Accepting final counts from repeat readings of otoliths: should a common criterion apply to the age estimation of fish?

Abstract

Multiple readings of otoliths are often carried out to assess the repeatability and reliability of increment counts for estimating fish age. Various criteria have been used to assign or discard age estimates from repeated counts when discrepancies occur although the reasons for doing so are usually not stated or justified. Trends in relative frequencies (percentage disagreement, PD) and magnitudes (inter-read discrepancy, IRD) of otolith-count discrepancies were explored for 15 species of fish collected from a range of locations around Australia to explore generality in the best explanatory model(s) for otolith-count discrepancies and, hence, the most appropriate criterion for accepting or rejecting age estimates from multiple-count data. Increasing discrepancies with increasing age, according to a constant per-increment probability of error, was the best-approximating model for 9 of the 15 species for PD data but for only two species for IRD data. Our results indicated disproportionately higher rates of rejection of estimates from older age groups if exact agreement between repeated counts was required for age acceptance. Results varied with the reader, region and the method of otolith reading, indicating that multiple criteria for accepting or rejecting counts from multiple readings may be required among or even within species.