Improved estimation of age composition by accounting for spatiotemporal variability in somatic growth

Abstract

Age composition is defined as the proportion of a fish population belonging to each age class and is an informative input to stock assessment models. Variations in somatic growth rates may lead to larger errors in age composition estimates. To reduce this source of error, we compared the performance of four methods for estimating age compositions of a simulated fish population: two methods based on age–length keys (ALK, pooled and annual) and two model-based approaches (generalized additive models (GAMs) and continuation ratio logits (CRLs)). CRL was the most robust and precise method, followed by annual ALKs, particularly when significant growth variability was present. We applied these methods to survey age subsample data for Pacific cod (Gadus macrocephalus) in the eastern Bering Sea, estimating age compositions that were then incorporated in its stock assessment model. The model that included age compositions estimated by CRL displayed the highest consistency with other data in the model. CRL approach has utility for estimating age compositions employed in stock assessment models, especially when substantial variation in somatic growth is present.