To address many real-world natural resource management problems, ecologists need the ability to forecast the consequences of environmental change for populations of special concern. This challenge often includes the need to make predictions for novel conditions. Such forecasting may not be difficult where single factors drive outcomes, but outcomes may depend on substitutable resources and trade-off decisions by individuals. For example, ecologists recognize the widespread importance – at the individual, population and community levels of organization - of the trade-off between the need to eat and the need to avoid predators. But in the real world and in IBMs capable of addressing real-world problems, individual decisions can depend on: many alternatives, decisions by others, unknowable future conditions, and limits on decision-making capability. These complexities preclude the use of optimization approaches. Our solution to this problem relies on fitness-seeking individuals that make incorrect but useful predictions over a sliding time horizon. For example, in an IBM of stream trout, individuals select habitat on a daily time step that maximizes their probability of surviving over the next 70 – 90 days (considering both starvation and predation risk), using the prediction that they will occupy the same habitat under current conditions over that time horizon.
Results/Conclusions
A variety of results indicate this approach yields useful individual-based models. In simple models that allowed comparison of sub-optimal decision-makers like those in the trout model to omniscient optimal decision-makers, survival of the former approached that of the latter, except when survival probability was extremely low. In pattern-oriented analysis of the trout model, individuals using imperfect prediction to select habitat exhibited a variety of individual-level patterns observed in real fish, including avoidance of extreme physical conditions, reasonable responses to changes in the intensity of competition, and the ability to trade-off food availability and predation risk. The trout model has also reproduced commonly observed population-level patterns and yielded site-specific, population-level results that correspond with empirical observations. Finally, the trade-off behavior included in the trout model allows it to reproduce trait-mediated trophic interactions now recognized as widespread and strong. In many instances, the adaptive behavior of individuals will be important to population-level responses to environmental change. IBMs with individuals that make incorrect but useful predictions can incorporate this fact while addressing real-world natural resource management problems.