Estimating mortality from times-at-large: testing accuracy and precision using simulated single tag-recovery data

Abstract

Abstract McGarvey, R., Matthews, J. M., and Feenstra, J. F. 2009. Estimating mortality from times-at-large: testing accuracy and precision using simulated single tag-recovery data. – ICES Journal of Marine Science, 66: 573–581. Single tag-recovery data are available for many commercial and recreational fish stocks. However, the use of tag-recovery data has been limited by the unknown proportion of tagged recaptured animals that fishers report back to researchers. Tag underreporting causes underestimation bias in commonly applied Petersen-based estimators of fraction harvested. Using simulated data, we tested the claim that estimates of total instantaneous mortality rate (Z) based on mean time-at-large are unbiased by tag non-reporting, and by short-term (type 1) tag shedding and tag-induced mortality. The Chapman time-at-large estimator produced unbiased estimates of Z. Non-reporting and short-term tag losses produced no detectable bias in estimates of Z. Standard errors of the Chapman Z-estimate decreased with , as predicted. Non-constant-Z scenarios of fishery closure and seasonal cycling gave time-at-large Z-estimates that closely approximated the true population time averages when tag releases were spread evenly across time. Ongoing (type 2) tag shedding and tag-induced mortality biased Z-estimates upwards by the respective ongoing loss rates. When tag-recovery experiments are cut short, accurate estimates of Z are provided by Deemer and Votaw's ‘truncated’ estimator, which had wide standard errors for short experimental length, precision becoming acceptable for tag-recovery experiments run longer than approximately two or three times Z−1.