Francis et al : Quantifying annual variation in catchabllity 



301 



and months for these series, it is unclear how much consis- 

 tency in catchability could be expected. 



In comparing the trawl survey and CPUE series, we 

 found only 12 matches, and the data were consistent at 

 8 of these i6T7c). This is not significantly different from 

 the value of 507( that we would expect if the data were 

 uncorrelated (P=0.39). Another place one might look for 

 consistency is between biomass trends for the same species 

 in different survey series. We made plots of biomass trends 

 for every instance where there were at least three survey 

 series with that species and at least three years in common. 

 A few of the 16 such plots showed strong consistency but it 

 was difficult to judge the significance of this because of the 

 possibility of obtaining agreement by chance. 



Discussion 



It is difficult to make inferences about catchability because 

 we cannot measure it directly. Instead, we must estimate it 

 indirectly with stock assessment models. These estimates 

 are compromised by the weakness of our models, which 



I I I I ml 1 — I I I 1 1 Ml 



0.01 01 



Estimated catchability 



Figure 6 



Estimates of trawl survey catchability by species. 

 Each point corresponds to an assessment data set; 

 estimates for different stocks of the same species 

 are plotted in the same horizontal line. 



provide only crude representations of population dynam- 

 ics (because the data to develop more complex models are 

 not available). With trawl survey data alone we cannot 

 estimate catchability; we can only detect years when 

 catchability was extreme for many species. Nevertheless, 

 the large data sets we have assembled do allow us to draw 

 some conclusions about New Zealand catchabilities. 



Are the assessment CVs the right size? 



Our results imply that, on average, the CVs used for CPUE 

 in New Zealand are too large, and those used for trawl 

 surveys are too small. For CPUE, the common (but usually 

 tacit) assumption that catchability varies from year to year 

 is supported. It is clear from Figure 3 and Table 3 that, had 

 the CPUE CVs been set equal to the observational error 

 (typically less than 0.1, Francis^), the associated stock 

 assessment residuals would have been much too large. 

 However, too much allowance for annual variability seems 

 to have been made: the CVs that are used in stock assess- 

 ments are, more often that not, too large. In other words, 

 the annual variability in CPUE catchability is not as large 

 as is implied by these CVs. Where the use of a default CV 

 is appropriate, it would seem that values around 0.15-0.2 

 would be better than the values of 0.3-0.35 that are cur- 

 rently used. This implies that annual variability in CPUE 

 catchability is less that 0.2. For trawl survey indices, the 



Francis, R. I. C. C. 1999. The impact of correlations in stan- 

 dardised CPUE indices. N.Z. Fish. Assess. Res. Doc. 99/42, 30 p. 

 National Institute of Water and Atmospheric Research, P.O. Box 

 14901, Wellington. New Zealand. 



