Data-rich but model-resistant: an evaluation of data-limited methods to manage fisheries with failed age-based stock assessments
Christopher M. Legault, John Wiedenmann, Jonathan J. Deroba, Gavin Fay, Timothy J. Miller, Elizabeth N. Brooks, Richard J. Bell, Joseph A. Langan, Jamie Cournane, Andrew W. Jones, Brandon Muffley
Abstract
Age-based stock assessments are sometimes rejected by review panels due to large retrospective patterns. When this occurs, data-limited approaches are often used to set catch advice, under the assumption that these simpler methods will not be impacted by the problems causing retrospective patterns in the age-based assessment. This assumption has never been formally evaluated. Closed-loop simulations were conducted where a known source of error caused a retrospective pattern in an age-based assessment. Twelve data-limited methods, an ensemble of a subset of these methods, and a statistical catch-at-age model with retrospective adjustment were all evaluated to examine their ability to prevent overfishing and rebuild overfished stocks. Overall, none of the methods evaluated performed best across the scenarios. A number of methods performed consistently poorly, resulting in frequent and intense overfishing and low stock sizes. The retrospective adjusted statistical catch-at-age assessment performed better than a number of the alternatives explored. Thus, using a data-limited approach to set catch advice will not necessarily result in better performance than relying on the age-based assessment with a retrospective adjustment.