Plant biodiversity correlates with productivity at large scales. However, global evidence from natural systems at local scales often contradicts those findings despite large number of theories and experiments supporting such correlations.
A common limitation in the explanatory models is the assumption that all plant species in a community follow a similar set of assembly rules, and therefore the community should be analyzed as a unit. However, recent evidence indicates that in a community different subsets of species might actually follow different rules.
A second limitation to analyses linking biodiversity and productivity is the direction of causality: some models assume that productivity drives diversity, while for others diversity drives productivity. This problem is partially caused by a confusion of “potential productivity” driven by environmental constraints and species pool, and “realized productivity” (biomass) which is the result of the species’ interactions given the environmental constraints.
Using unmanipulated plots of a global grassland experiment (the Nutrient Network), global climatic descriptors, as well as phylogentic descriptors of species similarity and total evolutionary history, we quantified the correlation and causal links between environmental conditions, biodiversity and biomass with linear models and path analysis. We tested those links splitting the community in the most prevalent functional groups (legumes, forbs and graminoids).
Results/Conclusions
At a global scale and using a multiyear average per site, (1) the biomass of functional groups within a site was uncorrelated among groups (p-value>0.1). (2) This lack of correlation was underlain by functional groups responding to different environmental drivers. (3) After accounting for site-level environmental conditions, intra-group phylogenetic diversity increased biomass of legumes and forbs (p-values<0.05), but graminoid phylogenetic diversity decreased forb and graminoid biomass (p-values<0.05).
At the plot scale, we fitted a hierarchical linear model, to represent year zero biomass with site as a random intercept. Within plots, (4) biomass of graminoids and legumes (p-value<0.05) as well as of graminoids and forbs (p-value<0.001) were negatively correlated, but forbs and legumes were uncorrelated (p-value>0.1). Also (5) intra-functional group phylogenetic diversity increased forbs (p-value<0.05) and legumes biomass (p-value<0.05). Finally, (6) biomass residual values of forbs, legumes and graminoids were larger when graminoids biomass was around 44, 50 and 86% of the total biomass, respectively, indicating biomass overyielding (p-value<0.001).
These results show that species interact and respond to the environment in heterogeneous ways. Therefore, partitioning the community to account for those differences provides a strong way to solve historically long-standing debates in community ecology.