Identification of recruitment regime shifts with a hidden Markov stock-recruitment model

Abstract Stock-recruitment relationships (SRRs) may differ substantially among environmental regimes. We developed a methodology including a Hidden Markov Stock-recruitment Model (HMSM), the maximum likelihood approach and a model selection procedure to identify abrupt changes of stock-recruitment (SR) dynamics. This method allows us to objectively identify the unobserved regimes, estimate regime-specific parameters, and predict the transition probabilities among regimes. First, we used simulation to verify that our method could identify the correct number of regimes and estimate the model parameters well. Then, we applied the models to an Atlantic cod stock on the southern Grand Bank off Newfoundland, Canada. Results indicated that the HMSM assuming 2 regimes performed the best, and the cod stock shifted to a regime characterized with lower productivity and higher density dependence in late 1980s. Additionally, the estimated probability to return to the previous high-productivity regime was very low, suggesting the cod stock may remain at the low-productivity regime for a prolonged period. Overall, we consider the methodology proposed in this paper as a useful tool to model regime shifts of SRRs in fisheries stock assessment.