From the perspective of landscape ecology, an extensively vegetated roofscape with minimal disturbance from management may serve to understand natural processes that occur in urban ecosystems. This research surveyed a sample of extensive green roofs (EGRs) two to three decades after installation in Germany, with the objective of characterising their vegetation and considering the services they can offer urban biodiversity and ecosystem function over the long-term. Some of the roofs sampled were prototypes of the systems that have since attained international commercial success. Species composition and abundance were quantified using a 1m2 quadrat and methods of applied plant ecology. Functional traits were allocated to all plant species, both current and original, and the realised niches were described using Ellenberg indicator values (EIVs). How have original species lists fared over the span of decades, in terms of persistence or loss, and in cover abundance? Does well-established EGR vegetation exhibit particular functional traits (competitor, stress-tolerator, ruderal)? Does the vegetation of unmanaged EGRs eventually converge at a certain character regardless of original lists, or does habitat heterogeneity lead towards divergence and random assemblages?
Contrasting the functional composition of original and current species lists revealed successional trajectories in which the vegetation shifted from initially diverse assemblages towards a resilient community of stress-tolerators and ruderals. This trend was reinforced by EIVs indicating that the surveyed vegetation comprised species associated with direct sunlight and extreme annual fluctuations in temperature and climate. While these results confirm the obvious, they also define the ecological novelty that shapes EGR ecosystems and that thereby determine the trophic complexity these systems can support. By integrating the general causes, specific mechanisms, ecological filters and feedbacks known to occur on EGRs, we propose a model that predicts natural succession on these systems. Testing this with empirical data, the model illuminates the effects of stress and disturbance on species and functional composition over the long-term. Pragmatically, these results suggest an approach for optimizing the habitat value of green roofs, which could be useful for bioregions in which EGRs are new. Designers should be encouraged to consider and specify the role of dynamism on EGR installations. A new frontier for green roof research lies in the context of collaboration and partnership with urban planning and development, whereby the construction of EGRs as replicated units could serve as designed experiments with provisions for access and for the long-term monitoring of actual roofs.