Rules for quantifying otolith chemical variability help expose nursery population structure, site fidelity and multiple origins in a fished herring stock

Abstract Reliable information on the structure of fished populations can help align management units with the ecology of target species. One way to attain such information lies in accurately defining the scales at which groups of fish vary in key attributes – the chemical markers within otoliths providing well known examples. Here, motivated by current uncertainties around stock status and population structure within the Icelandic summer spawning (ISS) herring (Clupea harengus) stock, we use simulations to derive quantitative rules for assessing the spatial and temporal scales of otolith chemical variation in a Bayesian modelling setting. We then apply these rules to empirical otolith chemistry data from ISS herring sampled across their full distributional range, over three years, to tackle some open questions on nursery connectivity, fidelity and contributions to the fishery. Bayesian multivariate linear models (BMLMs) exposed differences in otolith elemental (Li, Mg, Mn, Zn, Sr, Ba) and stable isotopic (δ13C, δ18O) concentrations in nursery-resident age-1 and age-2 juveniles that generally scaled with distance among nurseries, and time between sampling events. Our simulation-based rules built confidence in the application of BMLMs to characterise ‘source’ populations within the juvenile dataset. These sources were subsequently treated as baseline samples in mixture models that revealed 1) strong evidence for nursery-site fidelity between age-1 and age-2, and 2) multiple nurseries as contributors to a fished age-3 population, irrespective of presumed nursery quality. While additional sampling is required to confirm these results and their relevance for management of herring stocks in Iceland and further afield, we note that the rules and models presented here can be easily adapted for a broader suite of fisheries applications where reliable methods for evaluating differences in multiple attributes among individuals, populations or stocks are needed.