Best practices for modeling time-varying growth in state-space stock assessments
Abstract
Temporal variability in biological parameters (i.e., population processes) is common among fish stocks and may substantially impact the outputs of stock assessments. Growth variability is an important contributor to biomass fluctuations in several stocks, and its incorporation in assessment models has received more attention in recent years. State-space models (SSM) have also garnered attention recently due to their ability to estimate time variation in biological and fisheries processes like recruitment, natural mortality, catchability and selectivity. For example, the Woods Hole Assessment Model (WHAM) is a U.S. east coast age-structured SSM with flexible options to model time variation with random effects directly on the process, or via incorporation of environmental covariates, treating the true, unobserved values as random effects. However, current SSMs like WHAM cannot model time-varying growth internally, nor accept length data, and so the benefits of estimating time-varying growth cannot be explored for many stocks. In this study, we expand WHAM to incorporate new approaches to modeling changes in growth using a combination of parametric and non-parametric approaches while fitting to length and weight data. We implemented a simulation experiment to compare the performance of these modeling approaches under different scenarios. This study presents the first SSM that can be applied when length data are a key source of information, variation in growth is an essential part of the dynamics of the assessed stock, or when linking climate variables to growth in hindcasts or forecasts is relevant. Consequently, the importance of state-space approaches and growth variation can be tested across a broader range of fish stocks worldwide, helping to develop best practices and contribute to improving fisheries management goals.
