Landscape correlates of forest plant invasions: A high-resolution analysis across the eastern United States

Aim: Invasive species occurrence is often related to the anthropogenic context of a given area. Quantifying the effects of roads is of particular interest as roads are a major vector for invasion. Our objective was to further quantify the effects of roads on forest plant invasion through a macroscale, high-resolution investigation to assist effective invasion control and mitigation. Location: Eastern United States. Methods: Using invasive plant data from 23,039 forest inventory plots in 13 ecological provinces, we employed logistic regression to relate the odds of invasion to distance from a road, with adjustments for broadscale differences attributable to ecological provinces, and local scale differences in productivity, forest fragmentation and land use. Results: The overall proportion (P) of invaded plots was 0.58 (0.65 for plots within 50 m of a road), and the highest odds (P/1-P) of invasion were found in relatively more productive, fragmented forest in landscapes with more than 10% agriculture or developed land cover. Wald chi-square statistics indicated the best predictor of the odds of invasion was ecological province, followed by land use, productivity, forest fragmentation and distance from a road. Depending on the province, the adjusted odds of invasion decreased by up to 23% (typically 4%-10%) per 100 m distance from a road. The adjusted probability of invasion approached zero in only three provinces, for the least productive, least fragmented forest that was at least 2,000 m from a road in landscapes with less than 10% agricultural or developed land cover. Main conclusions: In the eastern United States, the existence of a nearby road is less important than the landscape context associated with the road. A purely road-mediated effect has little practical meaning because anthropogenic activities and roads are pervasive and confounded.