More than weeds_ Spontaneous vegetation in streets as a neglected element of urban biodiversity

[Audio] Landscape and Urban Planning 185 (2019) 163–172 Contents lists available at ScienceDirect Landscape and Urban Planning journal homepage: www.elsevier.com/locate/landurbplan Research Paper T More than weeds: Spontaneous vegetation in streets as a neglected element of urban biodiversity Sébastien Bonthouxa,⁎, Lolita Voisina, Sabine Bouché-Pillona, Simon Cholletb a CNRS UMR 7324 CITERES, Ecole de la nature et du paysage – INSA Centre Val de Loire, 9 rue de la chocolaterie, 41000 Blois, France b CNRS UMR 6553 ECOBIO, Rennes 1 University, France A R T I C L E I N F O A B S T R A C T Keywords: Pavement flora Urban plant assemblage Pavement design Urban grey space Manual weeding Compared to green spaces, the ecological value of built elements has been largely ignored despite representing the main land cover in cities. In streets, pavements are linear built elements which are omnipresent in all neighborhood types and which extend over large cumulative areas in cities. We assessed the capacity of pavements to promote spontaneous flora and investigated the drivers of pavement plant assemblages. Based on a plant survey along 48 km of pavements in a French city which no longer uses pesticides, we examined the relative importance of multi-scale factors (i.e. landscape context, pavement characteristics and frequency of manual weeding) on different aspects of pavement plant communities including total plant cover, species richness and beta-diversity. More than 300 species were recorded. Plant assemblages were mainly determined by the pavement type, plant cover and species richness being much higher on sandy than on asphalt pavements. There were marked differences in species composition between pavement types with many more species associated with sandy than with asphalt pavements. We found a higher species richness and plant cover on pavements located in commercial and industrial areas than in residential neighborhoods. The effects of weeding frequency and the presence of green space around pavements were marginally important. We demonstrate that pavements with a high level of permeability play a major role in promoting urban biodiversity which should be taken into account by urban planners. We recommend that ecologists work with civil engineers and landscape architects to develop new urban ecological designs. 1. Introduction Greulich, & Bouché-Pillon, 2014; Rupprecht & Byrne, 2014). These studies have shown that habitat size, vegetation structure, management practices and habitat isolation are the main determinants of urban biodiversity patterns and should be integrated in urban planning to foster the goal of biodiversity-friendly cities. Compared to green spaces, the ecological value of built elements (i.e. 'grey spaces') has largely been ignored despite the fact that they represent the largest land cover in cities (Pickett & Cadenasso, 2008). These built urban elements are usually characterized by hard and impermeable surfaces which have considerable effects on soil and hydrological properties and thus on plant and animal diversity (Lundholm, 2015). However, it has recently been demonstrated that new design methods of built elements such as roofs or walls can improve the colonization and maintenance of different taxa (e.g. for green roofs, Madre, Vergnes, Machon, & Clergeau, 2014; Williams, Lundholm, & Scott MacIvor,.

