Buzzing on top: Linking wild bee diversity, abundance and traits with green roof qualities

Buzzing on top: Linking wild bee diversity, abundance and traits with green roof qualities Green roofs are potentially valuable habitats for plants and animals in urban areas. Wild bees are important pollinators for crops and wild plants and may be enhanced by anthropogenic structures, but little is known about wild bees on green roofs in cities. This study investigates the effects of green roof qualities (floral resources, substrate character and depth, roof height and age) on wild bee diversity, abundance and traits (nesting type, sociality, pollen specialisation, body size) on green roofs in Vienna. Nine green roofs were sampled monthly between March and September 2014 by a semi quantitative approach. Wild bees were collected in pre-defined sub-areas for the same amount of time and floral resources were recorded. Over all green roofs, 992 individuals belonging to 90 wild bee species were observed. Wild bee diversity and abundance was strongly positively affected by increasing forage availability and fine substrates. Wild bees on roofs were characteristically solitary, polylectic and 8.3– 11.2 mm. Regarding nesting type, the percentage of above-ground nesting bees was higher compared to the common species composition in Middle Europe. Ground-nesting wild bees were mainly eusocial, smaller (6.4–9.6 mm) and positively affected by roofs with fine substrates. During June, when forage availability by wildflowers on roofs was Blow^ (5–15% flower coverage), flowering Sedum species were an important forage resource. We conclude that wild bee diversity and abundance on green roofs are enhanced by floral resources. Furthermore, the installations of areas with finer and deeper substrates benefit ground nesting and eusocial wild bees. . . . . . Keywords Vegetated roofs Apiformes Floral resources Urban biodiversity Pollinators Vienna Introduction colonize habitats successfully (Westrich 1996; Gathmann and Tscharntke 2002; Zurbuchen et al. 2010;Wojcikand The degradation and loss of habitats are seriously threaten- McBride 2012). Bees are ectothermic and thermophile or- ing wild bees (Winfree et al. 2009;LeFéonet al. 2010; ganisms, thus the warmer (micro-) climate of cities enables Potts et al. 2010; Kennedy et al. 2013), leading to declines favourable habitat conditions for many wild bee species in pollination services for crops (Kremen et al. 2002;Klein (Cane 2005; Hennig and Ghazoul 2012). Research on urban et al. 2007) and wild plants (Biesmeijer et al. 2006; sites likely to be populated by wild bees, provides impor- Fontaine et al. 2006). As important pollinators in gardens, tant information for urban planning to enhance wild bee parks and other green spaces (Matteson et al. 2008;Frankie diversity (Hernandez et al. 2009). et al. 2009; Matteson and Langellotto 2010), wild bees play Sustainable city planning considers green roofs as a vital a crucial role for urban ecosystems. The availability of pol- measure to compensate negative effects of sealed surfaces. len and nectar as well as nesting sites within species- Many abiotic benefits have been reported (reviewed in specific flying distances is important for wild bees to Oberndorfer et al. 2007) like the improvement of city’s water run off management (Mentens et al. 2006; Hathaway et al. 2008;Berndtsson 2010), the reduction of heat islands (Susca et al. 2011) and the increase of energy * Sophie Kratschmer efficiency by insulating the building’s indoor rooms against sophie.kratschmer@boku.ac.at heat and cold (Castleton et al. 2010; Zinzi and Agnoli 2012). Further benefits are aesthetic and amenity for urban Institute for Integrative Nature Conservation Research, University of citizens (van den Berg et al. 2007). From an ecological Natural Resources and Life Sciences, Vienna (BOKU), Gregor-Mendel-Straße 33, 1180 Vienna, Austria point of view, green roofs have high potential to restore 430 Urban Ecosyst (2018) 21:429–446 habitats for plants and animals in cities (Oberndorfer et al. (2) Characteristic wild bee traits (pollen specialization, 2007; Carter and Butler 2008; Dunnett and Kingsbury nesting type, sociality and body size) on green roofs 2008). During the last decade, studies on plants (Landolt and how these traits are affected by respective green roof 2001;Köhler 2005), birds (Gedge 2003;Baumann 2006; qualities. Fernandez-Canero and Gonzalez-Redondo 2010)and vari- (3) The importance of Sedum species compared to other ous invertebrate taxa like Araneae (Brenneisen 2003; wildflowers as foraging resource on green roofs. Kadas 2006), Collembola (Schrader and Böning 2006), Coleoptera (Kadas 2006) and Hymenoptera (Brenneisen 2003; Brenneisen 2005; Kadas 2006; Colla et al. 2009; MacIvor and Lundholm 2010; Tonietto et al. 2011; Methods Ksiazek et al. 2014;MacIvoretal. 2014) underpinned the importance of green roofs to contribute to biodiversity and Study sites nature conservation in cities. Green roofs are potentially important habitats for urban We studied wild bees on nine green roofs in Vienna (Fig. 1) wild bees, as they provide pollen and nectar resources during the vegetation period of 2014. In order to evaluate the throughout the year (Tonietto et al. 2011) and incorporate wild bee diversity and abundance on roofs with different qual- different nesting habitats (Brenneisen 2005; MacIvor et al. ities (i.e. substrate characteristics and depths, roof height and 2014). In addition, the higher insulation and thus warmer age), the roofs varied in greening type (extensive, semi- micro-climate at roof level result in favourable habitat con- extensive and intensive) leading to different planting possibil- ditions (Matteson and Langellotto 2010). Beside, a contin- ities (Dunnett and Kingsbury 2008) and floral resources (e.g. uous vegetation layer on green roofs is crucial to the abiotic Sedum species and other wildflowers). benefits discussed above and linked to the pollination by Detailed information about the roof’s qualities (Table 1) wild bees (Dunnett and Kingsbury 2008). was gathered during wild bee sampling or by interviewing Similar wild bee communities have been reported on the people responsible for the buildings. Substrate character- green roofs and ground level habitats in Toronto (Colla istics were determined in the field. Coarse substrates consisted et al. 2009) and Chicago (Tonietto et al. 2011) but higher of high amounts of expanded clay and crashed brick and are wild bee diversity on ground sites was related to higher typically used on extensive green roofs. Fine substrates entomophilous plant diversity. The habitat quality of consisted of a high amount of humus, earth and sand and are green roofs for wild bees was highlighted by observa- mainly used on intensive green roofs. Mixed substrate was tions of locally and/or nationally rare species (Kadas present on roofs where coarse and fine substrates were used 2006) and of wild bee species newly recorded on green in different areas. Substrate depths, height and age were roofs (Ksiazek et al. 2014;MacIvoretal. 2014). Green roof types support wild bee diversity differently because of the type of substrate and its character and plant diver- sity (Brenneisen 2005). Furthermore, Sedum species are characteristic green roof plants and play an important role as a foraging resource for wild bees (MacIvor et al. 2014). Whilst there has been research in Western Europe on green roofs and wild bees, knowledge about wild bees on green roofs in the eastern part of Europe, characterized by a warmer and dryer climate and therefore comprising a dif- ferent wild bee species composition is scarce. In this paper, we report the wild bee diversity, abundance and traits in relation to green roof qualities on green roofs in Vienna. The main focus of this study was to determine: (1) The effect of temporal variable floral resources (forage availability and entomophilous plant diversity) and con- stant green roof qualities (substrate characteristics and Fig. 1 Locations of the nine green roofs in Vienna surveyed in 2014 depths, roof heights and ages) on wild bee communities. (basic map: ViennaGIS 2016) Urban Ecosyst (2018) 21:429–446 431 Table 1 Green roof qualities for data analyses: Height: measured from humus, earth or sand; mixed = areas with coarse and fine substrate; street level (m); Age: years since first greening until 2014; Substrate: Sub-areas: pre-defined for sampling by types of plants or structural char- Depth (cm) and characteristics expressed by components: coarse = high acteristics. Additional information: Site features and % of sealed area in a percentage of expanded clay/crashed bricks; fine = high amount of 500 m radius (Stadt Wien Vienna GIS 2005) Site Coordinates Area (m ) Height (m) Age (year) Substrate (cm) Sub-areas Site features Sealed area (%) 1 400 9.5 4 10 BWildflowers^ Inner-city; 79.6 48.17935; coarse BSedum^ higher buildings around 16.326241 full sun until 4 pm 2 750 15 9 8 BWildflowers^ Inner-city; 68.6 48.23655; coarse BSedum^ higher buildings around; 16.379753 full sun until 4 pm 3 200 25 16 8–90 BWildflowers^ Inner-city; 75.6 48.189551; mixed BLawn^ no higher buildings around; 16.372002 BShrubs^ full sun all day BSedum^ 4 231 16 23 20–25 BWildflowers^ Inner-city; 83.9 48.216574; fine BLawn^ no higher buildings around; 16.329165 BShrubs^ full sun all day; BUnmaintained area^ 5 1000 12 18 20–50 BAlpine^;BBBQ spot^ Inner-city; 65.4 48.195187; fine BFruit trees^; BWildflowers^; no higher buildings around; 16.304639 BLawn^; BVegetables^ full sun all day BPannonic^; BShrubs^; 6 8800 11 16 5–7 BSedum^ Outskirt; 46.7 48.139076; coarse no higher buildings around; 16.366834 full sun all day 7 400 5 6 10–35 BWildflowers^ Inner-city; 68.5 48.236503; mixed BShrubs^ higher buildings on S/W; 16.379654 BPond^ full sun 10 am to 2 pm 8 1500 8 16 8–12 BBare substrate^ Outskirt; 47.1 48.266850; coarse BWildflowers^ no higher buildings around; 16.466844 BSedum^ full sun all day BDead wood^ 9 420 8 2 10 BWildflowers^ Inner-city; 83.3 48.222610, mixed BSedum^ higher buildings around; 16.333602 BShrubs in pots^ full sun until 4 pm determined from planning documents or by interviewing peo- or vegetation types (e.g. wildflowers, Sedum, lawn, shrubs, ple responsible for the building. Pannonian plants), structural characteristics (e.g. bare sub- Vienna is characterised by a temperate Pannonian climate, strate, dead wood elements) or functionality (e.g. BBQ spot, with the potential for precipitation throughout the year (Auer unmaintained area). To assess the value of Sedum as foraging and Böhm 2011). The average annual precipitation in 2014 resource for wild bees, it was defined as a distinct sub-area. was 756 mm and the average monthly temperature was Patches with spontaneous flora and herbaceous garden plants, 12.5 °C. The highest average daily temperature was recorded except for Sedum, were classified as BWildflowers^. Sampling during July and early August and ranged between 21 and time was adapted to the size of the sub-areas and ranged be- 23 °C (ZAMG - Zentralanstalt für Meterologie und tween 3 min for small objects like dead wood elements and Geodynamik 2015). 15 min for large areas like lawns, shrubs, Sedum or wild- flowers. Each roof was sampled monthly, seven times from March to September 2014 (Schindler et al. 2013). Sampling Wild bee sampling was conducted between 10 am and 4 pm on days with warm (22.8 ± 4.9 °C), windless and dry weather conditions. On each Wild bee sampling was conducted by a semi-quantitative sampling date, information on floral resources was carried out. method, hand netting wild bees in each of the pre-defined The forage availability (=flower coverage) of entomophilous sub-areas (Table 1) for the same period of time. The sub- plants, was assessed in each sub-area (Table 1) using five areas were distinct patches characterized by dominating plants 432 Urban Ecosyst (2018) 21:429–446 categories: 1 = Bvery low^ (< 5%), 2 = Blow^ (5–15%), 3 = B Data analyses medium^ (15–25%), 4 = Bhigh^ (25–50%) and 5 = Bvery high^ (> 50%). To assess the number of entomophilous plant We excluded Apis mellifera from analysis, because it is a species flowering at the time of sampling, plants were domestic species and the abundance could be biased by near- photographed and identified to species level. This was sepa- by hives (Kennedy et al. 2013). We only found a weak, insig- rately documented for each sub-area. On average (± SD), the nificant positive relation between honey bee and wild bee roofs comprised Bvery low^ to Bmedium^ forage availability abundance on the roofs (Spearman’s rank correlation; rho = (2 ± 0.6) and 46.8 (± 39) entomophilous plant species. 0.08; p = 0.8). Roof size was neither related to wild bee spe- Evidence of nesting activity was gathered qualitatively during cies richness (rho = −0.12; p = 0.7) nor to wild bee abundance wild bee sampling. Bare substrate patches were observed for (rho = −0.23; p = 0.5). Statistical analyses were performed in ground nesting activity and old plant stems, cavities in walls R 3.3.2 (R Core Team 2016) using R Studio V 0.99.903 or dead wood branches for above-ground nesting activity. (RStudio Team 2015). Wild bees were identified to species level (Ebmer 1969, The effects of temporal variable floral resources on wild 1970, 1971; Dathe 1980; Mauss 1994; Schmid-Egger and bee species richness and abundance were analysed by gener- Scheuchl 1997; Amiet et al. 1999, 2001;Scheuchl 2000, alized linear mixed models (GLMMs) with Poisson error dis- 2006; Gokcezade et al. 2010) by the authors (SK, BP) and tribution using the R-packages Blme4^ (Bates et al. 2015)and Karl Mazzucco was consulted for validation of some speci- BMatrix^ (Bates and Maechler 2016). We formulated null and mens mainly of the genus Lasioglossum and Hylaeus. All candidate models for each response variable with two random specimens are housed in the collection of the Institute for factors (Bmonth^ and Broof^) and either forage availability or Integrative Nature Conservation Research at BOKU Vienna. entomophilous plant species richness as fixed factor. We did Wild bee traits (Table 2)weresummarized by categories not formulate models containing both fixed factors because and determined by literature research (Westrich 1989a; prior data exploration revealed them as collinear (cor = 0.6), Michener 2007; Scheuchl and Willner 2016). Information on which results in unreliable parameter estimation (Zuur et al. body size was derived from identification literature (see 2010). For GLMMs with forage availability as fixed factor, above; bumblebees: von Hagen and Aichhorn 2003)which we used the category Bvery low^ as baseline for parameter give the range of body size within a species. For trait analysis estimation. Model selection was carried out by the second the average body size was calculated from the female and order Akaike Information Criterion (AICc) using the R pack- male values. The nesting type of Hylaeus imparilis remained age BAICcmodavg^ (Mazerolle 2016). TheAICcisusedfor unclear, but was attributed to the above-ground nesting group, modelling data with small sample size (Motulsky and like all other Hylaeus species (Falk 2015; Scheuchl and Christopoulos 2003). The cut-off to decide whether a model Willner 2016). is more likely to be correct than the next one was set at ΔAICc Table 2 Definitions, variable structure of wild bee traits and explanatory variable selection for trait analysis with generalized linear models (GLM); primitively eusocial: Beusocial^ hereafter Traits Variable type Definition Explanatory variables for GLM trait analysis Nesting type Ground nesting Excavate nests in the ground Substrate quality Above-ground Nesting in cavities, plant stems, dead wood or build on structures Substrate depths (cm) (incl. Bombus spp.) Roof height (m) Roof age (years) Sociality Solitary fertile ♀ nest and breed alone Substrate quality Primitively eusocial fertile ♀ establish the nest and 1st generation of workers Substrate depths (cm) initializing division of labour Roof height (m) Roof age (years) Parasitic fertile ♀ lay their eggs in nests of certain host species Body size Continuous variable Mean body size was averaged from range values in identification Substrate quality literature (see section BWild bee sampling^) Substrate depths (cm) Roof height (m) Roof age (years) Pollen specialization polylectic no pollen specialization Forage availability oligolectic pollen specialization on a plant family or genus No.of flowering plant species Urban Ecosyst (2018) 21:429–446 433 <2. We did not use p-values for interpretation because they averaged forage availability along a timeline plot created with are not reliable in GLMMs (Zuur et al. 2013). Microsoft Excel 2010. The effects of uniform green roof qualities (substrate char- acteristics and depths, roof height, age) on wild bee species richness and abundance were fitted with Poisson generalized Results linear models (GLMs). For GLMs the wild bee species num- bers and abundance were aggregated to total amounts per roof. Excluding