Heales et a\: Effect of size of subsamples on estimates of catch composition and abundance. 



795 



catches had been sorted, only two abundance cat- 

 egories (from >10 to <50, and >50 per subsample) 

 had mean sampling error rates below 25%. 



For the "rare" species (<one per subsample), 

 the gradient of the mean sampling error curve 

 was close to constant (Fig 6). The 95% upper con- 

 fidence interval was over 100% until more than 

 40% of the catches had been sorted. Even when 

 90% of the catches had been sorted, the mean 

 sampling error was just below 10%, and the 95% 

 upper confidence interval remained above 25%. 



For the "abundant" species (five or more per 

 subsample), the mean sampling error curve 

 started just below 25% after 10% of catches had 

 been sorted, and fell below 10% when more than 

 40% of the catches had been sorted (Fig 6). The 

 95% confidence interval did not fall below 25% 

 until 50% of catches had been sorted. 



Discussion 



This study shows that a large subsample is 

 required to accurately represent the species 

 composition of a large multispecies catch from 



° s 

 «■§ 



^% 



3 

 O 

 Q 

 O 



100 -, 

 90 - 

 80 - 

 70 

 60 

 50 

 40 

 30 

 20 

 10 

 



<1 

 "Rare" 



2 3 4 

 "Common" 



7 8 9 

 "Abundant" 



JZL 



10 10+ 



Abundance category 

 (no. of individuals per 10 kg subsample) 



Figure 2 



The percentage frequency of occurrence of 2333 cases of (species x 

 trawl ) relative abundance for bycatch species recorded from 20 trawl 

 catches. The cases are grouped into 11 categories of abundance indices 

 based on the average number of a species recorded per 10-kg subsample 

 of catch. 