[Audio] S. Bonthoux, et al. Landscape and Urban Planning 185 (2019) 163–172 each pavement section, we first recorded the percentage of total plant cover. We then recorded the presence of all vascular species. Due to their small size or vegetative state, a few plants were difficult to determine and were recorded at the genus level: Sagina (apetala or procumbens), Erigeron (canadensis or sumatrensis), Crepis (bursifolia, capillaris or setosa). Plant nomenclature followed the taxonomic repository TAXREF (Gargominy et al., 2015). When ornamental trees were present on pavements, plants located at their base were not included in the survey. 2.2. Environmental variables In order to explain variations of plant communities among pavement sections, we considered three types of explanatory variables reflecting pavement characteristics, management intensity and landscape characteristics (Fig. 1a, Table 1). 2.2.1. Pavement scale variables We considered four variables collected during the field campaign to characterize pavements (Table 1, Fig. 1a). We first measured the surface area ('Pav.area') of pavement sections where vegetation surveys was conducted. Then we considered the age and the quantity of available soil substrate of the pavement section through the 'Pav.type' variable which was divided into three classes: (1) recent pavements made with asphalt or paving stones with recent cement-based pointing, (2) old asphalt pavements with small cracks or paving stone pavements with sandy or cracked cement-based pointing, (3) pavements made with compacted sand (Fig. 1b). Finally, we noted the presence of ornamental trees on the pavement ('Pav.tree') and the use of land at the edge of the pavement the opposite side to the road ('Pav.edge' in three classes: asphalt (e.g. parking lot), wall or vegetation (e.g. urban grassland)). These two last variables reflect potential sources of propagules in the vicinity of pavements. a major challenge for applied urban ecologists. Among urban built elements, pavements are linear surfaces which are omnipresent in all neighborhood types and which extend along large cumulative areas at the city scale. Pavements are generally only designed for facilitating safe pedestrian circulation with no ecological aims and are built of impermeable materials like asphalt or concrete (but see Säumel, Weber, & Kowarik, 2016). Spontaneous wild vegetation is generally negatively perceived by city dwellers (Nassauer, 1995, but see Bonthoux, Chollet, Balat, Legay, & Voisin, 2019), therefore herbicides are intensely used on pavements to remove weeds leading to low potential for biodiversity conservation. However, growing public and local authority concerns about the negative impacts of pesticides on health and the environment are leading to an increasing number of cities banning the use of pesticides in public spaces (e.g. in France: www.ecophyto-pro.fr; in Europe: www.pesticide-free-towns.info). This change provides a challenge for urban management which has to find alternative weed-control methods, but it also offers an opportunity to develop more biodiversity-friendly cities. Until now biodiversity has been poorly integrated in reflections on pavement design and management due to a lack of scientific knowledge. In fact, whilst some studies have proposed alternative methods to herbicides to limit weed development (De Cauwer et al., 2014a; Fagot et al., 2011; Melander et al., 2009), no study has investigated the potential of pavements to improve urban biodiversity conservation. In this study, based on plant data collected along 48 km of pavements in a French city which no longer uses pesticide, we assessed the capacity of urban pavements to promote spontaneous flora in cities. We examined the relative importance of multi-scale factors on pavement plant assemblages (i.e. total plant cover, species richness and beta-diversity). We.

[Audio] S. Bonthoux, et al. Landscape and Urban Planning 185 (2019) 163–172 Fig. 1. (a) Illustration of variables used in the models to explain plant cover, species richness and beta-diversity of pavement plant communities (n = 884). See Table 1 for details about these variables and (b) Appearance of the three pavement types where plants were recorded. 165.

[Audio] S. Bonthoux, et al. Landscape and Urban Planning 185 (2019) 163–172 Table 1 Description of explanatory variables used in the models to explain plant cover, species richness and beta-diversity of pavement plant communities (n = 884). Level scale Variable name Description Mean ± SD Range Day Julian day of sampling 118 ± 20 91–156 Pavement scale Pav.area Area of the pavement section (m2) 111 ± 118 1.50–1089 Pav.type 3 Classes: recent (n = 323), cracked (n = 438), sandy (n = 123) Pav.tree 2 Classes: presence of trees (n = 93), absence (n = 791) Pav.edge 3 Classes: asphalt (n = 43), wall (n = 598), vegetation (n = 243) Management Management frequency of the pavement 2.97 ± 1.03 1 – 5 Landscape scale Green.500 % Green space in a radius of 500 m 53 ± 13 27–85 Neib.type 6 Neighbourhood classes (see SM 2) City.edge Distance to the city edge (m) 712 ± 397 65–1624 Table 2 Description of the six neighborhood types composing the 'Neib.type' variable (see Supplementary Material 2 for a map). 