Global evidence for non-random dynamics in fish recruitment

Abstract

Understanding what controls apparently random fluctuations in fish recruitment is a major challenge in fisheries science. Our current inability to anticipate recruitment failures has led to costly management actions and in some cases complete fishery collapse. Time series observations of fish recruitment reflect an interplay of underlying population processes, environmental forcing, measurement error, and process error. Given that the error component is often very strong, an important unresolved question is whether any non-random signal can be uncovered in the annual fluctuations of recruitment time series. Here, we address this fundamental question in an analysis of 569 fish populations from a global database of recruitment. Using a nonparametric time series analysis method, we find overwhelming evidence for non-random dynamics operating in the recruitment process. Unlike previous explorations of this topic, our approach does not require the specification of a stock-recruitment model, and it can be used on recruitment estimates derived from a wide-range of abundance estimation methods. The evidence for non-randomness is robust across a wide range of fish families and abundance estimation methods. We also find that the statistical support for non-randomness increases consistently with the number of observations of a fish stock. This result provides the encouraging news that with continued observations, and the appropriate covariates, we should ultimately be able to uncover the mechanistic drivers of fish recruitment.