298 



Fishery Bulletin 101(2) 



in biomass indices, expressed as ratios. This was done for 

 all species and for each pair of consecutive surveys that 

 included one extreme year. 



Are there consistencies between data sets? 



Three types of consistency were sought in the data. First, 

 is the range of estimated trawl survey catchabilities plau- 

 sible? Second, is there any consistency, between trawl 

 survey series, in the years that are labeled as having high 

 or low catchability? Third, is there consistency between 

 the extreme years in the trawl survey data and the CPUE 

 indices in the assessment data? To address the latter ques- 

 tion, each person who provided CPUE data was asked, for 

 each series, which, if any, of the trawl survey data sets were 

 "comparable"' in that they related to similar areas, depths, 

 and seasons. For "comparability" it was not necessary that 

 the CPUE species be a target for the trawl survey. We were 

 interested in knowing whether the person thought that 

 the fact that the catchability seemed to be extreme for 

 many species in the trawl survey in some year would be 

 reasonable grounds to believe that this would affect their 

 CPUE index in a similar way (but this person was not told 

 which trawl survey years were considered extreme). For 

 each match that was found between a CPUE index and 

 a trawl survey extreme year we asked whether the two 

 were consistent: that is, whether high (or low) trawl survey 

 catchability corresponded to a positive (or negative) CPUE 

 residual. 



Results 



Are the assessment CVs the right size? 



Results were different for the two types of assessment data 

 ( Fig. 3 ). For those with CPUE indices, there was a tendency 

 for CVs to be too large: K'was negative for 21 of the 30 data 

 sets (this is significantly more than half, P=0.02) and was 

 less than -2 for 9 of them. In contrast, K was positive for 

 13 out of the 18 data sets with trawl survey indices (again, 

 significantly more than half, P=0.02) and was greater than 

 2 for 2 of them. Median CVs for data sets for which the CVs 

 were found to be significantly too large ranged from 0.3 to 

 0.5; where CVs were significantly too small the median CVs 

 were between 0.02 and 0.24. 



If a default CV is to be used for all CPUE series, the 

 best value lies between 0.15 and 0.2; values in this range 

 give approximately equal numbers of positive and negative 

 values of k (Table 3). The best default value for a trawl 

 survey annual variation CV appears to be about 0.2; this 

 gives approximately equal numbers of positive and nega- 

 tive values of Kr(Table 4). 



Can we detect years of extreme trawl survey 

 catchability? 



Our informal graphical procedure showed that, for some 

 trawl survey scries, the biomass indices for many species 

 fluctuate synchronously, which suggests annual variation 



y Residuals too small Residuals too large v 

 ^ CVs too large CVs too small ^ 



0.6- 



0.5- 



0.4- 



0.3- 



0.2- 



0.1- 



> 



c cc 



cccccc c 



CE 



c c c 



Cc 

 c 



CPUE 



-4 



-4-2024 



Residual statistic, kappa 



Figure 3 



The residual statistic, v, plotted against assumed CV for 

 each of the assessment data sets: those with CPUE indices 

 in the upper panel, those with trawl survey indices in the 

 lower panel. Each point represents one data .set; where dif- 

 ferent indices in a data set had different assumed CVs. the 

 median of these is plotted. 



in catchability. Two clear examples are shown in Figure 4: 

 for series 5, the biomass indices for many species follow the 

 same up-down-up pattern; for series 6, the opposite pat- 

 tern (down-up-down) is followed by many species (but not 



