Premium
A DISPERSAL‐CONSTRAINED HABITAT SUITABILITY MODEL FOR PREDICTING INVASION OF ALPINE VEGETATION
Author(s) -
Williams Nicholas S. G.,
Hahs Amy K.,
Morgan John W.
Publication year - 2008
Publication title -
ecological applications
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.864
H-Index - 213
eISSN - 1939-5582
pISSN - 1051-0761
DOI - 10.1890/07-0868.1
Subject(s) - biological dispersal , habitat , ecology , vegetation (pathology) , environmental science , seed dispersal , geography , biology , population , medicine , demography , pathology , sociology
Developing tools to predict the location of new biological invasions is essential if exotic species are to be controlled before they become widespread. Currently, alpine areas in Australia are largely free of exotic plant species but face increasing pressure from invasive species due to global warming and intensified human use. To predict the potential spread of highly invasive orange hawkweed ( Hieracium aurantiacum ) from existing founder populations on the Bogong High Plains in southern Australia, we developed an expert‐based, spatially explicit, dispersal‐constrained, habitat suitability model. The model combines a habitat suitability index, developed from disturbance, site wetness, and vegetation community parameters, with a phenomenological dispersal kernel that uses wind direction and observed dispersal distances. After generating risk maps that defined the relative suitability of H. aurantiacum establishment across the study area, we intensively searched several locations to evaluate the model. The highest relative suitability for H. aurantiacum establishment was southeast from the initial infestations. Native tussock grasslands and disturbed areas had high suitability for H. aurantiacum establishment. Extensive field searches failed to detect new populations. Time‐step evaluation using the location of populations known in 1998–2000, accurately assigned high relative suitability for locations where H. aurantiacum had established post‐2003 (AUC [area under curve] = 0.855 ± 0.035). This suggests our model has good predictive power and will improve the ability to detect populations and prioritize areas for on‐going monitoring.