Date of Award


Degree Type


Degree Name

Master of Science in Statistics


Computer Science and Statistics

First Advisor

Gavino Puggioni


Fisheries management requires regularly assessing stock status and setting catch levels for the coming years. Although data-rich stock assessment models incorporating demographic and biological information are generally preferred, such approaches are often prohibited by either insufficient data or a history of poor performance. In such cases, simpler Index-Based Methods (IBMs) are often used to generate catch advice for a fishery. However, these approaches do not typically forecast future abundance levels or quantify scientific uncertainty, making it difficult to assess the performance of different candidate methods prior to implementation. As a result, there is not a consensus as to which IBMs may be best suited to a particular situation and available biological information is often under-utilized in the management process.

To address this shortcoming, this work developed a novel Index-Based Method framework using dynamic linear models (DLMs), a flexible Bayesian state-space approach. Using simulated population data mimicking member species of the Northeast Multispecies groundfish complex, the predictive performance of candidate DLM structures were evaluated via retrospective forecasting. In both an Index-Based (age-aggregated) and Age-Based formulation constructed to demonstrate how the modular nature of these models can make fuller use of available data, the tested DLMs displayed promising predictive ability. While further testing is needed, this preliminary evaluation suggests that DLMs may become a valuable approach in the management of fisheries for which a data-rich stock assessment approach is not possible.

Available for download on Friday, July 22, 2022