A. mellifera (1470 individuals), which was present Substrate depths were averaged if a range was indicated on all roofs, we identified 90 wild bee species belonging to 19 (Table 1) and the factor level Bcoarse^ was the baseline for genera. A total of 992 wild bee individuals were caught on the parameter estimation of substrate characteristics. green roofs (see Table 6 in Appendix for species list). The Characteristic traits on green roofs were examined by cal- most abundant species was Halictus subauratus with 94 indi- culating community weighted means (CWM) with the R- viduals (9.5%), whereas 25 wild bee species (27.8%) were package BFD^ (Laliberté et al. 2014). To analyse how traits represented by only one individual. responded to green roof qualities, aggregated species numbers On average (± SD), green roofs hosted 29 (± 16.2) wild bee and abundance per trait (nesting type, sociality and pollen species and 110 (± 95.3) individuals. The most diverse roof specialization) per roof were used as response variables in with 65 wild bee species (337 individuals) contrasted with a Poisson GLMs. The CWM of body size was used as response roof that had only six species and six individuals. in GLMs with Gaussian error distribution. To reduce the num- ber of GLMs for trait analysis, we chose the most interesting Green roof qualities green roof qualities as explanatory variables for each trait (Table 2) based on our expertise. Because almost all explana- Wild bee species richness and abundance was best predicted tory variables were collinear (cor. ranged from 0.3 to 0.8), we by forage availability (Table 3). only formulated GLMs with one explanatory variable. Model Forage availability had a strong positive effect on wild bee selection was done on AICc bases as described above. The species richness and abundance (Fig. 2a, b; Table 7 in explained deviance (R ) was calculated for the most ac- Appendix). Species richness and abundance increased from GLM curate GLMs to assess how much variation of the response is Bvery low^ to Blow^ and further to Bmedium^ forage avail- explained by the explanatory variable (Zuur et al. 2013). ability, but from Bmedium^ to Bhigh^ the positive effect was Model validation of GLMMs and GLMs was performed by minimal. diagnostic plots and dispersion values (Zuur et al. 2013). The The models with substrate characteristics as an explanatory package Beffects^ (Fox 2003) was used to visualize the effects variable were the most accurate for predicting wild bee species of the most accurate models. richness and abundance in relation to uniform green roof qual- To assess whether Sedum and other wildflowers provided ities (Table 4). crucial floral resources, a data sub-set of the sub-areas Roofs with fine substrates represented the highest mean (± BSedum^ and BWildflowers^ from five roofs (Table 1)was SD) wild bee diversity (47.5 ± 24.7) and abundance (232.5 ± analysed. The other green roofs had to be omitted from this 147.8) and affected wild bee species richness (Fig. 3a) and analysis because they did not include both sub-areas. Due to abundance (Fig. 3b) distinctively positive compared to roofs small sample size this was done descriptively by comparing with mixed or coarse substrates. Roofs with mixed substrates the aggregated wild bee species richness, abundance and increased moderately and represented a higher mean (± SD) Table 3 Candidate GLMMs for GLMM response variables Explanatory variables K AICc ΔAICc predicting wild bee species richness and abundance in Wild bee species richnness ~ forage availability + Broof^ + Bmonth^ 7894.38 0.000 dependence of floral resources on green roofs in Vienna ~ flowering plant species + Broof^ + Bmonth^ 4 1049.48 155.09 ~1+ Broof^ + Bmonth^ 3 1176.31 281.93 Wild bee abundance ~ forage availability + Broof^ + Bmonth^ 7 1195.79 0.000 ~ flowering plant species + Broof^ + Bmonth^ 4 1422.91 227.12 ~1+ Broof^ + Bmonth^ 3 1601.02 405.23 Quotation marks signify random factors K Number of estimated parameters, AICc Second order Aikaike Information Criterion, ΔAICc Difference between AICc to next accurate model 434 Urban Ecosyst (2018) 21:429–446 On average, floral resources were highest on roofs with fine (a) substrates (forage availability: Bmedium^, flowering plant species: 87 ± 69.3), less on roofs with mixed substrates (forage availability: Blow^, flowering plant species: 55.3 ± 20.6) and lowest on roofs with coarse substrate (forage availability: Bvery low^ to Blow^, flowering plant species: 20.3 ± 9.3). Models including substrate depth, roof height and age as explanatory variables were less accurate in predicting the total wild bee species richness and abundance. However, substrate characteristics and depths were highly positive collinear (cor = 0.8) and it should be mentioned here, that roofs with deep substrates affected wild bees positively too (Fig. 5a, b in very low low medium high very high Appendix). (b) Wild bee traits on roofs 12 A majority of 81 polylectic wild bee species (97% individuals) contrasts with a minority of ten oligolectic species (3.4% in- dividuals). As reflected by the CWM (Table 5), the wild bee community was composed of ground nesting (43.3% species; 6 43.3% individuals) and above-ground nesting wild bees (43.3% species; 53.5% individuals). Solitary species (62.2%) accounted for 54.7% of the individuals and eusocial species (24.4%) for 42.4% of the individuals. Parasitic species (13.3%) were represented by 3.1% of the total abundance and very low low medium high very high not included in the CWM results (Table 5). The CWM of body size from wild bees on the studied roofs ranged between 8.3 Forage availability and 11.2 mm. Individuals of ground nesting species ranged in Fig. 2 Effects of forage availability on (a) wild bee species and (b) CWM of body size between 6.4 and 9.6 mm and were mainly abundance per monthly collection period (n = 7) on each roof (n = 9). BVery low^ forage availability was used as baseline for parameter eusocial (68%). On the other hand, individuals of above- estimation. Error bars = 0.95 confidence intervals for fitted effects ground nesting species were larger (8.4–14.5 mm) and pre- dominantly solitary (74%). wild bee diversity (26 ± 9.2) and abundance (89.3 ± 42.8) Substrate character was the most important predictor for compared to roofs with coarse substrates, which comprised the species richness and abundance of ground nesting, the lowest mean (± SD) wild bee species richness (21.5 ± above-ground nesting and solitary wild bees as well as for 11.3) and abundance (64.7 + 50.8). the abundance of eusocial species (Table 8 in Appendix). Table 4 Candidate GLMs for GLM response variables Explanatory variables K AICc ΔAICc R GLM predicting wild bee species richness and abundance in Wild bee species richness ~ substrate character 3 98.04 0.00 41.3% dependence of uniform green roof qualities on green roofs in Vienna ~ substrate depths (cm) 2 108.89 10.85 ~ height (m) 2 120.70 22.65 ~ age (years) 2 121.41 23.37 Wild bee abundance ~ substrate character 3 357.08 0.000 51.3% ~ substrate depths (cm) 2 497.38 140.31 ~ age (years) 2 620.55 263.47 ~ height (m) 2 628.98 271.90 K Number of estimated parameters, AICc Second order Aikaike Information Criterion, ΔAICc Difference between AICc to next accurate model, R GLM explained deviance Wild bee abundance per observation Wild bee species per observation Urban Ecosyst (2018) 21:429–446 435 of Lasioglossum laticeps and Halictus subauratus nests on (a) the roofs with fine substrate underpin that result. The positive 50 effect of fine substrate on ground nesting wild bee diversity (Appendix Fig. 6a) was higher (R =35.8%) compared to GLM the effect of substrate depths (R =9.3%; Appendix GLM Fig. 6b). Substrate depths affected eusocial species richness positively (R = 42.4%; Appendix Fig. 6f). The positive GLM effect of fine substrate was weaker on the abundance of above- ground nesting species and solitary species (R =30.3%; GLM Appendix Fig. 6e, i). The observation of a Bombus lapidarius colony in the fine substrate of a roof underlines its positive effect on above-ground nesting and eusocial species. The coarse mixed fine abundance of parasitic wild bees was slightly positively af- fected by substrate depth (R = 6.7%, Appendix Fig. 6k). 250 GLM (b) Further, the effect of roof height on parasitic species richness 2 2 (R = 4.14%) and abundance (R = 10.1%) was also GLM GLM weakly positive (Appendix Fig. 6j, l). The species richness and abundance of polylectic wild bees was highly positively affected by the number of flowering plant species (R = 58.3%; 64.1%; Appendix Fig. 6m, GLM n). Oligolectic wild bee species richness was enhanced by an increasing number of flowering plant species (R = GLM 68.2%; Appendix Fig. 6o). Moreover, the abundance of oligolectic species (Appendix Fig. 6p, q) was even higher positively affected by forage availability (R = 74.6%) GLM coarse mixed fine 2 than by flowering plant species richness (R = 66.9%). GLM Substrate character Increasing roof height decreased the body size (Appendix Fig. 3 Effects of substrate character on (a) wild bee species and (b) Fig. 6r) of wild bees moderately (R =23.9%). GLM abundance. The factor level Bcoarse^ wasusedasbaselinefor parameter estimation. Error bars = 0.95 confidence intervals for fitted effects BSedum^ vs. BWildflowers^ sub-areas Fine substrate had the greatest positive effect on the abun- Thefiveroofswith BSedum^ as well as BWildflowers^ dance of ground nesting and eusocial species (R = sub-areas comprised 55 wild bee species (16.7 ± 7) and GLM 75.5% and 72.2%; Appendix Fig. 6c, g). The observations 421 individuals (42.2 ± 28.3). On average, the sub-areas Table 5 Community weighted means (CWM) per roof for nesting type, sociality, body size (mm) and pollen specialization (pl = polylectic) over all species and species (spp.) per nesting type (ground nesting, above ground nesting) separately CWM All species Ground nesting spp. Above-ground nesting spp. Green roof Nesting type Sociality Body size (mm) Pollen special. Sociality Body size (mm) Sociality Body size (mm) 1 Ground nesting eusocial 8.41 pl eusocial 7.27 solitary 9.43 2 Above-ground solitary 8.31 pl eusocial 7.50 solitary 8.39 3 Above-ground solitary 8.87 pl eusocial 6.38 solitary 9.44 4 Ground-nesting eusocial 8.81 pl eusocial 8.06 solitary 9.87 5 Above-ground solitary 9.55 pl eusocial 8.26 solitary 10.71 6 Ground-nesting solitary 10.50 pl solitary 9.25 eusocial 13.0 7 Ground-nesting eusocial 10.69 pl eusocial 7.89 eusocial 14.47 8 Ground-nesting solitary 11.16 pl eusocial 9.59 solitary 12.89 9 Above-ground solitary 8.68 pl solitary 8.12 solitary 9.01 Wild bee abundance per roof Wild bee species per roof 436 Urban Ecosyst (2018) 21:429–446 30 forage availability and only 13 wild bee species (23 indi- (a) viduals) were sampled. In July conditions in the sub-areas were reversed: BSedum^ decreased to Blow^ forage avail- ability, whereas BWildflowers^ increased to Bmedium^ for- age availability and 26 wild bee species (62 individuals) exceeded the 21 wild bee species (42 individuals) in BSedum^ sub-areas. Thedatasub-setof BSedum^ and BWildflowers^ sub- March April May June July August Sept. areas revealed that some species were only observed in BSedum^ sub-areas (Table 6 in Appendix) but these spe- (b) cies (e.g. Bombus pascuorum or Megachile lagopoda) were also documented on other roofs and in other sub- areas. Therefore, no species was exclusively observed in BSedum^ sub-areas. The highly abundant group of Halictidae showed no preference for a distinct sub-area (Table 6 in Appendix). March April May June July August Sept. Discussion (c) The studied green roofs hosted 90 wild bee species, which represent 20% of Vienna’s 456 recorded wild bees (Zettel et al. 2016). Compared to other studies, the roofs showed a considerable wild bee diversity: Depending on the num- ber of sampled roofs, the amount of study years and the March April May June July August Sept. study’s scope, other authors documented between 17 and 77 wild bee species on green roofs (Brenneisen 2005; Sampling period 2014 Kadas 2006; Tonietto et al. 2011; Ksiazek et al. 2012, Fig. 4 Timelines to compare (a) total wild bee species richness, (b) total 2014; MacIvor et al. 2014;MacIvor 2015). Due to its abundance and (c) mean forage availabilities in BSedum^ and geographical position, Vienna is influenced by the BWildflowers^ sub-areas on green roofs during the sampling period 2014; Lines: full = BSedum^,dashed= BWildflowers^,dottedsquare= Pannonian climate, which could be one reason that the highlights Sedum main flowering time studied green roofs emerged with higher species numbers compared to the studies mentioned above. The authors in Switzerland (Brenneisen 2005) and London (Kadas 2006) ranged from Bvery low^ to Bmedium^ forage availability used yellow-pan traps for wild bee sampling, which may (1.7 ± 0.6 vs. 2.5 ± 0.5) and differed in flowering plant be another reason for the difference in wild bee diversity species richness (3.2 ± 0.8 vs. 21.4 ± 7.1). Wild bee spe- found on the roofs and makes a comparison of the studies cies richness (17.6 ± 5.6) and abundance (44.8 ± 26.0) difficult. was slightly higher in BWildflowers^ sub-areas than in The roofs showed lower wild bee diversity compared BSedum^ (15.8 ± 8.7; 39.6 ± 33.4). to ground level habitats in Vienna, where recently 119 to The picture becomes clearer by looking at the whole 144 species were reported from the Danube Island sampling period: Except for May, the wild bee occurrence (Pachinger and Hölzler 2006), theVienneseBotanical and diversity harmonized well with the forage availability Garden (Hölzler 2004) and the Danube Park (Zettel in the sub-areas (Fig. 4a-c). For example, in April BSedum^ et al. 2013). Similarly, authors from Chicago reported comprisedonaverage Bvery low^ forage availability and higher wild bee diversity and abundance on ground level only one wild bee species (3 individuals) was sampled, habitats than on green roofs (Tonietto et al. 2011). Green whereas BWildflowers^ offered higher floral resources and roofs and ground level habitats in Toronto (Colla et al. 3 to 9 wild bee species (5 to 14 individuals) were observed. 2009) and Switzerland (Brenneisen 2005)comprised sim- Further, BSedum^ reached the main flowering time in June ilar wild bee diversity. However, these comparisons have with Bmedium^ forage availability and 21 wild bee species little informative value because of the different sample (62 individuals), but BWildflowers^ decreased to Blow^ sizes and study designs realised in these studies. Total bee abund. Mean forage avail. Total bee species per sub-area per sub-area per sub-area Urban Ecosyst (2018) 21:429–446 437 The results of our study are interpreted in the light of the the abundance of the respective oligolectic species. small sample size and heterogeneous green roof qualities. This However, compared to studies carried out on ground level and the assessment of model quality by diagnostic plots reveal habitats in Vienna, recording 16 to 24% of oligolectic wild that the presented results are difficult to generalise. However, bee species (Hölzler 2004; Pachinger and Hölzler 2006;Zettel by interpreting and discussing the results carefully we would et al. 2013), we observed only 11% oligolectic species. also like to point out open questions regarding wild bees on The positive effect of fine substrate on the abundance of green roofs. ground nesting as well as eusocial species is reflected in the traits of the most abundant wild bee on the roofs: Halictus Green roof qualities subauratus. This Sweat bee is a ground nesting and eusocial species, requires fine and sandy substrates and nests are Wild bee diversity and abundance on green roofs are positive- established at 10–15 cm below the surface (Scheuchl and ly affected by enhanced floral resources and fine substrates. Willner 2016). Therefore, the substrate conditions that posi- Although the most accurate GLMM indicated that forage tively affected wild bees in this study are beneficial for ground availability affected wild bee diversity and abundance primar- nesting species, which require fine substrates but build their ily, the importance of entomophilous plant species richness nests shallowly. Different ground nesting species require var- was evident by the relationship of these variables. The positive ious substrate properties (e.g. compaction, content of sand or effect of high and ongoing forage availability through a high humus, Scheuchl and Willner 2016) and build nests in differ- plant diversity on wild bee communities is known from other ent depths (Cane and Neff 2011). This was also revealed in ecosystems (e.g. Potts et al. 2003; Zurbuchen and Müller our study, because fine substrates as well as substrate depth, 2012; Braun-Reichert 2013) and was found on green roofs affected the diversity of ground nesting wild bees on green in Switzerland (Brenneisen 2005) and Toronto (Tonietto roofs. Like in other studies (Brenneisen 2005; Colla et al. et al. 2011). 