Combining the effects of surrounding land-use and propagule pressure to predict the distribution of an invasive plant

• Published in: Biological invasions 2015-01, Vol.17 (1), p.477-495
• Main Authors: Thomas, Shyam M, Moloney, Kirk A
• Format: Article
• Language: English
• Subjects: anthropogenic activities; Anthropogenic factors; Aquatic habitats; Aquatic plants; Bayesian analysis; Biomedical and Life Sciences; Developmental Biology; ecological invasion; Ecology; edge effects; Flowers & plants; Freshwater & Marine Ecology; Geographic information systems; habitats; Invasive plants; invasive species; land cover; Land use; landscapes; Life Sciences; Lythrum salicaria; Nonnative species; Original Paper; Plant Sciences; prediction; Prediction models; Principal components analysis; regression analysis; risk; surveys; uncertainty; wetland plants; Wetlands
• ISSN: 1387-3547; 1573-1464
• DOI: 10.1007/s10530-014-0745-7

The distribution of invasive plants across a landscape is largely governed by disturbance invoking anthropogenic land-use practices and propagule pressure. However, spatial variability associated with anthropogenic disturbances and propagule pressure is seldom used to develop distribution models of invasive plants. This study makes use of large-scale survey data to develop a spatially explicit predictive model for the invasive wetland plant—purple loosestrife. Using loosestrife presence data and land use land cover information, we first predicted loosestrife occurrences in two types of wetland habitat, namely herbaceous wetlands and open-water edges, with a series of logistic regression models that incorporated surrounding land-use at three different neighborhood scales. The best-fitting surrounding land-use model was then combined with three different distance constraint models that simulated propagule pressure. Loosestrife occurrence as a function of surrounding land-use showed best fit at a neighborhood radius of 400 m. Predictions made from the surrounding land-use model at the 400 m scale were fairly accurate and loosestrife invasion of wetland locations were correlated with the proportion of anthropogenic land-use conditions. Inclusion of an autocovariate simulating propagule pressure improved model fit and performance significantly. Model findings suggest that spatially explicit incorporation of surrounding land-use yields an ecologically realistic projection of invasion risk wherein disturbance prone habitat edges tend to be more invasible. Combining this prediction with location specific estimates of propagule pressure further reduces uncertainty by spatially constraining areas of high invasion risk. Our approach is applicable to other invasive plants since it is based on two key drivers of plant invasion: disturbance and propagule-pressure.