Invasive forest pest species pose a serious threat to the biodiversity and economy of the United States, and the level of devastation they have caused continues to rise. The spread of these pests must be forecasted in order to take appropriate management actions. Yet, it is not possible to build spread models for every species. We developed and examined the feasibility of using a single common predictive model for the spread of all damaging invasive forest pests presently found in the United States. We determined if there are general factors that can be used across species to predict spread, based on pest life history, environmental factors, and human demographic data. These factors were integrated into a process-based dispersal kernel model. We fit this model to our dataset to determine the factors important for United States pest spread using the Minimum Energy Test (MET), a metric of goodness-of-fit that relates the mean distance among presences within a range (predicted or observed) to the mean distance among presences across ranges (predicted and observed).
Results/Conclusions
Our theoretical analysis indicated that we could detect the effect of most predictor variables across their entire range of possible magnitudes if they were indeed influencing the spread process. The application of this model to the United States data indicated that both habitat invasibility (forested land) and propagule pressure variables (proximity to a major port, human population density, road length) had important influences on pest spread. We developed three dispersal models that could explain 72-80% of the variation in extent of invasion, and where predicted locational pest presences deviated from true occurrences by 47-80km across the entire pest range, on average across all pests. Thus, these results support the generalizability of a single predictive pest spread model.