'n' indicates the number of pavement sections in each type. Neighborhood types 1. Downtown (n = 292): High density of blocks of flats built from the 15th century 2. Western and Eastern Blois (n = 230): Mixture of detached houses and small block of flats built from 1950 to 2000 3. Blocks of flats (n = 128): Low density of blocks of flats built in 1960 4. Detached houses (n = 64): Intermediate density of detached houses with private gardens built in 1990 5. Industrial and commercial area (n = 93): Low density of large industrial buildings built in 1960 6. Southern-Western Blois (n = 77): Mixture of detached houses built in the 19th century and agricultural spaces outside the city. 2.3. Analyses uncertainty in the model selection process (Burnham & Anderson, 2003). To this end, we fitted all possible models nested within the full model and ranked them on the basis of AICc and assigned them Akaike weights (wi). The 'Day' variable was fixed in all candidate models. We then averaged the parameters of the 95% confidence set of models (sum of wi > 0.95) weighted by wi. We considered variables as significant when confidence intervals did not overlap zero. We conducted post hoc pairwise comparisons between classes of significant qualitative variables using Tukey contrasts. Finally, we calculated the percentage of explained variance (explained deviance for GLM) for the most parsimonious model (i.e. with the smallest AIC) of the confidence set. Spatial autocorrelation of residuals was tested for each best model by computing non-parametric spline correlograms using the ncf package for R. Neighboring pavement sections were considered up to a distance of 2000 m. Because a small positive inherent autocorrelation was observed only for very small distances (spatial autocorrelation < 0.2 within 100 m, Supplementary Material 4), we considered the potential effect as negligible. The relative importance of explanatory variables is often calculated from the sum of AIC weights of models included in the confidence set. This approach quantifies the probability that each variable is included in the best model, but it is not useful to quantify effect sizes. For example, an explanatory variable moderately correlated with the response variable can be included in the best models and have a sum of AIC weights close to 1 (Cade, 2015). Alternatively, we adopted a hierarchical partitioning approach to determine the proportion of variance explained independently by each explanatory variable (Chevan & Sutherland, 1991). We performed the analyses in three steps. We firstly examined relationships between three components of pavement plant communities (total plant cover, species richness.

[Audio] S. Bonthoux, et al. Landscape and Urban Planning 185 (2019) 163–172 an area between 0 m2 and 200 m2 (recent pavement n = 283, cracked pavement n = 270, sandy pavement n = 96). The frequency distribution of area was thus comparable between the three types of pavement, indicating that the probability of drawing one pavement section of X m2 was approximatively the same for each pavement type. To obtain robust values of composition comparisons between pavement types, we used a bootstrap approach. We randomly sampled 75 pavement sections of each pavement type. We then summed the number of presences for each species (frequency ranging from 0 to 75) to obtain the species pool associated to each pavement type. We first compared species pools by calculating two components of Bray-Curtis dissimilarity for each pair of pavement types: (1) balanced variation in abundance (Beta.bal), whereby the individuals of some species in one pool were substituted by the same number of individuals of different species in the other pool, (2) abundance gradients (Beta.gra), whereby some individuals of some species were lost from one pool to the other (Baselga, 2013). We then identified plant species associated with a particular pavement type by using the Indicator Value index (IndVal) developed by Dufrene and Legendre (1997). The statistical significance of the indicator values was tested using a permutation test (n permutations = 999). Finally, we used a Wilcoxon test to compare and control areas of pavement sections between the three pavement types. A high Pvalue for this test indicates that the difference of area was low and that the results of comparing species pools were not influenced by differences in size of pavement sections. We repeated this procedure 1 000 times and calculated means and standard deviations of Beta.bal and Beta.gra and the number of times (IndVal analysis) each species was associated with a particular pavement type. Analyses were performed using R 3.2.3 and the following packages: car (Fox, Weisberg, Adler, Bates, Baud-Bovy, Ellison, Firth, Friendly, Gorjanc, & Graves, 2016), ecodist (Goslee, Urban, & Goslee, 2013), MuMIn (Barton, 2015), multicomp (Hothorn et al., 2016), hier.part (Walsh & Mac Nally, 2013), ncf (Bjornstad, 2016), indicspecies (De Caceres & Jansen, 2016) and beta.part (Baselga, Orme, Villeger, De Bortoli, & Leprieur, 2013). 3. Results level of importance for 'Pav.type' and 'Neib.type' with 70% and 20% of importance (I) respectively (Table 3). Plant cover was significantly the highest on sandy pavements and in the industrial neighborhood (Neib.type 5, Fig. 2, Supplementary Material 5). 'Pav.tree' and 'Green.500' effects were also significant but with a much lower importance level (I < 3%; Table 3). Plant cover was significantly higher when trees were present on pavements and it increased significantly with the amount of green space in a radius of 500 m around the pavement sections (Fig. 2). For species richness, four of the 256 candidate models were retained in the confidence set of models to compute the averaged model. Similarly to plant cover, we found a high importance level of 'Pav.type' and 'Neib.type' (I = 41% and 16% respectively; Table 3). Species richness ranked the highest on sandy pavements and in the industrial and mixed residential-agricultural areas (Neib.type 5 and 6, Fig. 3, Supplementary Material 5). Species richness was also strongly positively influenced by 'Pav.area' (Table 3). Species richness was significantly influenced by 'Pav.edge', 'City.edge' and 'Management' but with a much lower importance level (I < 3%; Table 3). Species richness was the.

