688 



Fishery Bulletin 91(4), 1993 



20 



3 



o 



15 



3 



o 



9 2 



89/90 



'85/86 '87/88 '89/90 



SELECTIVITY TRANSITION 



Figure 9 



Residual sum of squares resulting from applications of 

 CAGEAN with changes in selectivity. The CAGEAN program 

 was applied to the simulated widow rockfish data in which 

 selectivity shifted to older ages and fishing mortality was 

 constant (Table 2A). In the upper panel the selectivity coeffi- 

 cients were allowed to vary abruptly between adjacent years, 

 thereby dividing the data into two periods of constant selec- 

 tivity. The minimum in the residual sum of squares corre- 

 sponds to the true change in selectivity that occurred between 

 1983 and 1984. In the lower panel, selectivity was constant 

 for the first three years but was allowed to vary between 

 adjacent years in the remaining period. The zero in the re- 

 sidual sum of squares corresponds to the true change in se- 

 lectivity that occurred between 1983 and 1984. In these 

 analyses the age at 100% selection was fixed at age 7 for the 

 first selectivity period, at age 8 for the second period, and at 

 age 9 for the third. 



situation is analogous to an application of two-way 

 analysis of variance (ANOVA). In fitting a two-way 

 ANOVA model, one should test for a significant inter- 

 action term before drawing inferences about the main 

 effects. By the same logic, in fitting a catch-at-age 

 model, one should test for changes in selectivity before 

 concluding that stock size has been increasing or 

 decreasing. 



I have no real evidence of changes from year to year 

 in the selectivity for widow rockfish off the coasts of 

 Washington, Oregon, and California. However, the work 

 described in this paper demonstrates that an incorrect 



assumption of constant selectivity can seriously dis- 

 tort an assessment of widow rockfish stock size. Fur- 

 thermore, there is at least one reason to suspect that 

 selectivity for widow rockfish has varied through time. 

 During the early years of the directed fishery for widow 

 rockfish, vessels targeted schools of fish using midwater 

 trawls. With the rapid expansion of the fishery, the 

 Pacific Fishery Management Council began imposing 

 increasingly restrictive limits on the amounts of widow 

 rockfish that could legally be landed from any single 

 fishing trip (Gunderson, 1984). One result of these "trip 

 limits" was a reduction in the landings by midwater 

 trawlers relative to the landings by bottom trawlers. 

 Midwater trawlers accounted for roughly 75% of the 

 widow rockfish landings in Oregon during 1984 through 

 1988, but they accounted for only 60% in 1990, and for 

 less than 50% in 1991. It seems quite probable that 

 the midwater trawls have different selection charac- 

 teristics than do bottom trawls, and that the shift from 

 a midwater fishery to a bottom fishery would cause 

 changes in selectivity. 



Any stock assessment model will have to make sim- 

 plifying assumptions to summarize succinctly the ma- 

 jor features of the data. However, in my view the as- 

 sumption of constant selectivity is an unnecessary 

 and misleading oversimplification, use of which can 

 result in catch quotas that are either needlessly con- 

 servative, resulting in immediate losses to the fishing 

 industry, or that are excessively liberal, producing 

 losses in recruitment and catches at a more distant 

 time. 



Acknowledgments 



I am grateful to staff at the Oregon Department of 

 Fish and Wildlife facility at Newport, Oregon, for an- 

 swering numerous questions about the fishery for 

 widow rockfish and the Stock Synthesis program. Also, 

 this paper benefited greatly from helpful suggestions 

 by Ronald Hardy, Linda Jones, John Shepherd, and 

 two anonymous referees. Funds for this research were 

 provided by the Oregon Trawl Commission, the 

 Fishermen's Marketing Association, the Oregon De- 

 partment of Fish and Wildlife, and the Agricultural 

 Experiment Station of Oregon State University. I ap- 

 preciate the support and encouragement of these 

 institutions. 



Literature cited 



Baker, R. J., and J. A. Nelder. 



1985. The GLIM System Release 3.77. Numerical 

 Algorithms Group Ltd, Oxford, 305 p. 



