Error patterns in age estimation of harp seals (Pagophilus groenlandicus): results from a transatlantic, image-based, blind-reading experiment using known-age teeth

Abstract

Abstract Frie, A. K., Fagerheim, K-A., Hammill, M. O., Kapel, F. O., Lockyer, C., Stenson, G. B., Rosing-Asvid, A., and Svetochev, V. 2011. Error patterns in age estimation of harp seals (Pagophilus groenlandicus): results from a transatlantic, image-based, blind-reading experiment using known-age teeth. – ICES Journal of Marine Science, 68: 1942–1953. Blind readings of known-age samples are the ultimate quality control method for age estimates based on hard tissues. Unfortunately, this is often not feasible for many species because of the scarcity of known-age samples. Based on a unique collection of known-age teeth of harp seals (age range: 1–18 years), ageing errors were evaluated in relation to true age, reader experience, sex, and tooth format (images vs. originals). Bias was estimated by linear models fitted to deviations from true age, and precision was estimated as their residual standard error. Image-based blind readings of 98 tooth sections by 14 readers, representing different levels of experience, generally showed high accuracy and precision up to a seal age of ∼8 years, followed by an increasingly negative bias and increased variance. Separate analyses were therefore conducted for young seals (1–7 years) and older seals. For young seals, moderate associations were found between reader experience and levels of bias, precision, and proportions of correct readings. For older seals, only precision levels showed a significant association with reader experience. Minor effects of sex and tooth format are unlikely to affect these main patterns. Observed errors, even for highly experienced readers, may affect important age-related parameters, emphasizing the importance of known-age calibration of the output from all readers.