Search

The most recent assessment of NAFO Division 4TVn spring and fall spawning Atlantic herring stocks was conducted in 2018 using Virtual Population Analysis models (VPA) with time-varying catchability. The increase in magnitude of a retrospective pattern in the spawning stock biomass (SSB) of the fall spawning herring models suggested that the model failed to incorporate one or more non-stationary processes in the population dynamics of this stock or in the observation model relating indices of abundance to population abundance. The 2018 spring spawning herring VPA also showed an increase in residual patterns in the catch-per-unit-effort (CPUE) and acoustic indices. Abundances of major predators of herring have changed drastically in the sGSL in the last decades, potentially generating important changes in herring natural mortality. Failure to account for changes in natural mortality due to changes in predator-prey interactions can result in biased estimates of population parameters and vital rates in stock assessments. Hence, estimating natural mortality was another motivation to re-explore 4T herring population models. The objectives of this paper are to perform a comparison of the VPA used in the most recent stock assessment and a series of Statistical Catch-at-Age (SCA) models with different assumptions about temporal variation in population processes (i.e., natural mortality) and/or observation processes (i.e., catchability in the fixed gear fishery). The aim is to determine the best performing model for the 2020 stock assessments. SCA models performed better than VPA models for both herring stocks. For the spring spawning herring stock, the SCA model estimating time-varying natural mortality and catchability to the CPUE index in the gillnet fishery (the qmSCA model) was the best performing model. In the fall spawning stock, the qSCA and the qmSCA models performed best, but the qmSCA was selected as the best model as it provided natural mortality estimates, an important parameter in 4TVn herring stock assessment. Retrospective patterns in SSB from this model must be monitored and the source of the pattern will be investigated using new data sources. Overall, the selected models offered improvements over VPA models used in previous assessments. [ABSTRACT FROM AUTHOR]