[Audio] S. Bonthoux, et al. Landscape and Urban Planning 185 (2019) 163–172 Table 3 Estimates and relative importance of explanatory variable effects on plant cover and species richness. Models used were linear model for plant cover and generalized linear model and Poisson distribution for species richness using a model averaging approach (n = 884). The estimated coefficients and standard errors are indicated for quantitative variables. 'X' and 'o' indicate significant and non significant effects respectively for qualitative variables, all modality estimates are given in Supplementary Material 5. 'I' represents the independent effect of each explanatory variable using a hierarchical partitioning procedure. D2 indicates the percentage of total deviance explained by the best candidate model in the model averaging procedure. Plant cover Species richness Estimate ± SE P-value I (%) Estimate ± SE P-value I (%) Intercept 0.357 ± 0.169 < 0.05 2.186 ± 0.069 < 0.001 Day 0.000 ± 0.001 0.62 0.3 −0.001 ± 0.000 < 0.01 0.3 Pav.area Not included in the model 0.235 ± 0.007 < 0.001 34.4 Pav.type X 70.4 X 41.2 Pav.tree X 2.6 o 1.2 Pav.edge o 1.4 X 2.5 Management 0.026 ± 0.029 0.36 0.8 −0.037 ± 0.010 < 0.001 1.1 Green.500 0.156 ± 0.040 < 0.001 1.4 −0.009 ± 0.014 0.50 0.8 Neib.type X 20.6 X 16.3 City.edge 0.063 ± 0.034 0.06 1.6 −0.029 ± 0.011 < 0.01 2.2 D2 total 0.66 0.56 Bold values indicate significant estimates. Fig. 2. Relationships between Plant cover and the significant variables of the averaged model (except for the 'Day' variable, Table 2). 'I' indicates the importance of each variable for explaining Plant cover. For qualitative variables, different letters indicate that effects are significantly different between modalities. In cities, as in other ecosystems, environmental filtering strongly influences the possibility of plants to disperse and recruit successfully, resulting in the opportunity to improve conditions for biodiversity. In fact, factors conditioning spontaneous pavement vegetation could be used by architects, urban planners and managers to promote biodiversity in cities. davidii, Solidago canadensis, Ailanthus altissima in Paris; Muratet, Machon, Jiguet, Moret, & Porcher, 2007) probably have their development limited by soil availability, pedestrian trampling and weed management of pavements. The variation in community parameters are surprisingly predictable since our model for species richness and plant cover explained 56% and 66% respectively of the total recorded variability. These results are particularly unexpected since urban communities are often considered to be highly influenced by stochastic events due to regular unpredictable disturbances through human activities (Sattler et al., 2010). 168.

