© 2006 International Council for the Exploration of the Sea
Seasonal variation in trawl codend selection of northern North Sea haddock
a Ege University, Fisheries Faculty, Fish Capture Department Bornova-Izmir 35100, Turkey
b FRS Marine Laboratory PO Box 101, Victoria Road, Aberdeen, AB11 9DB, Scotland, UK
c Biomathematics and Statistics Scotland, Rowett Research Institute Bucksburn, Aberdeen, AB21 9SB, Scotland, UK
d University of the Highlands and Islands, Millenium Institute, Moray College Moray Street, Elgin, IV 3O IJU, Scotland, UK
*Correspondence to R. S. T. Ferro: tel: +44 1224 876544; fax: +44 1224 295511. e-mail: ferro{at}marlab.ac.uk.
We investigate the variation in three separate months of trawl codend size selection for haddock, paying attention to changes in the length/girth relationship, and size of the total catch in the codend. Three cruises were conducted on the same fishing grounds using the same fishing gear on board the same commercial trawler in April 1995, September 1995, and February 1996. The periods coincided with postspawning in April, post-summer feeding in September, and prespawning in February. The seawater temperature at the seabed on the fishing grounds was near its annual maximum in September (12.2°C) and annual minimum in February (7.2°C) and in April (7°C). There was significant variation in selectivity of haddock with month and total catch in the codend. Fish length and, for two of the three cruises, fish maximum girth were used as variables in selectivity analysis. The probability of being retained was lowest in September, when the fish were in good condition and water temperature was highest, but girth was also largest (50% retention length L50 = 33.5 cm; girth at L50 = 172 mm). The retention probability was intermediate in February when the water was cold and the fish were prespawning, again with relatively large girths (L50 = 31.3 cm; girth at L50 = 158 mm). The retention probability was highest in April when the fish were postspawning, the water was cold, and the fish had the smallest girth (L50 = 27.3 cm; girth at L50 = 129 mm). Although care was taken to maintain similar conditions during all trips, there were some differences between April and the other months. Overall, retention was not related simply to girth and length. In September and February, at constant girth, codend retention increased with length, suggesting that within a given month, fish condition has an effect on selectivity. However, at constant length an increase in girth in February led to an increase in retention, but in September to a decrease in retention. A possible explanation is that, in February, the girth of a fish increases owing to its sexual maturity, whereas in September, increased girth is due to maximum volume of somatic tissue. Water temperature and volume of fish muscle affect the maximum rate and power of swimming. Swimming ability may be a significant factor in determining selectivity.
Keywords: haddock, Melanogrammus aeglefinus, North Sea, seasonal variation, selectivity
Received 1 April 2004; accepted 11 January 2005.
 |
Introduction |
|---|
 Top Introduction Material and methodsResultsDiscussionAppendixReferences |
|---|
Often, the objective of regulations on mesh size and other features of fishing gear design is to reduce fishing mortality of undersized fish. Undersized fish have the potential to grow to marketable size and to recruit to the spawning stock. The retention or escape of a fish clearly depends, inter alia, on its size in a well-defined way, as demonstrated by the selectivity curve for a single haul (Wileman et al., 1996). Length is usually used to characterize fish size because it is easy to measure. Length is well correlated with maximum girth (Margetts, 1957; Messtorff, 1958), which would be expected to be the critical body dimension governing escape through meshes. However, girth measurements of the same length of fish change throughout the year. The stage of gonad development, stomach volume, and adipose deposit of the body are the main factors influencing the girth of a fish (Kawamura, 1970). Consequently, at a given length, fish are expected to have maximum and minimum girths at different times of year related to feeding and spawning periods.
The length/weight data from Coull et al. (1989) indicate variation of weight for haddock, Melanogrammus aeglefinus, in the northern North Sea. Minimum weight is in April (after spawning) and maximum weight in September (after summer feeding). Girth can also be enhanced when gonad development is advanced, which is in the period January–March for haddock in the northern North Sea.
