414 



Fishery Bulletin 94(3). 1996 



exactly like the outside nets (see Andrew et al., 

 1991), it was excluded from our analyses. 



The codends used in the experiments mea- 

 sured 75 meshes long (3 m) and were con- 

 structed from 40-mm mesh netting (Fig. 1). 

 They comprised two panels. The anterior panel 

 was 100 meshes in circumference and con- 

 structed of 25 meshes of 48-ply twine attached 

 to 25 meshes of 35-ply twine. The posterior 

 panel was 200 meshes in circumference and con- 

 structed of 3-mm diameter braided twine. 



Experiment 1 — evaluation of codends 



Three codend designs were compared. The con- 

 trol codend was made entirely of diamond- 

 shaped meshes (Fig. 1A). The second codend 

 (termed the 85-mm-long codend) had a panel of 

 85-mm netting cut on the bar (7 bars x 11 bars) 

 inserted lengthwise into the top of the anterior 

 section (Fig. IB). The third codend (termed the 

 85-mm-wide codend) had the same panel in- 

 serted sideways into the top of the anterior sec- 

 tion (Fig. 1C). 



All three codends were compared against each 

 other in independant, paired trials. That is, in 

 separate tows, the 85-mm-long codend was compared 

 against the control; the 85-mm-wide codend was com- 

 pared against the control; and the 85-mm-long and 

 85-mm-wide codends were compared against each 

 other. The particular pair to be compared in a given 

 tow were placed on each outside net of the triple- 

 rigged gear. The position and order of each codend 

 was determined randomly (to eliminate any biases 

 between different nets) and the codends were used 

 in normal commercial tows of 90-min duration (with 

 a haulback delay of 10—15 s) on established prawn- 

 trawl grounds. Over six nights we completed a total 

 of 10 replicate tows of each paired comparison. Prior 

 to the trials and at the end of each night, we rigged 

 the outside nets with normal commercial codends and 

 performed tows to ensure that there were no differ- 

 ences in the fishing characteristics of each net. 



Experiment 2 — effects of haulback delay 



In this experiment the 85-mm-long codend from ex- 

 periment 1 was tested against the control for pos- 

 sible effects due to haulback delay. The codends were 

 interchanged between tows, and each night we com- 

 pleted two replicate 90-min tows that included 1) a 

 10-15 s haulback delay (measured as the duration 

 between slowing the vessel and engaging the winch) 

 and 2) no haulback delay. Over 4 nights we completed 

 a total of 8 replicate tows for each type of haulback. 



To ensure independence among tows, the location and 

 direction of each tow was randomly selected from the 

 available locations on established grounds northeast 

 of Yamba. The grounds trawled varied in depth and 

 ranged between 20 and 25 fm. 



Data collected from the experiments 



After each tow in each experiment, the codends were 

 emptied onto a partitioned tray. Prawns and all com- 

 mercially important species larger than the mini- 

 mum legal size (retained commercials) were sepa- 

 rated from the remaining bycatch. The remaining 

 bycatch (termed discarded bycatch) was then sorted. 

 This included individuals of commercially important 

 species that were smaller than the minimum legal 

 size (discarded commercials). Data collected from 

 each tow were the total weight of prawns and their 

 size (to the nearest 1-mm carapace length); the 

 weight of the discarded bycatch; the weights, num- 

 bers, and sizes (to the nearest 0.5 cm) of discarded 

 commercial species and retained commercial species; 

 and the numbers of noncommercial species in the 

 assemblage. All individuals were counted without 

 subsampling. Several commercially important spe- 

 cies were caught in sufficient quantities to allow 

 meaningful comparisons. These were eastern king 

 prawn, Penaeus plebejus, cuttlefish, Sepia sp., Octo- 

 pus, Octopus sp., red spot whiting, Sillago flindersi , 



