ICES Journal of Marine Science: Journal du Conseil Advance Access originally published online on May 7, 2008
ICES Journal of Marine Science: Journal du Conseil 2008 65(6):930-936; doi:10.1093/icesjms/fsn076
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Bias in size composition of chum salmon (Oncorhynchus keta) caught by a gillnet with a geometric series of mesh sizes, and its correction using gear intercalibration
1 Hokkaido National Fisheries Research Institute, Fisheries Research Agency, 116 Katsurakoi, Kushiro 085-0802, Japan
2 School of Aquatic and Fishery Sciences, University of Washington, Box 355020, Seattle, WA 98195-5020, USA
Correspondence to M. Fukuwaka: tel: +81 154 921715; fax: +81 154 919355; e-mail: fukuwaka{at}fra.affrc.go.jp
Fukuwaka, M., Azumaya, T., Davis, N. D., and Nagasawa, T. 2008. Bias in size composition of chum salmon (Oncorhynchus keta) caught by a gillnet with a geometric series of mesh sizes, and its correction using gear intercalibration. – ICES Journal of Marine Science, 65: 930–936.Some research gillnets with size combinations based on a geometric series have been used for research surveys underpinning the stock assessment of fresh-water and marine fish. We assessed a bias in size composition of chum salmon caught using a research gillnet consisting of ten different mesh sizes based on a geometric series of factor 1.14. In all, 11 fishing operations were conducted for gear intercalibration between the research gillnet and a midwater trawl in the central Bering Sea. The best-fit selectivity model to pooled catch data included different fishing intensities among gillnet meshes. The pooled catch efficiency and the maximum catch efficiency of the gillnet increased with fish size. Estimated size composition of chum salmon was more similar to trawl catches than to research gillnet catches. Bias in size composition of research gillnet catches may be caused by the difference in encounter probability among mesh sizes, variability in fish swimming speed based on fish size, mesh visibility influencing fish behaviour, and diel vertical migration of chum salmon. When conducting multimesh gillnet surveys for stock assessment, researchers should correct a bias in size composition by performing gear intercalibrations.
Keywords: chum salmon, gear selectivity, geometric mesh size, research gillnet, size composition
Received 6 August 2007; accepted 13 April 2008; advance access publication 7 May 2008.
 |
Introduction |
|---|
 Top Introduction Material and methodsResultsDiscussionReferences |
|---|
Gillnets have been used extensively as a research fishing gear in fresh-water and marine environments (Hamley, 1980; Hovgård and Lassen, 2000). Research surveys are expected to collect less-biased data than are obtained from commercial catches. However, gillnets retain fish with a narrow size range, related to the mesh sizes employed (reviewed by Hamley, 1975). Therefore, when gillnets are used for stock assessment, usually several panels of different mesh sizes are incorporated into the net to catch fish over the full range of fish sizes.
To obtain less-biased size or age composition, catch efficiency of the sampling gear should be similar over a wide range of sizes in the target fish population. Catch efficiency pooled over all mesh sizes in a geometric series of meshes is expected to be flat over a wide range of fish size, whereas efficiency pooled over all mesh sizes in an arithmetic series of mesh sizes increases linearly with fish size (Jensen, 1986). Some authors have designed research gillnets consisting of a geometric series for which the pooled efficiency was expected to be flat over a wide size range of target fish species (Ishida et al., 1966; Takagi, 1975; Jensen, 1990).
Recently, large biases were observed in the size composition of salmonids caught using the Nordic multimesh gillnet, which consists of 12 panels comprising a geometric series of mesh sizes (Finstad et al., 2000; Finstad and Berg, 2004). The pooled catch efficiency of this gillnet design increased with body size for some fresh-water fish (Kurkilahti and Rask, 1996; Kurkilahti et al., 1998; Finstad et al., 2000), indicating a positive relationship between mesh size and the gillnet catch efficiency. Maximum height of the catch efficiency curve can also increase with gillnet mesh size (reviewed by Hamley, 1975), suggesting that the pooled efficiency of a research gillnet consisting of a geometric mesh series could increase with fish size, rather than remaining flat, as originally assumed.