Variation between sampling months in codend selectivity with both length and girth is here investigated for haddock. Three cruises (using the same fishing ground, fishing gear, and commercial trawler) were conducted in April 1995, when eight valid hauls were made, September 1995 (17 valid hauls), and February 1996 (12 valid hauls). These periods coincided with postspawning in April, very good physical condition after summer feeding in September, and prespawning in February.
|
Material and methods |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionAppendixReferences |
|---|
The 578-kW commercial trawler "Aalskere" (K373), with an overall length of 24.4 m, was chartered for seven-day sea trials during the period 8–14 April 1995, 15–21 September 1995, and 21–27 February 1996. During each cruise, trawling was carried out in the Start Point area, 5–20 miles off the east coast of the Orkney Islands. However, because of poor weather in April 1995, the fishing area was closer inshore than during the other two cruises, the average towing speed was 0.4 knots slower, water depth was slightly shallower and, during two hauls (9 and 10), considerable quantities of seaweed (Flustra foliacea and hydroid) were retained by the net.
A Scotnet white fish rockhopper trawl with 524 x 200-mm meshes round the fishing circle and a 10-m straight extension ahead of the codend was towed using 36.6-m single sweeps, 59.4-m double bridles, and Poly-Ice oval otter boards weighing 1300 kg each. The same two-seam 100-mm diamond mesh codend of 5.24 m stretched length, constructed from 4-mm double braided polyethylene twine with 100 open meshes on its circumference, was used during all three cruises. The codend mesh sizes, i.e. opening of mesh (ISO, 1974), were measured (when wet) using an ICES spring-loaded gauge set at 4 kg tension (Table 1).
|
The experiments used the hooped covered codend technique (Wileman et al., 1996). The 40-mm mesh cover 17.2 m long was held clear of the codend by means of two supporting alkathene hoops of 1.8 and 3.0-m diameter, the latter being permanently fixed inside the cover at the point where the maximum codend diameter was expected, to minimize the masking effect of the cover meshes on the codend.
The majority of tows made in September and February were of about 2 h duration, whereas in April towing duration of the valid hauls varied between 2.5 and 5.5 h, to ensure that sufficient fish were caught for statistical purposes but not too much to overload the small mesh cover. During all hauls, SCANMAR acoustic trawl instrumentation was used to measure headline height and wing-end spread of the net. Water temperature at the seabed was measured during the September and February cruises as 12.2°C and 7.2°C, respectively. Temperature was not recorded in April. However, the average 30-year bottom temperature in this area for April (7.01°C; Turrell and Slesser, 1992) has been assumed.
The total catch in the codend (all fish, shellfish, and debris) was put into containers and weighed. For large catches, random samples were taken from the containers and the rest was not measured. In those cases, the sampling fractions were calculated as the ratios of the weight of the samples to the weight of the total catches. The total catch in the cover was treated in a similar manner. To increase the sampling rate and hence to improve the accuracy of the selectivity parameters, extra samples were taken of the bigger (usually 22+ cm) length classes of some of the cover catches, which tended to be dominated by small fish. Haddock were then sorted from the rest of the catch. Two categories of haddock data were collected, as defined below.
- To estimate selectivity parameters in terms of length, total fish lengths were measured to the cm below for all valid hauls. Prior to analysis a correction of 0.5 cm was added to all fish lengths. The sampling fractions ranged from 1 to 0.23 for fish in the codend, from 0.16 to 0.012 for small fish in the cover, and from 0.77 to 0.054 for larger fish in the cover.
- To estimate the selectivity parameters in terms of girth, measurements of length and maximum girth were taken for six hauls during the September 1995 and the February 1996 cruises. The maximum girths were measured by hand by laying a flexible measuring tape (graduated in mm) in contact with the skin all round the fish, without applying additional tension to the tape. Maximum girth was usually around the black spot of haddock. This was referred to as natural girth by Margetts (1954). Random sampling procedures were adopted, but owing to time constraints, fewer fish could be measured. For fish in the codend the sampling fractions ranged from 0.11 to 0.5, and for fish >25 cm in the cover they ranged from 0.03 to 0.10 in September, and from 0.05 to 0.2 in February.
