Identifying fish populations prone to abrupt shifts via dynamical footprint analysis

Fish population biomass fluctuates through time in ways that may be either gradual or abrupt. While abrupt shifts in fish population productivity have been shown to be common, they are rarely integrated into stock assessment or fishery management, in part because of the difficulty of predicting when abrupt shifts may occur and which stocks are prone to such shifts. In this study, we address the latter challenge by designing a mechanism-agnostic context-specific approach that is based on exploiting the dynamical properties of fish population fluctuations for detecting potential abrupt shifts. We use time series of fish population biomass from three global datasets, first, to classify their shapes into abrupt and nonabrupt (linear, quadratic, or no change) classes, and, second, to predict classified shapes based only on their dynamical footprint (a set of metrics such as variance, autocorrelation, etc, of the time series). We find that populations prone to abrupt shifts can be detected with moderate accuracy in the three datasets in spite of data limitations. In total, we identified 50 populations at risk of future abrupt shifts across 11 different Large Marine Ecosystem regions. Our context-specific approach offers critical insights into population stability and enables the identification of stocks whose dynamical properties suggest that they would benefit from more precautionary management.