Abundance, not species richness, drives ecosystem service delivery at large spatial and temporal scales
The biodiversity-ecosystem functioning (BEF) relationship has been explored largely through small-scale experiments, leaving key questions unanswered at large spatio-temporal scales. Here we use large-scale datasets to explore the importance of wild pollinator species richness, composition and abundance in driving crop pollination services. We collected parallel data in four study systems: watermelon in the eastern USA (15 sites over 3 years), watermelon in the western USA (7 sites over 3 years), cranberry (16 sites over 2 years), and blueberry (16 sites over 3 years). We measured wild bee richness, composition and abundance by net-collecting bees from crop flowers, and pollen deposition per bee in experimental trials with un-pollinated flowers. We estimated pollination function as the product of bee individuals and pollen deposition per individual. We used a generalization of the Price equation from evolutionary biology to partition differences in pollination services over space and time into additive components attributable to each of the following terms: random loss/gain in pollinator species richness, non-randomness with respect to functional contribution of the species that are lost/gained (a species composition effect); and changes in pollinator abundance.
Across our four study systems we collected 8720 individual bees of 135 species, and conducted 1178 pollen deposition trials. We found that in three of four systems, and in analyses conducted over both space and time, the effect of species richness on pollination function was weak relative to that of abundance. Non-random changes in species composition mediated this effect: rare species that contributed relatively little for pollination function were the most likely to be lost and gained between sites and years. Our large-scale findings therefore contrast with the smaller-scale experimental BEF literature, which finds stronger effects of species richness. In our study, species-abundance distributions were highly skewed, with a small number of dominant species constituting the majority of individuals and contributing most of the pollination services. In contrast, experiments generally use more even species-abundance distributions. We conclude that the skewed species-abundance distributions typical of real-world ecological communities, which result in a long tail of rare species that contribute little to function, may cause the BEF relationship at large scales to diverge from the predictions of smaller-scale experiments.