493 



Abstract— Tagging experiments are 

 a useful tool in fisheries for estimat- 

 ing mortality rates and abundance 

 of fish. Unfortunately, nonreporting 

 of recovered tags is a common prob- 

 lem in commercial fisheries which, 

 if unaccounted for, can render these 

 estimates meaningless. Observers are 

 often employed to monitor a portion 

 of the catches as a means of estimat- 

 ing reporting rates. In our study, 

 observer data were incorporated into 

 an integrated model for multiyear 

 tagging and catch data to provide 

 joint estimates of mortality rates 

 (natural and fishing), abundance, 

 and reporting rates. Simulations were 

 used to explore model performance 

 under a range of scenarios (e.g., dif- 

 ferent parameter values, parameter 

 constraints, and numbers of release 

 and recapture years). Overall, results 

 indicated that all parameters can be 

 estimated with reasonable accuracy, 

 but that fishing mortality, reporting 

 rates, and abundance can be esti- 

 mated with much higher precision 

 than natural mortality. An example of 

 how the model can be applied to pro- 

 vide guidance on experimental design 

 for a large-scale tagging study is pre- 

 sented. Such guidance can contribute 

 to the successful and cost-effective 

 management of tagging programs for 

 commercial fisheries. 



Incorporating fishery observer data 

 into an integrated catch-at-age and 

 multiyear tagging model for 

 estimating mortality rates and abundance 



J. Paige Eveson (contact author)^ 

 Tom Polacheck^ 

 Geoff M. Laslett^ 



Email lor J. Paige Eveson: paigeevesoni'aicsiro.au 



' Commonwealth Scientific and Industrial Research Organization (CSIRO) 

 Marine and Atmospheric Research 

 Castray Esplanade 

 Hobart, Tasmania 7000, Australia 

 Postal address: GPO Box 1538 

 Hobart, Tasmania 7001, Australia 



^ CSIRO Mathematical and Information Sciences 

 Private Bag 33 

 Clayton South, Victoria 3169, Australia 



Manuscript submitted 10 December 

 2006 to the Scientific Editor's Offiee. 



Manuscript approved for publication 

 24 May 2007 by the Scientific Editor. 



Fish. Bull. 105:493-508 (2007). 



Tagging experiments are becom- 

 ing increasingly important in large 

 pelagic fisheries as a means of pro- 

 viding estimates of stock abundance 

 and fishing mortality rates that are 

 independent of catch-rate data (Pola- 

 check and Hearn, 2003). In Polacheck 

 et al. (2006), we developed a maxi- 

 mum likelihood model that combines 

 two traditional, but fundamentally 

 different, approaches for analyzing 

 tagging data with a single, terminal 

 recapture (note that we refer to this 

 as "tag-recapture" data, but the term 

 "tag-recovery" data is often used in 

 the literature). The first approach, 

 generally referred to as a Brownie 

 model (Brownie et al., 1985), uses tag- 

 recapture data from multiple years of 

 tagging to provide annual estimates of 

 mortality rates by comparing return 

 rates over time from the releases in 

 consecutive years. Only the numbers 

 of tag releases and returns by year 

 are required, not the number of ani- 

 mals examined for tags. The standard 

 Brownie model is formulated in terms 

 of rates of survival and tag recovery, 

 but can also be expressed in terms of 

 instantaneous rates of natural mor- 

 tality and exploitation (Pollock et al., 

 1991; Hoenig et al., 1998a). This latter 

 formulation is particularly useful in 

 fishery applications (e.g., Hampton, 



2000; Frusher and Hoenig, 2001; 

 Polacheck et al., 2006). The second 

 approach, known as a Petersen model 

 (e.g., Seber, 1982), uses data from a 

 single release event to provide an 

 estimate of population size at the 

 time of tagging based on the ratio of 

 the number of tags returned from a 

 sample of the population to the total 

 number of tags in the population. In 

 fishery applications, commercial catch 

 data usually constitute the sample 

 from which tags are returned. 



The model developed by Polacheck 

 et al. (2006) integrates catch data 

 with data from a multiyear tagging 

 experiment and. in essence, incor- 

 porates a Petersen estimator into a 

 Brownie model; we will refer to it as 

 the Brownie-Petersen (BP) model. 

 The BP model involves a likelihood 

 for the tag-recapture data and a like- 

 lihood for the catch data, which can 

 be jointly maximized to provide esti- 

 mates of natural mortality rates, fish- 

 ing mortality rates, and abundance. 

 The addition of catch data to the 

 traditional Brownie model not only 

 allows for the population size at the 

 time of first tagging to be estimated 

 but also improves the precision of the 

 mortality-rate estimates (Polacheck et 

 al., 2006). For readers familiar with 

 multiple-recapture tagging models, 



