Changing bee and hoverfly pollinator assemblages along an urban-rural gradient

Abstract

Background: The potential for reduced pollination ecosystem service due to global declines of bees and other pollinators is cause for considerable concern. Habitat degradation, destruction and fragmentation due to agricultural intensification have historically been the main causes of this pollinator decline. However, despite increasing and accelerating levels of global urbanization, very little research has investigated the effects of urbanization on pollinator assemblages. We assessed changes in the diversity, abundance and species composition of bee and hoverfly pollinator assemblages in urban, suburban, and rural sites across a UK city.

Methodology/principal findings: Bees and hoverflies were trapped and netted at 24 sites of similar habitat character (churchyards and cemeteries) that varied in position along a gradient of urbanization. Local habitat quality (altitude, shelter from wind, diversity and abundance of flowers), and the broader-scale degree of urbanization (e.g. percentage of built landscape and gardens within 100 m, 250 m, 500 m, 1 km, and 2.5 km of the site) were assessed for each study site. The diversity and abundance of pollinators were both significantly negatively associated with higher levels of urbanization. Assemblage composition changed along the urbanization gradient with some species positively associated with urban and suburban land-use, but more species negatively so. Pollinator assemblages were positively affected by good site habitat quality, in particular the availability of flowering plants.

Conclusions/significance: Our results show that urban areas can support diverse pollinator assemblages, but that this capacity is strongly affected by local habitat quality. Nonetheless, in both urban and suburban areas of the city the assemblages had fewer individuals and lower diversity than similar rural habitats. The unique development histories of different urban areas, and the difficulty of assessing mobile pollinator assemblages in just part of their range, mean that complementary studies in different cities and urban habitats are required to discover if these findings are more widely applicable.

Figure 1. Relationships between site treatment and total abundance of bees, hoverflies and total pollinators.

Error bars  = +/− 1SE. Bars that do not share a letter showed significant differences (P<0.05) between treatments. Ns  =  no significant difference.

Figure 2. Relationships between site treatment and total species richness of bees, hoverflies, and total pollinators.

Error bars  =  +/− 1SE. Bars that do not share a letter showed significant differences (P<0.05) between treatments.

Figure 3. Co-plot showing relationships between explanatory variables used in the generalized linear models.

Pairwise scatterplots are shown with locally weighted loess smoothing to aid visual interpretation and panels with a Pearson correlation >0.3 highlighted. Variable codes are abbreviated for clarity (exp.  = % exposure, alt.  =  altitude in metres, flowR  =  flowering forb species richness, flowA  =  flowering forb flower abundance, tree  =  number of flowering trees, % gard.  = % gardens within 500m, and% built  = % built space within 500m). Percentage gardens and built space at 500m scale was used as an example because this was most representative of these variables at all wider landscape scales.

Figure 4. Redundancy analysis (RDA) showing species abundance responses to Built 2500m and Forb flower abundance.

The position of bee and hoverfly species in the ordination space are shown with filled circles and triangles respectively. For clarity, only species with the best model fit are illustrated.

Figure 5. Output from the ordination generalized linear models for six pollinator species.

Contours on the plots refer to the predicted abundance values of each species. Vectors are related to the significant explanatory variables.

Figure 6. Species incidences of all species at rural, and urban and suburban combined.

Data for urban and suburban sites were combined for clarity as they were very similar. Certain species are highlighted, but patterns for all species can be found using the species number in Table S1.

Figure 7. Distribution of the urban, suburban and rural survey sites.

Red dots mark the position of each survey site: nine sites in 1km squares classified by Owen et al. (2006) as ‘urban’ or ‘urban transport’ (pink transparency), nine sites in 1km squares classified as ‘suburban’ (blue transparency), and nine rural sites in villages outside Birmingham City limits (black line). Green space is shown as pale green, woodland as dark green, and large still water bodies as blue. Data Crown Copyright/database right 2008 and 2010, an Ordnance Survey/EDINA supplied service.