Georeferencing and species distribution modelling of invasive plants in the state of Florida, USA

Invasive exotic plants have wide ranging impacts. They contribute to the decline of native plant species (Mack et al. 2000; Hulme 2003; Collingham et al. 2000), negatively impact hydrology of certain habitats (Parker et al. 1999; Graf 1978 in Hobbs and Humphries 1995), and create problems for wildlife by altering vegetation structure and food sources (Mack et al. 2000; Hulme 2003). Researchers continue to look for new strategies and tools to learn about the ecology of these organisms to reduce the negative impacts of these species. Species occurrence data and species distribution models are two tools that can assist researchers studying invasive exotic species (Chapman 2005; Peterson and Vieglas 2001).;The objectives of this research were to (1) georeference all of the records in the Florida Exotic Pest Plant Council (FLEPPC) database using a geographic information system (GIS) and a protocol adapted from the MaNIS guidelines to produce maps of invasive exotic plants of Florida (2) analyze the relationships within the records and between these records and other spatial patterns (e.g., population density) in Florida and (3) predict the future distribution of three pairs of invasive exotic plant congeners in Florida using the species distribution model OM-GARP while addressing two questions: (1) What are the impacts of the precision of the species occurrence records on the distribution models? and (2) Is OM-GARP's accuracy in predicting current species occurrences related to features of the species distributions, in particular the areal extent of the distribution? The six species I investigated were: Lygodium microphyllum (old world climbing fern), Lygodium japonicum (Japanese climbing fern), Dioscorea alata (wild yam), Dioscorea bulbifera (air potato), Ardisia crenata (coral Ardisia) and Ardisia elliptica (shoebutton Ardisia).;A total of 4,349 (80.4%) of the records in the FLEPPC database were georeferenced. The majority of the records were in coastal counties of the peninsula. The panhandle contained far fewer records and several counties in the north central part of Florida did not have any records. The numbers of records per county were strongly correlated with population density per square mile, number of cities, and total miles of roads.;Models generated for 5 of the 6 invasive exotic species had greater than 72% accuracy (Table 2.5). The model generated for Lygodium japonicum had the lowest accuracy (48%) and two models generated for A. elliptica had the highest accuracy (100%). The models presented suggest that, for at least one species, the precision of the records used for modelling can impact the accuracy of the predicted distribution. The features of the species distributions may also influence the accuracy of predicted distributions. Across all pairs of congeners, models with higher accuracy were consistently generated for the species with the narrower known distribution (A. elliptica > A. crenata; D. alata > D. bulbifera; L. microphyllum > L. japonicum; Table 2.5). I suggest that a reason for this relationship may be that species with smaller niche hypervolumes are more likely to be adequately represented in small sample sizes of species occurrence records than species with larger niche hypervolumes.