Overlapping scales: Bias and variability in Alaska Chinook salmon age estimates

Alaska’s Chinook salmon (Oncorhynchus tshawytscha) populations have declined in size at age and age at maturity over the past 30 years, and productivity and adult abundance have decreased during the past decade. Chinook salmon scale age data are used to understand population demography, but methods for scale age estimation may have changed over time and may differ across the state. To begin addressing this, we examined variability and bias associated with age estimates from 10,200 scales, one scale per fish, from five Chinook salmon stocks (Stikine, Copper, Karluk, Nushagak, and Kuskokwim rivers), 1980–2015. This set of scales had an original assigned age, and, from this study, ages estimated by 10 participants (readers) with a range of experience. Variability was assessed using standard deviation (SD) of reader age estimations. We compared SD among stocks, marine ages, time periods (1980–1992, 1993–2005, and 2006–2015), and reader experience. Standard deviation increased by 0.014 for each 10.0 mm difference in fish length from expected length at age. Standard deviation differed among marine ages, with marine age-1 and age-5 scales having 155% and 188% of the variability of marine age-3 scales. Standard deviation did not differ among stocks or time periods. Reader bias was the percentage of scales over or underestimated relative to the median age assignment of 10 readers (the best approximation to known age). Reader bias differed among marine ages with the oldest ages tending to be overestimated and the youngest ages tending to be underestimated. Reader bias decreased with reader experience and was not different among stocks, time periods, or between familiar (those previously examined by a reader) and unfamiliar (new to the reader) stocks. Original ages were underestimated by 3% and overestimated by 9%. Bias in original ages did not differ among stocks, time periods, or marine ages. We provide recommendations, based on our results, for improving the quality of Chinook salmon scale age estimates in the future. •Variability in Chinook salmon scale age estimates decreases with greater marine age.•Scale readers tend to assign dominant age classes.•Scale patterns by age are comparable across stocks with the stream-type life history.•Stock-specific training is not essential for consistent age estimates among stocks.•Experience decreases bias and variability in Chinook salmon scale ages.