Accounting for movement in spatial surplus production models: a case study of redfish on the eastern Grand Banks of Newfoundland

Spatial surplus production models (SSPMs) are an important type of spatial state-space model for cases lacking reliable spatial age or length data. However, fish movements pose computational challenges for SSPMs and can confound process errors, complicating parameter estimation. We use a Gaussian Markov random field with a Matérn covariance structure to account for spatiotemporal variation in dynamics and population production, thereby circumventing the computational and confounding issues. Through simulation studies, wherein data are generated explicitly considering movements, our novel random field model outperforms alternative methods (i.e., non-spatial and explicit movement models) in estimating fish spatial abundance, as evaluated using statistical metrics including Akaike information criterion, Bayesian information criteria, and correlation between simulated and estimated populations. We also applied our method to reveal the spatial distribution of redfish in NAFO 3LN Divisions based on survey and commercial catch data. Model validation confirms a good fit. Our model’s ability to fit a SSPM to a relatively short time series data (i.e., 8 years) demonstrates the benefits of using the random field approach in data-poor fisheries stock assessments.