Analysis of length-based selectivity
Fish were grouped into 1-cm length classes, and for each length class the total numbers of fish sampled in the codend and the cover were determined. A binomial model with logistic link (McCullagh and Nelder, 1989) was used to describe the data. The selectivity curve r(l), which describes the probability that a fish of length l (cm) is retained in the codend, given that it entered the gear, is given by:
|
| (1) |
1 and 2 are selectivity parameters. Estimation of 1 and 2 was based on maximum likelihood, using the sampled numbers of fish, as opposed to raised numbers, because it provides more reliable estimates of the standard errors (Millar, 1994; Wileman et al., 1996). Goodness-of-fit was assessed by inspection of residual plots (Wileman et al., 1996). The need for an overdispersion parameter was tested by comparing the sum of Pearson residuals to a Chi-squared distribution, with degrees of freedom equal to the total number of length classes minus the number of model parameters (McCullagh and Nelder, 1989), but this was found not to be significant.
As the 50% retention length (L50) and length selection range (LSR) have a more practical interpretation than 1 and 2, the selectivity vector 
is introduced, where 
and 
are the estimated 50% retention length and selection range, respectively. Let W be the covariance matrix of 
(details given in Appendix). To account for unexplained variation between hauls, the following between-haul variation model, based on Fryer (1991), was employed (details in Appendix):
|
| (2) |
and Wsj are the estimated selectivity vector and its respective covariance matrix for haul j in season (month) s (s = A, S, and F refers to April, September, and February, respectively), Xsj is a 2 x p design matrix containing explanatory variables for haul j in season (month) s, 
s is a vector of length p for season (month) s, and D is the between-haul variation matrix. Assuming that 
, Wsj, and Xsj are known, s and D were estimated by maximizing the combined likelihood of formulation (2). For formulation (2) to hold, it is necessary that sufficient fish are sampled from both codend and the cover, for an adequate number of length classes. This was the case for all valid hauls.
Explanatory variables included in the design matrix Xsj were month, towing speed, water depth, distance covered, and total codend catch (including seaweed and bycatch). Distance covered is the product of haul duration and mean speed over the ground. Sea state was not included because there was little variation during September and February, 88% of the observations being either 0.5 or 1 m. Linear relationships were explored for all variables, whereas for catch size, quadratic relationships were also included. As an alternative to catch, log(catch) was also tried as a variable. Further, seasonal effects were investigated by allowing s to vary over months. Initially, formulation (2) was estimated by maximum likelihood. Variable selection was based on backward elimination, and variables were only included if they were significant at the 5% significance level. The best model was then fitted again using residual maximum likelihood, which gives unbiased estimates of the between-haul variation matrix and standard errors (see also Fryer, 1991).
Analysis of girth-based selectivity
In September 1995 and February 1996, measurements were made on random samples from codend and cover to estimate selectivity in terms of girth. Fish were grouped in 10-mm girth classes, and the numbers of fish sampled from the codend and the cover were determined for each class. For individual hauls, model (1) was fitted to the subsampled girth data, length l being replaced by girth g (mm). Let G50 be the 50% retention girth and GSR the girth selection range. Taking between-haul variation into account, the overall girth selectivity for September and February was obtained from fitting model (2) to the individual haul estimates of G50 and GSR. Individual haul characteristics (month, towing speed, water depth, distance covered, and total codend catch) were included via the design matrix Xsj.
Analyses of girth, length, and retention probability
Detailed data on length and girth measurements from the six hauls in September and February were used to investigate the effects of length and girth on retention probability (i.e. the proportion retained). For each haul, the sampled fish were grouped in 1-cm length and 10-mm girth classes. Appropriate raising factors were applied (given by the inverse of the subsampling fraction), and the raised data were then pooled over six hauls from both September and February cruises. Let r(l,g) be the probability that a fish of length l and girth g is retained in the codend. Analogous to the usual length analysis, r(l,g) was assumed to be logistic:
|
| (3) |
This model allows for investigation of the effects of length and girth on the proportions retained. Maximum likelihood was used to obtain estimates of ß0, ß1, ß2, and ß3. Overdispersion was significant, and was included in the model to account for the use of raising factors and pooling of data over hauls (Wileman et al., 1996). The relationship between length and girth was investigated through linear regression, using data from April, September, and February, as girth (mm) = 
0 + 1 length (cm).