2009; Tonietto et al. 2011), we found Lasioglossum and The finding that roofs with fine substrates enhanced wild Halictus species in high abundance (340 individuals) and spe- bee diversity and abundance has to be interpreted with cau- cies richness (20 species) compared to Andrena species with tion, because only two of the nine roofs represented this factor only few or single individuals per species. Two explanations level. Furthermore, planting possibilities on green roofs are seem reasonable for this observation: Firstly, Andrena species limited to substrate properties (Dunnett and Kingsbury occurrence is more scattered because the females breed soli- 2008). The reason for the positive effect of fine substrates on tary. Secondly, Andrena was represented by large species, the overall wild bee diversity and abundance in this study which probably require deeper substrates for nesting than could be the relation of higher floral resources on roofs with present on the studied green roofs and just forage here. fine substrates. But wild bee traits, especially the occurrence Similarly, Tonietto et al. (2011) concluded that large ground of ground nesting and eusocial species may better explain this nesting species nested in adjacent ground level habitats and effect. used green roofs for foraging. Since we did not sample wild bees in adjacent ground level habitats, our data cannot support Wild bee traits these findings directly. However, more Andrena species where documented on ground level habitats in Vienna (Pachinger The studied green roofs mainly attracted pollen generalists. and Hölzler 2006;Zetteletal. 2013) than on the studied green Although polylectic, these wild bees were affected by roofs. flowering plant diversity because some of them showed dis- Based on the CWM, solitary wild bees were determined to tinct preferences to certain plant taxa. For example, the benefit be characteristically on the studied green roofs. It was difficult of Sedum for polylectic species with pollen preferences for to determine whether ground nesting or above-ground nesting Sedum (Anthidium oblongatum, A. strigatum, Hylaeus was the most typical nesting trait, but the latter were more punctatus, Megachile leachella, M. pilidens, M. rotundata, abundant. Above-ground nesting wild bee species are typical M. willughbiella)(Westrich 1989b) was pointed out by their of urban areas (reviewed in Hernandez et al. 2009), because higher abundance on roofs with BSedum^ sub-areas compared the high density of vertical structures offers many potential to other roofs (Appendix Table 6). Oligolectic wild bee spe- nesting sites (Cane 2005). Further, we reported above- cies richness was enhanced by increasing floral diversity, be- ground nesting wild bees in higher diversity (43.3% of spe- cause it could raise the probability that different host plants cies) compared to ground level habitats in Vienna, comprising occur on a green roof for species with different pollen special- 20 to 32% of above-ground nesting species (Hölzler 2004; izations. Also, increasing forage availability could lead to Pachinger 2008;Zettel etal. 2013). This could be explained higher flower coverage of distinct host plants and enhance by differences in site features because the mentioned studies 438 Urban Ecosyst (2018) 21:429–446 were conducted in Vienna’s botanical garden or big recrea- exclusively sampled in BWildflowers^ sub-areas (Table 6 tional areas (e.g. Danube Island) where the density of vertical Appendix), and floral resources were only present in this structures within species-specific flying distances maybe low- sub-area (Fig. 4c). er than on roofs in urban areas. Further research has to be carried out to identify the effect of green roofs’ surrounding structures on above-ground nesting wild bees. Conclusion In contrast to MacIvor (2015),wefoundnostrongevi- dence, that increasing roof height negatively altered the It was demonstrated that the wild bee community on nine wild bee community. The CWM of body size from wild green roofs in Vienna was strongly positively affected by flo- bees tended to decrease with increasing roof height. A pos- ral resources and substrate characteristics. We conclude that sible explanation is that smaller species populate high fine substrates enhance ground nesting and eusocial wild bee green roofs without utilising ground level habitats. This species. This attributes to some guidance (Gedge et al. 2008, trend is probably reflected by the weak positive effect of 2012) and the proposition by Brenneisen (2006) that the cre- roof height on parasitic wild bees, because the occurrence ation of small areas of mounds, consisting of finer and deeper of brood parasitic species indicates a vital host population substrates should be a design consideration in the planning (Hudson et al. 2006). For example, the appearance of four processes of (extensive) green roofs in order to support ground Coelioxys species was in line with the occurrence of their nesting wild bee species. Above-ground nesting and solitary host, namely Megachile species, which were recorded in species are characteristic in urban environments, but further high abundance on the roofs (Table 6 in Appendix). research is needed to assess, which parameters surrounding green roofs affect them primarily. BSedum^ vs. BWildflowers^ Honey bees were observed on all roofs, but their abun- dance did not interfere with the occurrence of wild bees. Based on the study of five roofs, Sedum is an important tem- Oligolectic wild bee species occurrence was low, but strongly poral floral resource because during its main flowering period positively affected by increasing floral diversity. Sedum that is in June higher wild bee diversity and abundance was observed typically planted on green roofs promotes polylectic wild bee in this sub-area. Similar observations were reported by species, of which some preferably forage on this plant. The MacIvor et al. (2014) who found high proportions of Sedum study of BSedum^ and BWildflower^ sub-areas on five roofs pollen in palynological samples from wild bees gathered on a indicated that Sedum species can compensate temporal lacks green roof during the main flowering period of Sedum. of resources on green roofs during its main flowering period. However, in our study qualitative comparison showed no On the other hand, various wildflowers were important forag- big difference of forage availability between these sub-areas, ing resources in spring (March, April) and summer (July, which may suggest complementary resource availability by August). Sedum and wildflowers. The concurrent trend of forage avail- We conclude that increasing floral resources (flower abun- ability, wild bee diversity and abundance in the sub-areas over dance and floral diversity) and the installation of patches with the season underpins the results discussed earlier that forage fine and deeper substrates should be considered during the availability enhances the wild bee community on green roofs. planning process to enhance the wild bee diversity and abun- This is supported by similar findings in other studies dance on green roofs. (Brenneisen 2005; Kadas 2006; Tonietto et al. 2011). The opposite trend in May is explained by observations during Acknowledgements Open access funding provided by University of Natural Resources and Life Sciences, Vienna (BOKU). We would like fieldwork that Sedum was already flowering and favourably to thank the responsible people of the buildings (Mrs. Arlt, Mrs. Ehs, Mrs. visited by wild bees compared to other flowering plants. All Haimer, Mrs. Kapui, Mrs. Leidinger, Mr. Möseler, Mr. Schatovits, Mr. Hylaeus species were exclusively sampled on BSedum^ sub- Steinbauer, Mr. Ziemak,) and the municipal council 22 – Environmental areas during this month, probably because of a preference for Protection in Vienna (especially Mrs. Doppler and Mr. Preiss) for obtaining flexible access to the study roofs. Further, we thank Vera Enzi Sedum as forage resource, which is already known for some for her helpful suggestions of green roofs during the selection process of Hylaeus species (Westrich 1989b). Sedum promoted mainly study sites. Special thanks go to Karl Mazzucco for his help with bee generalist species because the abundance peaks in June species identification, to Luise Kratschmer and Martin Wittner for help and August (Fig. 4b) are dominated by a few generalist with bee preparation, to Matthias Kropf and Michael Kopetzky for help with plant identification, to Dusty Gedge for proof reading and to species (e.g. Anthidium oblongatum, Halictus subauratus, Bernhard Kratschmer for financial support. The baseline of this work Hylaeus punctatus, H. hyalinatus, Megachile rotundata, was done in the context of the first author’s master thesis (BSummen M. willughbiella). During spring (March, April), early oc- auf den Dächern Wiens^) approved by the University of Natural curring wild bees (Andrena, Anthophora and Osmia)were Resources and Life Sciences, Vienna (BOKU) in May 2015. Urban Ecosyst (2018) 21:429–446 439 Appendix Table 6 Bee species and abundance on nine green roofs in Vienna (2014) and abundance in BSedum^ (Sed.) and BWildflower^ (Wildf.) sub-areas from a study of five roofs. Apis mellifera abundance represents field counts Bee families, genera and species Green roofs Ecological traits Abundance 123456 789N S PS MBS Sed. Wildf. Andrenidae Andrena bimaculata (Kirby 1802) 1 1 t sol pl 13 1 danuvia Stoeckhert 1950 1 t sol pl 14 dorsata (Kirby 1802) 11t sol pl 9,5 flavipes Panzer 1799 3 t sol pl 10,5 gravida Imhoff 1832 6 1 t sol pl 13 minutula (Kirby 1802) 1 t sol pl 6 nigroaenea (Kirby 1802) 1 t sol pl 14 1 ovatula (Kirby 1802) 2 t sol pl 9,5 pilipes Fabricius 1781 1tsolpl13,5 1 tibialis Kirby 1802 1 t sol pl 13 varians (Kirby 1802) 1 t sol pl 10 Apidae Anthophora crinipes Smith 1854 1 t sol pl 12 plumipes (Pallas 1772) 2 12 12 2 3 t sol pl 15 2 quadrimaculata (Panzer 1798) 1 7 1 t sol pl 10,5 Apis mellifera Linnaeus 1758 115 91 108 72 448 501 45 43 47 fb hs pl 14,5 227 97 Bombus bohemicus Seidl 1801 1 p p p 20 1 hortorum (Linnaeus 1761) 1 c eus pl 16 humilis Illiger 1806 3 3 3 3 1 c eus pl 13,5 3 1 hypnorum (Linnaeus 1758) 4 2 5 2 c eus pl 14 5 lapidaries(Linnaeus1758) 4311 516 1 337c eus pl 17 9 16 lucorum (Linnaeus 1761) 1 2 1 4 1 c eus pl 15 1 1 pascuorum (Scopoli 1763) 2 1 21 8 1 c eus pl 13,5 1 pratorum (Linnaeus 1761) 1 c eus pl 13 rupestris (Fabricius 1793) 1 p p p 20 1 terrestris(Linnaeus1758) 22111 11 c eus pl 17 3 3 Eucera nigrescens Pérez 1879 3 t sol ol 18,5 3 Melecta albifrons Forster 1771 2 p p p 12,5 Nomada goodeniana (Kirby 1802) 1 2 p p p 12 Colletidae Colletes daviesanus Smith 1846 3 2 4 t sol ol 8,5 1 5 Hylaeus cardioscapus Cockerell 1924 1 r sol pl 6,5 communis Nylander 1852 2 5 17 6 9 4 c sol pl 5,5 5 12 gredleri Förster 1871 2 1 r sol pl 5 1 1 hyalinatus Smith 1842 15 10 12 4 12 c sol pl 6,5 23 13 imparilis Förster 1871 5 u sol pl 4,5 3 2 leptocephalus (Morawitz 1870) 9 5 4 13 c sol pl 5 7 11 pictipesNylander1852 6423 9c sol pl 4,5 5 13 punctatus (Brullé 1832) 5 10 7 11 4 c sol pl 5,5 13 13 sinuatus (Schenck 1853) 1 c sol pl 7,5 1 styriacus Förster 1871 1 1 1 c sol pl 4,5 Halictidae Halictus kessleri Bramson 1879 5 t eus pl 7 3 2 maculatus Smith 1848 1 1 t eus pl 8 2 rubicundus (Christ 1791) 1 13 1 t eus pl 10 1 seladonius (Fabricius 1794) 5 7 t eus pl 7 1 4 simplexBlüthgen1923 3 51 112t sol pl 9,5 3 3 subauratus(Rossi1792) 19 9 22 29 537t eus pl 7,5 15 22 tumulorum (Linnaeus 1758) 5 4 8 5 t eus pl 7 3 6 440 Urban Ecosyst (2018) 21:429–446 Table 6 (continued) Bee families, genera and species Green roofs Ecological traits Abundance 123456 789N S PS MBS Sed. Wildf. Lasioglossum calceatum (Scopoli 1763) 1 8 2 t eus pl 9 1 laticeps (Schenck 1868) 2 4 8 1 t eus pl 7 3 2 leucozonium (Schrank 1781) 1 3 1 1 1 t sol pl 9 2 malachurum (Kirby 1802) 1 t eus pl 8,5 marginatum(Brullé1832) 11 43 1t eus pl 8 2 1 minutulum (Schenck 1853) 1 t sol pl 6,5 morio(Fabricius1793) 42423 12t eus pl 5,5 7 5 nigripes (Lepeletier 1841) 4 t eus pl 9,5 nitidulum(Fabricius1804) 17429 6t eus pl 6 9 6 pauxillum (Schenck 1853) 1 9 2 t eus pl 5,5 2 1 politum (Schenck 1853) 4 1 11 25 15 8 2 t eus pl 4,5 5 12 sabulosum (Warncke 1986) 1 t sol pl 6,5 villosulum (Kirby 1802) 2 t sol pl 6,5 Sphecodes albilabris (Fabricius 1793) 1 p p p 12 monilicornis (Kirby 1802) 1 1 1 3 p p p 8,5 1 4 ruficrus (Erichson 1835) 1 pp p 9 1 Megachilidae Anthidium manicatum (Linnaeus 1758) 1 5 c sol pl 14.5 oblongatum (Illiger 1806) 2 1 4 2 c sol pl 9 6 2 strigatum (Panzer 1805) 1 1 fb sol pl 6.5 1 Chelostoma florisomne (Linnaeus 1758) 1 x sol ol 9,5 rapunculi (Lepeletier 1841) 7 1 c sol ol 9 Coelioxys conoidea (Illiger 1806) 1 p p p 14 1 echinata Förster 1853 1 1 2 p p p 8,5 2 elongata Lepeletier 1841 3 2 1 p p p 12,5 4 1 mandibularis Nylander 1848 1 1 p p p 10 1 Heriades crenulatus Nylander 1856 2 1 r sol ol 6,5 2 rubicola Pérez 1890 1 9 1 r sol pl 6 1 truncorum (Linnaeus 1758) 5 r sol ol 7,5 Megachile apicalis Spinola 1808 2 4 4 2 4 c sol pl 10 3 6 centuncularis (Linnaeus 1758) 1 1 c sol pl 10 dorsalisPérez1879 221 2 16t sol pl 9,5 6 5 ericetorum Lepeletier 1841 2 c sol ol 12 lagopoda (Linnaeus 1761) 1 3 2 1 c sol pl 15,5 4 pilidensAlfken1924 22333 12c sol pl 10 5 4 rotundata(Fabricius1787) 19233 7c sol pl 8,5 15 3 versicolor Smith 1844 4 r sol pl 10,5 willughbiella(Kirby1802) 310 4318 167c sol pl 14 16 11 Osmia adunca (Panzer 1798) 1 c sol ol 10,5 bicornis (Linnaeus 1758) 3 4 1 7 1 c sol pl 11,5 5 caerulescens (Linnaeus 1758) 2 4 2 c sol pl 9 1 cornuta (Latreille 1805) 2 2 c sol pl 13,5 leucomelana (Kirby 1802) 1 r sol pl 8 Stelis punctulatissima (Kirby 1802) 2 p p p 9 Mellitidae Melitta haemorrhoidalis (Fabricius 1775) 1 t sol ol 12 leporina (Panzer 1799) 1 t sol ol 12 Bee abundance/roof 97 116 111 128 337 6 40 40 117 Beespecies/roof 30302830656 162034 Flowering plant species 32 23 36 38 136 11 53 15 77 Mean forage availability (± SD) 3 (0.9) 2 (0.7) 3 (0.5) 3 (1.2) 4 (0.9) 1 (1) 2 (0.9) 2 (1.6) 3 (1.1) Ecological traits (Westrich 1989a; Scheuchl and Willner 2016): N Nesting type: Ground nesting: t terricol, Above-ground nesting: c cavity nesting, r rubicol, fb free-building on structures, x xylicol, p parasitic, S Sociality: sol solitary, eus primitively eusocial, hs highly eusocial, p parasitic, PS Pollen Specialization: ol oligolectic, pl polylectic. MBS Mean body size: average species specific body size from ♂ and ♀ (mm) values given by identification literature listed in Methods and References. Nomenclature after Gusenleitner et al. (2012); taxonomic rank by Families after Michener (2007) Urban Ecosyst (2018) 21:429–446 441 Table 7 Parameters estimated and p-values for each response variable. BVery low^ forage availability served as a baseline for parameter estimation Response variable Explanatory variable Estimate ± SE Random effect SD (Wild bees) (Forage availability) Roof (N = 9) Month (N = 7) Species richness 0.140 0.435 Intercept −1.199 ± 0.267 Low 1.854 ± 0.223 Medium 2.588 ± 0.219 High 2.669 ± 0.232 Very High 3.096 ± 0.242 Abundance 0.214 0.535 Intercept −0.988 ± 0.283 Low 1.991 ± 0.198 Medium 2.667 ± 0.197 High 2.792 ± 0.208 Very High 3.292 ± 0.218 Table 8 Mean values ±SD for wild bee species richness, wild bee abundance, forage availability and entomophilous plant species richness among characteristic properties on nine green roofs in Vienna Response variables for traits Explanatory variables K AICc ΔAICc ω R i GLM Ground nesting Wild bee species richness ~ substrate character 3 63.10 0.00 0.46 35.8% ~ substrate depths (cm) 2 64.07 0.97 0.28 9.3% ~ height (m) 2 65.36 2.26 0.15 ~ age (years) 2 66.03 2.93 0.11 Wild bee abundance ~ substrate character 3 131.76 0.00 1 75.5% ~ substrate depths (cm) 2 255.47 123.71 0 ~ age (years) 2 303.34 171.58 0 ~ height (m) 2 350.54 218.78 0 Above-ground nesting Wild bee species richness ~ substrate character 3 79.63 0.00 0.97 42.8% ~ substrate depths (cm) 2 86.44 6.81 0.03 ~ age (years) 2 94.37 14.74 0.00 ~ height (m) 2 95.82 16.19 0.00 Wild bee abundance ~ substrate character 3 309.23 0.00 1 30.3% ~ substrate depths (cm) 2 344.55 35.32 0 ~ age (years) 2 382.35 73.12 0 ~ height (m) 2 407.49 98.26 0 Solitary Wild bee species richness ~ substrate character 3 79.29 0.00 0.99 42.4% ~ substrate depths (cm) 2 88.36 9.07 0.01 ~ height (m) 2 94.69 15.40 0.00 ~ age (years) 2 94.84 15.54 0.00 Wild bee abundance ~ substrate character 3 305.13 0.00 1 30.3% ~ substrate depths (cm) 2 358.37 53.23 0 ~ age (years) 2 384.23 79.10 0 ~ height (m) 2 402.01 96.88 0 Eusocial Wild bee species richness ~ substrate depths (cm) 2 53.45 0.00 0.64 42.4% ~ substrate character 3 55.82 2.38 0.19 ~ age (years) 2 57.35 3.91 0.09 442 Urban Ecosyst (2018) 21:429–446 Table 8 (continued) Response variables for traits Explanatory variables K AICc ΔAICc ω R i GLM ~ height (m) 2 57.58 4.13 0.08 Wild bee abundance ~ substrate character 3 137.96 0.00 1 72.2% ~ substrate depths (cm) 2 217.40 79.44 0 ~ age (years) 2 292.36 154.40 0 ~ height (m) 2 355.53 197.57 0 Parasitic Wild bee species richness ~ height (m) 2 40.47 0.00 0.67 4.14% ~ substrate depths (cm) 2 42.59 2.12 0.23 ~ age (years) 2 44.60 4.13 0.08 ~ substrate