608 



Fishery Bulletin 89(4), 1991 



The development of a population dynamics model for 

 tagged southern bluefin that incorporates movement 

 between the NSW and SA/WA fisheries and offshore 

 movement into the Japanese longline fishery, as well 

 as natural mortality and fishery-specific catchability, 

 appears to overcome these difficulties. Although the 

 unconstrained-M version of the model could not pro- 

 vide useful parameter estimates because of high param- 

 eter correlation, accurate estimates were obtained by 

 assuming that the natural mortality rate was the same 

 in all fisheries, thus reducing the number of parameters 

 to be estimated by two. For a large highly-mobile apex 

 predator, this may not be an unreasonable assumption. 



However, this simplification of the model was not 

 achieved without a small, though significant, loss in 

 likelihood of the data. 



The substantial differences in some of the parameter 

 estimates produced by the unconstrained-M and con- 

 strained-M versions of the model simply mean that the 

 observed data could have been produced in a number 

 of alternative and similarly likely ways. It is unlikely 

 that M could vary amongst the three fisheries as much 

 as was suggested by the unconstrained-M estimates in 

 analysis A. It is noteworthy, however, that the estimate 

 of M 3 was similar for both analyses and also compared 

 well with the constrained-M estimates. 



Compliance with model assumptions is usually diffi- 

 cult to assess, and this case is no exception. Tag-shed- 

 ding was incorporated into both models used; therefore, 

 in theory, the tag-shedding assumption was satisfied. 

 In practice, the estimated shedding rates are very 

 uncertain, particularly for the older recaptures. It 

 would be desirable in the future to develop a method 

 whereby uncertainties in tag-shedding rates were re- 

 flected in the standard errors of the estimated mortal- 

 ity and movement parameters. This might be achieved 

 by the simultaneous estimation of shedding, mortality, 

 and movement rates from double-tagging data. 



There are no data available on which the calculation 

 of reporting rates could be based; all estimations were 

 therefore carried out for a range of reporting rates. 

 Fortunately, the M estimates were largely insensitive 

 to assumed reporting rates greater than about 0.7 in 

 the case of the constrained-M estimates. It is worth 