The data used to establish these regressions were collected by taking from each haul a non-random sample of 30 fish over the whole length range present. Only in September and February were separate random samples taken from six hauls to produce girth-based selectivity curves (see above).
All statistical analyses were performed in Genstat 5, release 4.1 (Lawes Agricultural Trust, Rothamsted, Herts, UK), except for the between-haul variation models, for which the EC model software (Constat, Grønspættevej 10, DK9800 Hjørring, Denmark) was used, implementing the method of Fryer (1991). A result is regarded significant if the corresponding value of p is <0.05 (i.e. a significance level of 5%).
|
Results |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionAppendixReferences |
|---|
Gear performance and environmental conditions
Despite careful design of the experiment, there were differences in gear performance from month to month for the valid hauls (Table 2). A haul was considered invalid when the gear was damaged or when there were insufficient fish for the analysis.
|
The mean wing-end spread was higher in September (22.8 m) than in both April (19.5 m) and February (20.0 m). Mean towing speed was approximately 0.4 knots lower in April than in either September or February. Dahm et al. (2002) found that L50 for haddock reduced as towing speed increased, suggesting that if the speed in April had been equal to that in September and February, then the difference in L50 would have been even greater (by about 0.7 cm). The differences in speed and spread imply changes in swept area, which may influence either total catch or the rate of accumulation of fish. However, the rate of accumulation of fish varies considerably from haul to haul owing to changes in population on the grounds, so possibly masking changes attributable to swept area. The effect of variation in catch is studied in the analysis. Total weights of all fish, shellfish, and debris caught both in the codend and the cover were recorded for each valid haul of all three cruises (Table 3).
|
Fouling of the net by seaweed was observed in all three cruises, and was noticeably worse during haul 9 in April, although the 50% retention length is near the mean value for April. From TV observations in April, the weed did not appear to hinder fish escape through the open meshes just ahead of the catch.
Characteristics of the fish population
The fish population available to all three cruises was predominantly haddock and whiting (Merlangius merlangus), with some flatfish and sometimes significant quantities of Norway pout (Trisopterus esmarkii). Most whiting were <25 cm in April and <30 cm in September and February, so few were retained in the codend. Therefore, on all cruises, significant numbers of haddock (Table 3), whiting, and other species escaped through the meshes of the codend.
The length frequency distributions for the three months (Figure 1) show the main growth of the 1+ year class between April and September. Because haddock become vulnerable to fishing gear at around 20 cm long, it is difficult to interpret the differences between the September and February distributions. Girth frequency distributions are available for September and February only (Figure 2).
|
|
Linear regressions showed a positive relationship between length and girth for haddock (Table 4, Figure 3) for the April, September, and February cruises. Correlation between girth and length was high, from 0.80 (April) to 0.97 (September and February).
|
|
Length-based selectivity parameter estimates by haul
The estimated 50% retention length, estimated length selection range, and their standard errors are given for each haul in Table 5. Inspection of the residual plots indicated that the logistic model gave a good fit. Overdispersion was not significant. Correlations between
1 and 2, and between L50 and LSR, calculated as R12/
(R11R22) and W12/(W11W22), respectively, are much higher (in an absolute sense) for the former (average correlation of 0.999 vs. 0.654, respectively). Standard errors of L50 and LSR are given by the square root of W11 and W22, respectively. The estimated L50s tended to be smaller for April than for September and February. Selectivity ogives of the individual hauls are given in Figure 4. The April selectivity curves generally lie to the left of the February curves, whereas the September curves generally lie to the right.
|
|
Length-based selectivity parameter estimates by month
Of the explanatory variables, only total codend catch was significant. For L50, the linear effect with total codend catch was significant (see also Figure 5a):
|
| (4) |
|
LSR and the intercept of L50 are month-dependent, while the catch effect on L50 is the same across months. An increase in catch of 100 kg results in a decrease in L50 of 0.32 cm (Table 6).
|
The model that gives the marginally better fit (Akaike's Information Criterion = 231.4 for model (5) compared with 232.6 for model (4)) is given by
|
| (5) |
This model assumes that there is a linear catch effect for LSR which is the same over months (see also Figure 5b), and that the intercept of LSR is the same for April and September, but different for February. Here, an increase in codend catch by 100 kg results in an increase in LSR of 0.29 cm (Table 7). Both models indicate that L50 and LSR are lowest for February.
|
Models including log(catch) instead of catch gave similar results. As a relationship with catch is easier to interpret than with log(catch), only models including catch (where significant) are presented. Figure 5a suggests that, after correcting L50 for month effect, a possible quadratic relationship between catch and L50 may exist, but was not significant.
