Sensitivity analysis of a state space fish stock assessment model compared to conventional approaches

Abstract

Surplus production models provide simple analytical methods of assessing fish populations by taking the annual biomass, the growth rate and the carrying capacity into account. However, these simple models may not adequately reflect fish stock dynamics that can be substantially more complex with age and length specific birth, growth, and death processes at play. To account for this, process errors can be included in the production model in a state-space modelling framework, which is used frequently in ecological modelling in recent years. In this study, we compare the sensitivity of estimators of state-space and conventional non-linear production models (without process errors) using both the traditional case deletion diagnostic method and the local influence analysis method introduced by R.D. Cook, 1986 [12]. We apply these diagnostics to different fish stocks to assess how estimated parameters respond to small perturbations of the data.