Modeling the effect of roads and topoclimate on plant invasions in mountains : the case of Trifolium repens

Abstract: Non-native species are a main cause for the global loss of biodiversity and ecosystem services. Mountain regions have been relatively spared from plant invasions up to now, mostly due to climatic restrictions and low human influence, with roads being the main pathways. But the invasion risk is increasing due to climate change and intensified land use. This is problematic, because mountain regions generally have high conservation value and are hard to manage. Therefore, prevention is crucial. The aim of my study is to improve predictions of plant invasions by including roads and topoclimate in the species distribution models and to provide suggestions for adequate roadside management. First, I downscaled bioclimatic variables according to the topography to a resolution of 50 x 50 m, applying a geographically weighted regression. Then, I fitted a species distribution model on both the original and downscaled bioclimate (‘topoclimate’), with a generalized linear mixed model. As response variable, I used presence-absence data for Trifolium repens (n=7683), which had been collected by the Mountain Invasion Research Network in 11 mountain regions worldwide. Furthermore, I fitted three species distribution models, based on ‘topoclimate’, ‘topoclimate and roads’ and ‘roads’. I then evaluated all models with the area under the receiver operating characteristic curve, focusing especially on sensitivity values. For validation, I used an independent dataset from Victoria, Australia. Both downscaling the bioclimate and including roads improved the species distribution models, with roads being an even more robust predictor than bioclimate. However, the overall predictive power of all models was very low, with moderate sensitivity values. This limited predictive power on a regional level (in Victoria, Australia) can be partly explained by general issues regarding invasive species but also by local peculiarities of the validation area. More local information would be needed in order to make accurate predictions for regional management. However, the global importance of mountain roads as pathways for plant invasions was confirmed by my study, which emphasizes the need for adequate roadside management. Generally, management should focus on both preventive measures as well as controlling further spread, especially in high conservation value areas.