Girth-based selectivity parameter estimates
Individual haul estimates of G50 and GSR are given in Table 8. Inspection of the residuals suggested an adequate fit. Table 9 shows the results from between-haul variation analysis. The model that gives the best fit is given by
|
| (6) |
|
|
No significant effects of individual haul characteristics were found, except for a negative linear relationship of catch size with G50 (p = 0.008, see also Figure 5c and d). An increase in the codend catch by 100 kg resulted in a 4.6-mm reduction in G50. Further, G50 and GSR are significantly different for September and February (p < 0.019), with GSR in September being 50% larger than in February. Selectivity ogives of the individual and combined hauls are shown as dashed and full lines, respectively, for September and February cruises in Figure 6. The average mesh perimeter is 201 mm, which is twice the average mesh size plus 4 mm, to allow for the thickness of the mesh-measuring gauge. In September, unlike in February, some curves do not reach 100% retention until girths are greater than this mesh perimeter, suggesting that fish escape more energetically in September.
|
Proportions retained of fish in each length/girth class
Proportions retained by the codend were calculated for the fish measured in each length/girth class during the September and February cruises (Tables 10 and 11).
|
|
A length girth model (Equation (3)) was fitted to all the pooled data. Parameter estimates for the September cruise and the February cruise are given in Table 12. Inspection of the residuals showed an adequate fit.
|
The variation of retention probability with girth at constant length (columns in Tables 10 and 11) and length at constant girth (rows) may reveal important clues as to the mechanisms of selection. At constant girth, retention probability tends to increase with increasing length in both months. Although globally the proportion retained seems to increase with increasing girth, the model indicated that this was only significant for February. However, a significant (p = 0.005) negative length/girth interaction term was found for September. This agrees with observations from, for instance, length classes 31.5–36.5 cm, which show that within a length class, there is a tendency for the proportion retained to decrease with increasing girth (Table 10).
|
Discussion |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionAppendixReferences |
|---|
Variation of selectivity with codend catch
Codend catch size, which naturally varies between hauls, is a potential source of variation in codend selectivity, because it affects codend geometry and tension in the mesh bars. O'Neill and Kynoch (1996) reported a significant increase of the 50% retention length for haddock over the catch size range 113–407 kg in a similar codend. They suggested that this tendency may not continue and that with increasing catch size, L50 would begin to level out or decrease. In these experiments the catch size for valid hauls ranged from 107 to 423 kg in April, from 127 to 823 kg in September, and from 211 to 762 kg in February. The results of subsequent trials reported in the literature have revealed no consistent relation between catch and selection (L50). Dahm et al. (2002) and Graham et al. (2004) suggested positive relationships variously for haddock, cod, and saithe. Madsen et al. (1998, 1999), however, found negative relationships for cod and whiting, as found here for haddock. Other workers (Kynoch et al., 2004) report no significant relationship. A more detailed study of catch effects may be required.
Variation of retention probability by month and between months
For a given length, an increase in girth leads to an increase in retention in February, whereas it leads to a decrease in retention in September. A possible hypothesis is that, in September, the girth of a fish is determined mainly by volume of muscle (which is related to swimming ability), whereas in February, girth is determined mainly by the stage of gonad development, so that the increase in girth does not lead to improved swimming ability. Further, the GSR and LSR for February are only two-thirds those of September, so one interpretation could be that selection in February is more mechanical – a fish can either pass through the meshes or it cannot. This evidence suggests that, within a given month, ability to escape may be determined by the condition of the fish as well as its physical dimensions in relation to the mesh opening.
