How does the landscape context of occurrence data influence models of invasion risk? A comparison of independent datasets in Massachusetts, USA

Abstract

The spatial distribution of non-native, invasive plants on the landscape is strongly influenced by human action. People introduce non-native species to new landscapes and regions (propagule pressure) as well as increase ecosystem invasibility through disturbance of native ecosystems. However, the relative importance of different landscape drivers of invasion may vary with landscape context (i.e., the types and amounts of surrounding land cover and land use). If so, data collected in one context may not be appropriate for predicting invasion risk across a broader landscape. To test whether independent occurrence datasets suggest similar landscape drivers of invasion, we compared landscape models based on data compiled by the Invasive Plant Atlas of New England (IPANE), which are contributed opportunistically by trained citizen scientists, to models based on Forest Stewardship plans (FSPs), which are located in privately owned and relatively undisturbed forests. We evaluated 16 landscape variables related to propagule pressure and/or disturbance for significant predictors of invasive plant presence based on presence/absence and count regression models. Presence and richness of invasive plants within FSPs was most influenced by proportion of open land and proximity to residential areas, which are both sources of propagules in forest interiors. In contrast, IPANE invasive plant presence and richness for the same area was influenced by distance to roads and streams. These results suggest that landscape drivers of invasion vary considerably depending on landscape context, and the choice of occurrence dataset will strongly influence model results.