, which was confirmed by the mixed model (p = 0.55). The mixed model also confirmed the significant difference between the relative catch rates of the two gears (p = 0.005), estimating the relative catch rate of the 300mm and 600mm gears to be 72% and 47% respectively (with 95% confidence intervals of 62-84% and 3760%). Ling: the smoother analysis found that the catch rates of both test gears differed from that of the control gear (p < 0.001 in both cases). Both test gears caught significantly fewer ling than the control gear at all lengths up to 100 cm (pointwise 95% level). The relative catch rates of the two gears differed (p < 0.001) with the relative catch rate of the 600mm gear below that of the 300mm gear at all length below 80 cm (pointwise 5% level). The relative catch rates showed no suggestion of length dependence, which was confirmed by the mixed model (p = 0.62). The mixed model also confirmed the significant difference between the relative catch rates of the two gears (p < 0.001), estimating the relative catch rate of the 300mm and 600mm gears to be 64% and 32% respectively (with 95% confidence intervals of 57-72% and 25-42%). Megrim: the smoother analysis found that the catch rate of the 300mm gear differed from that of the control gear (p < 0.001), with the 300mm gear catching fewer megrim at all lengths (pointwise 5% level). The relative catch rate showed no suggestion of length dependence, which was confirmed by the mixed model (p = 0.61). The mixed model could be fitted to the data from both test gears. There was a significant difference between the relative catch rates of the two gears (p < 0.001), with the relative catch rate of the 300mm and 600mm gears estimated to be 20% and 7% respectively (with 95% confidence intervals of 18-23% and 4-13%). Monkfish: the smoother analysis found that the catch rates of both test gears differed from that of the control gear (p < 0.001 in both cases). The 300mm and 600mm gears caught significantly fewer monkfish than the control gear below 72 and 80 cm respectively (pointwise 5% level). The relative catch rates of the two gears differed (p < 0.001) with the relative catch rate of the 600mm significantly below that of the 300mm gear at all lengths below 76 cm (pointwise 5% level). The mixed model showed that the relative catch rates increased with length (p < 0.001) and estimated that the 300mm and 600mm gears caught significantly fewer monkfish than the control gear below 76 and 83 cm respectively (pointwise 5% level). The mixed model also confirmed the significant difference between the relative catch rates of the two gears (p < 0.001), with the relative catch rate of the 600mm gear significantly below that of the 300mm gear at all lengths below 78 cm. Saithe: the smoother analysis found that the catch rates of both test gears differed from that of the control gear (p = 0.003, 0.008 for the 300mm and 600mm gears respectively). Both gears caught significantly fewer saithe than the control gear at lengths above 53 cm (pointwise 5% level). The relative catch rates of the two gears did not differ significantly (p = 0.18). The mixed model showed that the relative catch rates decreased with length (p < 0.001) and again estimated that the 300mm and 600mm gears caught significantly fewer saithe than the control gear above 53 cm (pointwise 5% level). However, the mixed model also estimated that the 600mm gear caught significantly more saithe than the control gear below 34 cm (pointwise 5% level). This is probably spurious as there were few saithe at these lengths, and the fitted values were being driven by the linear fit to the rest of the data. The mixed model found marginal evidence that the relative catch rates of the two gears differed (p = 0.054), with the relative catch rate of the 600mm gear significantly below that of the 300mm gear at all lengths above 61 cm. REFERENCES Campbell R, Harcus T, Weirman D, Fryer RJ, Kynoch RJ and O’Neill FG., 2010. The reduction of cod discards by inserting 300mm diamond mesh netting in the forward sections of a trawl gear. Fisheries Research, 102, 221-226. Fryer RJ, Zuur AF, Graham N, 2003. Using mixed models to combine smooth sizeselection and catch-comparison curves over hauls. Canadian Journal of Fisheries and Aquatic Sciences 60: 448-459. Holst R, Revill A, 2009. A simple statistical method for catch comparison studies. Fisheries Research 95: 254-259. Millar RB, Fryer RJ, 1999. Estimating the size-selection curves of towed gears, traps, nets and hooks. Reviews in Fish Biology and Fisheries 9: 89-116. Wileman, D.A., Ferro, R.S.T., Fonteyne, R., Millar, R.B., 1996. Manual of methods of measuring the selectivity of towed fishing gears. ICES Coop. Res. Rep. 215, p. 126.
© Copyright 2026 Paperzz