Our objectives in this study were (i) to test whether chum salmon (Oncorhynchus keta) catch efficiency in the gillnet increases with mesh size, (ii) to describe the pooled catch efficiency curve of the "non-selective salmon research gillnet" (Takagi, 1975) consisting of ten panels of a geometric series of mesh sizes, and (iii) to assess the bias in chum salmon size composition caught by the salmon research gillnet. Japanese research vessels have continued to survey salmon stocks in the Pacific Ocean and Bering Sea using the same salmon research gillnet since 1972 (Takagi, 1975, 1996). From 2002 to 2006, scientists of North Pacific Anadromous Fish Commission (NPAFC) member nations cooperatively surveyed salmon stocks in the Bering Sea under the Bering–Aleutian Salmon International Survey (BASIS) plan (NPAFC, 2001). In the BASIS plan, the standard salmon fishing gear was determined to be a trawl. Involvement in the BASIS plan provided us with an opportunity to compare catches of chum salmon from the research gillnet with catches from the research trawl at the same location, with the aim of assessing gillnet size selectivity.
|
Material and methods |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
Study area and fishing operations
We conducted 11 fishing operations using a salmon research gillnet mesh series on board the RV "Wakatake maru" (958 grt) and with a midwater trawl on board the RV "Kaiyo maru" (2942 grt) at four fishing stations in early July of 2002 and 2003 and at three stations in 2004 (Table 1). The fishing stations were set along 180° longitude in the international waters of the central Bering Sea (55°30'N, 56°30'N, 57°30'N, and 58°30'N). Each vessel completed its fishing operations at those stations within the same 5-d period each year.
|
The gillnet was suspended at the sea surface and was set at 16:00 in the afternoon and retrieved at 04:00 the following morning (local mean time = GMT + 12 h). The gillnet configuration consisted of a variable-mesh research gillnet representing a geometric mesh series of factor 1.14 (150 m by ca. 7-m panels each of 48, 55, 63, 72, 82, 93, 106, 121, 138, and 157 mm mesh; Takagi, 1975). To maintain the stretch of the research gillnet while fishing, additional panels of 115-mm mesh were attached at both ends of the research gillnet (total = 950 m). The trawl was towed at 5 knots in the surface layer from the sea surface to
50 m deep for 1 h during daylight. The trawl opening was 50 m in height and width, trawl length was 222 m, the maximum mesh size in the body was 26 m, and the codend liner was 12-mm mesh. Chum fork length (FL) from a maximum of 60 fish per trawl and from 60 chum salmon per gillnet mesh size was measured to the nearest millimetre after removal from the fishing gear (Table 1).
Gear selectivity models
We fitted gear selectivity models to length frequencies of chum salmon caught by each gear type. The length frequency distribution in 20-mm intervals by gear was weighted by the number of chum salmon caught in each fishing operation, then pooled over the total of 11 fishing operations conducted by trawl and gillnet. Because there was no catch in the 48-mm-mesh panels of the gillnet, we did not estimate gear selectivity for that size.
The number of chum salmon of each size class caught by each gear type was modelled as the product of fishing intensity of the gear, the selectivity curve of the gear on fish size, and the abundance of the size class (Millar and Fryer, 1999):
|
| (1) |
l the abundance of fish of length class l, and rg(l) the contact selection curve for gear g. The relative fishing intensity represents the probability that a fish of length l contacts gear g. Although relative fishing intensities are usually defined as the probability that a fish of a specific length contacts the gear and will sum to unity across gear types (Millar and Fryer, 1999), we define relative fishing intensity as relative to the fishing intensity of the 157-mm-mesh panels of the gillnet (restricting to unity) to avoid complications in estimation. Here, we define catch efficiency of gear as the product of the selection curve and relative fishing intensity:
|
| (2) |
The pooled efficiency of the research gillnet series was the sum of the catch efficiency of each mesh size of the gillnet series.
The gillnet selection curve is usually bell-shaped and assumes that selection depends only on the relative geometry of mesh size and fish size (Millar and Fryer, 1999). Here, we assumed the gillnet selection curve to be the probability function of normal, lognormal, or bi-normal distributions:
|
| (3) |
, Ra, a, 
, Rb, and b are constants. The normal probability function expresses a selectivity curve with the maximum selectivity at a particular size, and symmetrically decreases selectivity towards larger and smaller fish sizes. The lognormal probability function expresses a selectivity curve with a maximum selectivity at a particular size, steeply decreasing selectivity at smaller fish sizes, and gently decreasing selectivity at larger fish sizes. The bi-normal function has either two peaks in selectivity or a single wide peak over a range of sizes of the target species.
