Un modèle état-espace pour l’évaluation du stock de morue (Gadus morhua) dans la sous-division 3Ps de l’OPANO

Abstract

The paper describes the development of a state-space model (HYBRID) for the stock assessment of Northwest Atlantic Fisheries Organization (NAFO) Subdivision 3Ps cod stock. The HYBRID model fits to the DFO RV survey (1983–2005, 2007–19) as well as the following additional survey time series: the IFREMER (Institut Français de Recherche pour l'Exploitation de la Mer, English: French Research Institute for Exploitation of the Sea) ERHAPS (Evaluation des Ressources Halieutiques de la région 3PS) survey (1978–91), the Groundfish Enterprise Allocation Committee (GEAC) survey (1998–2005), and the sentinel gillnet and line-trawl surveys (1995–2018). The HYBRID model also fits to fisheries data with the expectation that the model can separate fishing mortality from natural mortality. Fisheries catch-at-age is fit using continuation ratio logits, and the fisheries landings are fit via censored likelihood. The use of censored likelihood for fitting landings allows the inclusion of expert opinion on reliability of landings throughout the model time series. The HYBRID model presented here starts in 1959, which is the first year for which landings data are available. The HYBRID model allows exploration of different forms for parameterization of time-varying fisheries selectivity, natural mortality, and approaches for estimation of catch-at-age data. The model formulations also differ depending on the survey series used in fitting the model. Seventeen model formulations are presented. Alternate formulations based on fishing mortality included: 1. a logistic flat topped or dome selectivity with time-blocks 2. SAM (State-space Assessment Model, Nielsen and Berg, 2014) style Multivariate Normal random walk 3. a separable age-year correlated process for fishing mortality rate (F) similar to NCAM (Cadigan 2016), and 4. option to break the processes for F in years where important events (like a moratorium) have impacted the fishery. Alternate formulation based on natural mortality included: 1. invariant natural mortality across age and years, and 2. time-varying natural mortality related to fish condition. The different model formulations are compared and evaluated on the basis of AIC (Akaike Information Criterion) values and their performance in retrospective analyses. The meeting concluded that time-varying natural mortality is an important consideration for modelling the dynamics of 3Ps cod and selecting a population model for this stock. Following the evaluations of the alternate model formulations, model formulations 11 which includes fit to sentinel data series, has F modelled using an MVN Random walk with year break for the moratorium, and has time-varying natural mortality rate (M) was considered to have utility as an assessment model for 3Ps cod stock assessment.