Realistic species losses have greater effects than randomized biodiversity treatments on ecosystem functioning: a meta-analysis

Background/Question/Methods

Species losses from communities do not occur at random.  A growing body of literature has emerged in the last decade to explore the effects of realistic species losses on ecosystem processes and to compare it to the effects of the randomized biodiversity treatments most commonly used in biodiversity-ecosystem functioning studies.

We used keyword and forward and backward citation approaches to exhaustively search the peer-reviewed journal literature for experimental and empirically based modeling studies of the effects of realistic species losses on one or more ecosystem functions.  To meet the criterion of “realistic,” the compositions of different species richness levels had to be based on a specific identified driver of loss, an observed richness gradient, or another criterion (such as rarity) that in our judgment reflects a plausible result of one or more specific drivers.  We (1) calculated effect sizes by study based on the ratio of functioning between the highest and lowest richness levels, and (2) fit curves to the data across richness levels for each study. For studies that included both realistic and randomized species loss treatments, we computed the same metrics for both sets of treatments.

Results/Conclusions

We identified 96 cases from 28 studies examining the effects of realistic species losses, representing four tropical and eight temperate nations; plants, invertebrates, fishes, mammals, birds and fungi; and marine, freshwater, forest, grassland, estuarine, and montane ecosystems.  Biodiversity-loss drivers addressed included direct harvest, pollution, exotics, fragmentation and climate change, although many studies used proxies of vulnerability such as rarity rather than directly observed species loss orders.  Realistic species losses had much larger effects on invasion (N=14) and disease resistance (N=10) than on productivity (N=28) or resource use (N=24), and had greater per-species effects in lower-diversity systems (grand mean 26% change in functioning/species, range 0 to >700%).

We found 25 cases from 9 studies directly comparing realistic and randomized species loss effects on one or more functions, spanning nearly the same range of taxonomies, regions, ecosystems and drivers as the larger set.  In direct comparisons, realistic species losses had higher per-species effects on ecosystem functioning than did randomized losses (mean pairwise effect size difference 0.066+/-0.019).  Our findings confirm that across a wide range of taxa, systems, and regions, the ecological consequences of realistic species losses exceed those of the randomized biodiversity treatments that dominate the literature.