Indicators for assessing the robustness of metapopulations against habitat loss

Habitat loss and fragmentation resulting from environmental changes are main drivers of global biodiversity loss, as the survival of metapopulations relies on the ability of individuals to disperse among suitable habitat patches. To prioritize conservation efforts, methods are needed for evaluating the robustness of metapopulations against habitat loss. We therefore investigate this robustness for different degrees of habitat loss, for different types of habitat loss (random, peripheral, and contagious) and of habitat networks, and for species differing in their local-extinction risks and dispersal ranges. In particular, we analyse several standard network types (with random, regular, small-world, or scale-free structure) and compare them with several alternative network types derived from real-world two-dimensional habitat landscapes (with random, clustered, or contiguous habitat allocation). Furthermore, we investigate how well 29 different graph-theoretic metrics of habitat networks can serve as indicators of metapopulation robustness against habitat loss – as this approach, where feasible, allows replacing complex simulation-based predictions with simple indicator-based predictions. We find that responses of species to habitat loss on the considered landscape-based habitat networks qualitatively differ from those on the considered standard habitat networks. This suggests that results obtained for the latter, albeit widely examined in the literature, can be unrepresentative and misleading. As expected, species with high risks of local extinction and short dispersal ranges are particularly vulnerable to habitat loss, across all considered types of habitat loss and habitat networks. The graph-theoretic network metric that best explains the robustness of metapopulations against habitat loss depends on the considered types of species, habitat networks, and habitat loss. None of the examined metrics give consistently reliable predictions under all circumstances. For sensitive species, characterized by high local-extinction risks and short dispersal ranges, a network’s average clique size, redundancy, average degree, connectance, clustering coefficient, and average closeness centrality are the best indicators of metapopulation robustness. For landscape-based habitat networks, a network’s average clique size, beta coefficient, clustering coefficient, redundancy, and cyclomatic number work best. For contagious habitat loss, the network type has a particularly strong impact on the species-specific robustness against habitat loss. In summary, our study introduces a method for evaluating the robustness of metapopulations against habitat loss and shows that a network’s clustering coefficient, under a wide range of circumstances, is a particularly reliable indicator of this robustness