Community assembly on extensive green roofs: Effects of dispersal‐, abiotic‐ and biotic filtering on the spontaneous species‐ and functional diversity

QUESTIONS: Extensive green roofs (EGRs) are novel ecosystems and essential tools in mitigating the negative effects of urbanization. However, the extent to which traditional community assembly insights apply to spontaneous plant and trait diversity and composition on EGRs and novel ecosystems in general is unclear: (a) is a dispersal filter present because of EGRs’ fragmented nature (e.g., roof height, urban environment); (b) can we confirm the strong abiotic filtering due to extreme EGR conditions (e.g., limited substrate depth); and (c) does the presence of the planted vegetation lead to biotic filtering? LOCATION: The Flanders and Brussels region of Belgium. METHODS: The vegetation of 129 EGRs was sampled during a field campaign. Information on local (e.g., substrate, roof height) and regional characteristics (e.g., connectivity indices) were collected and species traits were collected from multiple databases. All data were used to construct (generalized) linear mixed models that tested the effects of the dispersal‐, local abiotic‐ and local biotic filters on plant community formation. Redundancy analysis was used to investigate predictor effects on functional composition and null models were constructed to assess trait divergence/convergence. RESULTS: Our results show that the spontaneous plant species diversity is not affected by their landscape context (dispersal filtering), while their functional composition is. Additionally, the importance of abiotic filtering is confirmed, as substrate depth and exposure have a strong impact on species diversity and decreasing substrate depth negatively impacts functional diversity. Finally, we show that biotic filtering by increased planted species cover leads to trait divergence, changes in functional composition and reduced cover of spontaneous species. CONCLUSIONS: We find that a broad perspective that includes the regional and local biotic and abiotic environment is needed to assess all factors that influence the vegetation on EGRs. As biotic interactions are present, EGRs should not be considered as static systems.