440 



Fishery Bulletin 97(3), 1999 



had high deviance residuals (Fig. 7, see also Appen- 

 dix). Although all selectivity parameter estimates 

 were slightly greater in the composite-square-3 

 codend, the 95% confidence limits for these estimates 

 indicated that there were no statistically significant 

 differences between the two modified codends. 



Two-sample Kolmogorov-Smirnov tests comparing 

 the commercial size compositions of prawns retained 

 in each codend showed significant differences be- 

 tween the composite square-mesh and control 

 codends (Fig. 8 ). Both composite square-mesh designs 

 caught proportionally more larger prawns (<9/12 per 

 pound) and fewer smaller individuals (> 16/20 per 

 pound). 



Discussion 



The work described in this paper illustrated the ef- 

 fectiveness of a strategically located panel of square 

 mesh in codends for reducing catches of juvenile fish 

 (see also Robertson and Stewart, 1988; Fonteyne and 

 M'Rabet, 1992; Briggs, 1992; Broadhurst and 



Kennelly, 1996, 1997) and quantified, for the first 

 time in commercial penaeid prawn-trawls, the util- 

 ity of large panels of square mesh for improving the 

 size-selectivity of the targeted prawns. 



The composite square-mesh codends were equally 

 effective in excluding large quantities of small indi- 

 viduals of fusiform species, including sand trevally, 

 red mullet, school whiting, and southern sand flat- 

 head and there was no significant reduction in 

 weights of prawns or total retained bycatch (Fig. 4; 

 Table 1). Although the square mesh used through- 

 out the main sections of both codends may have been 

 large enough to permit some smaller fish (e.g. red 

 mullet) to escape, the differences in the sizes of many 

 fish retained between the composite square-mesh 

 codends and the control (Fig. 5) indicated that to 

 achieve the significant and quite large reductions in 

 catch, most fish probably escaped through the larger 

 85-mm mesh panel located in the tops of the poste- 

 rior sections of the codends. 



The escape of large numbers of these individuals 

 through the 85-mm square-mesh panel in both 

 codends may be attributed to the influences of water 



