Study, Results, and Data

Last modified on Aug 15, 2022

Variability and bias in Chinook salmon scale ages were measured in a federal grant funded study. In this study, 10,200 Chinook salmon scales were sampled from five stocks, Stikine, Copper, Karluk, Nushagak, and Kuskokwim rivers, 1980-2015. In 2016-2019, images were scanned of these scales and a web application was developed to view scale images and for age data entry. Then, ten ADF&G personnel with a range of experience with reading Chinook salmon scales used the application to view scale images and enter age estimates. Both the original sampling data and each reader's age data were put into an Oracle database. With these age data, we measured variability, or how spread out or inconsistent age estimates were, using standard deviation (SD). We also measured bias, or the difference between the reader age and the "true" age, by calculating percent of scales with bias. We compared these variability and bias percentages among groups to form recommendations for improving age data.

Results and data


This study allowed us to review common practices among scale age readers and examine how we might improve these processes to achieve more consistent, less variable results to improve the quality of the data used in fisheries management.

  • Our study demonstrated that age estimates of the youngest and oldest fish tend to be most inconsistent, so we recommend that readers study growth patterns on scales of fish that mature young (after their first or second year at sea).
  • Our results suggest that it is important for readers learn to interpret the growth patterns on the scales before considering other auxiliary information (e.g., length) when estimating age.
  • We recommend that new and experienced readers study known-age and consensus-aged scales and that training include blind tests on scales.
  • We recommend that offices work to retain experienced readers. Experienced readers have less bias and they were better at identifying the rare ages that accurately reflect the age composition.
  • We also recommend documenting procedures used for scale age estimation. This is especially helpful when an experienced reader leaves before being replaced, and a new reader does not have a mentor.
  • Finally, we recommend that scale images be regularly exchanged across offices in order to facilitate consistent age estimation criteria, which is especially important where few known-age scales are available, such as in Alaska.


Chang, W. Y. 1982. A statistical method for evaluating the reproducibility of age determination. Canadian Journal of Fisheries and Aquatic Sciences 39(8): 1208-1210.

Cribari-Neto, F., and A. Zeileis. 2010. Beta Regression in R. Journal of Statistical Software 34(2): 1-24. external site link

Ospina, R., and S. L. Ferrari. 2010. Inflated beta distributions. Statistical Papers 51(1): 111.

Spanos, A. 1999. Probability theory and statistical inference: Econometric modeling with observational data. Cambridge University Press


This project received funding under award NA16NMF4270252 from NOAA Fisheries Service, in cooperation with the Saltonstall-Kennedy Grant Program. The statements, findings, conclusions, and recommendations are those of the author(s) and do not necessarily reflect the views of NOAA Fisheries.

ADF&G retains intellectual property rights to data collected by or for ADF&G. Any dissemination of the data must credit ADF&G as the source, with a disclaimer that exonerates the department for errors or deficiencies in reproduction, subsequent analysis, or interpretation.