Site reoccupation in fragmented landscapes: testing predictions of metapopulation theory