character 3 47.78 7.31 0.02 Wild bee abundance ~ height (m) 2 50.39 0.00 0.49 10.1% ~ substrate depths (cm) 2 51.29 0.90 0.31 6.7% ~ age (years) 2 53.06 2.67 0.13 ~ substrate character 3 54.37 3.97 0.07 Polylectic Wild bee species richness ~ flowering plant species 2 76.84 0.00 0.98 58.3% ~ forage availability 3 84.71 7.87 0.02 Wild bee abundance ~ flowering plant species 2 266.84 0.00 1 64.1% ~ forage availability 3 326.45 59.61 0 Oligolectic Wild bee species richness ~ flowering plant species 2 27.12 0.00 0.91 68.2% ~ forage availability 3 31.71 4.59 0.09 Wild bee abundance ~ forage availability 3 43.98 0.00 0.51 74.6% ~ flowering plant species 2 44.06 0.07 0.49 66.9% Body size CWM (mm) ~ height (m) 3 34.24 0.00 0.58 23.9% ~ age (years) 4 36.23 2.00 0.22 5.1% ~ substrate depths (cm) 4 36.43 2.20 0.19 ~ substrate character 4 43.67 9.43 0.01 K Number of estimated parameters, AICc Second order Aikaike Information Criterion, ΔAICc Difference between AICc to next accurate model, ωi Akaike’s weight, R GLM explained deviance Fig. 5 Effects of substrate depths (cm) on (a) wild bee species and (b) abundance. Grey band = 0.95 confidence band for fitted effects Urban Ecosyst (2018) 21:429–446 443 Fig. 6 Effects of green roof qualities on wild bee species richness and abundance per traits. Error bars and grey bars = 0.95 confidence intervals for fitted effects 444 Urban Ecosyst (2018) 21:429–446 Fig. 6 (continued) Open Access This article is distributed under the terms of the Creative Amiet F, Herrmann M, Müller A, Neumeyer R (2001) Apidae 3: Halictus, Commons Attribution 4.0 International License (http:// Lasioglossum. Schweizerische Entomologische Gesellschaft, creativecommons.org/licenses/by/4.0/), which permits unrestricted use, Neuachtel distribution, and reproduction in any medium, provided you give Auer I, Böhm R (2011) Wetter und Klima in Wien. Vielfalt auf engstem appropriate credit to the original author(s) and the source, provide a link Raum. In: Berger R, Ehrendorfer F (eds) Ökosystem Wien: Die to the Creative Commons license, and indicate if changes were made. Naturgeschichte einer Stadt. Böhlau Verlag GesmbH&Co.KG, Wien, pp 88–105 Bates D, Maechler M (2016) Matrix: sparse and dense matrix classes and methods. R package version 1.2-8. https://CRAN.R-project.org/ package=Matrix. Accessed 10 Nov 2016 References Bates D, Maechler M, Bolker B, Walker S (2015) Fitting Linear Mixed- Effect Models Using lme4. J Stat Softw 67:1–48 Baumann N (2006) Ground-Nesting Birds on Green Roofs in Amiet F, Müller A, Neumeyer R (1999) Apidae 2: Colletes, Dufourea, Switzerland: Preliminary Observations. Urban Habitats 4: Hyleaus, Nomia, Nomioides, Rhophitoides, Rophites, Sphecodes, 37–50 Systropha. Schweizerische Entomologische Gesellschaft, Neuachtel Urban Ecosyst (2018) 21:429–446 445 van den Berg AE, Hartig T, Staats H (2007) Preference for Nature in Gathmann A, Tscharntke T (2002) Foraging ranges of solitary bees. J Urbanized Societies : Stress, Restoration, and the Pursuit of Anim Ecol 71:757–764 Sustainability. J Soc Issues 63:79–96 Gedge D (2003) From rubble to redstars. Green Rooftops for Sustainable Berndtsson JC (2010) Green roof performance towards management of Communities. Report. Chicago runoff water quantity and quality: A review. Ecol Eng 36:351–360. Gedge D, Newton J, Cradick K, et al (2008) Living Roofs and Walls. https://doi.org/10.1016/j.ecoleng.2009.12.014 Technical Report for the Greater London Authority’s London Plan Biesmeijer JC, Roberts SPM, Reemer M et al (2006) Parallel Declines in and Environment Teams. London Pollinators and Insect-Pollinated Plants in Britain and the Gedge D, Grant G, Kadas G, Dinham C (2012) Creating Green Roofs for Netherlands. Science 313:351–354. https://doi.org/10.1126/ Invertebrates. Buglife, Peterborough, pp 1–29 science.1127863 Gokcezade JF, Gereben-Krenn BA, Neumayer J, Krenn HW (2010) Braun-Reichert R (2013) Der Einfluss unterschiedlicher Beweidung auf Feldbestimmungsschlüssel für die Hummeln Österreichs, die Wildbienen- und Wespenfauna von Kalkmagerrasen Zeitpunkt, Deutschlands und der Schweiz (Hymenoptera, Apidae). Linzer Frequenz und Kontinuität der Beweidung. Galathea-Berichte des Kr Biol Beitr 42:5–42 Nürnberger Entomolgen 29:7–22 Gusenleitner F, Schwarz M, Mazzucco K (2012) Apidae (Insecta: Brenneisen S (2003) Ökologisches Ausgleichspotenzial von Extensiven Hymenoptera). In: Schuster R (ed) Biosystematics and Ecology Dachbegrünungen. Bedeutung des Ersatz-Ökotops für den Arten- Series No. 29: Checkliste der Fauna Österreichs, No. 6. ÖAW- und Naturschutz und die Stadtentwicklungsplanung. University of Verlag der Österreichischen Akademie der Wissenschaften, Basel, Basel Vienna, pp 1–129 Brenneisen S (2005) The Natural Roof (NADA): Research Project Report Hathaway AM, Hunt WF, Jennings GD (2008) A field study of green roof on the use of Extensive Green Roofs by Wild Bees. Wädenswil hydrologic and wanter quality performance. Trans ASABE 51:37– Brenneisen S (2006) Space for Urban Wildlife: Designing Green Roofs as Habitats in Switzerland. Urban Habitats 4:27–36 Hennig EI, Ghazoul J (2012) Pollinating animals in the urban environ- Cane JH (2005) Bees, pollination, and the challanges of sprawl. In: ment. Urban Ecosyst 15:149–166. https://doi.org/10.1007/s11252- Johnson EA, Klemens MW (eds) Nature in Fragments: The legacy 011-0202-7 of sprawl. Columbia University Press, New York, pp 109–124 Hernandez JL, Frankie GW, Thorp RW (2009) Ecology of Urban Bees : Cane JH, Neff JL (2011) Predicted fates of ground-nesting bees in soil A Review of Current Knowledge and Directions for Future Study. heated by wildfire: Thermal tolerances of life stages and a survey of Cities Environ 2:1–15 nesting depths. Biol Conserv 144:2631–2636. https://doi.org/10. Hölzler G (2004) Die Wildbienen des Botanischen Gartens der 1016/j.biocon.2011.07.019 Universität Wien. In: Pernstich A, Krenn HW (eds) Die Tierwelt Carter T, Butler C (2008) Ecological impacts of replacing traditional roofs des Botanischen Gartens der Universität Wien. Eigenverlag Institut with green roofs in two urban areas. Cities Environ 1:1–17 für Angewandte Biologie und Umweltbildung, Vienna, pp 141–163 Castleton HF, Stovin V, Beck SBM, Davison JB (2010) Green roofs; building energy savings and the potential for retrofit. Energy Build Hudson PJ, Dobson AP, Lafferty KD (2006) Is a healthy ecosystem one 42:1582–1591. https://doi.org/10.1016/j.enbuild.2010.05.004 that is rich in parasites? Trends Ecol Evol 21:381–385. https://doi. Colla SR, Willis E, Packer L (2009) Can green roofs provide habitat for org/10.1016/j.tree.2006.04.007 urban bees (Hymenoptera: Apidae). Cities Environ 2:1–12 Kadas G (2006) Rare Invertebrates Colonizing Green Roofs in London. Dathe HH (1980) Die Arten der Gattung Hylaeus F. in Europa Urban Habitats 4:66–86 (Hymenoptera: Apoidea, Colletidae). Mitt Zool Mus Berl 56:207– Kennedy CM, Lonsdorf E, Neel MC et al (2013) A global quantitative synthesis of local and landscape effects on wild bee pollinators in Dunnett N, Kingsbury N (2008) Planting green roofs and living walls. agroecosystems. Ecol Lett 16:584–599. https://doi.org/10.1111/ele. Timber Press Inc., Oregon Ebmer PAW (1969) Die Bienen des Genus Halictus LATR. S. L. im Klein A-M, Vaissière BE, Cane JH et al (2007) Importance of pollinators Großraum von Linz (Hymenoptera, Apidae): Systematik, in changing landscapes for world crops. Proc Biol Sci 274:303–313. Biogeographie, Ökologie und Biologie mit Berücksichtigung aller https://doi.org/10.1098/rspb.2006.3721 bisher aus Mitteleuropa bekannten Arten. Teil I. Mit neun Bildtafeln. Köhler M (2005) Long-Term Vegetation Research on Two Extensive Natkdl Jb Stadt Linz 1969:133–183 Green Roofs in Berlin. Urban Habitats 4:3–26 Ebmer PAW (1970) Die Bienen des Genus Halictus LATR.S.L. im Kremen C, Williams NM, Thorp RW (2002) Crop pollination from native Großraum von Linz (Hymenoptera, Apidae): Teil II Mit neun bees at risk from agricultural intensification. Proc Natl Acad Sci U S Bildtafeln. Natkdl Jb Stadt Linz 1970:19–82 A 99:16812–16816. https://doi.org/10.1073/pnas.262413599 Ebmer PAW (1971) Die Bienen des Genus Halictus LATR.S.L. im Ksiazek K, Fant J, Skogen K (2012) An assessment of pollen limitation Großraum von Linz (Hymenoptera, Apidae): Teil III mit 19 on Chicago green roofs. Landsc Urban Plan 107:401–408 Bildtafeln. Natkdl Jb Stadt Linz 1971:63–156 Ksiazek K, Tonietto R, Ascher J (2014) Ten bee species new to green Falk S (2015) Field Guide to the Bees of Great Britain and Ireland. British roofs in the Chicago area. Gt Lakes Entomol 47:87–92 Wildlife Publishing Bloomsbury, London Laliberté E, Legendre P, Shipley B (2014) FD: measuring functional Fernandez-Canero R, Gonzalez-Redondo P (2010) Green Roofs as a diversity from multiple traits, and other tools for functional ecology. habitat for birds: A review. J Anim Vet Adv 9:2041–2052 R package version 1.0-12. Fontaine C, Dajoz I, Meriguet J, Loreau M (2006) Functional Diversity of Landolt E (2001) Orchideen-Wiese in Wollishofen (Zürich) - ein Plant–Pollinator Interaction Webs Enhances the Persistence of Plant erstaunliches Relikt aus dem Anfang des 20. Jahrhunderts. Communities. PLoS Biol 4:129–135. https://doi.org/10.1371/ Vierteljahrsschrift der Naturforschenden Gesellschaft Zürich 146: journal.pbio.0040001 41–51 Fox J (2003) Effect Displays in R for Generalised Linear Models. J Stat So Le Féon V, Schermann-Legionnet A, Delettre Y et al (2010) ftw 8(15):1–27. http://www.jstatsoft.org/v08/i15/. Accessed 21 Intensification of agriculture, landscape composition and wild bee Feb 2017 communities: A large scale study in four European countries. Agric Frankie GW, Thorp RW, Hernandez J et al (2009) Native bees are a rich Ecosyst Environ 137:143–150. https://doi.org/10.1016/j.agee.2010. natural resource in urban California gardens. Calif Agric 63:113– 01.015 120. https://doi.org/10.3733/ca.v063n03p113 446 Urban Ecosyst (2018) 21:429–446 MacIvor JS (2015) Building height matters: nesting activity of bees and der Arten der Schweiz Band 3: Schlüssel der Arten der Familie Andrenidae. Eigenverlag, Velden/Vils wasps on vegetated roofs. Isr J Ecol Evol 9801:1–9. https://doi.org/ 10.1080/15659801.2015.1052635 Schrader S, Böning M (2006) Soil formation on green roofs and its MacIvor JS, Lundholm J (2010) Insect species composition and diversity contribution to urban biodiversity with emphasis on Collembolans. on intensive green roofs and adjacent level-ground habitats. Urban Pedobiologia (Jena) 50:347–356. https://doi.org/10.1016/j.pedobi. Ecosyst 14:225–241. https://doi.org/10.1007/s11252-010-0149-0 2006.06.003 MacIvor JS, Ruttan A, Salehi B (2014) Exotics on exotics: Pollen anal- Stadt Wien Vienna GIS (2005) Data: wiener umweltschutzabteilung – ysis of urban bees visiting Sedum on a green roof. Urban Ecosyst 18: MA 22. Basic data: mehrzweckkarte, stadtvermessung wien – MA 419–430. https://doi.org/10.1007/s11252-014-0408-6 41. Further Information: Themenplan Wien Umweltgut. Vienna: Matteson KC, Langellotto GA (2010) Determinates of inner city butterfly Stadt Wien and bee species richness. Urban Ecosyst 13:333–347. https://doi. Susca T, Gaffin SR, Dell’Osso GR (2011) Positive effects of vegetation: org/10.1007/s11252-010-0122-y Urban heat island and green roofs. Environ Pollut 159:2119–2126. Matteson KC, Ascher JS, Langellotto GA (2008) Bee Richness and https://doi.org/10.1016/j.envpol.2011.03.007 Abundance in New York City Urban Gardens Bee Richness and Tonietto R, Fant J, Ascher J et al (2011) A comparison of bee communi- Abundance in New York City Urban Gardens. Ann Entomol Soc ties of Chicago green roofs, parks and prairies. Landsc Urban Plan Am 101:140–150 103:102–108. https://doi.org/10.1016/j.landurbplan.2011.07.004 Mauss V (1994) Bestimmungsschlüssel für Hummeln. In: Deutscher ViennaGIS (2016) Vienna City Map. https://www.wien.gv.at/stadtplan/. Jugendbund für Naturbeobachtungen (ed), 5th edn. Hamburg, p 51 Accessed 5 Dec 2016 Mazerolle MJ (2016) AICcmodavg: model selection and multimodel in- von Hagen E, Aichhorn A (2003) Hummeln: bestimmen, ansiedeln, ference based on (Q)AIC(c). R package version 2.1-0. https://cran.r- vermehren, schützen, 5. überarb. Fauna Verlag, Nottuln project.org/package=AICcmodavg. Accessed 13 Feb 2017 Westrich P (1989a) Die Wildbienen Baden-Württembergs: Spezieller Mentens J, Raes D, Hermy M (2006) Green roofs as a tool for solving the Teil. Die Gattungen und Arten. Eugen Ulmer GmbH&Co., Stuttgart rainwater runoff problem in the urbanized 21st century ? Landsc Westrich P (1989b) Die Wildbienen Baden-Württembergs: Allgemeiner UrbanPlan77:217–226. https://doi.org/10.1016/j.landurbplan. Teil. Lebensräume, Verhalten, Ökologie und Schutz. Eugen Ulmer 2005.02.010 GmbH&Co., Stuttgart Michener CD (2007) The Bees of the World, 2nd edn. The Johns Hopkins Westrich P (1996) Habitat requirements of central European bees and the University Press, Baltimore problems of partial habitats. In: Matheson A, Buchmann SL, Motulsky H, Christopoulos A (2003) Fitting Models to Biological Data O’Toole C, Westrich P and Williams IH (eds) The conservation of using Linear and Nonlinear Regression: A practical guide to curve bees. Academic Press, London pp 1–16 fittin, 4th edn. GraphPad Software, Inc, San Diego Winfree R, Abuilar R, Vázquez DP et al (2009) A meta-analysis of bees’ Oberndorfer E, Lundholm J, Bass B et al (2007) Green Roofs as Urban response to anthropogenic disturbance. Ecology 90:2068–2076. Ecosystems: Ecological Structures, Functions, and Services. https://doi.org/10.1890/07-1861.1 Bioscience 57:823–833 Wojcik VA, McBride JR (2012) Common factors influence bee foraging Pachinger B (2008) Der Hohlweg am Johannesberg (Wien, Unterlaa) in urban and wildland landscapes. Urban Ecosyst 15:581–598. Lebensraum und Trittstein für Wildbienen (Hymenoptera:Apidae). https://doi.org/10.1007/s11252-011-0211-6 Beiträge zur Entomofaunistik 8:69–83 ZAMG - Zentralanstalt für Meterologie und Geodynamik (2015) Pachinger B, Hölzler G (2006) Die Wildbienen (Hymenoptera, Apidae) Klimaspiegel Wien Hohe Warte für 2014. http://www.zamg.ac.at/ der Wiener Donauinsel. Beiträge zur Entomofaunistik 7:119–148 cms/de/klima/klima-aktuell/klimaspiegel/jahr/wien_hohe_warte/? Potts SG, Vulliamy B, Dafni A et al (2003) Linking bees and flowers: jahr=2014. Accessed 5 Dec 2016 How do floral communities structure pollinator communities? Zettel H, Zimmermann D, Wiesbauer H (2013) Die Bienen und Ecology 84:2628–2642. https://doi.org/10.1890/02-0136 Grabwespen (Hymenoptera: Apoidea) im Donaupark in Wien Potts SG, Biesmeijer JC, Kremen C et al (2010) Global pollinator de- (Österreich). Sabulosi 3:1–23 clines: trends, impacts and drivers. Trends Ecol Evol 25:345–353. Zettel H, Zimmermann D, Wiesbauer H (2016) Ergänzungen zur https://doi.org/10.1016/j.tree.2010.01.007 Bienenfauna (Hymenoptera: Apidae) von Wien, Österreich. R Core Development Team (2016) R: A Language and Environment for Beiträge zur Entomofaunistik 17:85–107 Statistical Computing. V 3.3.2. R Foundationi for Statistical Zinzi M, Agnoli S (2012) Cool and green roofs. An energy and comfort Computing, Vienna comparison between passive cooling and mitigation urban heat is- RStudio Team (2015) RStudio: Integrated Development for R. V land techniques for residential buildings in the Mediterranean re- 0.99.903. RStudio Inc., Boston gion. Energy Build 55:66–76. https://doi.org/10.1016/j.enbuild. Scheuchl E (2000) Illustrierte Bestimmungstabellen der Widbienen 2011.09.024 Deutschlands und Österreichs Band 1: Schlüssel der Gattungen Zurbuchen A, Müller A (2012) Wildbienenschutz - von der Wissenschaft und der Arten der Familie Anthophoridae. Eigenverlag, Velden/Vils zur Praxis. Bristol-Stifung, Bern Scheuchl E (2006) Illustrierte Bestimmungstabellen der Wildbienen Zurbuchen A, Landert L, Klaiber J et al (2010) Maximum foraging ranges Deutschlands und Österreichs Band 2: Megachilidae - Melittidae, in solitary bees: only few individuals have the capability to cover 2nd edn. Eigenverlag, Velden/Vils long foraging distances. Biol Conserv 143:669–676. https://doi.org/ Scheuchl E, Willner W (2016) Taschenlexikon der Wildbienen 10.1016/j.biocon.2009.12.003 Mitteleuropas: Alle Arten im Porträt. Quelle & Meyer Verlag, Zuur AF, Ieno EN, Elphick CS (2010) A protocol for data exploration to Wiebelsheim avoid common statistical problems. Methods Ecol Evol 1:3–14. Schindler M, Diestelhorst O, Härtel S et al (2013) Monitoring agricultural https://doi.org/10.1111/j.2041-210X.2009.00001.x ecosystems by using wild bees as environmental indicators. BioRisk 71:53–71. https://doi.org/10.3897/biorisk.8.3600 Zuur AF, Hilbe JM, Ieno EN (2013) A Beginner’s Guide to GLM and Schmid-Egger C, Scheuchl E (1997) Illustrierte Bestimmungstabellen der GLMM with R: A frequentist and Bayesian perspective for ecolo- Wildbienen Deutschlands und Österreichs unter Berücksichtigung gists. Highland Statistics Ltd., Newburgh http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Urban Ecosystems Springer Journals

Buzzing on top: Linking wild bee diversity, abundance and traits with green roof qualities

Free
18 pages

Loading next page...
