The escape window as a management option to improve the size selectivity of the Baltic Sea cod fishery

Transcription

1 International Council for the Exploration of the Sea CM 19991R:1 The escape window as a management option to improve the size selectivity of the Baltic Sea cod fishery Niels Madsen\ Rene Holsrand Leslie Foldager 2 I. DFU, Danish Institute for Fisheries Research, The North Sea Centre, DK-985 Hirtshais, Denmark. 2. ConStat, The North Sea Centre, DK-985 Hirtshais, Denmark. Abstract The square mesh window has been implemented in the legislation for the Baltic Sea cod fishery. This paper reviews the history of the Baltic cod escape windows and presents preliminary results of a recent experiment. The objectives of this experiment were to make a comparison of an escape window with very good selectivity properties to a conventional diamond mesh codend. Three different mesh sizes were tested for each codend. A fixed and random effect model was formulated to analyse the results incorporating recorded variables. The results of this analysis are presented and the potential in escape windows as a management option compared to standard codends is finally discussed. Keywords: Size selectivity; Cod; Baltic Sea; Escape windows; Gadus morhua. 1. INTRODUCTION 1.1 Objectives From the mid-eighties a rapid decrease of the stock and catches of BaItic cod and a high fishing mortality (Bagge at ai., 1994) provided the incentive for a more sustainable exploitation pattern. Danish and Swedish experiments were initiated to assess the potential for escape windows to improve the size selectivity (Tschernij et ai., 1996; Lowry et ai., 1995). Codends with escape windows where the meshes have a square mesh configuration are from a theoretical point of view superior to standard diamond mesh codends, because meshes do not close with accumulating catch. The window concept as a tool to improve the selectivity of trawls is not a recent invention but was already tested by Ridderstad (1915) in the Kattegat and the BaItic Sea. This paper reviews some of the history of the experimental progress and the use of Baltic cod escape windows. Based on previous Danish experiments an experiment with a very improved escape window was designed. The objectives were to compare the performance of the window codend to a conventional diamond mesh codend. It is then possible to assess the benefit of the escape window as a management option. The preliminary results of this experiment are presented in this paper. 1.2 History of Baltic cod escape windows Different codend types, which have all been tested in Danish sea trials, are shown in Figure I. The bottom window codend (codend 1, which is also known as the "Danish" codend) and the side panel codend (codend 4, also known as the "Swedish" codend) were adopted into legislation from I. June 1995, by the European Communities [EC council regulation No. 3362/94, 2 December 1994] and for the whole Baltic Sea by the IBSFC. The legislation allowed the use of these codends with a stipulated 15 mm minimum mesh size for the both window designs as an alternative to a traditional 12 mm diamond mesh codend. The main technical differences between the two window types used by the commercial fleet today are that the net material in the side panel window is patented stiff plastic coated knotted single twine netting hung such that the meshes maintain the conventional fore-aft orientation but have a wide

