334 
Fishery Bulletin 115(3) 
§ O O iJ S » !J 
g.2>c p™i — 
* 
db 
□□□□□ 
Combined 
iliiill 
Species (region) 
Figure 2 
Proportion of total age-composition sample size for the (A, C, E) combined species types and (B, 
D, G) individual species investigated under simple random sampling (SRS), proportional alloca- 
tion (PA), and fixed allocation (FA) sampling by using data from the NOAA Alaska Fisheries Sci- 
ence Center bottom trawl surveys for the Gulf of Alaska (GOA, 1984-2011), Aleutian Islands (AI, 
1980-2010), and Bering Sea (BS, 1982-2011). Sampling goal 4 (SG4) is used for illustration. Spe- 
cies acronyms are explained in Table 1. Note the different scale values on the y axis: the left side 
designates the proportion of total sample size combined across species types; the right side desig- 
nates values for individual species. 
approached a minimum value for the index uncertainty 
cases E0-E3 (Fig. 6). For each of the index uncertainty 
cases the minimum CV obtained, or baseline, was re- 
lated to the underlying magnitude of the CV in survey 
biomass. In uncertainty cases E0 and El (Fig. 6, A and 
B), the baseline CV of the final year’s total biomass 
was smallest for flatfish, intermediate for roundfish, 
and largest for rockfish, which followed the relative 
magnitude of the underlying uncertainty in the survey 
index data for these 3 species types. In all index un- 
certainty cases, CV reduction resulting from increased 
age-composition sample size (the maximum CV ob- 
tained compared to the baseline) was greatest for the 
roundfish group (text in top right comer of each plot 
in Fig. 6) and the smallest for the flatfish group, with 
rockfish intermediate. 
The sample size for which the CV in the total bio- 
mass of final year changed by less than 2.5% for all 
species types, which we define as the point of diminish- 
ing returns, and was larger in cases with smaller sur- 
vey index uncertainty than in cases with larger survey 
index uncertainty (vertical lines with arrows in Fig. 
6). For example, the sample size at the point of dimin- 
ishing returns for case E0 of 2500 samples was larger 
