Model Behavior: Can GIS identify areas of potential non-target feeding by introduced biological control agents of invasive weeds?

The efficacy of two weevils (Rhinocyllus conicus and Trichosirocalus horridus) introduced in the U.S. against the Eurasian musk thistle (Carduus nutans) has been diminished by their impact on some native Cirsium thistles. Non-target feeding of these weevils on native thistles may impact plant reproduction and plant populations. Because the biologies of these weevils are closely linked with musk thistle, the distribution of musk thistle populations may influence non-target activity. The use of geographic information systems (GIS) could be useful in analyzing the spatial relationships among musk thistle, weevils and native thistles by characterizing suitable habitats for each thistle species over a large spatial scale. In Spring 2004 a study was initiated to: 1) characterize habitats where exotic thistles and native thistles can occur and 2) identify potential areas of non-target feeding using spatial analysis. The study area (ca. 4,800 km²) consisted of four counties in eastern Tennessee. Four thistles (native: C. carolinianum and C. discolor; introduced: C. nutans and C. vulgare) were selected as model species. Population localities of these model thistle species, along with spatial data of eight topographic and physiographic variables were analyzed using Mahalanobis distance (D²) to identify suitable habitats for each thistle species. Cumulative frequency analysis indicated that threshold D² values of as low as 6.58 correctly classified over 65% of suitable habitats for all thistles observed in this study. Comparatively, random locations would require many more localities and greater D² values to correctly classify similar percentages of suitable habitats. The area of overlap of all species occupied ca. 4% of the total study area. Further measures of model predictiveness and application will be discussed.