[Audio] S. Bonthoux, et al. Landscape and Urban Planning 185 (2019) 163–172 Fig. 3. Relationships between plant species richness and the significant variables of the averaged model (except for the 'Day' and 'Area' variable, Table 2). 'I' indicates the importance of each variable for explaining species richness. For qualitative variables, different letters indicate that effects are significantly different between modalities. 4.1. Major levers of pavement flora and applications for design and planning hosted the most diverse species pool. Indeed, we found strong differences in species composition (beta-diversity partitioning) between pavement types, with asphalt pavement species pools nested in the sandy pavement species pool. This result is confirmed by IndVal analysis which indicated that 43 species were specifically associated with sandy pavements, while only two species were associated with old asphalt and none with recent asphalt pavements. Many of these species associated with sandy pavements are also found in other urban ecosystems such as urban grasslands or wastelands (Godefroid, Monbaliu, & Koedam, 2007; Rudolph et al., 2017), but also in ecosystems outside cities (e.g. field margins, Alignier & Baudry, 2015). Among these species some offer pollen and nectar resources and could help to support the populations of pollinators in cities (e.g. Cirsium arvense, Hypochaeris radicata, We found pavement type to be the strongest determinant of plant cover and species richness. Sandy pavements had an average plant cover 30 times greater than and double the richness of asphalt pavements. These results concur with experimental studies (De Cauwer et al., 2014a, 2014b) which suggest that permeability level and consequently soil availability are important parameters to explain plant distribution on pavements. On sandy pavements, vegetation can grow on the entire non-trampled surface, while it can only develop along walls or in cracks of asphalt pavements. In addition to higher species richness and cover, sandy pavements 169.

[Audio] S. Bonthoux, et al. Landscape and Urban Planning 185 (2019) 163–172 Table 4 Results of the multiple regression model between plant beta-diversity and differences in explanatory variables among the pavement sections (n = 884). Estimate P-value R2 Intercept 0.585 ns Day 0.001 < 0.001 0.03 Pav.area −0.004 0.60 0.00 Pav.type 0.044 < 0.001 0.02 Pav.tree −0.001 0.95 0.00 Pav.edge 0.038 < 0.001 0.02 Management 0.015 < 0.001 0.01 Green.500 0.077 < 0.001 0.01 Neib.type 0.029 < 0.001 0.01 City.edge 0.000 0.73 0.00 R2 total 0.10 Bold values indicate significant estimates. variables in our models we suggest two alternative hypotheses that may explain our result. First, during the field campaign we observed several people weeding pavements in front of their homes. This behavior probably reflects an individual appropriation of the pavement that does not exist in industrial areas. Secondly, in commercial areas where the use of cars is prevalent, pavements are little frequented by pedestrians resulting in lower trampling intensity and higher vegetation development. This result agrees with a study in Paris in which the authors suggest that great opportunities exist to develop habitat availability and connectivity through green spaces on business sites (Serret et al., 2014). The effects of all the parameters linked to dispersal limitation and connectivity (i.e. presence of trees on pavements, pavement edges, percentage of green space around pavements and distance to the city edge) were very small. These findings suggest that compared to habitat quality represented by pavement type, the ability of seeds and propagules to disperse in urban mosaics is not a limiting factor for pavement plants. A similar low effect of landscape context on local communities has been found in other types of urban elements such as wastelands, lawns and parks (Bertoncini et al., 2012, 2014; Shwartz, Muratet, Simon, & Julliard, 2013). Promoting high quality habitats should thus be the main priority in urban planning to conserve spontaneous flora in cities. 4.2. Lessons for urban management Achillea millefollium, Trifolium repens, Bellis perennis, Hicks et al., 2016; Verboven, Aertsen, Brys, & Hermy, 2014). Pavements are often wider than the width required for pedestrians (e.g. in Blois, 28 km of the 48 km of pavements inventoried are wider than 1.50 m which is the recommended width). On these large pavements, a new design including some permeable soil (e.g. sand or natural soil) will, in addition to a capacity to host spontaneous vegetation, provide multiple advantages such as increasing water infiltration (Säumel et al., 2016). Pavements with an appropriate design could also host meadow-like areas sown with grassland species as has been proposed for wastelands (Fischer, von der Lippe, Rillig, & Kowarik, 2013). The location of pavements within the city is another factor that urban planners should take into account when designing new biodiversity-friendly cities. We found a higher species richness and plant cover on pavements located in commercial and industrial areas than in any other neighborhood type (which were all residential, Supplementary Material 2). Melander et al. (2009) found a comparable result and explained this by a low frequency of management by technical services and a high concentration of green spaces in these areas. However, due to the fact that we explicitly integrated these two Despite the large human and financial investment in manual plant control set up by urban managers, the effect of weeding frequency on the plant community was marginal compared to the pavement type and to a lesser degree compared to the neighborhood type. In addition, although species richness decreased with higher weeding frequency, we found no effect on total plant cover, while the main objective of the technical services is still to reduce plant cover. Management practices are strong.

