Precision of the age-length increments of three cyprinids: effects of fish number and sub-sampling strategy

The effects of number of fish that are aged and scale sub-sampling strategies on the precision of estimates of mean age-length increments from populations of Rutilus rutilus, Leuciscus leuciscus and Leuciscus cephalus were tested. Analyses used data derived from river fish communities in eastern England, U.K.. Regarding the number of fishes analysed in each age group, for each species and mean fork-length increment at age, significant relationships were detected between sample size (n) and the coefficient of variation of the mean (Z) and mean length increment x? and measured variance (s2). This enabled calculation of the number of scales for producing a mean length increment at age according to n=ax?b-2Zx?-2. Outputs indicated that the number of scales requiring ageing increased substantially as precision increased, but with little variation between species per age category. Ageing between seven and 12 scales per age group would thus provide estimates at 10% precision. As the ages of fishes are not known in advance of scale ageing, the effect of scale sub-sampling regime on precision was also tested using randomized strategies of 10 fish per 5 mm, five per 5 mm, three per 5 mm, 10 per 10 mm, five per 10 mm and three per 10 mm. These were applied to the datasets and the consequences of their reduction in the number of scales for precision were determined using Z=a0 center dot 5x?b/2-1n-0 center dot 5. When compared to no sub-sampling, three per 10 mm always significantly reduced data precision, whereas 10 per 5 mm never significantly reduced precision. These outputs can thus be applied to the design of fish sampling protocols where age and growth estimates are required, with the randomized sub-sampling likely to be the most useful strategy. (C) 2014 The Fisheries Society of the British Isles