Scandol et a\. Experiments to reduce bycatch in an estuanne squid-trawl fishery 



537 



were pooled and catches were transformed to natural 

 logarithms to stabilize variances (Underwood, 1997). 



A second analysis compared the diamond and 90- 

 CSMP with the conventional 41/50 codend in two ar- 

 eas (in estuarine waters off Patonga, and Flint and 

 Steel, see Fig. 1), to check consistency across fish- 

 ing-grounds. A balanced set of replicated hauls with 

 each type of gear was obtained. Again, the two BRDs 

 caused large and significant reductions in bycatch 

 compared with the conventional net. Although catches 

 of squid were smaller, they were not significantly so 

 and there was a large reduction in the ratio of weight 

 of bycatch to weight of squid caught. Strikingly, al- 

 though there was considerable variation from day to 

 day and between locations in the amount of bycatch, 

 there were no interactions with experimental gear 

 used. Therefore, reductions in bycatch were consistent 

 under the different conditions that are typical to the 

 fishery. 



Experiment 2 compared catches with the four codends 

 without BRDs, as in experiment 1, with catches with 

 the square-mesh codend. There were complex differ- 

 ences in the amount of bycatch caught with different 

 types of gear, despite many catches being smaller than 

 those in experiment 1. More bycatch was caught with 

 the 45/150 codend than with the 41/150 codend (i.e., 

 an effect of mesh size). This difference was, however, 

 detected only for the 150 codend. There was also an ef- 

 fect of circumference of the codends (41/100 caught more 

 bycatch than did 41/150), but only for the commercial 

 (41-mm mesh) codend. This finding explains the signifi- 



cant interaction between mesh-size and circumference 

 in the analysis in Table 2. 



Again, although there were differences in amounts 

 of bycatch, there were no significant or detected differ- 

 ences in catches of squid (Table 2, A and B); therefore 

 it was possible to achieve reductions in bycatch without 

 any notable effect on the commercial catch. 



During experiment 3 catches were generally small 

 and patchy. Data were very variable, but it was pos- 

 sible to complete five days of trawling with the 41/100 

 and square-mesh codends with or without their square- 

 mesh BRDs. There was a very marked reduction in 

 bycatch when a BRD was added to a codend, regardless 

 of whether there was a square-mesh panel in the codend 

 or not (Table 3A). Again, there were no influences on the 

 catches of squid, but the ratio of weight of bycatch to 

 weight of squid was smaller for both types of mesh when 

 a BRD was added (analysis of ratios in Table SB). 



Sizes of species in bycatch 



The secondary hypotheses of this study, that devices to 

 reduce bycatch would 2A) not cause disproportionate 

 bycatch of small or large fish; 2B) not cause differential 

 effects; and, 2C) not cause relatively larger catches of 

 particular species, were tested with multivariate sta- 

 tistical methods. 



Analyses of sizes of bycatch were uniform in their 

 outcome — there was no measurable difference on the 

 size composition of species caught for any of the devices 

 tested. As examples, two species offish were present in 



