Combined phylogenetic and geographic data can predict plant-pest interactions with high accuracy

Biological invasions are a major threat to biodiversity, often having destructive ecological and economic consequences. It is therefore essential to concentrate the limited resources available on managing the most serious risks. We build a robust Bayesian model to predict host species at risk from Agrilus, a genus of plant-feeding beetles that includes one of the worlds most destructive tree pests. The model uses the interplay of phylogenetic and geographic distance between known and potential hosts to estimate the probability of new Agrilus-host interactions. We assess the risk these beetles pose to Quercus species (oaks), their most common host, and find that our model can assign the status of known hosts with an accuracy of 83.6% in leave-one-out analyses. We employ this model to identify oak species at risk from Agrilus, providing data to inform more targeted efforts to prevent novel Agrilus introductions. Our model correctly identifies recently acquired novel hosts of Agrilus species that have invaded new areas, providing further evidence of its ability to accurately predict future potential interactions. Our approach makes use of readily available phylogenetic and geographic occurrence data and could be widely implemented to assess other potential invasive species, and help to prioritise counter measures against different threats worldwide.