[Audio] S. Bonthoux, et al. Landscape and Urban Planning 185 (2019) 163–172 Bjornstad, O.N. (2016). Package 'ncf'. Bonthoux, S., Brun, M., Di Pietro, F., Greulich, S., & Bouché-Pillon, S. (2014). How can wastelands promote biodiversity in cities? A review. Landscape and Urban Planning, 132, 79–88. Bonthoux, S., Chollet, S., Balat, I., Legay, N., & Voisin, L. (2019). Improving nature experience in cities: What are people's preferences for vegetated streets? Journal of Environmental Management, 230, 335–344. Borcard, D., Legendre, P., & Drapeau, P. (1992). Partialling out the spatial component of ecological variation. Ecology, 73, 1045–1055. Burnham, K. P., & Anderson, D. R. (2003). Model selection and multimodel inference: A practical information-theoretic approach. Springer Science & Business Media. Cade, B. S. (2015). Model averaging and muddled multimodel inferences. Ecology, 96, 2370–2382. Cadotte, M. W., Yasui, S. L. E., Livingstone, S., & MacIvor, J. S. (2017). Are urban systems beneficial, detrimental, or indifferent for biological invasion? Biological Invasions, 19, 3489–3503. Carboneras, C., Genovesi, P., Vilà, M., Blackburn, T. M., Carrete, M., Clavero, M., ... Wynde, R. (2018). A prioritised list of invasive alien species to assist the effective implementation of EU legislation. Journal of Applied Ecology, 55, 539–547. Chevan, A., & Sutherland, M. (1991). Hierarchical partitioning. The American Statistician, 45, 90–96. Cordier, J., Dupré, R., & Vahrameev, P. (2010). Catalogue de la flore sauvage de la région Centre. Symbioses, 26, 36–84. Weeding of public spaces can represent a large part of the annual municipal budget. In many countries the use of herbicides is being greatly reduced and implies developing new approaches for urban plant management. Furthermore, a recent social study shows that people can accept the moderate presence of spontaneous vegetation on pavements (French respondents, Bonthoux et al., 2019). That is why the previous objective of having no plant growth on pavements achieved by using herbicides must be reconsidered. In line with our results indicating that pavements could host significant plant diversity combined with the limited potential for manual weeding to control plant development, indicate that differentiated management strategies could be set up. These strategies are developed in green spaces and propose a globally biodiversity-friendly management with management frequency varying between spaces according to the urban dwellers' uses (Shwartz et al., 2013). For pavements, urban managers could for example concentrate efforts on reducing plant growth in areas of specific importance such as historical centers, but promote installation of plants by developing extensive management (i.e. mowing/weeding two or three times per year) in commercial and industrial areas. 5. Conclusion: using pavement webs to spread vegetation through the urban matrix Cornelis, J., & Hermy, M. (2004). Biodiversity relationships in urban and suburban parks in Flanders. Landscape and Urban Planning, 69, 385–401. De Caceres, M., & Jansen, F. (2016). Package 'indicspecies'. De Cauwer, B., Fagot, M., Beeldens, A., Boonen, E., Bulcke, R., & Reheul, D. (2014b). Effectiveness of water permeable joint filling materials for weed prevention in paved areas. Weed Research, 54, 532–540. De Cauwer, B., Fagot, M., Beeldens, A., Boonen, E., Bulcke, R., & Reheul, D. (2014a). Reduced weed growth with different paving constructions. Weed Research, 54, 151–161. Dormann, C. F., Elith, J., Bacher, S., Buchmann, C., Carl, G., Carré, G., ... Leitão, P. J. (2013). Collinearity: A review of methods to deal with it and a simulation study evaluating their performance. Ecography, 36, 27–46. Dufrene, M., & Legendre, P. (1997). Species assemblages and indicator.

