Reporting from 30 ecosystem functions collected over 82 large plots in a grassland biodiversity experiment, we tested how far the effect of plant species richness on dynamical stability translates when functionality is characterized at the population, the community, or the ecosystem level of organization.

We predicted that natural environmental perturbations should dampen up the hierarchy from the population to the ecosystem level, leading to both a reversal (from destabilizing to stabilizing) and a weakening of the biodiversity effect on dynamical stability at higher organizational levels. A lower coefficient of variation in ecosystem functions is stabilizing since it implies a dampening of environmental perturbations leading to reduced variances, or that multiple facets of ecosystem functionality complement in time leading to reduced co-variances.

Results/Conclusions The effect biodiversity on ecosystem stability is weakened as we integrate processes from the population to the ecosystem, but a net stabilizing effect on the variance of ecosystem functions remains at the higher level of organization. The effect of plant species richness on the co-variances of ecosystem functions was destabilizing at the population level, stabilizing at the community level, and neutral at the ecosystem level of organization. Plant species richness however showed a significant stabilizing effect on the variances of ecosystem functions at all organizational levels.

The present synthesis of empirical diversity-stability relationships supports an intermediate view between the endogenous population-level interactions and the exogenous influence of environmental perturbations, where both types of processes are constantly played. Since nowadays nobody would cast doubts to the role environmental perturbations play on ecosystem functionality at the global scale, theoretical and empirical ecologists need to coordinate for providing stakeholders a clear answer to the role of biodiversity in stabilizing ecosystem functions at the local scale.