The importance of integrating landscape ecology in habitat models: isolation-driven occurrence of north island robins in a fragmented landscape

Although the role of habitat fragmentation in species declines is well recognised, the effect of habitat quality on species distributions is often studied using presence–absence models that ignore metapopulation dynamics. We compared three approaches to model the presence–absence of North Island robins in 400 sites among 74 fragments of native forest in a 15,000-ha agricultural landscape in New Zealand. The first approach only considered local habitat characteristics, the second approach only considered metapopulation factors (patch size and isolation), and the third approach combined these two types of factors. The distribution of North Island robins was best predicted by patch isolation, as their probability of occurrence was negatively correlated with isolation from neighbouring patches and from the closest major forests, which probably acted as a source of immigrants. The inclusion of habitat factors gave only a slight increase in predictive power and indicated that robins were more likely to occur in areas with tall canopy, tall understory and low density of young trees. We modelled the effect of isolation using an index of functional patch connectivity based on dispersal behaviour of radio-tracked juveniles, and this functional index greatly improved the models in comparison to classical indices relying on Euclidean distances. This study highlights the need to incorporate functional indices of isolation in presence–absence models in fragmented landscapes, as species occurrence can otherwise be a misleading predictor of habitat quality and lead to wrong interpretations and management recommendations.