Predicting novel interactions between plants and herbivorous insects

It would be highly useful to be able to predict which local herbivore species might colonize a given introduced plant, even before that plant’s introduction. Recent trophic interaction models suggest that most novel interactions between Lepidoptera and introduced plants can be predicted by extrapolating from host-use patterns in the native food web. These initial models were mathematically simple and used very little information about the introduced plants or herbivores involved to make predictions. The models primarily used plant evolutionary relationships to define herbivore host breadth. Methods: Here we ask whether additional information about plant traits, regional geography, and history of introduction improve model predictions in a European plant-Lepidoptera food web. We also test whether more advanced modeling approaches better represent herbivore host affiliations than simple models. Results/Conclusion: We found that additional information about plants, in particular their regional geography, improves predictions of novel host use; however, phylogenetic host constraint was still the primary driver of novel herbivore-plant interactions. Novel herbivore-plant interactions are largely predictable by extrapolating from herbivores’ native hosts, and including more information about those hosts improves predictions.