318 



Fishery Bulletin 90(2), 1992 



lations (Paloheimo 1980, Collie and 

 Sissenwine 1983). These estimates do 

 not seem highly correlated. 



The unobserved change rate was a 

 random variable in the second test, so 

 its estimation error was not computed. 

 Error characteristics-of-abundance 

 estimates were extremely similar to 

 those of example one; apparently high 

 process variability does not adversely 

 affect estimation even in the presence 

 of sampling variance. 



The contracted seasonal recruitment 

 pattern (Fig. 3), conventionally inter- 

 preted as age-specific cohorts, was 

 used in the last two examples. Growth 

 parameters were the same as the two 

 previous examples and growth varia- 

 tion was moderate (cv 0.1). Sampling 

 efficiencies were also unchanged and 

 their variability set at that of example 

 1 (cv[q] = 0.2). The unobserved change 

 rate randomly varied five-fold (zt~ 

 U(0.05,0.25)). Catching was continu- 

 ous so each period's catch was as- 

 signed to midperiod for estimation. Overfishing was 

 simulated by rapidly increasing exploitation enough to 

 decrease stock abundance 36% during the four periods 

 of sampling (last four). The fishing mortality rates for 

 periods 6-19 were: 0.05, 0.1, 0.15, 0.2, 0.25, 0.8, 0.6, 



0.4, 0.5, 0.8, 0.6, 0.8, 1.0, and 1.2. Example 3 simulated 

 very low sampling levels and example 4, high levels. 

 It was of interest to find if abundance would be cor- 

 rectly estimated during overfishing under either sam- 

 pling condition. 