The selection curve for the trawl is usually a monotonic, non-decreasing function of fish length, with the upper asymptote of unity (Millar and Fryer, 1999). However, we thought that the trawl selection curve was possibly bell-shaped because large chum salmon could avoid the mouth of the net as a consequence of their fast swimming speed, as described by Beverton and Holt (1957, Section 8.1.2). Therefore, we assumed that the trawl selection curve could be either the uniform, probit, lognormal, or bi-normal probability function:
|
| (4) |
T, RT, RTa, Ta, T, RTb, and Tb are constants. The uniform function expresses a flat and even selection curve over the whole size range of the target fish species. The probit function expresses an asymptotically increasing selectivity curve. For the bell-shaped function, we selected the lognormal and bi-normal distribution functions because they are more flexible than the normal distribution function.
Parameter estimation and model selection
We used the maximum likelihood method, assuming the Poisson error to estimate parameter values in gear selectivity models. The log-likelihood function was
|
| (5) |
function was the analytical continuation of the factorial function for non-integer real numbers and, for integer x, (x + 1) = x! The maximum likelihood estimate of l was derived as |
| (6) |
To test the difference in fishing intensity among gillnet meshes, we compared the goodness-of-fit between a model with a common intensity fixed to unity for each gillnet mesh (i.e. pg for gillnets = 1, not estimated as parameters) and a model with different relative intensities (i.e. all pg were estimated as parameters). We used the Akaike Information Criterion (AIC) to compare goodness-of-fit and to select the best model fit to the pooled catch data:
|
| (7) |
We used the replication estimation of overdispersion (REP) to evaluate how well the expected catch, calculated using the selectivity model fitted to the combined-haul data, fitted the observed catches in individual hauls (Millar and Fryer, 1999; Millar et al., 2004). We tested the null hypothesis of no extra-Poisson variation using the Pearson 
2 statistic for model goodness-of-fit summed over an individual haul i:
|
| (8) |
If the null hypothesis was rejected, standard errors of estimated parameters should be multiplied by (REP)1/2. The REP was given by calculating the Pearson 2 statistic for model goodness-of-fit and dividing by its degrees of freedom (McCullagh and Nelder, 1989):
|
| (9) |
|
Results |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
The best-fit selectivity model to pooled catch data with the minimum AIC value included different values of relative fishing intensity among gillnet meshes, a bi-normal function of gillnet selection curve, and a probit function of trawl selection curve (Table 2, Model 22). The AIC value for this model (Model 22) was an improvement over a similar model in which the fishing intensity of each gillnet mesh was fixed to unity, whereas the selection curve functions were identical (Model 10). The AIC values for models including a fixed fishing intensity for every gillnet mesh (Models 1–12) were larger than the AIC values for models using the same shape of selection curve, but incorporating estimated relative fishing intensities (Models 13–24). This indicates that fishing intensities were different among gillnet meshes.
|
Fishing intensity of each mesh size, relative to fishing intensity of the 157-mm mesh, increased with gillnet mesh size (Figure 1). Fishing intensity of small mesh sizes was much lower than that of the 157-mm mesh. Relative fishing intensity of the 55–106-mm meshes was <0.1. The relative fishing intensity increased steeply in the 121–157-mm meshes. In addition, the REP was 0.956 for the best-fit model (Model 22). However, the probability of the null hypothesis of no extra-Poisson variation for model goodness-of-fit was too high to be rejected (p = 0.158). Therefore, we did not multiply the estimated standard errors of parameters by the REP value.
|
Estimated catch efficiency for each gillnet mesh size from the best-fit model (Model 22) was bell-shaped with a single peak, although Model 22 assumed a bi-normal function for the selection curve (Figure 2). The parameter estimates for Rb and
b were much larger than for Ra and a, indicating that the height of the second peak of the gillnet selection curve was much lower than the first, but that the tail of the curve covered a wide range of fish sizes. The maximum height of the efficiency curve, which was determined by estimating fishing intensity, increased with mesh size. Fish length at the peak of the efficiency curve was larger for larger mesh sizes. The pooled efficiency of the gillnet mesh size series increased with fish length. This is because of the increased fishing intensity of the gillnet at larger mesh sizes. In contrast, the estimated efficiency curve of the trawl was sigmoidal and flat over a wide range of fish lengths.