 
/lp/springer_journal/buzzing-on-top-linking-wild-bee-diversity-abundance-and-traits-with-oPXk4XAWRN
Publisher
Springer Journals
Copyright
Copyright © 2018 by The Author(s)
Subject
Life Sciences; Urban Ecology; Environmental Management; Ecology; Nature Conservation
ISSN
1083-8155
eISSN
1573-1642
D.O.I.
10.1007/s11252-017-0726-6
Publisher site
See Article on Publisher Site

Abstract

Green roofs are potentially valuable habitats for plants and animals in urban areas. Wild bees are important pollinators for crops and wild plants and may be enhanced by anthropogenic structures, but little is known about wild bees on green roofs in cities. This study investigates the effects of green roof qualities (floral resources, substrate character and depth, roof height and age) on wild bee diversity, abundance and traits (nesting type, sociality, pollen specialisation, body size) on green roofs in Vienna. Nine green roofs were sampled monthly between March and September 2014 by a semi quantitative approach. Wild bees were collected in pre-defined sub-areas for the same amount of time and floral resources were recorded. Over all green roofs, 992 individuals belonging to 90 wild bee species were observed. Wild bee diversity and abundance was strongly positively affected by increasing forage availability and fine substrates. Wild bees on roofs were characteristically solitary, polylectic and 8.3– 11.2 mm. Regarding nesting type, the percentage of above-ground nesting bees was higher compared to the common species composition in Middle Europe. Ground-nesting wild bees were mainly eusocial, smaller (6.4–9.6 mm) and positively affected by roofs with fine substrates. During June, when forage availability by wildflowers on roofs was Blow^ (5–15% flower coverage), flowering Sedum species were an important forage resource. We conclude that wild bee diversity and abundance on green roofs are enhanced by floral resources. Furthermore, the installations of areas with finer and deeper substrates benefit ground nesting and eusocial wild bees. . . . . . Keywords Vegetated roofs Apiformes Floral resources Urban biodiversity Pollinators Vienna Introduction colonize habitats successfully (Westrich 1996; Gathmann and Tscharntke 2002; Zurbuchen et al. 2010;Wojcikand The degradation and loss of habitats are seriously threaten- McBride 2012). Bees are ectothermic and thermophile or- ing wild bees (Winfree et al. 2009;LeFéonet al. 2010; ganisms, thus the warmer (micro-) climate of cities enables Potts et al. 2010; Kennedy et al. 2013), leading to declines favourable habitat conditions for many wild bee species in pollination services for crops (Kremen et al. 2002;Klein (Cane 2005; Hennig and Ghazoul 2012). Research on urban et al. 2007) and wild plants (Biesmeijer et al. 2006; sites likely to be populated by wild bees, provides impor- Fontaine et al. 2006). As important pollinators in gardens, tant information for urban planning to enhance wild bee parks and other green spaces (Matteson et al. 2008;Frankie diversity (Hernandez et al. 2009). et al. 2009; Matteson and Langellotto 2010), wild bees play Sustainable city planning considers green roofs as a vital a crucial role for urban ecosystems. The availability of pol- measure to compensate negative effects of sealed surfaces. len and nectar as well as nesting sites within species- Many abiotic benefits have been reported (reviewed in specific flying distances is important for wild bees to Oberndorfer et al. 2007) like the improvement of city’s water run off management (Mentens et al. 2006; Hathaway et al. 2008;Berndtsson 2010), the reduction of heat islands (Susca et al. 2011) and the increase of energy * Sophie Kratschmer efficiency by insulating the building’s indoor rooms against sophie.kratschmer@boku.ac.at heat and cold (Castleton et al. 2010; Zinzi and Agnoli 2012). Further benefits are aesthetic and amenity for urban Institute for Integrative Nature Conservation Research, University of citizens (van den Berg et al. 2007). From an ecological Natural Resources and Life Sciences, Vienna (BOKU), Gregor-Mendel-Straße 33, 1180 Vienna, Austria point of view, green roofs have high potential to restore 430 Urban Ecosyst (2018) 21:429–446 habitats for plants and animals in cities (Oberndorfer et al. (2) Characteristic wild bee traits (pollen specialization, 2007; Carter and Butler 2008; Dunnett and Kingsbury nesting type, sociality and body size) on green roofs 2008). During the last decade, studies on plants (Landolt and how these traits are affected by respective green roof 2001;Köhler 2005), birds (Gedge 2003;Baumann 2006; qualities. Fernandez-Canero and Gonzalez-Redondo 2010)and vari- (3) The importance of Sedum species compared to other ous invertebrate taxa like Araneae (Brenneisen 2003; wildflowers as foraging resource on green roofs. Kadas 2006), Collembola (Schrader and Böning 2006), Coleoptera (Kadas 2006) and Hymenoptera (Brenneisen 2003; Brenneisen 2005; Kadas 2006; Colla et al. 2009; MacIvor and Lundholm 2010; Tonietto et al. 2011; Methods Ksiazek et al. 2014;MacIvoretal. 2014) underpinned the importance of green roofs to contribute to biodiversity and Study sites nature conservation in cities. Green roofs are potentially important habitats for urban We studied wild bees on nine green roofs in Vienna (Fig. 1) wild bees, as they provide pollen and nectar resources during the vegetation period of 2014. In order to evaluate the throughout the year (Tonietto et al. 2011) and incorporate wild bee diversity and abundance on roofs with different qual- different nesting habitats (Brenneisen 2005; MacIvor et al. ities (i.e. substrate characteristics and depths, roof height and 2014). In addition, the higher insulation and thus warmer age), the roofs varied in greening type (extensive, semi- micro-climate at roof level result in favourable habitat con- extensive and intensive) leading to different planting possibil- ditions (Matteson and Langellotto 2010). Beside, a contin- ities (Dunnett and Kingsbury 2008) and floral resources (e.g. uous vegetation layer on green roofs is crucial to the abiotic Sedum species and other wildflowers). benefits discussed above and linked to the pollination by Detailed information about the roof’s qualities (Table 1) wild bees (Dunnett and Kingsbury 2008). was gathered during wild bee sampling or by interviewing Similar wild bee communities have been reported on the people responsible for the buildings. Substrate character- green roofs and ground level habitats in Toronto (Colla istics were determined in the field. Coarse substrates consisted et al. 2009) and Chicago (Tonietto et al. 2011) but higher of high amounts of expanded clay and crashed brick and are wild bee diversity on ground sites was related to higher typically used on extensive green roofs. Fine substrates entomophilous plant diversity. The habitat quality of consisted of a high amount of humus, earth and sand and are green roofs for wild bees was highlighted by observa- mainly used on intensive green roofs. Mixed substrate was tions of locally and/or nationally rare species (Kadas present on roofs where coarse and fine substrates were used 2006) and of wild bee species newly recorded on green in different areas. Substrate depths, height and age were roofs (Ksiazek et al. 2014;MacIvoretal. 2014). Green roof types support wild bee diversity differently because of the type of substrate and its character and plant diver- sity (Brenneisen 2005). Furthermore, Sedum species are characteristic green roof plants and play an important role as a foraging resource for wild bees (MacIvor et al. 2014). Whilst there has been research in Western Europe on green roofs and wild bees, knowledge about wild bees on green roofs in the eastern part of Europe, characterized by a warmer and dryer climate and therefore comprising a dif- ferent wild bee species composition is scarce. In this paper, we report the wild bee diversity, abundance and traits in relation to green roof qualities on green roofs in Vienna. The main focus of this study was to determine: (1) The effect of temporal variable floral resources (forage availability and entomophilous plant diversity) and con- stant green roof qualities (substrate characteristics and Fig. 1 Locations of the nine green roofs in Vienna surveyed in 2014 depths, roof heights and ages) on wild bee communities. (basic map: ViennaGIS 2016) Urban Ecosyst (2018) 21:429–446 431 Table 1 Green roof qualities for data analyses: Height: measured from humus, earth or sand; mixed = areas with coarse and fine substrate; street level (m); Age: years since first greening until 2014; Substrate: Sub-areas: pre-defined for sampling by types of plants or structural char- Depth (cm) and characteristics expressed by components: coarse = high acteristics. Additional information: Site features and % of sealed area in a percentage of expanded clay/crashed bricks; fine = high amount of 500 m radius (Stadt Wien Vienna GIS 2005) Site Coordinates Area (m ) Height (m) Age (year) Substrate (cm) Sub-areas Site features Sealed area (%) 1 400 9.5 4 10 BWildflowers^ Inner-city; 79.6 48.17935; coarse BSedum^ higher buildings around 16.326241 full sun until 4 pm 2 750 15 9 8 BWildflowers^ Inner-city; 68.6 48.23655; coarse BSedum^ higher buildings around; 16.379753 full sun until 4 pm 3 200 25 16 8–90 BWildflowers^ Inner-city; 75.6 48.189551; mixed BLawn^ no higher buildings around; 16.372002 BShrubs^ full sun all day BSedum^ 4 231 16 23 20–25 BWildflowers^ Inner-city; 83.9 48.216574; fine BLawn^ no higher buildings around; 16.329165 BShrubs^ full sun all day; BUnmaintained area^ 5 1000 12 18 20–50 BAlpine^;BBBQ spot^ Inner-city; 65.4 48.195187; fine BFruit trees^; BWildflowers^; no higher buildings around; 16.304639 BLawn^; BVegetables^ full sun all day BPannonic^; BShrubs^; 6 8800 11 16 5–7 BSedum^ Outskirt; 46.7 48.139076; coarse no higher buildings around; 16.366834 full sun all day 7 400 5 6 10–35 BWildflowers^ Inner-city; 68.5 48.236503; mixed BShrubs^ higher buildings on S/W; 16.379654 BPond^ full sun 10 am to 2 pm 8 1500 8 16 8–12 BBare substrate^ Outskirt; 47.1 48.266850; coarse BWildflowers^ no higher buildings around; 16.466844 BSedum^ full sun all day BDead wood^ 9 420 8 2 10 BWildflowers^ Inner-city; 83.3 48.222610, mixed BSedum^ higher buildings around; 16.333602 BShrubs in pots^ full sun until 4 pm determined from planning documents or by interviewing peo- or vegetation types (e.g. wildflowers, Sedum, lawn, shrubs, ple responsible for the building. Pannonian plants), structural characteristics (e.g. bare sub- Vienna is characterised by a temperate Pannonian climate, strate, dead wood elements) or functionality (e.g. BBQ spot, with the potential for precipitation throughout the year (Auer unmaintained area). To assess the value of Sedum as foraging and Böhm 2011). The average annual precipitation in 2014 resource for wild bees, it was defined as a distinct sub-area. was 756 mm and the average monthly temperature was Patches with spontaneous flora and herbaceous garden plants, 12.5 °C. The highest average daily temperature was recorded except for Sedum, were classified as BWildflowers^. Sampling during July and early August and ranged between 21 and time was adapted to the size of the sub-areas and ranged be- 23 °C (ZAMG - Zentralanstalt für Meterologie und tween 3 min for small objects like dead wood elements and Geodynamik 2015). 15 min for large areas like lawns, shrubs, Sedum or wild- flowers. Each roof was sampled monthly, seven times from March to September 2014 (Schindler et al. 2013). Sampling Wild bee sampling was conducted between 10 am and 4 pm on days with warm (22.8 ± 4.9 °C), windless and dry weather conditions. On each Wild bee sampling was conducted by a semi-quantitative sampling date, information on floral resources was carried out. method, hand netting wild bees in each of the pre-defined The forage availability (=flower coverage) of entomophilous sub-areas (Table 1) for the same period of time. The sub- plants, was assessed in each sub-area (Table 1) using five areas were distinct patches characterized by dominating plants 432 Urban Ecosyst (2018) 21:429–446 categories: 1 = Bvery low^ (< 5%), 2 = Blow^ (5–15%), 3 = B Data analyses medium^ (15–25%), 4 = Bhigh^ (25–50%) and 5 = Bvery high^ (> 50%). To assess the number of entomophilous plant We excluded Apis mellifera from analysis, because it is a species flowering at the time of sampling, plants were domestic species and the abundance could be biased by near- photographed and identified to species level. This was sepa- by hives (Kennedy et al. 2013). We only found a weak, insig- rately documented for each sub-area. On average (± SD), the nificant positive relation between honey bee and wild bee roofs comprised Bvery low^ to Bmedium^ forage availability abundance on the roofs (Spearman’s rank correlation; rho = (2 ± 0.6) and 46.8 (± 39) entomophilous plant species. 0.08; p = 0.8). Roof size was neither related to wild bee spe- Evidence of nesting activity was gathered qualitatively during cies richness (rho = −0.12; p = 0.7) nor to wild bee abundance wild bee sampling. Bare substrate patches were observed for (rho = −0.23; p = 0.5). Statistical analyses were performed in ground nesting activity and old plant stems, cavities in walls R 3.3.2 (R Core Team 2016) using R Studio V 0.99.903 or dead wood branches for above-ground nesting activity. (RStudio Team 2015). Wild bees were identified to species level (Ebmer 1969, The effects of temporal variable floral resources on wild 1970, 1971; Dathe 1980; Mauss 1994; Schmid-Egger and bee species richness and abundance were analysed by gener- Scheuchl 1997; Amiet et al. 1999, 2001;Scheuchl 2000, alized linear mixed models (GLMMs) with Poisson error dis- 2006; Gokcezade et al. 2010) by the authors (SK, BP) and tribution using the R-packages Blme4^ (Bates et al. 2015)and Karl Mazzucco was consulted for validation of some speci- BMatrix^ (Bates and Maechler 2016). We formulated null and mens mainly of the genus Lasioglossum and Hylaeus. All candidate models for each response variable with two random specimens are housed in the collection of the Institute for factors (Bmonth^ and Broof^) and either forage availability or Integrative Nature Conservation Research at BOKU Vienna. entomophilous plant species richness as fixed factor. We did Wild bee traits (Table 2)weresummarized by categories not formulate models containing both fixed factors because and determined by literature research (Westrich 1989a; prior data exploration revealed them as collinear (cor = 0.6), Michener 2007; Scheuchl and Willner 2016). Information on which results in unreliable parameter estimation (Zuur et al. body size was derived from identification literature (see 2010). For GLMMs with forage availability as fixed factor, above; bumblebees: von Hagen and Aichhorn 2003)which we used the category Bvery low^ as baseline for parameter give the range of body size within a species. For trait analysis estimation. Model selection was carried out by the second the average body size was calculated from the female and order Akaike Information Criterion (AICc) using the R pack- male values. The nesting type of Hylaeus imparilis remained age BAICcmodavg^ (Mazerolle 2016). TheAICcisusedfor unclear, but was attributed to the above-ground nesting group, modelling data with small sample size (Motulsky and like all other Hylaeus species (Falk 2015; Scheuchl and Christopoulos 2003). The cut-off to decide whether a model Willner 2016). is more likely to be correct than the next one was set at ΔAICc Table 2 Definitions, variable structure of wild bee traits and explanatory variable selection for trait analysis with generalized linear models (GLM); primitively eusocial: Beusocial^ hereafter Traits Variable type Definition Explanatory variables for GLM trait analysis Nesting type Ground nesting Excavate nests in the ground Substrate quality Above-ground Nesting in cavities, plant stems, dead wood or build on structures Substrate depths (cm) (incl. Bombus spp.) Roof height (m) Roof age (years) Sociality Solitary fertile ♀ nest and breed alone Substrate quality Primitively eusocial fertile ♀ establish the nest and 1st generation of workers Substrate depths (cm) initializing division of labour Roof height (m) Roof age (years) Parasitic fertile ♀ lay their eggs in nests of certain host species Body size Continuous variable Mean body size was averaged from range values in identification Substrate quality literature (see section BWild bee sampling^) Substrate depths (cm) Roof height (m) Roof age (years) Pollen specialization polylectic no pollen specialization Forage availability oligolectic pollen specialization on a plant family or genus No.of flowering plant species Urban Ecosyst (2018) 21:429–446 433 <2. We did not use p-values for interpretation because they averaged forage availability along a timeline plot created with are not reliable in GLMMs (Zuur et al. 2013). Microsoft Excel 2010. The effects of uniform green roof qualities (substrate char- acteristics and depths, roof height, age) on wild bee species richness and abundance