Estimates of the tag loss rates for single and double tagged toothfish (Dissostichus mawsoni) fishery in the Ross Sea

The rate at which tags are lost from tagged toothfish is an important parameter in modelling of the Antarctic toothfish (Dissostichus mawsoni) populations in Subarea 88.1 and 88.2. The toothfish stocks in these areas have been assessed using data from tag-release and recapture experiments within a CASAL integrated stock assessment model, using tag loss rates derived by Dunn et al. (2005). We update their estimates and calculate tag loss rates for both single and double tagged fish for use within the CASAL stock assessment models from the available data.

Revised estimates of the rate at which individual tags are lost from tagged toothfish in Subareas 88.1 and 88.2 from a sample of 969 double tagged and subsequently recaptured fish suggested that the loss rate was about 3.5% (95% C.I.s 0.020–0.054) of individual tags were lost almost immediately, and then there was an ongoing rate of about 0.039 (95% C.I.s 0.027–0.052) tags per year. For double tagged fish this corresponds to 99.5% of double tagged fish having at least one tag remaining after one year at liberty; 98.9% after two years at liberty; declining to 94.6% after six years; and to 88.4% after ten years.

Comparison of the different loss rate models suggested that there was evidence of immediate failure of tags (α = 0.035) and an ongoing constant rate of failure (λ = 0.039 y-¹), but no evidence of a change in the failure rate over time. It is plausible that there could be a catastrophic failure of tags or some other systematic change in the tag loss rate after some long period at liberty, but there was no evidence of such failure in these data for periods of up to six years at liberty.

The loss rate for double tags had been incorrectly derived and applied in the assessment models of Dunn & Hanchet (2009a, 2009b), with the loss rates slightly over-estimated for double tagged fish in the first four years and under-estimated after that. However, the tag loss rate and the double tag approximation rates calculated in this study suggest that the change in value of the tag loss rate parameter combined with the incorrect double tag model had very little impact on the assessment estimates of biomass in the assessment models.

The equivalent tag loss rate that can be used to provide a close approximation of the true tag loss rate in the Subarea 88.1 and 88.2 assessment models is either λ = 0.0071 y-¹ (where we exclude recapture events that occur after four years) or λ = 0.0084 y-¹ (where we exclude recapture events that occur after six years). Simulations showed that the impact on the assessment of ignoring tag recapture data after a six year period was to introduce negligible bias of less than 0.5% with less than 1% change in the overall estimated variance (mean squared error).  Similarly, a simulation experiment that excluded earlier years of release data from the assessment models  also suggested that the removal of early data had little impact on the assessment models. Estimates of bias were negligible (<0.5%) and the increased variability (as measured by mean squared error) was less than 3% even if all tag release and recapture data before 2005 were removed.