|
The size composition of chum salmon caught using the gillnet mesh series was relatively flat over a wide range of fish lengths (Figure 3a). In contrast, the size composition of the trawl catch was bimodal (Figure 3b). Size composition estimated as
l was multimodal, which expressed the bias-corrected size composition estimated from catch efficiencies of both gillnet and trawl (Figure 3c). Chum salmon size composition of the trawl catch was more similar to the estimated composition than the chum salmon in the gillnet catch because the estimated efficiency curve of the trawl was flat over a wide range of fish lengths (Figure 2).
|
|
Discussion |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
Relationship between relative catch efficiency and gillnet mesh size
In this study, we identified that the relative fishing intensity of the gillnet increased with mesh size. Some studies using direct methods with known fish size composition showed that the maximum height of the selection curve of gillnets often increase with mesh size (e.g. Hamley and Regier, 1973; Borgström, 1989; Jensen, 1995; Fujimori et al., 1996). However, Takagi (1975) and Jensen (1986) assumed a common maximum height of the selection curve for different mesh sizes to estimate gillnet selectivity. They concluded that the pooled efficiency of a gillnet with a geometric mesh-size series was expected to be flat over a wide range of fish sizes. Although we used an indirect method of comparing gillnet and trawl catches, we demonstrated that the pooled efficiency of a gillnet comprising a geometric mesh-size series could increase with fish length.
The gillnet is a passive gear and catch efficiency may depend heavily on fish behaviour. The association between fishing intensity and mesh size may be due to a correlation between fish size and encounter probability with the gillnet. Larger fish can swim faster so may have a greater encounter probability with the gillnet than smaller, slower fish (Rudstam et al., 1984). In addition, active piscivorous fish are likely to have an enhanced encounter probability (Finstad et al., 2000). Fine mesh sizes may be more visible to the fish, which decreases gear efficiency (Hartt, 1975). A continuous multimesh gillnet could lead fish along the net wall from the smaller meshes to gilling in the larger, less-visible meshes (Hartt, 1975; Hovgård and Lassen, 2000).
The diel vertical migration of fish could also influence their gillnet encounter probability. From previous fishing surveys and biotelemetry studies using ultrasonic transmitters or archival tags, the depth distribution of maturing chum salmon seems generally to be limited to the upper 10 m by night and to the upper 20 m and even more by day during summer (Machidori, 1966; Ogura and Ishida, 1995; Walker et al., 2000; Ishida et al., 2001; Azumaya and Ishida, 2005). In contrast, the depth distribution of immature chum salmon based on fishing depths appears to be 0–20-m depth by both day and night (Machidori, 1966). In this study, the gillnet fished at a depth of 0–7 m from late afternoon, through the night to early morning. The encounter probability of small immature chum salmon to the gillnet might be lower than that of large maturing chum salmon at night. In contrast, the encounter probability of immature and maturing chum salmon to the daylight trawl operations may be similar because the trawl opening was large enough to fish from the surface to 50 m and capable of fishing at the depths frequented by small and larger fish.
The trawl is an active gear and catch efficiency may depend heavily on fish swimming capability. The selectivity of otter trawls occurs mainly by herding by the sweeps (or bridles), when fish are swimming in front of the trawl mouth, and by mesh selection at the codend (He, 1993). Fish encountering a trawl are guided and herded by sweeps to the centre of the net mouth at slow swimming speeds, where they turn and swim forward near the net mouth at the same speed as the net. Eventually they become exhausted from swimming, and by keeping clear of the net wall are guided down the net body to the codend (Wardle, 1993). In the body of the net, some small fish swim through meshes and escape out of the net, but others are trapped in the codend. The selection curve for the trawl was sigmoidal and flat over a wide range of fish size, indicating that fish size may have less of an effect on trawlnet avoidance than was the case with the gillnet. The sustained swimming speed of Pacific salmon was up to 3 x total fish length per s (Webb, 1995), or 2.4 m s–1 for a fish of 800-mm body length, which was near the largest fish size we caught. Such a sustained swimming speed did not exceed the towing speed of the trawl (2.57 m s–1). Burst swimming speed (up to 10 x total fish length) exceeds the trawl towing speed, but can only be maintained for <15 s (Webb, 1995). There may be relatively few chum salmon capable of swimming away from the net mouth. However, for small chum salmon, the towing speed may be too fast for the fish to keep clear of the net wall where they might escape through the large mesh. Once in the codend, the 12-mm mesh is sufficiently small to retain small chum salmon. Therefore, although some small fish could escape the trawl through meshes in the trawl body, a wide size range of fish will be trapped in the codend.
