State-space stock assessment models for data-moderate ﬁsheries
Fish populations are subject to natural growth, environmental pressures, and natural mortality. In addition, they may experience pressure from anthropic fishing mortality. Management of fish stocks requires the collection of suitable data from which population models can be built. State-space models (SSMs) are one modelling approach, and this project investigates their application to data-moderate stocks. We define data-moderate stocks as those for which there are no survey data, no information on age composition, and fisheries-dependent data are the only available source of information. We find that many existing state-space models are either too simple (e.g., state-space surplus production models) or too complex (e.g., state-space age-structured models) for these stocks, although many fisheries around the world face data-moderate situations.
A state-space model is becoming a favoured choice in modelling fish population dynamics, as it allows one to incorporate both measurement and process errors. However, several studies have found that separation of the two sources of variability can result in estimability problems even in simple state-space models. Using a state-space surplus production model as an example, we found that such estimability problems can occur even in a simple stock assessment model, especially when measurement error is large relative to process error. This problem even exists when constraints are imposed on most of the model parameters. Such findings suggest the limitations of SSMs and the importance of model diagnostics.
Using data collected from South Korean fish stocks as application examples, we developed two stock assessment models in state-space form. The first model is a state-space two-life stage-structured production model which can be applied to stocks where juvenile and adult fish have been separately exploited by different fisheries. The key feature of the model is that in the absence of any composition data (e.g., age and size), demographic relationships between juvenile and adult populations are incorporated using abundance indices collected from different fisheries, each of which selectively targets the two different life stages of fish. The second model is a state-space length-based age-structured model. This integrated model is developed to utilise length composition data to inform the age structure of a population. Such data are often available in many data-moderate stocks, instead of a direct measure of age composition, such as catch-at-age data. Separating age groups based on length compositions is not a new concept, but most existing models do not allow process error. Thus, the development of such a model in state-space form could provide a more reliable assessment tool for many data-moderate stocks.
This thesis research contributes to the better understanding of potential estimability issues in SSMs for fish stock assessments, as well as development of the two new state-space models for data-moderate fisheries. We also identified several issues associated with our findings which could be useful for future research.