Introduction Habitat loss is the leading cause of extinctions on the planet. However, negative effects of habitat loss and fragmentation on biodiversity can be reduced if resources in urban or semi-natural areas in the surrounding matrix can be used by wildlife. Methods We investigated the influence of floral and nesting resources in urban-and forest-associated oaksavannah fragments, surrounding urban and forest matrix, and urban areas spatially independent from oak-savannah habitat on pollinator community composition in a fragmented oak-savannah ecosystem. Results Both independent urban and urban matrix sites supported high abundance and richness of plants and pollinators relative to other fragment categories, especially towards the end of the season when plants and pollinators in oak-savannah fragments were scarce. A species of particular conservation concern in our region, Bombus occidentalis, was supported by late-flowering resources in our urban sites. Forestassociated oak-savannah fragments were missing lateseason species while urban-associated fragments supported high abundance and richness of mid-to lateseason pollinators, likely due to supplemental use of floral resources in the urban matrix. Female cavitynesting and ground-nesting bees were not restricted by the availability of natural nesting resources we expected them to require (e.g. small cavities, bare soil). Conclusion These results provide important information on native pollinators in a highly fragmented habitat, and suggest that we should consider matrix quality in conservation planning.

1. Fragmentation of natural habitats due to urban development is predicted to have negative impacts on species diversity. The surrounding landscape (or 'matrix') of urban or semi-natural habitats can sometimes support biodiversity, but the amount of support will depend on species-specific traits, and on the resources available in the fragment and the matrix.2. Using data on bees collected from 19 oak-savannah fragments, the question of whether bee communities differ when fragments are embedded in different landscapes (Douglas-fir forest vs. urban residential neighbourhoods) was investigated, and also whether these differences could be attributed to species-specific traits of bees (e.g. body size, specialization) and/or within-fragment floral resources.3. No differences were found in overall richness or abundance of bees, but there were distinct differences in plant and bee community composition between matrix types. Common wood-nesters and late-flying, small-bodied bees tended to be found in urban-associated fragments, which also had a lower availability of within-fragment floral resources.4. Forest-associated fragments, on the other hand, had a greater density and richness of early-flowering native plant species, and supported a higher abundance of largebodied bee species. Bumble bee abundance, in particular, increased with increasing proportion of forest cover in the surrounding landscape.5. Large-bodied bees appear to respond to increased availability of within-fragment floral resources, but it was also hypothesised that nesting and floral resources in matrix habitat drive the differences in bee community assemblages.

No abstract