2 opening. The net material in the bottom panel codend is conventional netting turned through 9 degrees to form square meshes. The regulations do not specify use of single or double twine and double twine is most commonly used. According to the above mentioned regulations the side panel windows must be 8 % of the length of the codend and end only 4-S cm from the codline. The windows are 5 cm wide but are inserted in an opening of 15-2 cm, giving the window an out-bending bulbous shape. The bottom panel windows end m from the codline and should be square meshes (bars) long giving a total length of about 3.S m. The bottom window must be 8 square meshes wide giving a width of about SO cm. Comparative experiments (see Madsen, 1999) as well as experiments measuring the selectivity parameters (Madsen et ai., 1998) indicated that the bottom window codend were not optimal compared to the side window codend. An ICES working group (Anon, 1996) reviewed experimental data sets from the Baltic Sea area which gave an overall estimated selection factor (SF = LSO% / window mesh size) of the bottom window codends to be only 3.6 (giving a LSO% of 32 cm at los mm window mesh size) and that of side window codends to be much higher 3.S2 (giving a LSO% of37 cm at los mm window mesh size). The selection range (SR) was estimated to be marginally lower for the side window codend, 8.2 cm compared to 8.6 cm for the bottom window codend. In Denmark codends with the bottom window are used by most vessels in the fleet. This codend is also very used by fleets from other countries as well. Comparative catch experiments have been conducted using a twin trawl rig to compare different locations of windows (Madsen, 1999). These results indicated that the selectivity was improved when the window was moved aft (Fig. 1, codend 2), ending at the same position as the Swedish window, close to the end of the codend. Further experiments indicated that the selectivity could be further improved in a codend with a single top window (Fig. I, codend 3). However the catches were sparse when this experiment was conducted. Nevertheless we decided to continue selectivity experiment with the top window because a single top window codend is easier to install and is used in other areas (see discussion). In the legislation valid from 1999 [EC council regulation No. 48/99, 18 December 1998] the bottom Godend window was extended to end 4-5 cm from the bottom of the codend (like codend 2 and 4, Fig. 1) and have a length of 8 % of the codend (like codend 4). There are still no specifications of the net material. The side window codend (Codend 4) and a standard 12 mm diamond mesh codend can still be legally used. 2. METHODS 2.1 Experimental procedures The experiment was conducted from 12'th June to lo'th July The vessel chosen for the sea trials was a 29 HP commercial stem trawler which is built in 1978 having an overall length of m and GRT of This trawler has been used in all recent Danish selectivity experiments in the Baltic Sea. The experiment were conducted on fishing grounds in the Baltic Sea around the Danish island Bornholm (ICES subdivision 24 and 25). Two identical typical commercial Baltic Sea cod trawls were fished in a twin trawl rig. Each codend was fished at both sides of the twin trawl rig. Three identical conventional 6 m long codends with a circumference on 96 open meshes and 4 meshes placed in the two selvedges were made of 4 mm green PET double twine netting. The nominal mesh size was los mm, 12 mm and 14 mm. Actual mesh sizes are shown in Table 1. Three window codends were constructed and inserted in three identical codends with a nominal mesh size of 15 mm. Ultra Cross netting (4.9 mm PET) was used for window material. The window was 3.5 long m and 1.4 m wide and inserted in the top panel (Codend 3, Fig 1). Nominal inside mesh sizes of 11 mm, 125 mm and 135 mm were chosen. Actual mesh sizes of the window are shown in Table 1. Codend covers, supported by kites to keep them clear of the codend, were used. Operational conditions were recorded for each haul. Wind, Depth, speed and sea state were recorded at the beginning and end ofthe haul and average values were calculated. The starboard codend was hauled first in every haul. It was often necessary to subsample from the cover because of large catches. The 2

3 whole cover catch was measured III baskets and baskets were randomly chosen for subsampling measurements. 2.2 Statistical analysis Data was applied to a fixed and random effect model (Wileman et ai., 1996). The model parameters were estimated by residual maximum likelihood (REML) fitting procedure to overcome problems with underestimation of variances (Fryer 1991). We have considered the following models: A. L5 = L5 + CMS;L5. CMS + O'CC;L5. ICC + lib;l5 + liw;l5 SR = CMS;SR CMS + lib;sr + liw,sr B. L5 =OL5 +OCMS,L5 CMS+liB,L5 +liw;1.5 SR = CMS;SR. CMS + li B.SR + liw;sr C. L5 = WMS-L5 WMS + s-l5. S + S_WMS L5 S WMS + lib li W -L5,,, ".., SR = WMS-SR. WMS + li B-SR + li W -SR D. L5 = WMS;1.5 WMS + li B;1.5 + li W ;1.5 SR = WMS;SR. WMS + li B;SR + liw-,sr The abbreviations stand for: CMS = Codend Mesh size ICC = Inverse Codend Catch CC = Codend Catch EB = Between-haul variation EW= Within-haul variation WMS = Window Meshsize S = Side of twin trawl rig where codend is located ( for starboard and 1 for port) Model A and B are used for the standard codends whereas models C and D for the codends with window. The reason for using inverse codend catch is that plots of the codend catch against L5% and SR show curvatures of the inverse (l/x) type. In the models A and C we have used the EC-model to analyse the full models with all potential covariates, and then removed insignificant covariates one at a time. The models shown above are the reduced versions of these full models. We reduced from the following list of effects: Intercept, CMS or WMS, Side, Side'CMS or Side'WMS, ICC, Depth, Speed and Seastate. In the models B and D we only used intercept and mesh size (CMS or WMS) as fixed effects in the "full" models, These models provides an identification of the meshsize effect. Another reason for using these models is that the precision and usefulness of side (S) and codend catch might be poor. In Table 2 the AlC value is a measure, which can be used to compare models that are not constraints or reduced versions of each other, selecting the model with the lowest Akaike's Information Criterion (AlC value) as the best one (Jones, 1993). We have however not used this value to make any decisions. The L5 % and SR are estimated at the actual (measured) meshsize, and for model A at the mean value of the codend catch, 3

