Sampson Constant selectivity and stock assessment for Sebastes entomelas 



681 



100 



UJ 40 



o 



CURVE B 



CURVE C 



SELECTION YEAR 



TREND CURVE 123456789 10 



INCREASING B 



C 



A 



DECREASING B 



C 



Figure 1 



Selectivity curves used in the simulations. To simulate the 

 effects of changing selectivity, the selectivity curve was either 

 shifted towards older fish through time (increasing selectiv- 

 ity) or towards younger fish (decreasing selectivity). 



both the Stock Synthesis and CAGEAN programs. Un- 

 like the other two assessment procedures, the multi- 

 plicative model estimates relative, rather than abso- 

 lute, abundance. To conduct the multiplicative 

 catch-at-age analyses, I used the GLIM statistical pro- 

 gram (Baker and Nelder, 1985) and assumed a log- 

 normal error structure. 



In the experiments with CAGEAN and the multipli- 

 cative catch-at-age model, I used a subset of the data 

 from the earlier experiments with the Stock Synthesis 

 program. Two sets of catch-at-age data were analyzed, 

 one from a population with selectivity shifting to older 

 ages (selectivity increasing) and fishing mortality con- 

 stant (Table 2A), the other from a population with 

 selectivity shifting to younger ages (selectivity decreas- 

 ing) and fishing mortality constant (Table 2B). I tuned 

 the CAGEAN program to the true fishing mortality 

 coefficients, and constrained the multiplicative catch- 

 at-age analysis to have a trend of zero in the annual 

 fishing mortality coefficients. The experiments here cor- 

 respond to the cases examined earlier in which the 

 Stock Synthesis program was tuned to fishing mortal- 

 ity and recruitment was estimated. 



Sensitivity of the Stock Synthesis program 

 when applied to data from a heavily 

 exploited stock 



To determine whether the results from the experiments 

 with a simulated stock of widow rockfish would apply 

 to fish stocks with different biological characteristics, 

 I generated two additional data sets, one from a popu- 

 lation with selectivity shifting to older ages (selectiv- 

 ity increasing), the other from a population with selec- 

 tivity shifting to younger ages (selectivity decreasing). 

 Both simulated populations, which suffered an instan- 

 taneous natural mortality rate of 0.30 per year and an 

 instantaneous fishing mortality rate of 0.60 per year, 

 had significantly fewer old animals compared to the 

 populations in the previous simulations. I analyzed 

 the two data sets with the Stock Synthesis program, 

 with tuning to the true fishing mortality coefficients,' 

 and with recruitment estimated. 



Results 



The assessment programs that I investigated all pro- 

 duce a wide variety of estimates, including selectivity 

 coefficients and matrices of abundance and catch by 

 age and year. Rather than evaluating bias for all esti- 

 mates, my analysis focussed on estimates of annual 

 stock biomass, numerical abundance, and recruitment. 

 Of special importance to a stock assessment scientist 

 or fishery manager is the bias in the estimate of aver- 

 age biomass for the final year of a data series. This 

 estimate is approximately the biomass estimate on 

 which the catch quota for the next year is based 2 . If 

 the estimate of average biomass in the final year is, 

 say, 20% too high, then the quota will be roughly 20% 

 too high; if the estimate is 10% too low, the quota will 

 also be about 10% too low. 



Selectivity of the Stock Synthesis program 

 when applied to data for widow rockfish 



The results of the experiments with the Stock Synthe- 

 sis program and the data for the simulated stock of 

 widow rockfish suggest that some of the assessment 

 results can be highly sensitive to slight trends in se- 

 lectivity. For example, when selectivity shifted towards 

 younger ages, the biomass estimate for the final year 

 of the series was 74% too high (Table 3 A; selectivity 

 decreasing, tuned to fishing mortality, fishing mortal- 



•The annual catch quota is derived from an estimate of the biomass 

 at the end of the previous year, plus an appropriate amount for the 

 new recruitment. In practice this differs little from the estimate of 

 average biomass in the final year. 



