PS 10-114 - Species distribution model for an invasive wetland plant – a hierarchical approach

Monday, August 8, 2011
Exhibit Hall 3, Austin Convention Center
Shyam M. Thomas, Ecology, Evolution & Organismal Biology, Iowa State University, IA and Kirk A. Moloney, Department of Ecology, Evolution, and Organismal Biology, Iowa State University, Ames, IA

Developing distribution models for invasive species is both a pressing conservation need and a formidable modeling challenge. We address this issue for Purple Loosestrife (Lythrum salicaria L.), an invasive wetland plant, by developing a hierarchical model of  its spatial distribution using land cover information and presence records from the Minnesota-DNR. Landscape level species distributions are the outcome of multiple hierarchical processes that play out at different scales. The model being developed tests if the purposeful incorporation of hierarchy results in a more coherent ecological model that is also capable of making robust predictions. To incorporate a hierarchy, the distribution of loosestrife was analyzed at three nested levels viz.; loosestrife’s preferred land use land cover types (habitats), its landscape context (habitat neighborhood) and, finally, proximity to other loosestrife populations (propagule pressure). 


Preferred habitats of loosestrife were identified as herbaceous wetlands, open water edges and developed open spaces. Loosestrife occurrence within these preferred habitats varied strongly with respect to neighborhood composition. Neighborhoods that tended towards higher proportion of light to moderately intense disturbance were more likely to contain loosestrife. However, in the case of developed open spaces only water rich neighborhoods mattered.  Loosestrife invaded habitats had, on average, more loosestrife as neighbors than uninvaded loosestrife habitats. Taken together, loosestrife distribution appears to be modified hierarchically by habitat preference, its neighborhood and finally by propagule pressure. A hierarchical auto-logistic model will be fitted to estimate the probability of loosestrife occurrence within a given habitat, given its neighborhood and proximity to extant loosestrife populations. The final expected outcome is a probabilistic map showing locations that are highly vulnerable to loosestrife invasion.

Copyright © . All rights reserved.
Banner photo by Flickr user greg westfall.