An examination of a tag-shedding assumption, with application to southern bluefin tuna

The technique of attaching two tags, rather than one tag, to each fish in a tagging experiment is a well-established means of measuring the extent to which tags are shed, by observing the proportion of recaptured tagged fish that have retained only one tag. This commonly requires an assumption that tag-shedding is a random process which is unaffected by the presence of the other tag on the same fish. In theory, this makes possible an estimate of the unobservable number of caught fish that have shed both tags. This paper examines biases that may follow if data are pooled (sometimes unwittingly) from experiments with different tag-shedding rates. It is found that the bias in tag-shedding estimates, due to pooling, can be lowered markedly by reducing either the rate of shedding and/or variability between experiments in the tagging technique. The proportions of fish recovered in different experiments can sometimes be compared to support (or otherwise) the existence of bias. Southern bluefin tuna (Thunnus maccoyii) data sub-sets are analysed to illustrate the potential seriousness of tagshedding bias.