Non-native plant species richness and influence of greenhouses and human populations in the conterminous United States.

Background: One issue in invasive plant ecology is identification of the factors related to the invasion process that increase number of non-native species. When invasion by non-native species increases, so does the probability that some non-native species will become harmful, or classified as invasive species, which disrupt natural ecosystems with attendant economic and social costs. I quantified patterns of how non-native species richness varied with vegetation types and human populations. To evaluate the relative importance of different predictor variables for invasion pathways in the conterminous United States, I modeled non-native plant species richness by county compared to current and historical human populations; greenhouses and nurseries; railroads, pipelines, transmission lines, and oil and gas wells; and land covers of impervious surface, development intensity categories, agriculture, and vegetation types. I also modeled these variables within vegetation types, excluding vegetation variables. Results: To summarize patterns, non-native plant species richness increased from 72 to 200 with increasing human population density classes. Forests and forest land use mosaics had the greatest mean number of non-native plant species, ranging from 121 to 166, whereas grasslands and grassland mosaics had the least number of non-native plant species, about 70. For modeling variable importance, all combined variables had R² values of 56% (random forests regressor) and 54% (cubist regressor) for predictions of withheld observations of non-native plant species richness, with greenhouse density and percent forestlands as most influential variables. Single variables of greenhouses (R² = 29%), historical and current human populations (R² = 27% and 23%), impervious surface (25%), and medium intensity development (23%) were most associated with non-native plant species richness. For vegetation types, greenhouse and historical human population densities were influential variables particularly in forestlands, shrublands, and wetlands. Conclusions: Based on these models, human population measures and horticultural locations of greenhouses and plant nurseries may have stronger relationships than measures of land use disturbance and transport with non-native plant species richness. [ABSTRACT FROM AUTHOR]