[Audio] S. Bonthoux, et al. Landscape and Urban Planning 185 (2019) 163–172 Potential contributions of green spaces at business sites to the ecological network in an urban agglomeration: The case of the Ile-de-France region, France. Landscape and Urban Planning, 131, 27–35. Mitchell, M. G., Wu, D., Johansen, K., Maron, M., McAlpine, C., & Rhodes, J. R. (2016). Landscape structure influences urban vegetation vertical structure. Journal of Applied Ecology, 53, 1477–1488. Muratet, A., Machon, N., Jiguet, F., Moret, J., & Porcher, E. (2007). The role of urban structures in the distribution of wasteland flora in the greater Paris area, France. Ecosystems, 10, 661. Seto, K. C., Güneralp, B., & Hutyra, L. R. (2012). Global forecasts of urban expansion to 2030 and direct impacts on biodiversity and carbon pools. Proceedings of the National Academy of Sciences, 109, 16083–16088. Shah, M. A., Callaway, R. M., Shah, T., Houseman, G. R., Pal, R. W., Xiao, S., ... Chen, S. (2014). Conyza canadensis suppresses plant diversity in its nonnative ranges but not at home: A transcontinental comparison. New Phytologist, 202, 1286–1296. Nassauer, J. I. (1995). Messy ecosystems, orderly frames. Landscape Journal, 14, 161–170. Pickett, S. T., & Cadenasso, M. L. (2008). Linking ecological and built components of urban mosaics: An open cycle of ecological design. Journal of Ecology, 96, 8–12. Rasool, N., Reshi, Z. A., Khasa, D. P., Roshan, M., & Shah, M. A. (2016). Invasion by Conyza sumatrensis alters soil microbial community structure in urban ecosystems. Ecological Processes, 5, 10. Shwartz, A., Muratet, A., Simon, L., & Julliard, R. (2013). Local and management variables outweigh landscape effects in enhancing the diversity of different taxa in a big metropolis. Biological Conservation, 157, 285–292. Stagoll, K., Lindenmayer, D. B., Knight, E., Fischer, J., & Manning, A. D. (2012). Large trees are keystone structures in urban parks. Conservation Letters, 5, 115–122. Rudolph, M., Velbert, F., Schwenzfeier, S., Kleinebecker, T., & Klaus, V. H. (2017). Patterns and potentials of plant species richness in high-and low-maintenance urban grasslands. Applied Vegetation Science, 20, 18–27. Vallet, J., Beaujouan, V., Pithon, J., Rozé, F., & Daniel, H. (2010). The effects of urban or rural landscape context and distance from the edge on native woodland plant communities. Biodiversity and Conservation, 19, 3375–3392. Rupprecht, C. D., & Byrne, J. A. (2014). Informal urban greenspace: A typology and trilingual systematic review of its role for urban residents and trends in the literature. Urban Forestry & Urban Greening, 13, 597–611. Vallet, J., Daniel, H., Beaujouan, V., & Rozé, F. (2008). Plant species response to urbanization: Comparison of isolated woodland patches in two cities of North-Western France. Landscape Ecology, 23, 1205–1217. Sattler, T., Borcard, D., Arlettaz, R., Bontadina, F., Legendre, P., Obrist, M. K., & Moretti, M. (2010). Spider, bee, and bird communities in cities are shaped by environmental control and high stochasticity. Ecology, 91, 3343–3353. Säumel, I., Weber, F., & Kowarik, I. (2016). Toward livable and healthy urban streets: Roadside vegetation provides ecosystem services where people live and move. Environmental Science & Policy, 62, 24–33. Schielzeth, H. (2010). Simple means to improve the interpretability of regression coefficients. Methods in Ecology and Evolution, 1, 103–113. Verboven, H. A., Aertsen, W., Brys, R., & Hermy, M. (2014). Pollination and seed set of an obligatory outcrossing plant.