Optimizing invasive species distribution models: a test with the giant hogweed in Switzerland

Invasive alien species are one of the primary drivers of biodiversity loss and cause great economic damage worldwide. They can be managed, but anticipating their invasions is paramount. This can be accomplished by modelling their current and future distributions. Species Distribution Models are a staple tool for this purpose and recent advances to use them to model invasive alien species have been made and are applied here. In this paper, two common species distribution modelling methods, MaxEnt and Random Forests, are employed to simulate the present and future potential distributions dangerous invasive species, Heracleum mantegazzianum Sommier and Levier, using opportunistically collected occurrence datasets, a nested modelling approach and background points sampled stratified randomly in environmental space. Possible bias in the occurrence datasets is treated in several ways and the differences in output distributions are discussed, as well as the uses of the outputs for guiding eradication efforts.