A mesoscale approach to extinction risk in fragmented habitats

Assessing the fate of species endangered by habitat fragmentation1,2,3 using spatially explicit and individual-based models4,5,6,7 can be cumbersome and requires detailed ecological information that is often unavailable. Conversely, Levins-like8 macroscale models9,10 neglect data on the distribution of local numbers, which are frequently collected by field ecologists11,12,13. Here we present an alternative, mesoscale approach for metapopulations that are subject to demographic stochasticity, environmental catastrophes and habitat loss. Starting from a model that accounts for discrete individuals in each patch and assumes a birth–death stochastic process with global dispersal14,15, we use a negative-binomial approximation16 to derive equations for the probability of patch occupancy and the mean and variance of abundance in each occupied patch17. A simple bifurcation analysis18 can be run to assess extinction risk. Comparison with both the original model and a spatially explicit model with local dispersal proves that our approximation is very satisfactory. We determine the sensitivity of metapopulation persistence to patch size, catastrophe frequency and habitat loss, and show that good dispersers are affected more by habitat destruction than by environmental disasters.