4 3. RESULTS 3.1 Catches Total catches in codends and covers are shown in Table 1. Codend catches were relatively high in the 15 and 12 mm standard codend whereas catches were relatively low in the 14 mm standard codend and the 11 mm, 125 mm and 135 mm window codend. The low catches in these codends is however mainly a product of a relatively good selectivity. Relatively many cod were caught in the cover. Catches of other species were relatively limited. There was some bycatch which mainly consisted of flounder in the codend. Minor quantities of herring and sprat were caught in the cover. 3.2 Selectivity analysis The 5 % retention length (L5%) of single hauls is plotted against selection range (SR) in Fig. 2. The mean REML values with confidence regions for individual models for each codend (no fixed effects) is shown in Fig. 3. Both Figures (Fig. 2 and 3) shows a positive correlation of the two parameters. The correlation is most pronounced for the window codend where it is relatively steep. The parameters of the single hauls (Fig. 2) as well as the mean REML values (Fig. 3) of the 12 mm codend are found relatively close to the 15 mm codend. Parameters of single hauls are relatively scattered for the 135 mm window codend and the confidence regions of this codend are wider than for the other codends. The overall impression is however that confidence regions for all codends are relatively narrow. The mesh size is a fixed parameter in all models. Model A predicts the catch size effect for the standard codends. Model B and D can predict the L5% and SR for the standard and window codend respectively when other fixed effects but the catch size are ignored. Normally results from this kind of selectivity experiments will be reported for further management use in this format. Model C predicts the side effect that refers to which side in the twin trawl rig the codend was fished. Parameters for all models are shown in Table 2 and 3. The models with fixed effects are shown in Figure 4. The top figure shows the effect of the codend catch in the standard codend on L5%. The minimum mean and maximum catch is indicated. The model predicted inverse relationship seems unlikely at catches below minimum observed catch. However, catches below the minimum observation are too low to be of significant importance when modelling commercial conditions. There is a relatively rapidly decrease of L5% from the minimum observed catch to the mean observed catch but a slow and steady decrease from the mean to the maximum observed catch. Observations larger than the maximum observed catch could have been very useful. The centred Figure (Fig. 4) shows the predicted LSO% as a function of the mesh size for all four models. Model A is shown with the mean catch. This Figure shows the span between L5% of the standard and window codend when applied to respective models. There is a very significant improvement in L5% comparing the window codend to the standard codend. The side effect decreases the L5% of the port codend with increasing mesh size. The bottom Figure (Fig. 4) shows SR versus Meshsize. Plots are only shown for model A and C since the parameters for model A and B, and for model C and D, are almost equal. We would therefore not be able to distinguish these lines. The plot demonstrates a somewhat lower SR for the window codend. 4. DISCUSSION This experiment indicated that a codend with a single top window had a higher L5% and lower SR when the window mesh size and the codend mesh size of the standard codend was the same. To avoid introducing one more parameter we decided to keep the mesh size of the window codend constant. Alternatively it could be considered to keep the windowfcodend mesh size ratio constant. However, this experiment demonstrated that the size selectivity could be efficiently adjusted by only changing the mesh 4

