Sampson Constant selectivity and stock assessment for Sebastes entomelas 



685 



TRUE CATCH-AT-AGE 



5 10 15 20 



AGE (YEARS) 



1.0 



0.5 - 



3 

 cc 

 u 

 [11 



0.0 



TUNED TO 

 PROPORTION AT-AGE 



/ 

 TRUE VALUES 



TUNED TO FISHING 

 MORTALITY 



'81 



'83 



'85 

 YEAR 



'87 



89 



Figure 4 



Trends in selectivity induce trends in recruitment. True 

 catches-at-age (upper panel I shift through time towards older 

 fish because of the changes in selectivity. Fishing mortality 

 was constant for the entire period. The Stock Synthesis pro- 

 gram attempted to mimic the changes in catch-at-age by gen- 

 erating estimates of recruitment that had a decreasing trend 

 (lower panel). 



earlier experiments with the simulated stock of widow 

 rockfish. With increasing selectivity, the bias in the 

 estimated average biomass in the final year was -86% 

 (Fig. 7) as opposed to the bias of -54% found earlier 

 (Table 3A; selectivity increasing, tuned to fishing mor- 

 tality, fishing mortality constant, and recruitment es- 

 timated). With decreasing selectivity, the estimated av- 

 erage biomass in the final year was 213% too high 

 (Fig. 8); in the earlier experiment the corresponding 

 estimate was only 74% too high (Table 3 A; selectivity 

 decreasing, tuned to fishing mortality, fishing mortal- 

 ity constant, and recruitment estimated). 



Discussion 



In the experiments described in this paper, the assess- 

 ment programs were unable to fit exactly the simu- 

 lated catch-at-age data because the catch model was 

 mis-specified. Selectivity was falsely assumed constant, 



Table 4 



Sensitivity analysis of the Stock Synthesis program. Bias in 

 the estimated average biomass in the final year when selec- 

 tivity varied randomly. 



Known 

 recruitment 



Estimated 

 recruitment 



MeanC7r) SD(%) Mean(7r) SD(%) 



Tuned to fishing mortality 



F increasing -1.4 0.8 -19.7 21.9 



F decreasing 11.3 4.4 -17.0 24.0 



F constant 7.8 3.5 -18.6 23.2 

 Tuned to proportion-at-age 



F increasing -0.1 0.8 -4.2 3.5 



F decreasing 13.3 4.6 -8.5 5.7 



F constant 8.9 3.8 -6.0 4.9 



and the biomass and abundance estimates were bi- 

 ased as a consequence. In any real application, not 

 only would the assessment program be ignorant of the 

 true model structure, but the program would also have 

 to contend with "noise" in the data due to measure- 



