Initial and ongoing tag loss rates were updated using double tagged Antarctic toothfish (Dissostichus mawsoni) released between 2005 and 2021 that were recaptured with at least one tag. Different tag loss models were explored, including assessing the effect of increasing the maximum time at liberty included in the analysis to those tags recaptured beyond 6 years at liberty, determining if ongoing tag loss was a function of time, determining the effect of initial tag loss on ongoing tag loss rates, and allowing tag loss rates to differ by size-class of tagged fish or season of tag release.
Models with an initial tag loss and a constant ongoing annual tag loss only were the most parsimonious based on likelihood ratio tests. Estimated rates of initial and ongoing tag loss from the 3555 double tagged fish from 2005 and subsequently recaptured suggested that about 5.7% (95% C.I.s 0.042–0.072) of individual tags were lost immediately, followed by an ongoing rate of 0.033 y-1 (95% C.I.s 0.027–0.038 y-1) tags per year. The loss rates were similar to previous estimates, with the initial tag loss rate slightly higher and the ongoing tag loss rate slightly lower. Ongoing tag loss failure rate did not change as a function of time but did increase with increasing time at liberty of the recaptured tags until up to 6 years at liberty, likely because incorrectly applied tags fall out and the tag loss parameters, as expressed in the probability formula, were highly correlated. Lower initial tag loss rates were estimated when fewer data (lower maximum time at liberty) was available.
No clear trend was apparent in initial or ongoing tag loss by size or season of tagged fish release for Antarctic toothfish in the Ross Sea region, suggesting that initial tag retention has remained consistent over time and that there was no evidence for a tagging cohort effect in the ongoing tag loss rates. A small increase in ongoing tag loss with a matching decrease initial tag loss was seen for fish larger than 130 cm or for release seasons after 2017 but this may be an artifact of correlations between the two loss rates. A key assumption of the tag loss rate estimation was that probability of a lost tag on an individual fish was independent of the loss of the other tag and that the two tags had identical tag loss rates.
We recommend using all years at liberty when estimating tag loss parameters, the tag loss rates estimates be updated periodically, and that the updated estimates be used in future stock assessments of the Ross Sea region. We further recommend alternate model formulations be investigated to evaluate the effect of the correlation between initial and ongoing tag loss rates and simulation studies carried out to investigate the potential effect of non-independence of tag loss with time. We suggest that collecting biological information about the tag placement and/or taking clear pictures of the tags in situ upon recapture is considered as part of the data collection of recaptured fish because this would allow incorporation of other information that may influence tag loss and help inform discussions on whether protocols for tagging of larger fish may need improvement.