Abstract.- Models are formu- 

 lated for estimating tag-shedding rates 

 from recapture records for eight dou- 

 ble-tagging experiments with south- 

 ern bluefin tuna (Thunniis maccoyii) 

 in three Australian fishing areas. 

 These models incorporate either a 

 constant or time-varying rate of tag 

 shedding and allow for the possibil- 

 ity of immediate tag shedding. Likeli- 

 hood ratio tests are used to select the 

 most parsimonious model for each 

 data set. The probability of a tag be- 

 ing shed after 4 years at liberty (the 

 length of the most recent experi- 

 ments) varied from around 0.2 for 

 the 1983 and 1984 experiments to 

 0.5-0.7 for the experiments carried 

 out in the 1960s and 1970s. Although 

 a single, best-fitting model was se- 

 lected in all but one experiment and 

 despite the large total numbers of 

 recoveries, precise estimates of long- 

 term shedding rates could not be ob- 

 tained because there were relatively 

 few long-term data. This has signifi- 

 cant implications for analyses of tag- 

 return data that give heavy weight- 

 ing to long-term recaptures. 



Tag Shedding by Southern 

 Bluefin Tuna Thunnus maccoyii 



John Hampton 

 Geoffrey P. Kirkwood 



CSIRO Division of Fisheries, Marine Laboratories 

 GPO Box 1538, Hobart, Tasmania 7001, Australia 



Manuscript accepted 11 December 1989. 

 Fishery Bulletin. U.S. 88:313-321. 



Southern bluefin tuna Thunnus mac- 

 coyii were tagged throughout the 

 1960s in three areas of major Austra- 

 Han commercial fishing activity: off 

 the south coast of Western Australia 

 (WA), in the Great Australian Bight 

 off South Australia (SA), and off the 

 south coast of New South Wales 

 (NSW). During the 1970s, tagging 

 took place mainly off WA. The aim 

 of these experiments was to provide 

 information on movements, growth, 

 and mortality of the southern bluefin. 

 In 1983 and 1984, new experiments 

 were initiated off WA and SA in 

 order to answer more specific ques- 

 tions relating to fishery interactions, 

 yield-per-recruit, and schooling be- 

 havior, as well as to update informa- 

 tion obtained from the previous ex- 

 periments (Majkowski and Murphy 

 1983). Apart from the period prior 

 to 1963, almost all fish were double 

 tagged. 



Quantitative analyses of tagging 

 data that require an estimate of the 

 total number of recaptures of tagged 

 fish (e.g., fishery interactions, yield- 

 per-recruit, and mortality) need to 

 take account of tag loss due to shed- 

 ding. As reviewed by Wetherall (1982), 

 this problem can be approached in 

 two ways. If estimation of mortality 

 rates is the main objective of the 

 analysis, one can attempt to model 

 the entire process of removal of tags 

 from the population directly by incor- 

 porating into the model parameters 

 that account for all souj-ces of removal, 

 i.e., natural mortality, fishing mor- 

 tality, permanent emigration, non- 

 reporting of recaptured tags, and tag 

 shedding. Alternatively, tag-shed- 



ding rates may be estimated indepen- 

 dently from a double-tagging experi- 

 ment and the recovery data adjusted 

 accordingly before proceeding with 

 further analyses (other losses, e.g., 

 non-reporting, may also require in- 

 dependent estimation). A disadvan- 

 tage of the former approach is that, 

 even if simple functions are chosen 

 to describe mortality, emigration, 

 and tag shedding, the model will 

 inevitably be parameter-laden, and 

 therefore difficult to estimate. In any 

 case, such models will not necessar- 

 ily lend themselves to analyses of 

 interactions and yields-per-recruit, 

 where the mortality and emigration 

 processes are embedded in the recap- 

 ture statistics and do not necessarily 

 require explicit resolution (Majkow- 

 ski et al. 1984). 



The aim of this paper is to estimate 

 tag-shedding rates for a number of 

 southern bluefin tuna double-tagging 

 experiments, with a view to using 

 these estimates as the basis for cor- 

 recting for tag shedding in other 

 analyses. Hynd (1969) made a prelim- 

 inary estimate of the tag-shedding 

 rate, assumed constant, based on 

 recoveries up to 1968 only. Subse- 

 quently, Kirkwood (1981) developed 

 generalized tag-shedding models and 

 analyzed recoveries of southern blue- 

 fin tuna that were double tagged dur- 

 ing the period 1962-76 but excluded 

 all tagging that was contracted to 

 fishermen. In both cases, these anal- 

 yses used tag-recovery data that 

 were grouped by time of recovery. A 

 reevaluation of southern bluefin tuna 

 tag shedding is now appropriate 

 because, (1) for a number of present 



313 



