465 



Estimates of commercial longline selectivity 

 for Pacific halibut (Hippoglossus stenolepis) 

 from multiple marking experiments 



William G. Clark 



Stephen M. Kaimmer 



International Pacific Halibut Commission 



250 Ocean Teaching Building 



University of Wasfiington 



Seattle, Washington 98145 



E-mail address (for W G Clark): bill(a/iphc Washington edu 



The term "selectivity" refers to the 

 relationship between the size (or age) 

 of a fish and its vulnerability to a 

 given kind of fishing gear. A selectiv- 

 ity schedule, along with other param- 

 eters, is normally estimated in the 

 course of fitting a stock assessment 

 model, and the estimated schedule 

 can have a large effect on both the 

 estimate of present stock abundance 

 and the choice of an appropriate har- 

 vest rate. The form of the relationship 

 is usually not known and not well 

 determined by the data, and equally 

 good model fits can often be obtained 

 with different plausible specifications 

 of selectivity. Choosing among the 

 model fits and associated abundance 

 estimates in this situation is prob- 

 lematic (Sigler, 1999; Sullivan et al., 

 1999). 



The selectivities of different gears 

 can be compared by fishing the gears 

 side by side, but without knowing the 

 size composition of the stock being 

 fished, it is impossible to determine 

 the form of the selectivity functions. 

 Therefore, one has to make some as- 

 sumptions about them in order to 

 locate estimates (Millar and Fryer, 

 1999, and references therein). In 

 this case, too, equally good fits can 

 often be obtained with a variety of 

 assumed forms (Huse et al., 2000; 

 Woll et al., 2001); therefore the true 

 form cannot be determined by simple 

 fishing experiments. 



Mark-recapture data can yield di- 

 rect and reliable estimates of selec- 

 tivity because in this situation the 

 size composition of the fished stock is 

 known (Myers and Hoenig, 1997). In 



this note, we report estimates of the 

 commercial longline selectivity of Pa- 

 cific halibut (Hippoglossus stenolepis) 

 based on the large number of mark- 

 recapture experiments conducted 

 by the International Pacific Halibut 

 Commission (IPHC) in the 1960s, 

 1970s, and 1980s. A similar analy- 

 sis was done by Myhre (1969), but 

 he used data from only two experi- 

 ments; the present study uses data 

 from more than 100 experiments. 



Materials and methods 



Kaimmer (2000) described all IPHC 

 tag data for all varieties of external 

 tags in setline and trawl catches 

 dating back to 1925. We also used tag 

 data for all varieties of tags (except 

 the small strap type); however, our 

 data were for tags released during 

 setline catches only and our data 

 dated back to 1960, the first year of 

 recorded data in the computer release 

 and recovery IPHC database. The 

 total number of tags released was 

 over 100,000, of which more than 

 13,000 were recovered in the commer- 

 cial longline fishery. About half of the 

 releases were at systematically placed 

 setline survey stations that covered 

 a large part of an IPHC regulatory 

 area (Fig. 1). The other half were at 

 "spot" fishing locations, deliberately 

 chosen to produce good catches, either 

 for marking or for gathering data on 

 the performance of different gear 

 types. For our study, an experiment 

 was defined as all releases of a given 

 tag type in a given regulatory area 



in a given year during either survey 

 or spot fishing operations (not both), 

 where at least 10 fish were released. 

 Between 1960 and 1990 there were 

 131 such experiments. 



These data are not usable for esti- 

 mating exploitation rates or migra- 

 tion rates because of uncertainty 

 concerning things like recovery ef- 

 fort and reporting rates, but they 

 can be used to estimate commercial 

 selectivity. In the case of a single ex- 

 periment, a straightforward plot of 

 short-term recovery rate by length 

 at release will show how selectivity 

 changes with length. The absolute 

 recovery rates will depend on usually 

 unknown factors (tagging, fishing, 

 and natural mortality rates; tag loss 

 and reporting rates), but the relative 

 recovery rates should depend mainly 

 on selectivity (barring large varia- 

 tions in length with any of the un- 

 known factors). 



Myers and Hoenig (1997) showed 

 how data from many experiments 

 can be combined to obtain a single 

 set of selectivity estimates. To sum- 

 marize their derivation, let n^ , be the 

 recovery rate of fish of length / in 

 experiment (. This rate is treated as 

 the product of a length-specific com- 

 mercial selectivity Sj, which is the 

 same for all experiments, and an 

 experiment-specific recovery rate r, 

 that combines all the unknown fac- 

 tors mentioned above. Thus n^j=r^-Sj 

 and log;r, /=log7-^-i-logs,. This has the 

 form of a generalized linear model 

 with a log link function and a bino- 

 mial variance; therefore the point 

 and variance estimates can be ob- 

 tained in standard fashion. 



Some rule has to be chosen for 

 scaling the selectivities to make the 

 model determinate. The most com- 

 mon rule is to require that the maxi- 

 mum selectivity be 1.0, but that can 

 involve using a scaling factor that is 

 poorly determined by the data if the 

 maximum occurs in a length group 

 with few releases and recoveries. To 

 avoid this problem, the rule used in 



Manuscript submitted 10 March 2005 

 to the Scientific Editor's Office. 



Manuscript approved for publication 



27 September 2005 by the Scientific Editor. 



Fish. Bull. 104:465-467 (2006). 



