Does increased complexity improve performance: Evaluating a constrained biomass dynamics model and an empirical harvest control rule

Abstract

Biomass dynamics models are a common stock assessment method applied in-data limited situations. We present a management strategy based on a new constrained biomass dynamics model (‘dynamic tier 4’) as an alternative to the currently applied empirical management strategy (‘empirical tier 4’) used in Australia’s Southern and Eastern Scalefish and Shark Fishery (SESSF). The dynamic tier 4 management strategy is constrained because the biomass dynamics model on which it is based involves the assumption that yield corresponding to the biomass target occurred during a pre-specified set of reference years. This is the same assumption that is made by the empirical tier 4 management strategy. Management strategy evaluation is used to evaluate performance among three ways of accounting for error when fitting the dynamic tier 4 approach, and to evaluate performance against the empirical tier 4 management strategy as well as a management strategy based on an integrated assessment implemented using Stock Synthesis (‘tier 1’). Finally, the sensitivity of the empirical tier 4 and dynamic tier 4 management strategies to an incorrect choice for the pre-specified reference years is explored. The results demonstrate improved performance by the dynamic tier 4 management strategy compared to the empirical tier 4 management strategy. As expected, the data-rich, tier 1 management strategy had the best performance. The dynamic tier 4 management strategy has a reduced probability of the stock falling below the limit reference point, reduced catch variability and is less sensitive to incorrect reference year choice than the empirical tier 4 management strategy. These results demonstrate that the dynamic tier 4 management strategy is a suitable alternative to the empirical tier 4 management strategy in the SESSF, and more generally, can be considered as a robust option for forming the basis for management recommendations for data-limited situations.