AbstractAim Uncertainty has been widely recognized as one of the most critical issues in predicting the expansion of ecological invasions. The uncertainty associated with the introduction and spread of invasive organisms influences how pest management decision makers respond to expanding incursions. We present a model‐based approach to map risk of ecological invasions that combines two potentially conflicting goals: (1) estimating the likelihood of a new organism being established at a given locale and (2) quantifying the uncertainty of that prediction.Location Eastern and central Canada.Methods Our methodology focuses on the potential for long‐distance, human‐assisted spread of invasive organisms. First, we used a spatial simulation model to generate distributions of plausible invasion outcomes over a target geographical region. We then used second‐degree stochastic dominance (SSD) criteria to rank all geographical locations in the target region based on these distributions. We applied the approach to analyze pathways of human‐assisted spread (i.e., with commercially transported goods) of the emerald ash borer (EAB) (Agrilus planipennisFairmaire), a major pest of ash trees in North America.Results The projected potential of the pest to establish at remote locations is significantly shaped by the amount of epistemic uncertainty in the model‐based forecasts. The estimates based on the SSD ranking identified major ‘crossroads’ through which the movement of the EAB with commercial transport is most likely to occur. The system of major expressways in Ontario and Quebec was confirmed as the primary gateway of the pest’s expansion throughout the Canadian landscape.Main conclusions Overall, the new approach generates more realistic predictions of long‐distance introductions than models that do not account for severe uncertainties and thus can help design more effective pest surveillance programmes. The modelling technique is generic and can be applied to assess other environmental phenomena when the level of epistemic uncertainty is high.
