Tomasz H. Szymura1, Magdalena Szymura2, Maria Zając3 and Adam Zając3
1Department of Ecology, Biogeochemistry and Environmental Protection, University of Wrocław, Maksa Borna Sq. 9, 50-328 Wrocław, Poland
2Department of Agroecosystems and Green Areas Management, Wrocław University of Environmental and Life Sciences, Grunwaldzki Sq. 24A, 50-363 Wrocław, Poland
3Jagiellonian University in Kraków, Faculty of Biology and Earth Sciences, Institute of Botany, Kopernika 27, 31-501 Kraków, Poland
The realistic modeling of invasion risk and/or invasion level requires data with appropriate quality regarding their spatial resolution. Due to data availability, models at large (e.g. continental, national) extents, which derive data from atlases of plant species distribution, usually prepared in rather coarse scales (e.g. 10 × 10 km) are quite numerous. This allows for studying relationships between environment and invasion level at large spatial scales, while such examinations at finer scales (e.g. regional) are scarce. This is problematic because the effect of a particular environmental factor on alien plant richness and its interactions with other environmental variables is context-dependent, and can vary from region to region. In our examination, we used an atlas of neophyte distribution prepared in 2 × 2 km grid covering 31.2 thousands square kilometers in Carpathian massif and its foreground, Central Europe. Using the boosted regression trees technique, we assessed the effect of anthropogenic factors, soil variables, land relief, climate and landscape structure on neophyte richness (NR). We found that each examined sphere of the environment explained the NR, but their explanatory ability varied more than two-fold. Climate explained the highest fraction of deviation, followed by anthropogenic factors, soil type, land relief, and landscape structure. A global model, which incorporated the crucial variables from all studied spheres of the environment, revealed that the deviation explained by variables representing particular environment spheres overlapped. When the variables representing landscape structure were not included in the global model as redundant, the climatic variables finally explained a smaller fraction of deviation compared to anthropogenic factors. We assessed the course of dependencies between NR and particular explanatory variables accounting for the average effect of the remaining variables. The relationships were usually curvilinear and revealed some threshold values of environmental variables (e.g. percentage of urbanized areas, human population density, and road network density) beyond which the NR changed rapidly.