Bias in size composition of chum salmon caught using the salmon research gillnet
Researchers usually expect that data obtained from research surveys are less biased than data obtained from commercial fishing. For a long time, the salmon research gillnet constructed with a geometric mesh series has been expected to sample fish with a small bias (e.g. Takagi, 1975). However, we found the pooled size composition of the research gillnet catch to be different from the estimated fish size composition in this study. Comparing the estimated pooled efficiency of the research gillnet, a 450-mm FL chum salmon (the approximate size of a fish having spent two winters at sea) was expected to be caught six times more effectively, and a 530-mm FL chum salmon (the approximate size of a fish having spent three winters at sea) was expected to be caught 11 times more effectively than a 350-mm FL chum salmon (the approximate size of a fish having spent one winter at sea). Recently, Finstad et al. (2000) and Finstad and Berg (2004) showed that the bimodal size composition of Arctic charr (Salvelinus alpinus) resulted from a strong bias in gillnet sampling. The size frequency distribution is often used to estimate somatic growth, mortality, age composition, or size-based VPA in fish stock assessments (Hilborn and Walters, 1992). Bias of these parameters may result in serious errors in stock assessment (Tanaka, 1989; Tyler et al., 1989).
The estimated parameters of gear selectivity were not affected by sampling and measurement errors in individual surveys or by differences in size composition between odd- and even-numbered years. In the survey area, Asian pink salmon abundance dominates in odd-numbered years (Ruggerone and Nielsen, 2004). The size composition of chum salmon in the ocean could be different between odd- and even-numbered years because scale growth, which reflects somatic growth, is negatively correlated with Asian pink and chum salmon abundance (Walker et al., 1998). However, overdispersion was not observed in the model-fitting, and REP was close to unity. Because the REP evaluates the size of the errors or the biases, this indicates that assumptions in the estimation, including a common catchability among surveys, were appropriate for our dataset.
Previous studies (Finstad et al., 2000; Finstad and Berg, 2004) have observed biased fish size composition in gillnet catches, and the gillnet selectivity parameters estimated here should be limited to chum salmon in the study area and season. The size composition of the population in the study area and season can be biased in terms of the whole population of chum salmon at other times of the year. Although chum salmon are distributed in the Bering Sea and elsewhere in the North Pacific Ocean in summer, smaller chum salmon are distributed to the south (Neave et al., 1976). Catch efficiency estimates in smaller size classes can be less precise because smaller chum salmon are less abundant in the gillnet and trawl catches. Active migration of maturing chum salmon might affect their vulnerability to a gillnet during the study season because at that time, maturing chum salmon are beginning to migrate to, or are en route to, natal rivers in their homing migration (Ogura and Ishida, 1995; Azumaya and Ishida, 2005; Tanaka et al., 2005). As gillnet selectivity and the precision of its estimates can be variable in space and time, our selectivity correction factors should be used only when fishing in areas or seasons when chum salmon of body size and maturity level similar to this study are abundant.
Correction of the bias in the size composition
We have shown that the size composition of fish caught using a research gillnet is strongly biased, but this bias can be corrected using known gear efficiency, or by intercalibrating the research gillnet with another fishing gear. By comparing gear efficiencies between a research gillnet and a trawl, we were able to estimate the pooled efficiency of the research gillnet and the size composition of fish using an indirect method. Recently, pelagic trawls have been used widely in research surveys for stock assessment of pelagic fish and salmon (e.g. NPAFC, 2001), because they are expected to be less selective than a research gillnet. However, pelagic trawls are limited by the capacity of the vessel to tow the trawl quickly to catch fast-swimming pelagic fish. Because research gillnets are easy to use, less expensive than trawls to manufacture, and a specific ship type (e.g. stern trawler) is not required, stock assessments based on research gillnet surveys are likely to continue in future. Therefore, when gillnets are used for stock assessments, we suggest that researchers enhance their results by estimating the bias in gillnet selectivity and its size-dependent catch efficiency by intercalibrating with other sampling gear.