5 size of the window. The advantage is that mesh size regulations only have to include the window and the remaining codend can still be used. L5% and SR are positively correlated. This is an important finding which has not been very well documented in previous experiments. When advising and using selectivity parameters it is often assumed that SR is a constant parameter. It is important to realise that the consequence is that the size selectivity will be less efficient with increasing mesh sizes. Even though it is a well known fact that standard meshes tend to close with accumulating catch the actual effect on the selectivity parameters has not been very well documented. The catch size effect that was demonstrated in this experiment is a good argument for preferring the window codend. If the stock and catches increases the L5% might be substantially reduced when using standard codends. The results indicated a side effect for the window codend. It is likely that cod escaping when the codend is tackled, at the surface along the boat side, caused this effect. However the highest L5% was found in the codend which was tackled and hauled on board first. It might be possible that the cod are more alive in the first tackled codend and make more escape attempts. Baltic cod escaping through the codend meshes during trawling has good possibilities of surviving (Suuronen et al., 1996) but the fate of surface escaping cod is still not known. Survival experiments of surface escaping cod from a Danish seine codend (Soldal and Isaksen, 1993) demonstrated however a negligible mortality. The full square mesh codend was originally considered to improve size selectivity of gadoids (see Robertsen and Stewart, 1988). The top window has some advantages compared to the full square mesh codend: the square mesh netting is protected from bottom damage, the codend is more flexible and easier to handle and less square mesh netting is needed which is also an important parameter because good netting for square meshes is rather expensive. Top windows have been successfully used to reduce gadoid bycatches in other areas (Robertson, 1993; Madsen and Moth-Poulsen, 1994) and are to be specified in EU legislation. The top window makes it possible to standardise legislation to some extent between fishing areas with regard to gear specification and inspection. This experiment was conducted from a single vessel. It has been demonstrated for the Baltic Sea cod fishery that the vessel effect on selectivity can be of substantial magnitude (Tschemij et ai., 1999). Vessel size, vessel-gear interaction and hauling technique were significant parameters. However this experiment was conducted with standard 12 mm codends. A codend with an escape window might however minimise the vessel effect. It is therefore important to repeat this experiment with different vessels types. It is also important to consider the pelagic cod fishery where bulk catches are often experienced. In this case it might be a good idea to use a longer window than used in this experiment. The same net material (4 mm double PET) as that used for codends is also used for the windows by the commercial Danish fleet. It is thought that this material is not well suited for windows because the double twines tend to separate, reducing the opening of the meshes and not giving fish an optimal chance of escape. Another problem with using conventional netting in square mesh windows is knot slippage, which leads to a high degree of uneven mesh sizes and shapes. This is particularly a problem because the window is placed behind the lifting strop were the force is very large when hauling the catch. To overcome these problems we decided to use the Ultra Cross netting. In this netting the twine threads are continuous in the all bars direction making it very strong and of stable mesh configuration when used as square mesh netting. Coated netting is used in the side panel codend to ensure uniformity of the meshes. These special net materials are however more expensive to purchase and more difficult to repair than traditional double PET. It is important to consider net material used for windows carefully in the legislation. Only allowing use of single twine would probably ensure satisfactory performance and standardisation of the escape window. Acknowledgements This project was carried out by the Danish Institute of Fisheries Technology and Aquaculture (DIFTA) with funds from DG XIV of the European Commission (the BACOMA project) and the Danish Ministry of Fisheries. Thanks are due to the crew on the vessel and the Danish Fishermen Organisation 5

