Informing network management using fuzzy cognitive maps

Modern conservation requires robust predictions about how management will affect an ecosystem and its species. The large uncertainties about the type and strength of interactions make model predictions particularly unreliable. In this paper, we show how fuzzy cognitive maps can produce robust predictions in complex and uncertain ecosystems. The use of fuzzy cognitive maps has been increasing markedly, but there are two critical issues with the approach: translation of expert knowledge into the FCM is often done incorrectly; and sensitivity analyses are rarely conducted. Translating expert knowledge is a constant challenge for ecological modellers, often because experts know about the behaviour of a system, but modellers need to know model parameters, which subsequently lead to system behaviour. We describe how to correctly incorporate expert knowledge into FCMs, and we describe how to appropriately conduct uncertainty and sensitivity analysis. We illustrate this process with a previously published network for feral cat and black rat control on Christmas Island. Perverse indirect effects of conservation management are a key concern, and methods to help us make informed decisions are required. Fuzzy cognitive maps are a promising approach for this, but it requires the methodological improvements that we present here.