Most ecosystems are affected by anthropogenic or natural pulse disturbances, which alter the community composition and functioning for a limited period of time. Whether and how quickly communities recover from such pulses is central to our understanding of biodiversity dynamics and ecosystem organization, but also to nature conservation and management. Here, we present a meta-analysis of 508 (semi-)natural field experiments globally distributed across marine, terrestrial and freshwater ecosystems. We found recovery to be significant yet incomplete. At the end of the experiments, disturbed treatments resembled controls again when considering abundance (94%), biomass (82%), and univariate diversity measures (88%). Most disturbed treatments did not further depart from control after the pulse, indicating that few studies showed novel trajectories induced by the pulse. Only multivariate community composition on average showed little recovery: disturbed species composition remained dissimilar to the control throughout the experiment. Still, when experiments showed a higher compositional stability, they tended to also show higher functional stability. Recovery was more complete when systems had high resistance, whereas resilience and resistance were negatively correlated. The overall results were highly consistent across studies, but significant differences between ecosystems and organism groups appeared. Future research on disturbances should aim to understand these differences, but also fill obvious gaps in the empirical assessments for regions (especially the tropics), ecosystems and organisms. In summary, we provide general evidence that (semi-)natural communities can recover from pulse disturbances, but compositional aspects are more vulnerable to long-lasting effects of pulse disturbance than the emergent functions associated to them.

This data set contains all effect sizes between a control and a disturbed treatment for 508 experiments used to analyze recovery across ecosystems and organisms. A full description of the methods is in the paper associated to this manuscript. The data contains 6 columns.

case.ID is a unique identifier for each of the 508 experiments. Appendix S1 at the paper contains full documentation for each of the experiments.

resp.cat categorizes the response variable as either biomass, abundance, an index of diversity or composition. The manuscript details the rationale behind these categries, and the detailed variables measures are given in Appendix S1.

Day of the experiment, with 0 being the pre-distrubance day.

RD: Relative duration within the experiment, i.e., normalizing the day of the experiment to the duration of the experiment, such that 1 is the final sampling date.

LRR is the effect size of the disturbance, in most cases a log-response ratio, only for composition mostly a dissimilarity metric. Again, for detauls see the paper's method section

var.lrr is the sampling variance of the effect sizes, if blank it could not be derived from the primary literature.