It is not known how the number of species needed to maintain an ecosystem service changes from small to large spatial scales. At the smaller scale of biodiversity-ecosystem functioning experiments, function increases with species richness due to within-community mechanisms such as niche complementarity among species. In real-world landscapes, across-community mechanisms will also be important. Few species might be needed, because real-world communities are dominated by a few common and widespread species, which might provide most of the ecosystem services. Conversely, many species might be needed to maintain an ecosystem service across an entire landscape because spatial turnover in species composition leads to different species providing the same service in different places. We collected data on wild bee diversity and the pollination services each species provides to crop plants at 48 sites throughout a 3700 km2 landscape, and used these data together with null model analysis to answer the following questions: (1) How many bee species are needed to provide a threshold level of ecosystem services at one site, and then at each of many sites as spatial extent expands? (2) To what extent does dominance decrease the number of species required, relative to the effect of species turnover in increasing it?
Over half of the bee species observed in our study (55 of 108 species) were needed to meet the pollination threshold at all 48 sites. As spatial extent expanded from 1 to 48 sites, the number of bee species needed increased by a factor of four. The cumulative proportion of bee species required, which we had expected to decrease strongly as rare and functionally unimportant species accumulated faster than the common and functionally dominant ones, in fact did not show consistent directionality with scale. Functional dominance consistently decreased the number of bee species needed to meet the pollination threshold, as compared to a null model in which functional dominance was removed. However, dominance had much weaker effects than species turnover, which caused the number of species needed to increase as spatial extent expanded. Furthermore, the effect of dominance decreased with increasing spatial extent, whereas the effect of turnover increased. Our findings contrast with recent studies finding that a few dominant species provide most of the ecosystem services across large spatial scales. Possible reasons for this discrepancy will be discussed.