were fitted with Poisson generalized Results linear models (GLMs). For GLMs the wild bee species num- bers and abundance were aggregated to total amounts per roof. Excluding A. mellifera (1470 individuals), which was present Substrate depths were averaged if a range was indicated on all roofs, we identified 90 wild bee species belonging to 19 (Table 1) and the factor level Bcoarse^ was the baseline for genera. A total of 992 wild bee individuals were caught on the parameter estimation of substrate characteristics. green roofs (see Table 6 in Appendix for species list). The Characteristic traits on green roofs were examined by cal- most abundant species was Halictus subauratus with 94 indi- culating community weighted means (CWM) with the R- viduals (9.5%), whereas 25 wild bee species (27.8%) were package BFD^ (Laliberté et al. 2014). To analyse how traits represented by only one individual. responded to green roof qualities, aggregated species numbers On average (± SD), green roofs hosted 29 (± 16.2) wild bee and abundance per trait (nesting type, sociality and pollen species and 110 (± 95.3) individuals. The most diverse roof specialization) per roof were used as response variables in with 65 wild bee species (337 individuals) contrasted with a Poisson GLMs. The CWM of body size was used as response roof that had only six species and six individuals. in GLMs with Gaussian error distribution. To reduce the num- ber of GLMs for trait analysis, we chose the most interesting Green roof qualities green roof qualities as explanatory variables for each trait (Table 2) based on our expertise. Because almost all explana- Wild bee species richness and abundance was best predicted tory variables were collinear (cor. ranged from 0.3 to 0.8), we by forage availability (Table 3). only formulated GLMs with one explanatory variable. Model Forage availability had a strong positive effect on wild bee selection was done on AICc bases as described above. The species richness and abundance (Fig. 2a, b; Table 7 in explained deviance (R ) was calculated for the most ac- Appendix). Species richness and abundance increased from GLM curate GLMs to assess how much variation of the response is Bvery low^ to Blow^ and further to Bmedium^ forage avail- explained by the explanatory variable (Zuur et al. 2013). ability, but from Bmedium^ to Bhigh^ the positive effect was Model validation of GLMMs and GLMs was performed by minimal. diagnostic plots and dispersion values (Zuur et al. 2013). The The models with substrate characteristics as an explanatory package Beffects^ (Fox 2003) was used to visualize the effects variable were the most accurate for predicting wild bee species of the most accurate models. richness and abundance in relation to uniform green roof qual- To assess whether Sedum and other wildflowers provided ities (Table 4). crucial floral resources, a data sub-set of the sub-areas Roofs with fine substrates represented the highest mean (± BSedum^ and BWildflowers^ from five roofs (Table 1)was SD) wild bee diversity (47.5 ± 24.7) and abundance (232.5 ± analysed. The other green roofs had to be omitted from this 147.8) and affected wild bee species richness (Fig. 3a) and analysis because they did not include both sub-areas. Due to abundance (Fig. 3b) distinctively positive compared to roofs small sample size this was done descriptively by comparing with mixed or coarse substrates. Roofs with mixed substrates the aggregated wild bee species richness, abundance and increased moderately and represented a higher mean (± SD) Table 3 Candidate GLMMs for GLMM response variables Explanatory variables K AICc ΔAICc predicting wild bee species richness and abundance in Wild bee species richnness ~ forage availability + Broof^ + Bmonth^ 7894.38 0.000 dependence of floral resources on green roofs in Vienna ~ flowering plant species + Broof^ + Bmonth^ 4 1049.48 155.09 ~1+ Broof^ + Bmonth^ 3 1176.31 281.93 Wild bee abundance ~ forage availability + Broof^ + Bmonth^ 7 1195.79 0.000 ~ flowering plant species + Broof^ + Bmonth^ 4 1422.91 227.12 ~1+ Broof^ + Bmonth^ 3 1601.02 405.23 Quotation marks signify random factors K Number of estimated parameters, AICc Second order Aikaike Information Criterion, ΔAICc Difference between AICc to next accurate model 434 Urban Ecosyst (2018) 21:429–446 On average, floral resources were highest on roofs with fine (a) substrates (forage availability: Bmedium^, flowering plant species: 87 ± 69.3), less on roofs with mixed substrates (forage availability: Blow^, flowering plant species: 55.3 ± 20.6) and lowest on roofs with coarse substrate (forage availability: Bvery low^ to Blow^, flowering plant species: 20.3 ± 9.3). Models including substrate depth, roof height and age as explanatory variables were less accurate in predicting the total wild bee species richness and abundance. However, substrate characteristics and depths were highly positive collinear (cor = 0.8) and it should be mentioned here, that roofs with deep substrates affected wild bees positively too (Fig. 5a, b in very low low medium high very high Appendix). (b) Wild bee traits on roofs 12 A majority of 81 polylectic wild bee species (97% individuals) contrasts with a minority of ten oligolectic species (3.4% in- dividuals). As reflected by the CWM (Table 5), the wild bee community was composed of ground nesting (43.3% species; 6 43.3% individuals) and above-ground nesting wild bees (43.3% species; 53.5% individuals). Solitary species (62.2%) accounted for 54.7% of the individuals and eusocial species (24.4%) for 42.4% of the individuals. Parasitic species (13.3%) were represented by 3.1% of the total abundance and very low low medium high very high not included in the CWM results (Table 5). The CWM of body size from wild bees on the studied roofs ranged between 8.3 Forage availability and 11.2 mm. Individuals of ground nesting species ranged in Fig. 2 Effects of forage availability on (a) wild bee species and (b) CWM of body size between 6.4 and 9.6 mm and were mainly abundance per monthly collection period (n = 7) on each roof (n = 9). BVery low^ forage availability was used as baseline for parameter eusocial (68%). On the other hand, individuals of above- estimation. Error bars = 0.95 confidence intervals for fitted effects ground nesting species were larger (8.4–14.5 mm) and pre- dominantly solitary (74%). wild bee diversity (26 ± 9.2) and abundance (89.3 ± 42.8) Substrate character was the most important predictor for compared to roofs with coarse substrates, which comprised the species richness and abundance of ground nesting, the lowest mean (± SD) wild bee species richness (21.5 ± above-ground nesting and solitary wild bees as well as for 11.3) and abundance (64.7 + 50.8). the abundance of eusocial species (Table 8 in Appendix). Table 4 Candidate GLMs for GLM response variables Explanatory variables K AICc ΔAICc R GLM predicting wild bee species richness and abundance in Wild bee species richness ~ substrate character 3 98.04 0.00 41.3% dependence of uniform green roof qualities on green roofs in Vienna ~ substrate depths (cm) 2 108.89 10.85 ~ height (m) 2 120.70 22.65 ~ age (years) 2 121.41 23.37 Wild bee abundance ~ substrate character 3 357.08 0.000 51.3% ~ substrate depths (cm) 2 497.38 140.31 ~ age (years) 2 620.55 263.47 ~ height (m) 2 628.98 271.90 K Number of estimated parameters, AICc Second order Aikaike Information Criterion, ΔAICc Difference between AICc to next accurate model, R GLM explained deviance Wild bee abundance per observation Wild bee species per observation Urban Ecosyst (2018) 21:429–446 435 of Lasioglossum laticeps and Halictus subauratus nests on (a) the roofs with fine substrate underpin that result. The positive 50 effect of fine substrate on ground nesting wild bee diversity (Appendix Fig. 6a) was higher (R =35.8%) compared to GLM the effect of substrate depths (R =9.3%; Appendix GLM Fig. 6b). Substrate depths affected eusocial species richness positively (R = 42.4%; Appendix Fig. 6f). The positive GLM effect of fine substrate was weaker on the abundance of above- ground nesting species and solitary species (R =30.3%; GLM Appendix Fig. 6e, i). The observation of a Bombus lapidarius colony in the fine substrate of a roof underlines its positive effect on above-ground nesting and eusocial species. The coarse mixed fine abundance of parasitic wild bees was slightly positively af- fected by substrate depth (R = 6.7%, Appendix Fig. 6k). 250 GLM (b) Further, the effect of roof height on parasitic species richness 2 2 (R = 4.14%) and abundance (R = 10.1%) was also GLM GLM weakly positive (Appendix Fig. 6j, l). The species richness and abundance of polylectic wild bees was highly positively affected by the number of flowering plant species (R = 58.3%; 64.1%; Appendix Fig. 6m, GLM n). Oligolectic wild bee species richness was enhanced by an increasing number of flowering plant species (R = GLM 68.2%; Appendix Fig. 6o). Moreover, the abundance of oligolectic species (Appendix Fig. 6p, q) was even higher positively affected by forage availability (R = 74.6%) GLM coarse mixed fine 2 than by flowering plant species richness (R = 66.9%). GLM Substrate character Increasing roof height decreased the body size (Appendix Fig. 3 Effects of substrate character on (a) wild bee species and (b) Fig. 6r) of wild bees moderately (R =23.9%). GLM abundance. The factor level Bcoarse^ wasusedasbaselinefor parameter estimation. Error bars = 0.95 confidence intervals for fitted effects BSedum^ vs. BWildflowers^ sub-areas Fine substrate had the greatest positive effect on the abun- Thefiveroofswith BSedum^ as well as BWildflowers^ dance of ground nesting and eusocial species (R = sub-areas comprised 55 wild bee species (16.7 ± 7) and GLM 75.5% and 72.2%; Appendix Fig. 6c, g). The observations 421 individuals (42.2 ± 28.3). On average, the sub-areas Table 5 Community weighted means (CWM) per roof for nesting type, sociality, body size (mm) and pollen specialization (pl = polylectic) over all species and species (spp.) per nesting type (ground nesting, above ground nesting) separately CWM All species Ground nesting spp. Above-ground nesting spp. Green roof Nesting type Sociality Body size (mm) Pollen special. Sociality Body size (mm) Sociality Body size (mm) 1 Ground nesting eusocial 8.41 pl eusocial 7.27 solitary 9.43 2 Above-ground solitary 8.31 pl eusocial 7.50 solitary 8.39 3 Above-ground solitary 8.87 pl eusocial 6.38 solitary 9.44 4 Ground-nesting eusocial 8.81 pl eusocial 8.06 solitary 9.87 5 Above-ground solitary 9.55 pl eusocial 8.26 solitary 10.71 6 Ground-nesting solitary 10.50 pl solitary 9.25 eusocial 13.0 7 Ground-nesting eusocial 10.69 pl eusocial 7.89 eusocial 14.47 8 Ground-nesting solitary 11.16 pl eusocial 9.59 solitary 12.89 9 Above-ground solitary 8.68 pl solitary 8.12 solitary 9.01 Wild bee abundance per roof Wild bee species per roof 436 Urban Ecosyst (2018) 21:429–446 30 forage availability and only 13 wild bee species (23 indi- (a) viduals) were sampled. In July conditions in the sub-areas were reversed: BSedum^ decreased to Blow^ forage avail- ability, whereas BWildflowers^ increased to Bmedium^ for- age availability and 26 wild bee species (62 individuals) exceeded the 21 wild bee species (42 individuals) in BSedum^ sub-areas. Thedatasub-setof BSedum^ and BWildflowers^ sub- March April May June July August Sept. areas revealed that some species were only observed in BSedum^ sub-areas (Table 6 in Appendix) but these spe- (b) cies (e.g. Bombus pascuorum or Megachile lagopoda) were also documented on other roofs and in other sub- areas. Therefore, no species was exclusively observed in BSedum^ sub-areas. The highly abundant group of Halictidae showed no preference for a distinct sub-area (Table 6 in Appendix). March April May June July August Sept. Discussion (c) The studied green roofs hosted 90 wild bee species, which represent 20% of Vienna’s 456 recorded wild bees (Zettel et al. 2016). Compared to other studies, the roofs showed a considerable wild bee diversity: Depending on the num- ber of sampled roofs, the amount of study years and the March April May June July August Sept. study’s scope, other authors documented between 17 and 77 wild bee species on green roofs (Brenneisen 2005; Sampling period 2014 Kadas 2006; Tonietto et al. 2011; Ksiazek et al. 2012, Fig. 4 Timelines to compare (a) total wild bee species richness, (b) total 2014; MacIvor et al. 2014;MacIvor 2015). Due to its abundance and (c) mean forage availabilities in BSedum^ and geographical position, Vienna is influenced by the BWildflowers^ sub-areas on green roofs during the sampling period 2014; Lines: full = BSedum^,dashed= BWildflowers^,dottedsquare= Pannonian climate, which could be one reason that the highlights Sedum main flowering time studied green roofs emerged with higher species numbers compared to the studies mentioned above. The authors in Switzerland (Brenneisen 2005) and London (Kadas 2006) ranged from Bvery low^ to Bmedium^ forage availability used yellow-pan traps for wild bee sampling, which may (1.7 ± 0.6 vs. 2.5 ± 0.5) and differed in flowering plant be another reason for the difference in wild bee diversity species richness (3.2 ± 0.8 vs. 21.4 ± 7.1). Wild bee spe- found on the roofs and makes a comparison of the studies cies richness (17.6 ± 5.6) and abundance (44.8 ± 26.0) difficult. was slightly higher in BWildflowers^ sub-areas than in The roofs showed lower wild bee diversity compared BSedum^ (15.8 ± 8.7; 39.6 ± 33.4). to ground level habitats in Vienna, where recently 119 to The picture becomes clearer by looking at the whole 144 species were reported from the Danube Island sampling period: Except for May, the wild bee occurrence (Pachinger and Hölzler 2006), theVienneseBotanical and diversity harmonized well with the forage availability Garden (Hölzler 2004) and the Danube Park (Zettel in the sub-areas (Fig. 4a-c). For example, in April BSedum^ et al. 2013). Similarly, authors from Chicago reported comprisedonaverage Bvery low^ forage availability and higher wild bee diversity and abundance on ground level only one wild bee species (3 individuals) was sampled, habitats than on green roofs (Tonietto et al. 2011). Green whereas BWildflowers^ offered higher floral resources and roofs and ground level habitats in Toronto (Colla et al. 3 to 9 wild bee species (5 to 14 individuals) were observed. 2009) and Switzerland (Brenneisen 2005)comprised sim- Further, BSedum^ reached the main flowering time in June ilar wild bee diversity. However, these comparisons have with Bmedium^ forage availability and 21 wild bee species little informative value because of the different sample (62 individuals), but BWildflowers^ decreased to Blow^ sizes and study designs realised in these studies. Total bee abund. Mean forage avail. Total bee species per sub-area per sub-area per sub-area Urban Ecosyst (2018) 21:429–446 437 The results of our study are interpreted in the light of the the abundance of the respective oligolectic species. small sample size and heterogeneous green roof qualities. This However, compared to studies carried out on ground level and the assessment of model quality by diagnostic plots reveal habitats in Vienna, recording 16 to 24% of oligolectic wild that the presented results are difficult to generalise. However, bee species (Hölzler 2004; Pachinger and Hölzler 2006;Zettel by interpreting and discussing the results carefully we would et al. 2013), we observed only 11% oligolectic species. also like to point out open questions regarding wild bees on The positive effect of fine substrate on the abundance of green roofs. ground nesting as well as eusocial species is reflected in the traits of the most abundant wild bee on the roofs: Halictus Green roof qualities subauratus. This Sweat bee is a ground nesting and eusocial species, requires fine and sandy substrates and nests are Wild bee diversity and abundance on green roofs are positive- established at 10–15 cm below the surface (Scheuchl and ly affected by enhanced floral resources and fine substrates. Willner 2016). Therefore, the substrate conditions that posi- Although the most accurate GLMM indicated that forage tively affected wild bees in this study are beneficial for ground availability affected wild bee diversity and abundance primar- nesting species, which require fine substrates but build their ily, the importance of entomophilous plant species richness nests shallowly. Different ground nesting species require var- was evident by the relationship of these variables. The positive ious substrate properties (e.g. compaction, content of sand or effect of high and ongoing forage availability through a high humus, Scheuchl and Willner 2016) and build nests in differ- plant diversity on wild bee communities is known from other ent depths (Cane and Neff 2011). This was also revealed in ecosystems (e.g. Potts et al. 2003; Zurbuchen and Müller our study, because fine substrates as well as substrate depth, 2012; Braun-Reichert 2013) and was found on green roofs affected the diversity of ground nesting wild bees on green in Switzerland (Brenneisen 2005) and Toronto (Tonietto roofs. Like in other studies (Brenneisen 2005; Colla et al. et al. 2011). 