Adaptive management of ecological systems under partial observability

Adaptive management has a long history in ecology and conservation. Uncertainty in both the state of a system and the model defining its dynamics are fundamental challenges in adaptive management of complex ecological systems. Traditional approaches in conservation biology often ignore one or both sources of uncertainty due to the computational complexity involved. Here, we show that underestimating the role of uncertainty in both model estimation and decision-making results in aggressive decision rules which can potentially lead to the dramatic decline and possible collapse of a population, species, or ecosystem. We propose an approximate solution to adaptive management of ecological systems under both model and state uncertainties that is computationally feasible and applicable to complex management problems and provide a software for detailed implementation of our method, http://doi.org/10.5281/zenodo.1161521. We apply the proposed method in a marine ecosystem management context and show that by learning from historical data and arrival of new observations, decision makers can adapt their policies to avoid decline in the population and reach a sustainable population stability.