Bertignac et al.: Estimates of exploitation rates for Thunnus alalunga from tagging data 



431 



In the absence of other informa- 

 tion, the tag-reporting rate was as- 

 sumed to be constant over time. It 

 is possible that reporting could have 

 varied over time, owing to variation 

 in the effectiveness of tag recovery 

 procedures or cooperation of fishing 

 fleets. Although it is not possible to 

 test conclusively such hypotheses 

 with the available data, we might 

 note that the time-related variabil- 

 ity in catchability estimated with 

 model 8 (Fig. 8) could equally have 

 been interpreted as variation in re- 

 porting rates, because catchability 

 and reporting rate are highly cor- 

 related in these models. 



Discussion 



U.S. baitboat Japan baitboat 



■T— I J"l"l I I ' I — I-T — I I I I I I I 



1234 1234 1234 



Troll Longline Other 



Figure 8 



Estimated relative catchability (seasonally averaged and normalized within fleet 

 categories! by multiyear time period, with 95'* confidence intervals. 1=1971 Q4 to 

 1974 Q4, 2=1975 Ql to 1979 Q4, 3=1980 Ql to 1984 Q4, 4=1985 Ql to 1992 Q4. 



12 



e 10- 



3 



The use of a tagged sample of the 

 population to infer characteristics 

 of the population in general is a 

 common technique in fisheries stock 

 assessment. For such inferences to 

 be valid, several assumptions, dis- 

 cussed in detail by Seber (1973), 

 need to hold. For North Pacific al- 

 bacore, the assumption of equal 

 probabilities of capture of tagged 

 and untagged fish is likely to be 

 critical. As noted earlier, both the 

 spatial distribution of the tagged 

 population in relation to recapture 

 effort, and the size distribution of the 

 tagged fish in relation to the size-se- 

 lective characteristics of the different 

 fishing gears are likely to result in 

 violation of this assumption. We have 

 developed a procedure to correct for these deficiencies 

 in an approximate way. The procedure uses time-at- 

 liberty information and knowledge of appropriate time 

 lags to provide some correction for nonuniform avail- 

 ability of the tagged population to the various fishing 

 fleets that is due to spatial and size effects. A more 

 elegant approach would be to develop a model that ex- 

 plicitly deals with the spatial and size structure of the 

 tagged population and with the spatial distribution and 

 size selectivity of fishing effort. Spatially disaggregated 

 tag models are now available (e.g. Kleiber and Hamp- 

 ton, 1994; Sibertetal., in press). An extension of such 

 a model to include size or age structure would pro- 

 vide an improved method for analyzing the North 

 Pacific albacore data. 



14 



T 



ii 



n 



T • - , T 



'•Ui 



1972 1974 1976 1978 1980 



1982 



Year 



1984 1986 1988 1990 1992 



Figure 9 



Time series and 95'* confidence intervals of estimated annual exploitation rates 

 of North Pacific albacore obtained for model 8 (see Table 4 for description). 



The natural mortality rate parameter estimated 

 from tagging data in this study is higher than some 

 previous estimates for albacore. We have shown that 

 M will be overestimated if the tag-reporting rate is 

 overestimated. Although tag reporting for the major 

 fleets is believed to have been high, it is possible that 

 unreported tag recaptures by the drift gillnet fleet, 

 which reported relatively few recaptures, have re- 

 sulted in a positive bias in our estimate of M. 



If the exploitation of tagged albacore is similar to 

 exploitation of the North Pacific population in gen- 

 eral, our results suggest that aggregate exploitation 

 rates declined from the mid-1970s to 1988, after 

 which some increase occurred. The results of our 

 preferred model (model 8) further suggest that an- 



