Understanding how organisms interact with landscape features during dispersal can be critical for species conservation. We addressed the question of how landscape features effect dispersal by conducting behavioural experiments and observational surveys of rare butterflies across a large landscape. Our primary study species were small, wetland butterflies: Neonympha mitchellii francisci (St. Francis satyr, endangered), as well as surrogates for it, including Neonympha areolata (rare) and Satyroides appalachia (non-listed). Our experiments fell into two groups: those that tested butterfly movement choices at habitat boundaries, and those that tested spatial aspects of movement paths in major habitat types on our landscape. During 2006-07, we collected movement data from 167 individuals by conducting experimental releases and observational surveys in 5 different habitat types. These experiments provided spatially explicit movement paths and movement choices at different habitat boundaries. We observed an additional 153 individuals in wetlands to test reactions to boundaries of preferred habitat.

Results/Conclusions

Results from our experiments at wetland boundaries illustrate species-specific responses to habitat boundaries (p=0.002, F=4.50, df=4). Of all species observed, the endangered butterfly showed the highest edge sensitivity and was least likely to exit a habitat patch. Choice experiments at different habitat boundaries indicate that butterflies avoid open and developed habitats and show a preference for wetter habitats over drier ones. Experiments from released surrogates in different habitat types show that movement path step lengths varied between habitats (F_2,104=6.04, p<0.001). For example, step lengths were greatest in open habitat (mean=15.4 m+/-6.22 m std), and shortest in riparian habitat (3.7 m +/- 2.54 m). Butterfly turn angles also varied among habitats. For example, there were significant differences in the turn angle distributions observed in wetland and those observed in either open or upland habitats (wetland:open D=2.76, p=<0.001; wetland:upland D=2.80, p<0.001). With these data, we calculated a metric to describe the overall resistance value of our landscape. First, dispersal values of each habitat were calculated as the product of the mean displacement rate in that habitat and the probability that a butterfly enters that habitat from any other habitat. Then, by incorporating land-cover data, a landscape dispersal value was calculated as the sum of habitat dispersal values weighted by the relative area of/edge adjacent to each habitat in our landscape. This approach has illuminated some counter-intuitive results, such as the poor dispersal value of breeding habitat, and provides a framework to theoretically test the effect of landscape manipulation on dispersal.