The main objective of the experiments was to measure variation in selectivity over months while variation attributable to other factors was minimized. Therefore, the same codend was tested on the same gear, towed on the same fishing ground by the same fishing boat. The weather was worse in April, resulting in a reduction of 0.4 knots in mean towing speed. There is little recent reported evidence of the effect of speed. Anon. (2002) concluded that, for both haddock and whiting, any effect of speed is likely to be small, but Dahm et al. (2002) report a linear decrease in haddock L50 of 1.9 cm for each knot increase in speed, which would increase the difference between our results for April and for the other months. Weed accumulated in the net in greater quantities during one valid haul in April, but the 50% retention length for that haul was close to the mean value for April. The months chosen for the experiments coincided with specific conditions of haddock in the northern North Sea. Fish were in good condition after summer feeding in September. Most of the mature fish were just about to spawn in February, and in April had recently spawned.
The 50% retention lengths from the three cruises in different months are significantly different from each other. In April, the 50% retention length is lowest despite the girth being smallest. L50 was highest in September when girth was largest. L50 and the associated girth in February are intermediate between those of April and September. If selection were solely dependent on the relationship between girth and mesh size, then we would expect the 50% retention girth (G50) to be equal for all months. However, girth-based selectivity curves also give a significantly higher G50 in September than in February (Figure 6). A summary of the overall results from all three cruises is given in Table 13, using models (4) and (6), which contain a catch effect in L50 and G50.
|
The lower retention probability in September than in February may be related to enhanced swimming performance. This may be due to the higher water temperature (12°C) in September having a beneficial effect on fish swimming ability (Videler and Wardle, 1991; Özbilgin and Wardle, 2002). In addition, the fish are in very good physical condition in September after summer feeding, with somatic tissue constituting a greater proportion of body weight. The higher retention probability of fish in April than in February cannot be due to water temperature, which was approximately 7°C during both these two months. The poor condition of fish after spawning in April may have a significant effect on the selectivity of the trawl codend. This study provides a strong indication that, between months and seasons, differences in selectivity may be explained by changes in swimming ability related to fish condition and water temperature (Özbilgin and Wardle, 2002), as well as by the physical dimensions of the fish relative to the mesh opening.
|
Appendix |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionAppendixReferences |
|---|
Calculation of the covariance matrix of L50 and LSR
Let
and 
be the maximum likelihood estimates of 1 and 2, having covariance matrix R. Estimates of L50 and LSR are then given by (Wileman et al., 1996): |
|
Let W be the covariance matrix of 
and 
. Based on the delta method (Stuart and Ord, 1994), the elements of W are given by
|
|
W is calculated from replacing L50, LSR, and 2 by their respective estimates. Because 
and 
are maximum likelihood estimates, this implies that 
and 
are the maximum likelihood estimates of L50 and LSR, so the following holds (Mood et al., 1974). Let 
= (L50, LSR), and let its estimate be denoted by 
. Then
|
| (A1) |
sj follows a normal distribution, with mean and between-haul variation given by
|
| (A2) |
|
|
|
Acknowledgements |
|---|
We thank Iain Harcus and Lee Groat, the skippers of the "Aalskere", and their crew for generously providing advice, expertise, and cooperation during the three cruises, and our colleagues Tadashi Tokai of Tokyo University of Fisheries, and Rab Hutcheon, Craig Davis, Bill Leiper, Peter Barkel, Martin Burns, and Scott McKay of the Marine Laboratory for support and encouragement. Two referees made valuable comments on an early draft. This experiment was one part of a three-year project partly funded by the European Commission under the AIR programme (Contract AIR2–CT94–1544). The project was undertaken jointly by the Danish Institute for Fisheries Technology and Aquaculture, Hirtshals, Institut fur Fischereitechnik Hamburg, the Institute of Marine Research, Bergen, and the FRS Marine Laboratory, Aberdeen. Thanks are due to the scientists from all three institutes who were actively involved in the planning and execution of the seasonal variation work, and to the European Commission for providing financial support. The views expressed in this paper do not necessarily reflect those of the European Commission. GH is employed by BioSS, which receives financial support from the Scottish Executive Environment and Rural Affairs Department (SEERAD).
|
References |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionAppendixReferences |
|---|
-
Anon. (2002) Report of Scottish trials assessing the selectivity of trawls and seines. Fisheries Research Services Report, Industry/Science Partnership 2001–2002, vol. 1.
