796 



Fishery Bulletin 101(4) 



150 



100 



50 



150 



100 



50 



10 smallest catches 



B 



10 largest catches 



<D^ ^(3^ 



o^^ 



rs^:^ <-ii 



r^ 'X.-' f^ 4> ^ 



^^ ^"^ 



Cumulative weight of catch sorted (l<g) 



Figure 3 



The cumulative numbers of bycatch species recorded as the 

 weight of catch analysed increased for (A) the 10 smallest 

 and (B) the 10 largest catches that were entirely sorted 



the Northern Prawn Fishery (NPF). As more of the catch 

 is sorted, more new species are encountered. On average, 

 50% of the catch weight needs to be sorted to record 80% of 

 the species in a single catch. Our data suggest that taking 

 subsamples of between 10% and 30% of catch weight can 

 result in highly variable percentages (from 31% to 88%) of 

 the total species in the catch. When estimating the rela- 

 tive abundance of individual species within a catch, sub- 

 sampling small percentages of catch weight (around 10%) 

 results in a sampling error around 80% for "rare" species, 

 and around 25% for "abundant" species. 



The sampling error (when estimating the relative abun- 

 dance of a species in a trawl) is a function of the total num- 

 ber of bycatch individuals caught in that trawl. For ex- 

 ample, 10 individuals of species "X" may occur in one trawl, 

 at a ratio of one in every 100 bycatch individuals. In the very 

 next trawl, 10 individuals of species "X" may occur at a ratio 

 of only one in every 1000 bycatch individuals because other 

 "abundant" species have swamped its occurrence ratio. 

 As a consequence, the sampling error for the same number 

 of individuals of species "X" varies greatly between trnwls. 



There are two sources of within-trawl variation to take 

 into account when calculating catch rates for individual 

 species. The first is due to changes in catchability at the 

 trawl-species interface {either on the sea floor or in the 

 water column). The second is due to on-deck subsampling 

 techniques. This study has been able to allocate percent- 

 ages of sampling error based on the occurrence rate of a 

 species of interest within individual catches. To do this, we 

 calculated the average occurrence ratios for the different 

 categories of relative abundance used in this study. For 

 example, "rare" species occurred at a rate of less than one 

 individual in every 389 bycatch individuals; "abundant" 

 species occurred at a rate of five or more individuals in 

 every 389 bycatch individuals. These ratios can now be ap- 

 plied to species of interest recorded in other trawl catches 

 and the average sampling errors can be calculated. 



Monitoring catch rates of a suite of "indicator" bycatch 

 species is a possible future option for measuring the health 

 of nontarget species in trawl fisheries such as the NPF. 

 Stobutzki et al. (2001) identified a suite of small NPF fish 

 species that arc more likely to be impacted by trawling. 



