A Comparison of Fleet Dynamics Models for Predicting Fisher Location Choice

Scientific advice for fisheries management rarely takes into account how fishers react to regulations, which can lead to unexpected results and unrealistic expectations of the effectiveness of the management measures. Short-term decisions about when and where to fish are one of the greatest sources of uncertainty in predicting management outcomes. Several models have been developed to predict how fishers allocate effort in space and time, including mechanistic methods such as gravity and dynamic state variable models, and statistical methods such as random utility and Markov models. These have been individually used to predict effort allocation for various fisheries, but there is no comparative synthesis of their structure and characteristics. We demonstrate strong theoretical links between utility and choice in gravity, random utility, Markov and dynamic state variable models. Using an advanced event-based simulation framework, we find that mechanistic models bias effort allocation to certain areas when applying commonly used strong assumptions about drivers of effort allocation; and conversely, statistical models accurately predict the distribution of fishing effort under business as usual. However, predictive performance degrades with previously unobserved dynamics, such as a spatial closure. Mechanistic models were less suited to general application under business as usual but provide a useful framework for testing hypotheses about a fishery system in response to policy change. Comparison of simple model formulations yielded significant insight into the characteristics of the models and how they could be used to evaluate alternative management approaches for mixed fisheries.