Substantial evidence suggests that biodiversity can stabilize ecosystem function, but how it does this is less clear. In very general terms, the hypothesis is that biodiversity stabilizes function because having more species increases the role of compensatory dynamics, which occur when species in a community show different responses to the environment. Here, we focus on two forms of compensatory dynamics, cross-scale resilience (CSR) and response diversity (RD). CSR occurs when species respond to a disturbance at different scales, such that scale-specific disturbances do not negatively affect all species. RD occurs when species contributing to the same function show different responses to an environmental change. We developed a new analytical approach that can compare the strength of CSR and RD in the same dataset, and used it to study native bee pollination of blueberry at 16 farms that varied in surrounding agricultural land use. We then asked whether CSR and RD among bee species are associated with the stability of blueberry pollination. Although CSR and RD were both present, only RD was associated with higher stability of pollination. Furthermore, the effects of RD on stability were due to a single widespread species, Andrena bradleyi, that is a specialist on blueberry and unlike other bee species, was highly abundant at farms surrounded by intensive blueberry agriculture. Thus, the stabilizing effect we observed was attributable to an 'identity effect' more than to species richness per se. Our results demonstrate how CSR and RD can be empirically measured and compared and highlight how the theoretical expectations of the biodiversity-ecosystem functioning field are not always upheld when confronted with real-world data.

We collected data from 16 commercial highbush blueberry farms in southern New Jersey from 2011-2012. At each farm, we assessed the percent of surrounding land cover in agriculture using spatial data from the New Jersey Department of Environmental Protection Geographic Information System (http://www.state.nj.us/dep/gis). Farm sites were chosen so that percent of agriculture surrounding the farm at small (300 m radius) and large (1500 m radius) spatial scales were not correlated.

Bees were hand-netted from blueberry flowers at each farm in 2011 and 2012. Collections took place during peak bloom (mid-April through mid-May in our study region), with netting taking place only on relatively warm sunny days with little wind. Within each year, we collected data at each farm on three different days; on each day, we collected bees twice, once in the morning and once in the afternoon.