466 



Fishery Bulletin 98(3) 



and southern calamari (which fishermen are legally per- 

 mitted to retain), all remaining species comprised those 

 that are considered commercially important bycatch (e.g. to 

 other fisheries) and that are normally discarded. 



Analysis of mesh sizes 



The mesh-size measurements collected from the trawl 

 bodies were analyzed for heteroscedasticity with Cochran's 

 test and a two-factor, balanced analysis of variance (ANOVA) 

 with trawl body and location of meshes considered fixed 

 and random factors, respectively. Significant differences 

 detected in these analyses were investigated by Tukey's 

 multiple comparisons of means test. 



Analysis of catch data 



Catch data for all replicates that had sufficient numbers of 

 each variable (i.e. 1 individual in at least 10 replicates) were 

 analyzed with paired ?-tests (P<0.05). Except for the weights 

 and numbers of blue swimmer crabs and small-toothed floun- 

 der (species that could not pass through the meshes in any 

 of the trawls ), all variables were analyzed with one-tailed 

 tests to test the hypothesis that the larger-mesh trawl bodies 

 retained fewer individuals than the control. Catches of blue 

 swimmer crabs and small-toothed flounder were compared 

 with two-tailed tests. To examine the relative effectiveness 

 of the spectra- 1-mm and momoi-1.7-mm trawl bodies, differ- 

 ences in catches (between each trawl and their respective 

 controls) for those variables that had data in all tows were 

 analyzed by using Cochran's test for homogeneity of variances 

 and a balanced, two-factor ANOVA. In these analyses, trawl- 

 t^pe and nights were considered fixed and I'andom factors, 

 respectively. Where there were sufficient data in>25 in each 

 trawl, pooled across all tows), size frequencies of commer- 

 cially or recreationally important fish (or both) were plotted 

 and compared with two-sample Kolmogorov-Smimov tests 

 (P=0.05i. 



Analysis of prawn sizes 



Size-frequencies of prawns retained in each of the three 

 trawls were combined across all tows. Using an estimated 

 split model (Millar and Walsh, 1992), we fitted logistic 

 curves to these data by maximum likelihood method (Pope 

 et al., 1975 ). Logistic cui-ve parameters, associated standard 

 errors and 95'r confidence limits were calculated for each 

 large-mesh trawl body. Model deviance values were deter- 

 mined for a goodness-of-fit hypothesis (i.e. to test //,,: that 

 the curves were logistic). Size categories of commercially 

 graded prawns from each trawl were plotted and analyzed 

 with two-sample Kolniogorov-Smirnov tests (P=0.05i. 



Results 



Analysis of mesh sizes 



There were significant differences detected in size of mesh 

 between the three trawl bodies (Table 1). Tukey's com- 



Table 1 



Summaries of F ratios from two-factor analysis of variance 

 to determine differences in size of mesh in the trawls and 

 at various locations and results of Tukey's comparison of 

 means test for the significant difference detected in size of 

 mesh between trawls (spectra-l-mm=momoi-1.7-mm>con- 

 trol). Data were treated in the raw form. '* = P<0.01. 



Treatment 



df 



Mesh size (mm) 



Trawls (T) 



Location of mesh in trawl (Li 



T ' L 



Residual 



2 2494.7** 



3 0.85 

 6 1-4 



346 



parison of means test showed no significant differences 

 between the spectra- 1-mm and momoi-1.7-mm trawl bodies 

 (mean mesh sizes ±SE of 52.43 ±0.08 mm and 52.96 ±0.09 

 mm. respectively). The mean size of mesh in the control 

 trawl was significantly less at 44.42 ±0.13 mm. Mesh size 

 was not significantly different among the various locations 

 (e.g. wing, footrope, head rope, and posterior body) exam- 

 ined in any of the trawl bodies. 



Analysis of catch data 



Compared with the control, the spectra-1-mm and momoi- 

 1.7-mm trawl bodies significantly reduced the numbers 

 of western king prawns caught (means reduced by 13.7'7f 

 and 15.6^'?-, respectively), without significantly reducing the 

 weights of prawns (although mean catch from the spectra- 

 1-mm was 6.3% lower than that from the control) (Table 

 2; Fig. 3, A and Bi. The spectra-1-mm and momoi-1.7-mm 

 trawl bodies also significantly reduced the weights of total 

 discarded bycatch (by 29.3'7f and 20.3% i; numbers of leath- 

 erjacket (by 32.5% and 23.7% ) and their weights (by 24.2% 

 and 19.6% ); numbers of sand trevally( by 56.2% and 40.4%) 

 and their weights (by 52.8% and 40.4% ); and the numbers 

 of southern sand flathead (by 59.8% and 40.2%) and red 

 mullet (by 67.2% and 59.3%) (Fig. 3, C, E, F, G, and I; 

 Table 2). The spectra-1-mm trawl body also significantly 

 reduced the weights of southern sand flathead (by 31.8%), 

 red mullet (by 57% ), and the numbers of southern calamari 

 (by 33.6% I (Fig. 3, H, J, and L; Table 2). ANOVA of the dif- 

 ferences in catches between the new trawl bodies and their 

 controls showed no significant interactions nor main effects 

 for any of the variables examined (Table 3i. 



Tu'o-sample Kolmogorov-Smirnov tests for comparing 

 the size-frequency distributions offish measured from the 

 control and new trawl bodies detected significant differ- 

 ences in the relative size compositions of southern sand 

 flathead retained by both new trawl bodies and in the rel- 

 ative size compositions of sand trevally retained by the 

 momoi-1.7-mm trawl body (the new trawl bodies retained 

 proportionally fewer small-size fish) (Fig. 4, A and B). 

 There were no other significant differences detected. 



