Reliable prediction of metapopulation persistence in fragmented landscapes has become a priority in conservation biology, with ongoing destruction of habitat confining increasing numbers of species into networks of small patches. A spatially realistic metapopulation model, which includes the first-order effects of patch area and isolation on extinction and colonization, has been tested, The distribution of an endangered butterfly was successfully predicted on the basis of parameter values estimated for a well-studied congeneric species. This modeling approach can be a practical tool in the study and conservation of species in highly fragmented landscapes.