Can we predict species invasions in the future?

Background/Question/Methods Climate change complicates the management of invasive species. It is already extremely difficult to predict which species are likely to become invasive; a rapidly changing climate makes those predictions even more tenuous. Species that are invasive today may not be tomorrow, while today's natives may be tomorrow's pests. Consequently, there is great interest in making predictions about the future given directional climate change, and there are pragmatic reasons for doing so: We need to prioritize limited resources toward controlling species that will show high fitness in a warmer world. Alternatively, if we can anticipate species losing fitness in a warmer world we can use those episodes for restoration opportunities. But are long-term predictions feasible? Can they be accurate and produce results that should be used to alter today's management actions? Species distribution modeling (also known as ecological niche modeling or bioclimatic modeling) is widely used for projecting possible changes in individual species ranges or communities due to climate change. While the desirability of models that can project the future is clear, the limitations of SDMs to achieve this goal are also evident and have been reviewed extensively. Invasive species, which are in effect species already in the process of shifting and expanding ranges, present an important case study because they test some of the fundamental assumptions in modeling approaches. In this paper, I review the findings and accuracy of studies using correlative SDMs to project species distributions across space and time. I examine how uncertainty is evaluated in those studies, and whether the focus is on temporal endpoints rather than on increments or transitions. These analyses are used to reflect on whether results of these studies are suitable for application to on-the-ground management. Results/Conclusions The results of this review highlight that models applied across space and time often perform poorly. This lack of fit challenges assumptions about controls on species distributions and niche conservation. For management, this review suggests that the use of SDM as a guiding tool for risk assessment of invasive as well as other species is not a straightforward exercise. Improvements in methodology and data availability are required, and the results will be more apt for some species than others. In conclusion, I discuss a few examples of best practices.