Population Responses to Habitat Fragmentation: Statistical Power and the Random Sample Hypothesis

Oikos 76: 235-242. The random sample hypothesis states that small habitat fragments are random samples from large ones. Therefore, the effect of habitat fragmentation on population size is only related to pure habitat loss, the area and isolation of habitat fragments do not influence the population size in the landscape. An alternative hypothesis is that probability of occupation of a habitat fragment is related to area and isolation of the habitat fragment. Simulation models that included effects of area and isolation were used to investigate the effect of habitat fragmentation on population survival in landscapes with different proportions of suitable habitat. Four different combinations of area sensitivity and dispersal ability were used. These models were used to estimate the statistical power to reject the random sample hypothesis. The statistical power to detect differences, between a model based on the random sample hypothesis and a model which included effects of habitat fragment size and isolation between habitat fragments on population survival, was low in landscapes with a high proportion of suitable habitat, but increased as the proportion of suitable habitat decreased. Therefore, the results of different tests that support or reject the random sample hypothesis might be due to differences in the proportion of suitable habitat in the landscape. In landscapes where there are no effects of fragment size and isolation the habitat is functionally still continuous and the two hypotheses will give very similar predictions about the effects of habitat fragmentation. The threshold value in proportion of suitable habitat in the landscape where there is a true effect of habitat fragmentation (threshold for fragmentation) will vary across species and landscapes. H.