Site reoccupation in fragmented landscapes: Testing predictions of metapopulation theory

Abstract

1. Populations of formerly continuously distributed species subdivided by habitat fragmentation may show distributions in space and time that are consistent with predictions of metapopulation theory. Local extinctions and recolonizations should result in the most fragmented sites being infrequently occupied and the least fragmented sites being continuously occupied by sensitive species. The probability of extinction is predicted to be negatively correlated with patch size and the amount of habitat in the landscape. Conversely, recolonization is predicted to be negatively correlated with the isolation of the patch, and positively correlated with the amount of habitat in the landscape. 2. Data from a 3-year study of the effect of fragmentation were used to test whether these predictions from metapopulation theory apply to populations of the long-distance migrant forest bird Piranga olivacea (Scarlet Tanager) in fragmented North American landscapes. 3. Principal components analysis was used to derive a composite measure of fragmentation. This measure was used in a logistic regression as a predictor of the number of years that territorial males would occupy a site, given that it was occupied at least once. More fragmented sites were more likely to be occupied only once; the least fragmented sites were more likely to be occupied in all three years. Data on fragmentation were necessary, but not sufficient, to predict site reoccupation, and were poor predictors at medium levels of fragmentation. 4. The univariate measures of fragmentation (patch size and isolation, proportion of forest, and forest/non-forest edge), were also used in logistic regressions to predict the separate, probabilities of local extinction or recolonization. Local extinctions were negatively correlated with patch size and amount of forest in the landscape, as predicted. Recolonizations were negatively correlated with isolation of the patch as predicted, and surprisingly. also with the amount of edge in the landscape. This suggests that stochasticity may drive, extinctions, but that habitat selection may play an important role in recolonization. 5. Demographic data are usually required to establish the suitability of habitat to support persistent populations, but multiple-year distributional data can provide information on habitat quality far above that obtained from single-year studies.