240 



Fishery Bulletin 92(2). 1994 



been scaled by percentage sampled to represent each 

 catch and further scaled by stratum biomass to ap- 

 proximate the population. Samples in 1985 and 1986 

 were scaled to represent solely the catch, as survey 

 design was inadequate for biomass estimation. 



Length-frequency data are difficult to compare 

 statistically and, for the purposes of this study, have 

 not been attempted. However, to enable a general 

 comparison, a single distribution was constructed 

 combining length-frequency data from all years 

 weighted by the number of tows each year. This 

 distribution is plotted together with those from each 

 year separately. 



Mean size by sex was calculated separately for 

 three main regions of spawning within the survey 

 area (strata 1, 4; 10; 9, 11) as it was unlikely these 

 areas had been fished equally (see later 'Commer- 

 cial Fishery' section). The sample sizes used in cal- 

 culating the standard error were number of tows, 

 not number of fish. Orange roughy can associate in 

 size groups; between-tow variance was greater than 

 within-'tow variance. Variance is represented by ±2.0 

 standard errors for all years except 1986, when ±2.2 

 standard errors was arbitrarily used because there 

 were only 10 trawls. 



Reproduction Macroscopic staging of reproductive 

 condition followed Pankhurst et al. (1987): 



Stage 



Female 



Male 



Relative frequency of gonad stages was examined. 

 Analyses were based on the samples taken. They 

 were not scaled in any way, as there were no appar- 

 ent differences between the length frequencies of the 

 samples and the distribution of the total population. 

 Only data from females are presented, as their 

 macroscopic gonad stages can be determined more 

 accurately than those from males. 



Size at maturity was determined from samples 

 taken over the total survey area using a 'probit 

 analysis' approach (after Pearson and Hartley, 

 1976). It was assumed that length at maturity is 

 normally distributed in the population. The regres- 

 sion part of the analysis was repeated 10 times to 

 ensure convergence of the estimate. 2 A standard lin- 



2 Francis, C, MAF Fisheries, pers. commun. 1991 



ear regression analysis was carried out on results 

 to investigate trends over time by using the SAS sta- 

 tistical package (SAS, 1988). 



Feeding Data on frequency of occurrence were 

 available from all surveys. Frequency of occurrence 

 was defined as the number of stomachs in which a 

 food item occurs, expressed as a proportion of the 

 total number of stomachs containing food. Only 

 stomachs with part-full or full classifications, and 

 with fresh or partly digested contents, were included 

 in analyses. 



Commercial fishing data 



Data on the catch and position of each tow and the 

 start and finish times have been collected since 

 1980. However, catch and effort information is dif- 

 ficult to standardize and interpret for orange roughy. 

 Fish can be highly aggregated at various times of 

 the year, and 'windows' or escape panels in the net 

 are frequently used to reduce catch size and mini- 

 mise damage to nets. Fishing performance varies 

 with experience of skipper and crew, and technol- 

 ogy has advanced considerably in recent years (in 

 particular, development of Global Positioning Sys- 

 tem navigation, which enabled improved accuracy 

 when fishing pinnacles). Fishing logbooks often do 

 not have accurate information on length of tow on 

 the bottom. Fishing for orange roughy on the Chal- 

 lenger Plateau occurs on a variety of bottom terrain: 

 on flat bottom, in troughs and steep slope, and on 

 the tops and sides of pinnacles. In each case, the 

 effective fishing time and fishing technique differ 

 greatly, and they are almost separate types of fish- 

 eries. In order to gain an indication of trends in 

 catch rates, data were examined on the basis of 

 mean catch per tow for two size classes of vessel (20— 

 60 m, generally domestic fresh fish boats; and 60- 

 90 m, domestic factory trawlers). Catch per unit of 

 effort (CPUE) values were similar for both classes, 

 and so data here are combined. Monthly data were 

 amalgamated into two time periods: first, 'winter' 

 (June, July, August) which covers the spawning 

 period; second, 'out of season' (all other months). 

 This division represents two distinct phases of or- 

 ange roughy distribution, as well as differences in 

 the mode of fishing (Clark, 1992). The former period 

 is characterized by the formation of relatively stable, 

 dense aggregations of fish, whereas in the latter 

 period the orange roughy are more dispersed and 

 widely distributed (Clark, 1991a). Fishing in win- 

 ter generally involves shorter tows, often with 

 smaller nets, than does out-of-season fishing. 