Coull K. A., Jermyn A. S., Newton A. W., Henderson G. I., Hall W. B. (1989) Length/weight relationships for 88 species of fish encountered in the north east Atlantic. Scottish Fisheries Research Report No 43/89.
Dahm E., Wienbeck H., West C.W., Valdemarsen J.W., O'Neill F.G. (2002) On the influence of towing speed and gear size on the selective properties of bottom trawls. Fisheries Research 55:103–119.[CrossRef][Web of Science]
Fryer R.J. (1991) A model of between-haul variation in selectivity. ICES Journal of Marine Science 48:281–290.
Graham N., O'Neill F.G., Fryer R.J., Galbraith R.D., Myklebust A. (2004) Selectivity of a 120 mm diamond codend and the effect of inserting a grid or a square mesh panel. Fisheries Research 67:151–161.[CrossRef][Web of Science]
ISO. (1974) Fishing nets – Netting – Basic terms and definitions. ISO 1107.
Kawamura G. (1970) Study on herring in the north-western Pacific Ocean, 2. Distribution and catch of korfo-karaginsk herring in 1967–1968. Bulletin of the Faculty of Fisheries, Hokkaido University 21:1–11.
Kynoch R.J., O'Dea M.C., O'Neill F.G. (2004) The effect of strengthening bags on codend selectivity of a Scottish demersal trawl. Fisheries Research 68:249–257.[CrossRef][Web of Science]
Madsen N., Moth-Poulsen T., Holst R., Wileman D. (1999) Selectivity experiments with escape windows in the North Sea Nephrops (Nephrops norvegicus) trawl fishery. Fisheries Research 42:167–181.[CrossRef][Web of Science]
Madsen N., Moth-Poulsen T., Lowry N. (1998) Selectivity experiments with window codends fished in the Baltic Sea cod (Gadus morhua) fishery. Fisheries Research 36:1–14.[Medline]
Margetts A.R. (1954) The length–girth relationship in haddock and whiting and their application to mesh selection. Journal du Conseil Permanent International pour l'Exploration de la Mer 20:56–61.
Margetts A.R. (1957) The length girth relationship in whiting and cod and their application to mesh selection. Journal du Conseil Permanent International pour l'Exploration de la Mer 23:64–71.
McCullagh P. and Nelder J.A. (1989) Generalised Linear Models 2nd edn (Chapman & Hall, London).
Messtorff J. (1958) Length–girth measurements of cod and their relationship to mesh selection. ICES Comparative Fishing Committee 23:.
Millar R.B. (1994) Sampling from trawl gears used in size selectivity experiments. ICES Journal of Marine Science 51:293–298.
Mood A.M., Graybill F.A., Boes D.C. (1974) Introduction to the Theory of Statistics 3rd edn (McGraw-Hill, Tokyo).
O'Neill F.G. and Kynoch R.J. (1996) The effect of cover mesh size and codend catch size on codend selectivity. Fisheries Research 28:2–11.
Özbilgin H. and Wardle C.S. (2002) Effect of seasonal temperature changes on the escape behaviour of haddock, Melanogrammus aeglefinus, from the codend. Fisheries Research 58:323–331.[CrossRef][Web of Science]
Stuart A. and Ord J.K. (1994) Kendall's Advanced Theory of Statistics. 1. Distribution Theory 6th edn (Edward Arnold, London).
Turrell W. and Slesser G. (1992) Annual cycles of physical chemical and biological parameters in Scottish waters. Scottish Fisheries Working Paper 5/92.
Videler J.J. and Wardle C.S. (1991) Fish swimming stride by stride: speed limits and endurance. Reviews in Fish Biology and Fisheries 1:23–40.[CrossRef]
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 Cooperative Research Report, 215.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  H. Ozbilgin, Z. Tosunoglu, A. Tokac, and G. Metin Seasonal variation in the trawl codend selectivity of picarel (Spicare smaris) ICES J. Mar. Sci., November 1, 2007; 64(8): 1569 - 1572. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||