Evaluating the Performance of Methods Used to Estimate Growth Parameters from Subsampled Age Data

Abstract

Due to time and resource limitations, the growth of many fish species is evaluated with age and length data in a length‐stratified subsample of a sample from a population. Some fisheries professionals draw inferences about growth using age and length data from only those fish that were actually aged (aged‐only method). Alternatively, growth may be evaluated by employing a method of incorporating the length‐stratified aged subsample with information from the entire sample of fish collected, such as the weighted‐mean and assigned‐age methods, which are commonly used in inland fisheries management. More recently, two prominent alternatives to those methods have emerged: the reweighting and empirical proportion methods. In this study, we used observed data from three freshwater fish species with different life history traits (short‐ to long‐lived) to inform a simulation model comparing estimated von Bertalanffy growth model parameters and predicted mean lengths at age calculated using the five aforementioned methods across a range of sample sizes per length‐group and a range of CVs. As expected, our results demonstrated that the aged‐only method is unreliable and biased when estimating von Bertalanffy parameters and mean lengths at age. The weighted‐mean method performed better than the aged‐only method but was not among the top‐performing methods. The assigned‐age, reweighting, and empirical proportion methods all performed well and produced similar estimates, although the empirical proportion method generally resulted in slightly more precise and less biased estimates. We recommend that fisheries professionals discontinue their use of the aged‐only and weighted‐mean methods and instead use the assigned‐age, reweighting, or empirical proportion method.

An assessment of sampling approaches for estimating growth from fishery-dependent biological samples

“…This may be due to the added logistical complexity of setting up and operating such a sampling approach or the need to understand population-level length composition in order to inform a POS sampling approach. Such length composition data are also needed to apply a weighted ALK which may correct for bias from FOS (Kimura, 1977;Goodyear, 2019;Lusk et al, 2021). Our models assumed that samples were available for the upper and lower size bins for FOS, but in reality, the largest and smallest individuals are not regularly encountered in many fisheries, and significant sampling effort would need to be allotted to obtain them.…”

“…FOS sets a predetermined number of samples for each length bin (Chang et al, 2019), a strategy which has been shown to lead to overestimates of mean length-at-age and L ∞ (Goodyear, 1995;Chih, 2009a;Coggins Jr et al, 2013;Goodyear, 2019). Biased growth parameter estimates associated with FOS may be corrected by applying an age-length key (ALK; Lusk et al, 2021;Goodyear, 2019). Problematically, ALKs require informed assumptions regarding the numbers at length in a fish population, which are likely lacking for data-poor fisheries in need of growth studies.…”