Managing sparse data in biological invasions: a simulation study

Despite recent progress, estimating species spread and the risk of different sites becoming invaded remains a major challenge in invasion biology. One of the most common problems is sparse data: we will have rarely sampled all areas for invaders and consequently do not know where all the invaded sites are. This is problematic because the entire system of sites determines invasion probability for any single site—unsampled sites can act as both sources and sinks of propagules. Thus, the difficulty is to predict invasions across the entire system with only sparse data. In this manuscript, we develop an approach to make use of partial data to forecast new invasions, and compare it with default ways of handling missing information in biological invasions. We demonstrate that it is possible to estimate spread with only a fraction of sites sampled. We find that reliability depends on the number of invaded sites sampled and that there is a tendency to underestimate the effect of propagule pressure when Allee effects are present.