6 for their good co-operation; to Thomas Moth-Poulsen and Mogens Andersen for their invaluable assistance when conducting the experiment and Dr. Josianne Stottrup for commenting the draft manuscript. This article does not necessarily reflect the views of the European Commission and in no way anticipates any future opinion of the Commission. 5. REFERENCES Anon., Report on advice on Baltic cod mesh selection. ICES Fishing Technology and Fish Behaviour Working Group. Woods Hole April Bagge,., Thurow, F., Steffensen, E. and Bay, J., The Baltic cod. Dana, vol. 1, pp Fryer, R. J., A model of the between-haul variation in selectivity. ICES J. Mar. Sci. 48: Jones, R.H., Longitudinal data with serial correlation: a state-space approach. Chapman & Hall. Lowry, N., Knudsen, L. H. and Wileman, D., Selectivity in Baltic cod trawls with square mesh codend windows. ICES CM 19951B:5. Madsen, N., & Moth-Poulsen, T., Measurements of the selectivity of Nephrops and demersal roundfish species in conventional and square mesh panel codends in the northern North Sea. ICES CM-19941B:14 Madsen, N. Moth-Poulsen, T. and Lowry, N., Selectivity experiments with window codends fished in the Baltic Sea cod (Gadus morhua) fishery. Fish. Res. 36: Madsen, N., Experimental adjustments of the escape window position in trawl codends - implications for Baltic Sea cod fishery. Symposium on selectivity research in the Baltic area March 1999, Sea Fisheries Institute, Gdynia, Poland. To be printed in Report Series of Institute of marine Research (Sweden). Ridderstad, G., A new construction of trawl-net intended to spare under-sized fish. Svenske Hydr.. BioI. Kommisionens skrifter Vol 6. Robertson, J.H.B. and Stewart, P.A.M., A comparison of size selection of haddock and whiting by square and diamond mesh codends. J. Cons. Explor. Mer. 44: Robertson, IH.B., Design and fitting of square mesh windows in whitefish and prawn trawls and seine nets. Scottish Fisheries Information Pamphlet. No. 2. Soldal, A.V. og Isaksen, B., Survival of cod (Gadus morhua) and Haddock (Melanogrammus aeglejinus) escaping from a Danish seine at the sea surface. ICES-FTFB-WG Meeting. Suuronen, P., Lehtonen, E., Tscernij, V. & Larsson, P.-O., Skin injury and mortality of Baltic cod escaping from trawl codends equipped with exit windows. Arch. Fish. Mar. Res. 44(3), 1996, Tschernij, V., Larsson, P-O., Suuronen, P. and Holst, R., Swedish trials in the Baltic Sea to improve selectivity in demersal trawls. ICES CM 19961B:25. Tschernij, V., Foldager, L., Holst, R. and Larsson, P-O., Evidence of factors at vessel-level affecting codend selectivity in Baltic cod demersal trawl fishery. ICES FTFB WG meeting, St. Johns, Canada April Wileman, D.A., Ferro, R.S.T., Fonteyne, R. and Millar, R.B. (editors), Manual of methods of measuring the selectivity of towed fishing gears. ICES Cooperative Research Report. No

7 Table 1. Total catches of codends and covers. Codend Mesh size'" No. hauls Total codend catch Total cover catch (mm) Cod Cod Bycatch Cod Cod Bycatch (Kg) (No.) (kg) (Kg) (No.) (kg) Standard 15 mm 11.7± Standard 12 mm 117.1± Standard 14 mm ± Window 11 mm 14.6 ±.8* Window 125 mm ± 1.2* Window 135 mm 131.2± \.3* * measured with 4 kg ICES gauge. Codend mesh sizes were: *1) 11.6 ± 1.6; *2) 11.2 ± 2.; *3)11.5 ± 1.6 Table 2. Estimated parameters for each of the four models applied. Model Parameter Estimate Std. Dev. t-value df p-value 8 LSO l A 8CMS;LSO ICC;LSO CMS;SR B 9CMS;LSO CMS;SR WMS,lSO S;L C 9S WMS;L WMS;SR D 8WMS;L5o WMS;SR Table 3. Selectivity parameters for all models. Model Meshsize AIC Estimated L-5 % Estimated SR SF ICES Gauge EECwedge Standard A Standard Codend Standard B Standard Standard Window C Window starboard Window Window C Window port Window Window D Window Window

8 Cadend 1: bottom Window Codend 2: bottom Window moved aft ~,.;;-.,;;...,;.r."';:"""""""'-"':'".~:~''' ~~~ ---'C:~-...:... :~~.. COdend 3: single top Window COdend 4: side window FigtlJ:e 1. COdend illustrations. 8

9 Standard IOmm nomm HOmm " L5D'IIo " " Windows W 11 mm W 125 mm W 135 mm " L5'llo -_ "1I"I!.- - ".. Figure 2. Selection range (SR) plotted against the 5 % retention length (L5%). : <f) ~..J I, i, I ~-1 w...! I, Standud 12 mm, [ St~ndBrd I i L~-- \J 15 mm Standard Window Window 11 mm,, :.l, '.' Stand~rd 14 mm o Window 125 mm., ' ' \. ~ Window 135 mm ' , , , r------~ Figure 3. Confidence regions (95%) for Selection range (SR) plotted against the 5 % retention length (L5%) for mean REML estimates without any fixed effects. 9

10 , t-' OJ (1 *- "' "' ~... ~~, min Catch 1 Nl mean'catch max catch Standarcl14 mm Standard 12 mm Siandard 15 mm Codend Catch t- o ( "' ~ :s... '" '" ~ Model A - mean Catch Model B Model C - Starboard Model C - Port ModelD Meshsize ~ > '" ~ U) t-, (- : co" Model A & B Model C& D Meshsize Figure 4, Illustration ofthe selectivity models, 1