2009; Tonietto et al. 2011), we found Lasioglossum and The finding that roofs with fine substrates enhanced wild Halictus species in high abundance (340 individuals) and spe- bee diversity and abundance has to be interpreted with cau- cies richness (20 species) compared to Andrena species with tion, because only two of the nine roofs represented this factor only few or single individuals per species. Two explanations level. Furthermore, planting possibilities on green roofs are seem reasonable for this observation: Firstly, Andrena species limited to substrate properties (Dunnett and Kingsbury occurrence is more scattered because the females breed soli- 2008). The reason for the positive effect of fine substrates on tary. Secondly, Andrena was represented by large species, the overall wild bee diversity and abundance in this study which probably require deeper substrates for nesting than could be the relation of higher floral resources on roofs with present on the studied green roofs and just forage here. fine substrates. But wild bee traits, especially the occurrence Similarly, Tonietto et al. (2011) concluded that large ground of ground nesting and eusocial species may better explain this nesting species nested in adjacent ground level habitats and effect. used green roofs for foraging. Since we did not sample wild bees in adjacent ground level habitats, our data cannot support Wild bee traits these findings directly. However, more Andrena species where documented on ground level habitats in Vienna (Pachinger The studied green roofs mainly attracted pollen generalists. and Hölzler 2006;Zetteletal. 2013) than on the studied green Although polylectic, these wild bees were affected by roofs. flowering plant diversity because some of them showed dis- Based on the CWM, solitary wild bees were determined to tinct preferences to certain plant taxa. For example, the benefit be characteristically on the studied green roofs. It was difficult of Sedum for polylectic species with pollen preferences for to determine whether ground nesting or above-ground nesting Sedum (Anthidium oblongatum, A. strigatum, Hylaeus was the most typical nesting trait, but the latter were more punctatus, Megachile leachella, M. pilidens, M. rotundata, abundant. Above-ground nesting wild bee species are typical M. willughbiella)(Westrich 1989b) was pointed out by their of urban areas (reviewed in Hernandez et al. 2009), because higher abundance on roofs with BSedum^ sub-areas compared the high density of vertical structures offers many potential to other roofs (Appendix Table 6). Oligolectic wild bee spe- nesting sites (Cane 2005). Further, we reported above- cies richness was enhanced by increasing floral diversity, be- ground nesting wild bees in higher diversity (43.3% of spe- cause it could raise the probability that different host plants cies) compared to ground level habitats in Vienna, comprising occur on a green roof for species with different pollen special- 20 to 32% of above-ground nesting species (Hölzler 2004; izations. Also, increasing forage availability could lead to Pachinger 2008;Zettel etal. 2013). This could be explained higher flower coverage of distinct host plants and enhance by differences in site features because the mentioned studies 438 Urban Ecosyst (2018) 21:429–446 were conducted in Vienna’s botanical garden or big recrea- exclusively sampled in BWildflowers^ sub-areas (Table 6 tional areas (e.g. Danube Island) where the density of vertical Appendix), and floral resources were only present in this structures within species-specific flying distances maybe low- sub-area (Fig. 4c). er than on roofs in urban areas. Further research has to be carried out to identify the effect of green roofs’ surrounding structures on above-ground nesting wild bees. Conclusion In contrast to MacIvor (2015),wefoundnostrongevi- dence, that increasing roof height negatively altered the It was demonstrated that the wild bee community on nine wild bee community. The CWM of body size from wild green roofs in Vienna was strongly positively affected by flo- bees tended to decrease with increasing roof height. A pos- ral resources and substrate characteristics. We conclude that sible explanation is that smaller species populate high fine substrates enhance ground nesting and eusocial wild bee green roofs without utilising ground level habitats. This species. This attributes to some guidance (Gedge et al. 2008, trend is probably reflected by the weak positive effect of 2012) and the proposition by Brenneisen (2006) that the cre- roof height on parasitic wild bees, because the occurrence ation of small areas of mounds, consisting of finer and deeper of brood parasitic species indicates a vital host population substrates should be a design consideration in the planning (Hudson et al. 2006). For example, the appearance of four processes of (extensive) green roofs in order to support ground Coelioxys species was in line with the occurrence of their nesting wild bee species. Above-ground nesting and solitary host, namely Megachile species, which were recorded in species are characteristic in urban environments, but further high abundance on the roofs (Table 6 in Appendix). research is needed to assess, which parameters surrounding green roofs affect them primarily. BSedum^ vs. BWildflowers^ Honey bees were observed on all roofs, but their abun- dance did not interfere with the occurrence of wild bees. Based on the study of five roofs, Sedum is an important tem- Oligolectic wild bee species occurrence was low, but strongly poral floral resource because during its main flowering period positively affected by increasing floral diversity. Sedum that is in June higher wild bee diversity and abundance was observed typically planted on green roofs promotes polylectic wild bee in this sub-area. Similar observations were reported by species, of which some preferably forage on this plant. The MacIvor et al. (2014) who found high proportions of Sedum study of BSedum^ and BWildflower^ sub-areas on five roofs pollen in palynological samples from wild bees gathered on a indicated that Sedum species can compensate temporal lacks green roof during the main flowering period of Sedum. of resources on green roofs during its main flowering period. However, in our study qualitative comparison showed no On the other hand, various wildflowers were important forag- big difference of forage availability between these sub-areas, ing resources in spring (March, April) and summer (July, which may suggest complementary resource availability by August). Sedum and wildflowers. The concurrent trend of forage avail- We conclude that increasing floral resources (flower abun- ability, wild bee diversity and abundance in the sub-areas over dance and floral diversity) and the installation of patches with the season underpins the results discussed earlier that forage fine and deeper substrates should be considered during the availability enhances the wild bee community on green roofs. planning process to enhance the wild bee diversity and abun- This is supported by similar findings in other studies dance on green roofs. (Brenneisen 2005; Kadas 2006; Tonietto et al. 2011). The opposite trend in May is explained by observations during Acknowledgements Open access funding provided by University of Natural Resources and Life Sciences, Vienna (BOKU). We would like fieldwork that Sedum was already flowering and favourably to thank the responsible people of the buildings (Mrs. Arlt, Mrs. Ehs, Mrs. visited by wild bees compared to other flowering plants. All Haimer, Mrs. Kapui, Mrs. Leidinger, Mr. Möseler, Mr. Schatovits, Mr. Hylaeus species were exclusively sampled on BSedum^ sub- Steinbauer, Mr. Ziemak,) and the municipal council 22 – Environmental areas during this month, probably because of a preference for Protection in Vienna (especially Mrs. Doppler and Mr. Preiss) for obtaining flexible access to the study roofs. Further, we thank Vera Enzi Sedum as forage resource, which is already known for some for her helpful suggestions of green roofs during the selection process of Hylaeus species (Westrich 1989b). Sedum promoted mainly study sites. Special thanks go to Karl Mazzucco for his help with bee generalist species because the abundance peaks in June species identification, to Luise Kratschmer and Martin Wittner for help and August (Fig. 4b) are dominated by a few generalist with bee preparation, to Matthias Kropf and Michael Kopetzky for help with plant identification, to Dusty Gedge for proof reading and to species (e.g. Anthidium oblongatum, Halictus subauratus, Bernhard Kratschmer for financial support. The baseline of this work Hylaeus punctatus, H. hyalinatus, Megachile rotundata, was done in the context of the first author’s master thesis (BSummen M. willughbiella). During spring (March, April), early oc- auf den Dächern Wiens^) approved by the University of Natural curring wild bees (Andrena, Anthophora and Osmia)were Resources and Life Sciences, Vienna (BOKU) in May 2015. Urban Ecosyst (2018) 21:429–446 439 Appendix Table 6 Bee species and abundance on nine green roofs in Vienna (2014) and abundance in BSedum^ (Sed.) and BWildflower^ (Wildf.) sub-areas from a study of five roofs. Apis mellifera abundance represents field counts Bee families, genera and species Green roofs Ecological traits Abundance 123456 789N S PS MBS Sed. Wildf. Andrenidae Andrena bimaculata (Kirby 1802) 1 1 t sol pl 13 1 danuvia Stoeckhert 1950 1 t sol pl 14 dorsata (Kirby 1802) 11t sol pl 9,5 flavipes Panzer 1799 3 t sol pl 10,5 gravida Imhoff 1832 6 1 t sol pl 13 minutula (Kirby 1802) 1 t sol pl 6 nigroaenea (Kirby 1802) 1 t sol pl 14 1 ovatula (Kirby 1802) 2 t sol pl 9,5 pilipes Fabricius 1781 1tsolpl13,5 1 tibialis Kirby 1802 1 t sol pl 13 varians (Kirby 1802) 1 t sol pl 10 Apidae Anthophora crinipes Smith 1854 1 t sol pl 12 plumipes (Pallas 1772) 2 12 12 2 3 t sol pl 15 2 quadrimaculata (Panzer 1798) 1 7 1 t sol pl 10,5 Apis mellifera Linnaeus 1758 115 91 108 72 448 501 45 43 47 fb hs pl 14,5 227 97 Bombus bohemicus Seidl 1801 1 p p p 20 1 hortorum (Linnaeus 1761) 1 c eus pl 16 humilis Illiger 1806 3 3 3 3 1 c eus pl 13,5 3 1 hypnorum (Linnaeus 1758) 4 2 5 2 c eus pl 14 5 lapidaries(Linnaeus1758) 4311 516 1 337c eus pl 17 9 16 lucorum (Linnaeus 1761) 1 2 1 4 1 c eus pl 15 1 1 pascuorum (Scopoli 1763) 2 1 21 8 1 c eus pl 13,5 1 pratorum (Linnaeus 1761) 1 c eus pl 13 rupestris (Fabricius 1793) 1 p p p 20 1 terrestris(Linnaeus1758) 22111 11 c eus pl 17 3 3 Eucera nigrescens Pérez 1879 3 t sol ol 18,5 3 Melecta albifrons Forster 1771 2 p p p 12,5 Nomada goodeniana (Kirby 1802) 1 2 p p p 12 Colletidae Colletes daviesanus Smith 1846 3 2 4 t sol ol 8,5 1 5 Hylaeus cardioscapus Cockerell 1924 1 r sol pl 6,5 communis Nylander 1852 2 5 17 6 9 4 c sol pl 5,5 5 12 gredleri Förster 1871 2 1 r sol pl 5 1 1 hyalinatus Smith 1842 15 10 12 4 12 c sol pl 6,5 23 13 imparilis Förster 1871 5 u sol pl 4,5 3 2 leptocephalus (Morawitz 1870) 9 5 4 13 c sol pl 5 7 11 pictipesNylander1852 6423 9c sol pl 4,5 5 13 punctatus (Brullé 1832) 5 10 7 11 4 c sol pl 5,5 13 13 sinuatus (Schenck 1853) 1 c sol pl 7,5 1 styriacus Förster 1871 1 1 1 c sol pl 4,5 Halictidae Halictus kessleri Bramson 1879 5 t eus pl 7 3 2 maculatus Smith 1848 1 1 t eus pl 8 2 rubicundus (Christ 1791) 1 13 1 t eus pl 10 1 seladonius (Fabricius 1794) 5 7 t eus pl 7 1 4 simplexBlüthgen1923 3 51 112t sol pl 9,5 3 3 subauratus(Rossi1792) 19 9 22 29 537t eus pl 7,5 15 22 tumulorum (Linnaeus 1758) 5 4 8 5 t eus pl 7 3 6 440 Urban Ecosyst (2018) 21:429–446 Table 6 (continued) Bee families, genera and species Green roofs Ecological traits Abundance 123456 789N S PS MBS Sed. Wildf. Lasioglossum calceatum (Scopoli 1763) 1 8 2 t eus pl 9 1 laticeps (Schenck 1868) 2 4 8 1 t eus pl 7 3 2 leucozonium (Schrank 1781) 1 3 1 1 1 t sol pl 9 2 malachurum (Kirby 1802) 1 t eus pl 8,5 marginatum(Brullé1832) 11 43 1t eus pl 8 2 1 minutulum (Schenck 1853) 1 t sol pl 6,5 morio(Fabricius1793) 42423 12t eus pl 5,5 7 5 nigripes (Lepeletier 1841) 4 t eus pl 9,5 nitidulum(Fabricius1804) 17429 6t eus pl 6 9 6 pauxillum (Schenck 1853) 1 9 2 t eus pl 5,5 2 1 politum (Schenck 1853) 4 1 11 25 15 8 2 t eus pl 4,5 5 12 sabulosum (Warncke 1986) 1 t sol pl 6,5 villosulum (Kirby 1802) 2 t sol pl 6,5 Sphecodes albilabris (Fabricius 1793) 1 p p p 12 monilicornis (Kirby 1802) 1 1 1 3 p p p 8,5 1 4 ruficrus (Erichson 1835) 1 pp p 9 1 Megachilidae Anthidium manicatum (Linnaeus 1758) 1 5 c sol pl 14.5 oblongatum (Illiger 1806) 2 1 4 2 c sol pl 9 6 2 strigatum (Panzer 1805) 1 1 fb sol pl 6.5 1 Chelostoma florisomne (Linnaeus 1758) 1 x sol ol 9,5 rapunculi (Lepeletier 1841) 7 1 c sol ol 9 Coelioxys conoidea (Illiger 1806) 1 p p p 14 1 echinata Förster 1853 1 1 2 p p p 8,5 2 elongata Lepeletier 1841 3 2 1 p p p 12,5 4 1 mandibularis Nylander 1848 1 1 p p p 10 1 Heriades crenulatus Nylander 1856 2 1 r sol ol 6,5 2 rubicola Pérez 1890 1 9 1 r sol pl 6 1 truncorum (Linnaeus 1758) 5 r sol ol 7,5 Megachile apicalis Spinola 1808 2 4 4 2 4 c sol pl 10 3 6 centuncularis (Linnaeus 1758) 1 1 c sol pl 10 dorsalisPérez1879 221 2 16t sol pl 9,5 6 5 ericetorum Lepeletier 1841 2 c sol ol 12 lagopoda (Linnaeus 1761) 1 3 2 1 c sol pl 15,5 4 pilidensAlfken1924 22333 12c sol pl 10 5 4 rotundata(Fabricius1787) 19233 7c sol pl 8,5 15 3 versicolor Smith 1844 4 r sol pl 10,5 willughbiella(Kirby1802) 310 4318 167c sol pl 14 16 11 Osmia adunca (Panzer 1798) 1 c sol ol 10,5 bicornis (Linnaeus 1758) 3 4 1 7 1 c sol pl 11,5 5 caerulescens (Linnaeus 1758) 2 4 2 c sol pl 9 1 cornuta (Latreille 1805) 2 2 c sol pl 13,5 leucomelana (Kirby 1802) 1 r sol pl 8 Stelis punctulatissima (Kirby 1802) 2 p p p 9 Mellitidae Melitta haemorrhoidalis (Fabricius 1775) 1 t sol ol 12 leporina (Panzer 1799) 1 t sol ol 12 Bee abundance/roof 97 116 111 128 337 6 40 40 117 Beespecies/roof 30302830656 162034 Flowering plant species 32 23 36 38 136 11 53 15 77 Mean forage availability (± SD) 3 (0.9) 2 (0.7) 3 (0.5) 3 (1.2) 4 (0.9) 1 (1) 2 (0.9) 2 (1.6) 3 (1.1) Ecological traits (Westrich 1989a; Scheuchl and Willner 2016): N Nesting type: Ground nesting: t terricol, Above-ground nesting: c cavity nesting, r rubicol, fb free-building on structures, x xylicol, p parasitic, S Sociality: sol solitary, eus primitively eusocial, hs highly eusocial, p parasitic, PS Pollen Specialization: ol oligolectic, pl polylectic. MBS Mean body size: average species specific body size from ♂ and ♀ (mm) values given by identification literature listed in Methods and References. Nomenclature after Gusenleitner et al. (2012); taxonomic rank by Families after Michener (2007) Urban Ecosyst (2018) 21:429–446 441 Table 7 Parameters estimated and p-values for each response variable. BVery low^ forage availability served as a baseline for parameter estimation Response variable Explanatory variable Estimate ± SE Random effect SD (Wild bees) (Forage availability) Roof (N = 9) Month (N = 7) Species richness 0.140 0.435 Intercept −1.199 ± 0.267 Low 1.854 ± 0.223 Medium 2.588 ± 0.219 High 2.669 ± 0.232 Very High 3.096 ± 0.242 Abundance 0.214 0.535 Intercept −0.988 ± 0.283 Low 1.991 ± 0.198 Medium 2.667 ± 0.197 High 2.792 ± 0.208 Very High 3.292 ± 0.218 Table 8 Mean values ±SD for wild bee species richness, wild bee abundance, forage availability and entomophilous plant species richness among characteristic properties on nine green roofs in Vienna Response variables for traits Explanatory variables K AICc ΔAICc ω R i GLM Ground nesting Wild bee species richness ~ substrate character 3 63.10 0.00 0.46 35.8% ~ substrate depths (cm) 2 64.07 0.97 0.28 9.3% ~ height (m) 2 65.36 2.26 0.15 ~ age (years) 2 66.03 2.93 0.11 Wild bee abundance ~ substrate character 3 131.76 0.00 1 75.5% ~ substrate depths (cm) 2 255.47 123.71 0 ~ age (years) 2 303.34 171.58 0 ~ height (m) 2 350.54 218.78 0 Above-ground nesting Wild bee species richness ~ substrate character 3 79.63 0.00 0.97 42.8% ~ substrate depths (cm) 2 86.44 6.81 0.03 ~ age (years) 2 94.37 14.74 0.00 ~ height (m) 2 95.82 16.19 0.00 Wild bee abundance ~ substrate character 3 309.23 0.00 1 30.3% ~ substrate depths (cm) 2 344.55 35.32 0 ~ age (years) 2 382.35 73.12 0 ~ height (m) 2 407.49 98.26 0 Solitary Wild bee species richness ~ substrate character 3 79.29 0.00 0.99 42.4% ~ substrate depths (cm) 2 88.36 9.07 0.01 ~ height (m) 2 94.69 15.40 0.00 ~ age (years) 2 94.84 15.54 0.00 Wild bee abundance ~ substrate character 3 305.13 0.00 1 30.3% ~ substrate depths (cm) 2 358.37 53.23 0 ~ age (years) 2 384.23 79.10 0 ~ height (m) 2 402.01 96.88 0 Eusocial Wild bee species richness ~ substrate depths (cm) 2 53.45 0.00 0.64 42.4% ~ substrate character 3 55.82 2.38 0.19 ~ age (years) 2 57.35 3.91 0.09 442 Urban Ecosyst (2018) 21:429–446 Table 8 (continued) Response variables for traits Explanatory variables K AICc ΔAICc ω R i GLM ~ height (m) 2 57.58 4.13 0.08 Wild bee abundance ~ substrate character 3 137.96 0.00 1 72.2% ~ substrate depths (cm) 2 217.40 79.44 0 ~ age (years) 2 292.36 154.40 0 ~ height (m) 2 355.53 197.57 0 Parasitic Wild bee species richness ~ height (m) 2 40.47 0.00 0.67 4.14% ~ substrate depths (cm) 2 42.59 2.12 0.23 ~ age (years) 2 44.60 4.13 0.08 ~ substrate character 3 47.78 7.31 0.02 Wild bee abundance ~ height (m) 2 50.39 0.00 0.49 10.1% ~ substrate depths (cm) 2 51.29 0.90 0.31 6.7% ~ age (years) 2 53.06 2.67 0.13 ~ substrate character 3 54.37 3.97 0.07 Polylectic Wild bee species richness ~ flowering plant species 2 76.84 0.00 0.98 58.3% ~ forage availability 3 84.71 7.87 0.02 Wild bee abundance ~ flowering plant species 2 266.84 0.00 1 64.1% ~ forage availability 3 326.45 59.61 0 Oligolectic Wild bee species richness ~ flowering plant species 2 27.12 0.00 0.91 68.2% ~ forage availability 3 31.71 4.59 0.09 Wild bee abundance ~ forage availability 3 43.98 0.00 0.51 74.6% ~ flowering plant species 2 44.06 0.07 0.49 66.9% Body size CWM (mm) ~ height (m) 3 34.24 0.00 0.58 23.9% ~ age (years) 4 36.23 2.00 0.22 5.1% ~ substrate depths (cm) 4 36.43 2.20 0.19 ~ substrate character 4 43.67 9.43 0.01 K Number of estimated parameters, AICc Second order Aikaike Information Criterion, ΔAICc Difference between AICc to next accurate model, ωi Akaike’s weight, R GLM explained deviance Fig. 5 Effects of substrate depths (cm) on (a) wild bee species and (b) abundance. Grey band = 0.95 confidence band for fitted effects Urban Ecosyst (2018) 21:429–446 443 Fig. 6 Effects of green roof qualities on wild bee species richness and abundance per traits. Error bars and grey bars = 0.95 confidence intervals for fitted effects 444 Urban Ecosyst (2018) 21:429–446 Fig. 6 (continued) Open Access This article is distributed under the terms of the Creative Amiet F, Herrmann M, Müller A, Neumeyer R (2001) Apidae 3: Halictus, Commons Attribution 4.0 International License (http:// Lasioglossum. Schweizerische Entomologische Gesellschaft, creativecommons.org/licenses/by/4.0/), which permits unrestricted use, Neuachtel distribution, and reproduction in any medium, provided you give Auer I, Böhm R (2011) Wetter und Klima in Wien. Vielfalt auf engstem appropriate credit to the original author(s) and the source, provide a link Raum. In: Berger R, Ehrendorfer F (eds) Ökosystem Wien: Die to the Creative Commons license, and indicate if changes were made. Naturgeschichte einer Stadt. Böhlau Verlag GesmbH&Co.KG, Wien, pp 88–105 Bates D, Maechler M (2016) Matrix: sparse and dense matrix classes and methods. R package version 1.2-8. https://CRAN.R-project.org/ package=Matrix. Accessed 10 Nov 2016 References Bates D, Maechler M, Bolker B, Walker S (2015) Fitting Linear Mixed- Effect Models Using lme4. J Stat Softw 67:1–48 Baumann N (2006) Ground-Nesting Birds on Green Roofs in Amiet F, Müller A, Neumeyer R (1999) Apidae 2: Colletes, Dufourea, Switzerland: Preliminary Observations. Urban Habitats 4: Hyleaus, Nomia, Nomioides, Rhophitoides, Rophites, Sphecodes, 37–50 Systropha. Schweizerische Entomologische Gesellschaft, Neuachtel Urban Ecosyst (2018) 21:429–446 445 van den Berg AE, Hartig T, Staats H (2007) Preference for Nature in Gathmann A, Tscharntke T (2002) Foraging ranges of solitary bees. J Urbanized Societies : Stress, Restoration, and the Pursuit of Anim Ecol 71:757–764 Sustainability. J Soc Issues 63:79–96 Gedge D (2003) From rubble to redstars. Green Rooftops for Sustainable Berndtsson JC (2010) Green roof performance towards management of Communities. Report. Chicago runoff water quantity and quality: A review. Ecol Eng 36:351–360. Gedge D, Newton J, Cradick K, et al (2008) Living Roofs and Walls. https://doi.org/10.1016/j.ecoleng.2009.12.014 Technical Report for the Greater London Authority’s London Plan Biesmeijer JC, Roberts SPM, Reemer M et al (2006) Parallel Declines in and Environment Teams. London Pollinators and Insect-Pollinated Plants in Britain and the Gedge D, Grant G, Kadas G, Dinham C (2012) Creating Green Roofs for Netherlands. Science 313:351–354. https://doi.org/10.1126/ Invertebrates. Buglife, Peterborough, pp 1–29 science.1127863 Gokcezade JF, Gereben-Krenn BA, Neumayer J, Krenn HW (2010) Braun-Reichert R (2013) Der Einfluss unterschiedlicher Beweidung auf Feldbestimmungsschlüssel für die Hummeln Österreichs, die Wildbienen- und Wespenfauna von Kalkmagerrasen Zeitpunkt, Deutschlands und der Schweiz (Hymenoptera, Apidae). Linzer Frequenz und Kontinuität der Beweidung. Galathea-Berichte des Kr Biol Beitr 42:5–42 Nürnberger Entomolgen 29:7–22 Gusenleitner F, Schwarz M, Mazzucco K (2012) Apidae (Insecta: Brenneisen S (2003) Ökologisches Ausgleichspotenzial von Extensiven Hymenoptera). In: Schuster R (ed) Biosystematics and Ecology Dachbegrünungen. Bedeutung des Ersatz-Ökotops für den Arten- Series No. 29: Checkliste der Fauna Österreichs, No. 6. ÖAW- und Naturschutz und die Stadtentwicklungsplanung. University of Verlag der Österreichischen Akademie der Wissenschaften, Basel, Basel Vienna, pp 1–129 Brenneisen S (2005) The Natural Roof (NADA): Research Project Report Hathaway AM, Hunt WF, Jennings GD (2008) A field study of green roof on the use of Extensive Green Roofs by Wild Bees. Wädenswil hydrologic and wanter quality performance. Trans ASABE 51:37– Brenneisen S (2006) Space for Urban Wildlife: Designing Green Roofs as Habitats in Switzerland. Urban Habitats 4:27–36 Hennig EI, Ghazoul J (2012) Pollinating animals in the urban environ- Cane JH (2005) Bees, pollination, and the challanges of sprawl. In: ment. Urban Ecosyst 15:149–166. https://doi.org/10.1007/s11252- Johnson EA, Klemens MW (eds) Nature in Fragments: The legacy 011-0202-7 of sprawl. Columbia University Press, New York, pp 109–124 Hernandez JL, Frankie GW, Thorp RW (2009) Ecology of Urban Bees : Cane JH, Neff JL (2011) Predicted fates of ground-nesting bees in soil A Review of Current Knowledge and Directions for Future Study. heated by wildfire: Thermal tolerances of life stages and a survey of Cities Environ 2:1–15 nesting depths. Biol Conserv 144:2631–2636. https://doi.org/10. Hölzler G (2004) Die Wildbienen des Botanischen Gartens der 1016/j.biocon.2011.07.019 Universität Wien. In: Pernstich A, Krenn HW (eds) Die Tierwelt Carter T, Butler C (2008) Ecological impacts of replacing traditional roofs des Botanischen Gartens der Universität Wien. Eigenverlag Institut with green roofs in two urban areas. Cities Environ 1:1–17 für Angewandte Biologie und Umweltbildung, Vienna, pp 141–163 Castleton HF, Stovin V, Beck SBM, Davison JB (2010) Green roofs; building energy savings and the potential for retrofit. Energy Build Hudson PJ, Dobson AP, Lafferty KD (2006) Is a healthy ecosystem one 42:1582–1591. https://doi.org/10.1016/j.enbuild.2010.05.004 that is rich in parasites? Trends Ecol Evol 21:381–385. https://doi. Colla SR, Willis E, Packer L (2009) Can green roofs provide habitat for org/10.1016/j.tree.2006.04.007 urban bees (Hymenoptera: Apidae). Cities Environ 2:1–12 Kadas G (2006) Rare Invertebrates Colonizing Green Roofs in London. Dathe HH (1980) Die Arten der Gattung Hylaeus F. in Europa Urban Habitats 4:66–86 (Hymenoptera: Apoidea, Colletidae). Mitt Zool Mus Berl 56:207– Kennedy CM, Lonsdorf E, Neel MC et al (2013) A global quantitative synthesis of local and landscape effects on wild bee pollinators in Dunnett N, Kingsbury N (2008) Planting green roofs and living walls. agroecosystems. Ecol Lett 16:584–599. https://doi.org/10.1111/ele. Timber Press Inc., Oregon Ebmer PAW (1969) Die Bienen des Genus Halictus LATR. S. L. im Klein A-M, Vaissière BE, Cane JH et al (2007) Importance of pollinators Großraum von Linz (Hymenoptera, Apidae): Systematik, in changing landscapes for world crops. Proc Biol Sci 274:303–313. Biogeographie, Ökologie und Biologie mit Berücksichtigung aller https://doi.org/10.1098/rspb.2006.3721 bisher aus Mitteleuropa bekannten Arten. Teil I. Mit neun Bildtafeln. Köhler M (2005) Long-Term Vegetation Research on Two Extensive Natkdl Jb Stadt Linz 1969:133–183 Green Roofs in Berlin. Urban Habitats 4:3–26 Ebmer PAW (1970) Die Bienen des Genus Halictus LATR.S.L. im Kremen C, Williams NM, Thorp RW (2002) Crop pollination from native Großraum von Linz (Hymenoptera, Apidae): Teil II Mit neun bees at risk from agricultural intensification. Proc Natl Acad Sci U S Bildtafeln. Natkdl Jb Stadt Linz 1970:19–82 A 99:16812–16816. https://doi.org/10.1073/pnas.262413599 Ebmer PAW (1971) Die Bienen des Genus Halictus LATR.S.L. im Ksiazek K, Fant J, Skogen K (2012) An assessment of pollen limitation Großraum von Linz (Hymenoptera, Apidae): Teil III mit 19 on Chicago green roofs. Landsc Urban Plan 107:401–408 Bildtafeln. Natkdl Jb Stadt Linz 1971:63–156 Ksiazek K, Tonietto R, Ascher J (2014) Ten bee species new to green Falk S (2015) Field Guide to the Bees of Great Britain and Ireland. British roofs in the Chicago area. Gt Lakes Entomol 47:87–92 Wildlife Publishing Bloomsbury, London Laliberté E, Legendre P, Shipley B (2014) FD: measuring functional Fernandez-Canero R, Gonzalez-Redondo P (2010) Green Roofs as a diversity from multiple traits, and other tools for functional ecology. habitat for birds: A review. J Anim Vet Adv 9:2041–2052 R package version 1.0-12. Fontaine C, Dajoz I, Meriguet J, Loreau M (2006) Functional Diversity of Landolt E (2001) Orchideen-Wiese in Wollishofen (Zürich) - ein Plant–Pollinator Interaction Webs Enhances the Persistence of Plant erstaunliches Relikt aus dem Anfang des 20. Jahrhunderts. Communities. PLoS Biol 4:129–135. https://doi.org/10.1371/ Vierteljahrsschrift der Naturforschenden Gesellschaft Zürich 146: journal.pbio.0040001 41–51 Fox J (2003) Effect Displays in R for Generalised Linear Models. J Stat So Le Féon V, Schermann-Legionnet A, Delettre Y et al (2010) ftw 8(15):1–27. http://www.jstatsoft.org/v08/i15/. Accessed 21 Intensification of agriculture, landscape composition and wild bee Feb 2017 communities: A large scale study in four European countries. Agric Frankie GW, Thorp RW, Hernandez J et al (2009) Native bees are a rich Ecosyst Environ 137:143–150. https://doi.org/10.1016/j.agee.2010. natural resource in urban California gardens. Calif Agric 63:113– 01.015 120. https://doi.org/10.3733/ca.v063n03p113 446 Urban Ecosyst (2018) 21:429–446 MacIvor JS (2015) Building height matters: nesting activity of bees and der Arten der Schweiz Band 3: Schlüssel der Arten der Familie Andrenidae. Eigenverlag, Velden/Vils wasps on vegetated roofs. Isr J Ecol Evol 9801:1–9. https://doi.org/ 10.1080/15659801.2015.1052635 Schrader S, Böning M (2006) Soil formation on green roofs and its MacIvor JS, Lundholm J (2010) Insect species composition and diversity contribution to urban biodiversity with emphasis on Collembolans. on intensive green roofs and adjacent level-ground habitats. Urban Pedobiologia (Jena) 50:347–356. https://doi.org/10.1016/j.pedobi. Ecosyst 14:225–241. https://doi.org/10.1007/s11252-010-0149-0 2006.06.003 MacIvor JS, Ruttan A, Salehi B (2014) Exotics on exotics: Pollen anal- Stadt Wien Vienna GIS (2005) Data: wiener umweltschutzabteilung – ysis of urban bees visiting Sedum on a green roof. Urban Ecosyst 18: MA 22. Basic data: mehrzweckkarte, stadtvermessung wien – MA 419–430. https://doi.org/10.1007/s11252-014-0408-6 41. Further Information: Themenplan Wien Umweltgut. Vienna: Matteson KC, Langellotto GA (2010) Determinates of inner city butterfly Stadt Wien and bee species richness. Urban Ecosyst 13:333–347. https://doi. Susca T, Gaffin SR, Dell’Osso GR (2011) Positive effects of vegetation: org/10.1007/s11252-010-0122-y Urban heat island and green roofs. Environ Pollut 159:2119–2126. Matteson KC, Ascher JS, Langellotto GA (2008) Bee Richness and https://doi.org/10.1016/j.envpol.2011.03.007 Abundance in New York City Urban Gardens Bee Richness and Tonietto R, Fant J, Ascher J et al (2011) A comparison of bee communi- Abundance in New York City Urban Gardens. Ann Entomol Soc ties of Chicago green roofs, parks and prairies. Landsc Urban Plan Am 101:140–150 103:102–108. https://doi.org/10.1016/j.landurbplan.2011.07.004 Mauss V (1994) Bestimmungsschlüssel für Hummeln. In: Deutscher ViennaGIS (2016) Vienna City Map. https://www.wien.gv.at/stadtplan/. Jugendbund für Naturbeobachtungen (ed), 5th edn. Hamburg, p 51 Accessed 5 Dec 2016 Mazerolle MJ (2016) AICcmodavg: model selection and multimodel in- von Hagen E, Aichhorn A (2003) Hummeln: bestimmen, ansiedeln, ference based on (Q)AIC(c). R package version 2.1-0. https://cran.r- vermehren, schützen, 5. überarb. Fauna Verlag, Nottuln project.org/package=AICcmodavg. Accessed 13 Feb 2017 Westrich P (1989a) Die Wildbienen Baden-Württembergs: Spezieller Mentens J, Raes D, Hermy M (2006) Green roofs as a tool for solving the Teil. Die Gattungen und Arten. Eugen Ulmer GmbH&Co., Stuttgart rainwater runoff problem in the urbanized 21st century ? Landsc Westrich P (1989b) Die Wildbienen Baden-Württembergs: Allgemeiner UrbanPlan77:217–226. https://doi.org/10.1016/j.landurbplan. Teil. Lebensräume, Verhalten, Ökologie und Schutz. Eugen Ulmer 2005.02.010 GmbH&Co., Stuttgart Michener CD (2007) The Bees of the World, 2nd edn. The Johns Hopkins Westrich P (1996) Habitat requirements of central European bees and the University Press, Baltimore problems of partial habitats. In: Matheson A, Buchmann SL, Motulsky H, Christopoulos A (2003) Fitting Models to Biological Data O’Toole C, Westrich P and Williams IH (eds) The conservation of using Linear and Nonlinear Regression: A practical guide to curve bees. Academic Press, London pp 1–16 fittin, 4th edn. GraphPad Software, Inc, San Diego Winfree R, Abuilar R, Vázquez DP et al (2009) A meta-analysis of bees’ Oberndorfer E, Lundholm J, Bass B et al (2007) Green Roofs as Urban response to anthropogenic disturbance. Ecology 90:2068–2076. Ecosystems: Ecological Structures, Functions, and Services. https://doi.org/10.1890/07-1861.1 Bioscience 57:823–833 Wojcik VA, McBride JR (2012) Common factors influence bee foraging Pachinger B (2008) Der Hohlweg am Johannesberg (Wien, Unterlaa) in urban and wildland landscapes. Urban Ecosyst 15:581–598. Lebensraum und Trittstein für Wildbienen (Hymenoptera:Apidae). https://doi.org/10.1007/s11252-011-0211-6 Beiträge zur Entomofaunistik 8:69–83 ZAMG - Zentralanstalt für Meterologie und Geodynamik (2015) Pachinger B, Hölzler G (2006) Die Wildbienen (Hymenoptera, Apidae) Klimaspiegel Wien Hohe Warte für 2014. http://www.zamg.ac.at/ der Wiener Donauinsel. Beiträge zur Entomofaunistik 7:119–148 cms/de/klima/klima-aktuell/klimaspiegel/jahr/wien_hohe_warte/? Potts SG, Vulliamy B, Dafni A et al (2003) Linking bees and flowers: jahr=2014. Accessed 5 Dec 2016 How do floral communities structure pollinator communities? Zettel H, Zimmermann D, Wiesbauer H (2013) Die Bienen und Ecology 84:2628–2642. https://doi.org/10.1890/02-0136 Grabwespen (Hymenoptera: Apoidea) im Donaupark in Wien Potts SG, Biesmeijer JC, Kremen C et al (2010) Global pollinator de- (Österreich). Sabulosi 3:1–23 clines: trends, impacts and drivers. Trends Ecol Evol 25:345–353. Zettel H, Zimmermann D, Wiesbauer H (2016) Ergänzungen zur https://doi.org/10.1016/j.tree.2010.01.007 Bienenfauna (Hymenoptera: Apidae) von Wien, Österreich. R Core Development Team (2016) R: A Language and Environment for Beiträge zur Entomofaunistik 17:85–107 Statistical Computing. V 3.3.2. R Foundationi for Statistical Zinzi M, Agnoli S (2012) Cool and green roofs. An energy and comfort Computing, Vienna comparison between passive cooling and mitigation urban heat is- RStudio Team (2015) RStudio: Integrated Development for R. V land techniques for residential buildings in the Mediterranean re- 0.99.903. RStudio Inc., Boston gion. Energy Build 55:66–76. https://doi.org/10.1016/j.enbuild. Scheuchl E (2000) Illustrierte Bestimmungstabellen der Widbienen 2011.09.024 Deutschlands und Österreichs Band 1: Schlüssel der Gattungen Zurbuchen A, Müller A (2012) Wildbienenschutz - von der Wissenschaft und der Arten der Familie Anthophoridae. Eigenverlag, Velden/Vils zur Praxis. Bristol-Stifung, Bern Scheuchl E (2006) Illustrierte Bestimmungstabellen der Wildbienen Zurbuchen A, Landert L, Klaiber J et al (2010) Maximum foraging ranges Deutschlands und Österreichs Band 2: Megachilidae - Melittidae, in solitary bees: only few individuals have the capability to cover 2nd edn. Eigenverlag, Velden/Vils long foraging distances. Biol Conserv 143:669–676. https://doi.org/ Scheuchl E, Willner W (2016) Taschenlexikon der Wildbienen 10.1016/j.biocon.2009.12.003 Mitteleuropas: Alle Arten im Porträt. Quelle & Meyer Verlag, Zuur AF, Ieno EN, Elphick CS (2010) A protocol for data exploration to Wiebelsheim avoid common statistical problems. Methods Ecol Evol 1:3–14. Schindler M, Diestelhorst O, Härtel S et al (2013) Monitoring agricultural https://doi.org/10.1111/j.2041-210X.2009.00001.x ecosystems by using wild bees as environmental indicators. BioRisk 71:53–71. https://doi.org/10.3897/biorisk.8.3600 Zuur AF, Hilbe JM, Ieno EN (2013) A Beginner’s Guide to GLM and Schmid-Egger C, Scheuchl E (1997) Illustrierte Bestimmungstabellen der GLMM with R: A frequentist and Bayesian perspective for ecolo- Wildbienen Deutschlands und Österreichs unter Berücksichtigung gists. Highland Statistics Ltd., Newburgh

Journal

Urban EcosystemsSpringer Journals

Published: Feb 20, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off