Human societies depend on services provided by ecosystems, but these are threatened by biodiversity loss. Our perception of the importance of biodiversity to ecosystem services has been strongly influenced by >600 small-scale experiments. However, because many factors that are controlled by experiments are fundamentally different in un-manipulated natural communities (e.g. evenness) or covary with richness (e.g. composition, aggregate community abundance) the importance of richness to ecosystem services remains unknown. Here, we use a powerful new approach, based on the Price equation, to analyze the relationship between biodiversity and ecosystem services. The Price equation partitions changes in ecosystem services across sites into three different components of biodiversity: species richness; composition; and the ‘context dependent effect’ which includes the effects of abundance and per-capita function. We completed 60 Price equation-based analyses of ecosystem services provided by both animals and plants in locations across six continents. Specifically, we ask: (1) To what extent does species loss, as opposed to changes in abundance and per-capita function of individuals, affect changes in ecosystem services; (2) does evenness predict the importance of species loss; and (3) is non-random species loss the underlying mechanism for effects of evenness?
The importance of species loss and gain to ecosystem services varied dramatically across the 60 analyses. Specifically, we found that species loss and gain explained as little as 14% or as much as 93% of variation in ecosystem service delivery between sites. The remainder in each case was attributable to changes in abundance and per-capita function. We also found that evenness (measured in terms of contribution to ecosystem services, rather than abundance) was a strong positive predictor of the importance of species richness to changes in services between sites (p << 0.001, r2 = 0.59). This was due to predictable changes in species loss patterns in low- and high-evenness communities. At low evenness, species loss consistently targeted low-functioning species, but this pattern dissipated as evenness increased. This is the first synthetic analysis of ecosystem service provisioning in real-world systems that precisely defines the importance of species loss relative to changes in the abundance and per-capita function of persisting species. We found wide variation in the importance of species loss, suggesting that biodiversity research should focus on understanding the circumstances under which species loss is likely to be especially important to ecosystem services in the real world.