|
Acknowledgements |
|---|
We thank the captains, officers, crews, and scientists on board the RV "Kaiyo maru" and RV "Wakatake maru" for their care in collecting data and samples. We also thank Trevor Branch of the University of Washington for valuable comments on an earlier version of the manuscript. The work was supported by the Promotion Programme for International Resources Surveys of the Fisheries Agency of Japan. Support for at-sea participation by the US co-author was provided by the Auke Bay Laboratory, National Marine Fisheries Service, NOAA Contract #50ABNF-1-00002.
|
References |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
-
Azumaya T., Ishida Y. Mechanism of body cavity temperature regulation of chum salmon (Oncorhynchus keta) during homing migration in the North Pacific Ocean. Fisheries Oceanography (2005) 14:81–96.[CrossRef][Web of Science]
Beverton R. J. H., Holt S. J. On the dynamics of exploited fish populations. (1957) Fishery Investigations, London, Series II, 19. 533.
Borgström R. Direct estimation of gill-net selectivity for roach (Rutilus rutilus (L.)) in a small lake. Fisheries Research (1989) 7:289–298.[CrossRef][Web of Science]
Finstad A. G., Berg O. K. Bimodal population size distributions and biased gillnet sampling. Canadian Journal of Fisheries and Aquatic Sciences (2004) 61:2151–2157.
Finstad A. G., Jansen P. A., Langeland A. Gillnet selectivity and size and age structure of an alpine Arctic char (Salvelinus alpinus) population. Canadian Journal of Fisheries and Aquatic Sciences (2000) 57:1718–1727.
Fujimori Y., Tokai T., Hiyama S., Matuda K. Selectivity and gear efficiency of trammel nets for kuruma prawn (Penaeus japonicus). Fisheries Research (1996) 26:113–124.[CrossRef][Web of Science]
Hamley J. M. Review of gillnet selectivity. Journal of the Fisheries Research Board of Canada (1975) 32:1943–1969.[Web of Science]
Hamley J. M. Sampling with gillnet. In: Guidelines for Sampling Fish in Inland Waters, pp. 37–53—Backiel T., Welcomme R. L., eds. (1980) Rome: FAO. 176. EIFAC Technical Paper, 33.
Hamley J. M., Regier H. A. Direct estimates of gillnet selectivity to walleye (Stizostedion vitreum vitreum). Journal of the Fisheries Research Board of Canada (1973) 30:817–830.[Web of Science]
Hartt A. C. Problems in sampling Pacific salmon at sea. International North Pacific Fisheries Commission Bulletin (1975) 32:165–231.
He P. Swimming speeds of marine fish in relation to fishing gears. ICES Marine Science Symposia (1993) 196:183–189.
Hilborn R., Walters C. J. Quantitative Fisheries Stock Assessment: Choice, Dynamics and Uncertainty. (1992) London: Chapman and Hall. 570.
Hovgård H., Lassen H. Manual on estimation of selectivity for gillnet and longline gears in abundance surveys. FAO Fisheries Technical Paper, 397 (2000) 84.
Ishida T., Ito J., Osako M. Design of the non-mesh-selective salmon gillnets for research use, and results of the experimental operations: preliminary report. Bulletin of the Hokkaido National Fisheries Research Institute (1966) 31:1–10. (in Japanese with English abstract).
Ishida Y., Yano A., Ban M., Ogura M. Vertical movement of a chum salmon Oncorhynchus keta in the western North Pacific Ocean as determined by the depth-recording archival tag. Fisheries Science (2001) 67:1030–1035.[CrossRef][Web of Science]
Jensen J. W. Gillnet selectivity and the efficiency of alternative combinations of mesh sizes for some freshwater fish. Journal of Fish Biology (1986) 28:637–646.[CrossRef][Web of Science]
Jensen J. W. Comparing fish catches taken with gill nets of different combinations of mesh sizes. Journal of Fish Biology (1990) 37:99–104.[CrossRef][Web of Science]
Jensen J. W. A direct estimate of gillnet selectivity for brown trout. Journal of Fish Biology (1995) 46:857–861.[CrossRef][Web of Science]
Kurkilahti M., Appelberg M., Bergstrand E., Enderlein O. An indirect estimate of bimodal gillnet selectivity of smelt. Journal of Fish Biology (1998) 52:243–254.[CrossRef][Web of Science]
Kurkilahti M., Rask M. A comparative study of the usefulness and catchability of multimesh gill nets and gill net series in sampling of perch (Perca fluviatilis L.) and roach (Rutilus rutilus L.). Fisheries Research (1996) 27:243–260.[CrossRef][Web of Science]
Machidori S. Vertical distribution of salmon (Genus Oncorhynchus) in the north-western Pacific. I. Bulletin of the Hokkaido National Fisheries Research Institute (1966) 31:11–17. (in Japanese with English abstract).
