Why model assumptions matter for natural resource management: interactions between model structure and life histories in fishery models

1. Bioeconomic models are increasingly used to provide benchmarks for harvest levels in wildlife and natural resource management, yet uncertainties related to model structure are underexplored. We investigate the importance of a range of uncertainties with a focus on model structure and life histories when estimating bioeconomic target reference point (TRPs) and assess the policy implications of ignoring these uncertainties. 2. We use three contrasting case studies to investigate the interactions between model, observational and process errors related to life-history parameters: the short-lived Japanese common squid Todarodes pacificus and Pacific saury Cololabis saira, and the slow-growing Patagonian toothfish Dissostichus eleginoides. We developed a simulation framework to test the harvest strategies resulting from bioeconomic TRPs under various assumptions about model structures and parameters. 3. We found the relative importance of different types of uncertainties affecting precision and accuracy of the model outputs varied according to the life-history traits. Little difference in TRP estimates was found between simple vs. complex population models for saury, while large differences were found for toothfish. The assumptions made about stock structure for squid not only resulted in different TRP estimates (generally, smaller for the multistock models), but also different economic outcomes depending on the balance of effort allocation between stocks. 4. Synthesis and applications. We use models similar to those used in the actual management of three case study species to explore the effects of interacting uncertainties on the management advice. We show that the interactions between structural elements of the models lead to very different management advice, depending on the life history of the species concerned. For the long-lived toothfish, life-history and gear selectivity parameters interacted strongly. For the short-lived squid which is managed as two stocks, spatial fishing effort allocation, correlation of environmental drivers between stocks and differential stock productivity interacted, producing very poor economic performance if assumptions about stock structure are incorrect. The key message for model-based natural resource management is that it is vital to investigate the major uncertainties related to model structure, process and estimation errors simultaneously, because they interact to produce non-intuitive results.