Spatial prediction of habitat overlap of introduced and native thistles to identify potential areas of nontarget activity of biological control agents.

Nontarget feeding of Rhinocyllus conicus Fröelich and Trichosirocalus horridus (Panzer) on native North American thistles in the genus Cirsium has been documented. Some species of these native thistles have shown greater infestation levels of R. conicus in populations that are in close proximity to the target plant species, Carduus nutans L. In 2005 a study was initiated to identify areas of potential nontarget feeding by R. conicus and T. horridus on thistle species by predicting habitats of two known introduced hosts [C. nutans and Cirsium vulgare (Savi) Tenore] and two native species [Cirsium carolinianum (Walter) Fernald and Schubert and C. discolor (Muhlenberg ex Willdenow) Sprengel] using Mahalanobis distance (D(2)). Cumulative frequency graphs showed that the D(2) models for all four plant species effectively identified site conditions that contribute to the presence of the respective species. Poisson regression showed an association between D(2) values and plant counts at field-test sites for C. nutans and C. carolinianum. However, negative binomial regression detected no association between D(2) values and plant counts for C. discolor or C. vulgare. Chi-square analysis indicated associations between both weevil species and sites where C. vulgare and Carduus nutans were found, but not between the weevil and native thistle species. Habitats of C. nutans and Cirsium carolinianum overlapped in ≈12% of the study area. Data-based habitat models may provide a powerful tool for land managers and scientists to monitor native plant populations for nontarget feeding by introduced biological control agents.