McCullagh P., Nelder J. A. Generalized Linear Models (1989) 2nd edn. London: Chapman and Hall. 511.
Millar R. B., Broadhurst M. K., Macbeth W. G. Modelling between-haul variability in the size selectivity of trawls. Fisheries Research (2004) 67:171–181.[CrossRef][Web of Science]
Millar R. B., Fryer R. J. Estimating the size-selection curves of towed gears, traps, nets and hooks. Reviews in Fish Biology and Fisheries (1999) 9:89–116.[CrossRef][Web of Science]
Neave F., Yonemori T., Bakkala R. G. Distribution and origin of chum salmon in offshore waters of the North Pacific Ocean. International North Pacific Fisheries Commission Bulletin (1976) 35:1–79.
NPAFC (North Pacific Anadromous Fish Commission). Plan for NPAFC Bering–Aleutian Salmon International Survey (BASIS) 2002–2006. (2001) Vancouver: NPAFC. 27. NPAFC Document 579, Rev. 2.
Ogura M., Ishida Y. Homing behavior and vertical movements of four species of Pacific salmon (Oncorhynchus spp.) in the central Bering Sea. Canadian Journal of Fisheries and Aquatic Sciences (1995) 52:532–540.
Rudstam L. G., Magnuson J. J., Tonn W. M. Size selectivity of passive fishing gear: a correction for encounter probability applied to gill nets. Canadian Journal of Fisheries and Aquatic Sciences (1984) 41:1252–1255.
Ruggerone G. T., Nielsen J. L. Evidence for competitive dominance of pink salmon (Oncorhynchus gorbuscha) over other salmonids in the North Pacific Ocean. Reviews in Fish Biology and Fisheries (2004) 14:371–390.[CrossRef][Web of Science]
Takagi K. A non-selective salmon gillnet for research operations. International North Pacific Fisheries Commission Bulletin (1975) 32:13–41.
Takagi K. A study of the mesh selectivity of salmon research gillnets. Bulletin of the National Research Institute of Far Seas Fisheries (1996) 33:17–122. (in Japanese with English abstract).
Tanaka H., Naito Y., Davis N. D., Urawa S., Ueda H., Fukuwaka M. First record of the at-sea swimming speed of a Pacific salmon during its oceanic migration. Marine Ecology Progress Series (2005) 291:307–312.[CrossRef][Web of Science]
Tanaka S. Influence of errors in parameter estimates on stock assessment. In: Effects of Ocean Variability on Recruitment and an Evaluation of Parameters used in Stock Assessment Models, pp. 37–42—Beamish R. J., McFarlane G. A., eds. (1989) 379. Canadian Special Publication of Fisheries and Aquatic Sciences, 108.
Tyler A. V., Beamish R. J., McFarlane G. A. Implications of age determination errors to yield estimates. In: Effects of Ocean Variability on Recruitment and an Evaluation of Parameters used in Stock Assessment Models, pp. 27–35—Beamish R. J., McFarlane G. A., eds. (1989) 379. Canadian Special Publication of Fisheries and Aquatic Sciences, 108.
Walker R. V., Myers K. W., Davis N. D., Aydin K. Y., Friedland K. D., Carlson H. R., Boehlert G. W., et al. Diurnal variation in thermal environment experienced by salmonids in the North Pacific as indicated by data storage tags. Fisheries Oceanography (2000) 9:171–186.[CrossRef][Web of Science]
Walker R. V., Myers K. M., Ito S. Growth studies from 1956–95 collections of pink and chum scales in the central North Pacific Ocean. North Pacific Anadromous Fish Commission Bulletin (1998) 1:54–65.
Wardle C. S. Fish behaviour and fishing gear. In: Behaviour of Teleost Fishes, pp. 609–643—Pitcher T. J., ed. (1993) 2nd edn. London: Chapman and Hall. 743.
Webb P. W. Locomotion. In: Physiological Ecology of Pacific Salmon, pp. 69–99—Groot C., Margolis L., Clarke W. C., eds. (1995) Vancouver: UBC Press. 510.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||