Green roofs in context: Understanding the impact of the local and regional environment on the diversity and performance of extensive green roofs

The 21st century will be characterized by urbanization, as the number of people living in cities will continue to rise. As cities continue to expand, their impact on the environment does as well. Rising urban temperatures, increased risk of flooding and the loss of habitat of local plants and animals are just a few of the problems with which we will be faced. Nature, however, can also deliver solutions to these issues by providing benefits (or ecosystem services) to mankind. Nature can for instance be re-integrated into cities by adding green roofs to our rooftops. These roofs consisting of plants and substrate ('soil') can deliver numerous benefits such as the cooling of buildings and the reduction of floods while not taking up space for other urban activities and land uses. Especially extensive green roofs are widely applied on new and existing buildings because of their benefits, limited weight (because of the substrate depths of less than 20 cm), low maintenance and relatively low cost.

While these extensive green roofs can play an important role in improving the urban environment, the extent to which they deliver benefits can be hard to predict because of two reasons. First off, extensive green roofs are also ecological test tubes, because they combine species that do not occur together in nature, in an environment that's completely man-made but does not need humans to continue to exist. These test tubes can also be described as 'novel' ecosystems. Because of these novel characteristics, the extent to which classical ecological knowledge applies is sometimes unclear and needs to be tested. Secondly, the vegetation (the plants on the roof, often consisting of mainly succulent species (e.g. Sedum)) can be considered as semi-natural as it develops spontaneously over time and natural processes are at play due to the limited maintenance. This can lead to unforeseen issues, such as the occurrence of weeds on the roofs or gaps in the vegetation. Owners and green roof firms can consider both weeds and gaps as problems and indicators of a reduced overall performance of the entire green roof. In this thesis, we therefore try to fill the gaps in our knowledge of extensive green roofs as novel ecosystems and try to understand weeds and gaps by looking at them from different perspectives and scientific fields.

We start off by trying to answer if (extensive green roof) looks matter, with a specific focus on gaps and weeds aside from other visual green roof characteristics. A discrete choice experiment, a specific type of questionnaire from which the individual effects of changes in characteristics can be derived, was sent out to 155 Flemish respondents to gather information about their preferences. The results of this experiment indicated that looks did indeed matter for extensive green roofs. Gaps were shown to have the highest overall importance out of all characteristics and a strong negative effect on preferences. Roofs with lots of gaps or large gaps are thus considered unwanted. The importance of weeds was lower but still had a negative effect on preferences. Finally, a mixed green roof vegetation, consisting of a mix of standard succulents and herbaceous plants, which can provide a colorful and structurally diverse vegetation was shown to be preferred by the respondents of this study.

Next, we shed light on understudied novel characteristics of extensive green roofs and focus on their community assembly, or how the collection of plants on extensive green roofs is formed over time. Within community assembly, species that occur in the region around the roof are put through a series of three filters before they can successfully reach the green roof and grow on it. The first filter is related to the plant being capable of reaching the green roof (landscape/dispersal filter), while the second filter is made up of environmental conditions such as drought and sunlight (local abiotic filter). The final, third filter is determined by interactions with plants that are already present on the green roof (local biotic filter). To investigate the importance of the different filters, we gathered data about the environmental characteristics and the vegetation on 129 extensive green roofs across Northern Belgium. Firstly, we saw that the spontaneously colonizing plants on green roofs were very similar to the weeds that we find in our gardens. These spontaneously colonizing plants generally showed no difficulty in reaching the roofs (dispersal filter). Secondly, local environmental conditions such as exposure to sunlight acted as a strong filter on species. Finally, a direct competition with planted species for space was found, with further analysis also suggesting that spontaneous species traits were spread out (functional divergence) to fill the ecological niches that are not occupied by planted species (local biotic filter).

Another classical ecological approach is the study of seed banks (the storage of seeds in the soil), which was still unexplored for extensive green roofs and has only seen limited study in other novel ecosystems (e.g. brownfields). In addition to the data collected in the previous chapter, we collected soil seed bank samples on a subset 109 roofs. Our results proved the presence of a seed bank on extensive green roofs and showed similarities in seed density, seed bank versus vegetation similarity and persistence with other novel (urban) systems. Older roof seed banks also contained more species and a higher density of seeds and can be considered as reservoirs of biodiversity. Finally, we showed that the seed bank develops at a slower pace than the vegetation over time.

Building on the data and the insights gathered in the previous chapters, we tried to understand gaps and weeds on extensive green roofs. We found that gaps cover more than one third of the studied roofs on average, while the area covered by weeds was more limited but their species number was higher than the number of species in the planted vegetation. Focusing on the impact of the local abiotic and biotic and the regional environment, we saw that weed cover was lower if roofs were farther from large potential seed source habitats (low proximity) or if local abiotic conditions were stressful (exposure to sunlight and low substrate productivity). For tree species, cover was limited by low proximity and roof height and stressful abiotic conditions. Annual species, however, showed no dispersal limitation and thrived in stressful conditions with a limited range of planted species traits. Finally, gaps were shown to decrease with roof age and increasing soil productivity. Based on these results, we generate recommendations for green roof design and maintenance.

Finally, an exploratory proof of concept for the use of hyperspectral remote sensing for the evaluation of extensive green roof performance was developed, focusing on vegetation cover and species (functional, or trait) diversity. While vegetation cover could easily be assessed with ground-based and airborne hyperspectral measurements using a range of different approaches, only low correlations of diversity with the variation in the spectral measurements were found. Overall, hyperspectral remote sensing could yield practical applications for the monitoring and maintenance of extensive green roofs in the future.

Overall, this thesis sheds light on several understudied fundamentally scientific and practically oriented aspects of extensive green roofs. By considering extensive green roofs multidisciplinary and from a range of perspectives, we conclude that their context is broad (e.g. spatially, temporally, psychologically) and see a need for continued research within this expanded framework to increase and optimize the role of extensive green roofs in ensuring the livability of our increasingly urbanized world.