Application of species distribution models for protected areas threatened by invasive plants

Local species distribution models (SDMs) were constructed for three of the most harmful invasive alien plant species (Fallopia spp., Solidago spp. and Heracleum mantegazzianum) in the Kokořínsko Protected Landscape Area (Czech Republic), using Natura 2000 habitat types and other environmental conditions as predictors for the SDMs. Presence or absence data (recorded by field mapping) was entered into the SDMs and used to predict the potential distribution of particular species. Here, we critically evaluate the accuracy of the models and assess their applicability for natural resource protection. Variables such as habitat and soil type tend to dictate the current distribution of Solidago spp., while distance from roads and water corridors and elevation are important for Fallopia spp. distribution. For Solidago spp., ‘generalised boosted models’ and ‘generalised additive models’ were considered the most suitable algorithms. For Fallopia spp., however, the predictive power of the models tended to be weak, while the number of localities was too low for SDMs in the case of H. mantegazzianum. The number of initial localities containing invasive alien species was an important factor for making significant predictions of potential distribution. In general, the predictive power of the models was too low when using less than 10 localities; for good predictive power at the local scale, we suggest that at least 100 localities/100 km2 are used.