Although the buffering and performance-enhancing mechanisms underpinning the insurance effects of biodiversity [4, 5] are generally accepted as being equally plausible, experimental evidence for the latter is scarce. Here we performed an experiment to test how bacterial biodiversity-respiration relationships were influenced by the amplitude and frequency of a temperature change. In congruence with our hypotheses, we found that environmental variability strengthened the effect of species richness on ecosystem functioning and therefore provided clear evidence of a performing-enhancing role of biodiversity under fluctuating conditions.
In general, complementarity and selection effects can explain positive effects of species richness on ecosystem functioning [29, 30]. Previous studies that have found an increase in ecosystem functioning under fluctuating environmental conditions due to niche complementarity and positive interactions [15, 16], indicating several mechanisms might simultaneously operate and be of importance. Our results are, on the contrary, more consistent with the findings of Steiner et al.  who could show that species richness increased resilience of total community biomass after a perturbation due to rapid growth of a few dominant species. In order to be able to clearly separate selection and complementarity effects, information on the performance of individual species within a mixed community is required, which we lack here [29, 30]. Nevertheless, our study does suggest ecosystem functioning increases under varying environmental regimes due to stronger selection effects, i.e. the increased relative performance of individual species. There was a marked increase in levels of ecosystem functioning with increasing amplitude of temperature fluctuation, associated with combinations involving strain A (Figure 5). Combinations without strain A had, on the contrary, similar or lower levels of ecosystem functioning under fluctuating compared to constant temperature conditions. Moreover, the activities of strain A in monoculture increased both with increasing temperatures (Additional file 2) as well as with increasing amplitudes of temperature fluctuation (Figure 6). On the other hand, there was a clear lack of effect, or decline, in activities of monocultures of all the other strains (Figure 6) as well as of any combinations that did not contain strain A (Figure 5). Positive Dmax values indicating overyielding and thus complementarity effects were generally sparse (Additional file 3). If they were found, it was primarily in treatments with the highest amplitudes of temperature fluctuation and primarily in communities without strain A, indicating that complementarity effects might be more important in communities that are not strongly dominated by a particular species. It is, however, important to note that the system in general had a high level of functional redundancy, with all species able to utilize glucose, which might have limited the potential for complementarity effects, such as niche differentiation and facilitation. Nevertheless, our findings suggest that selection effects were more important than complementarity effects and that environmental change may lead to a situation where (i) particular species, in this case strain A (Figures 5 and 6), are better suited to varying environments, or (ii) some species, in this case strains B, C and D (Figures 5 and 6), are negatively affected, resulting in a competitive release from species which are otherwise functionally equivalent (redundant). Both outcomes result in an increasing dominance of the best performing species and, hence, an increase in ecosystem functioning. This also implies that performance-enhancing effects of biodiversity are likely to be influenced by changes in species interactions imposed by environmental circumstance, not functional capacity alone. Consequently, the level of ecosystem functioning achieved will largely depend on the contribution of a particular subset of species and how they interact to specific biodiversity-environment futures , emphasizing the need to consider species-specific responses to particular components (alone and in combination) of environmental change [33, 34].
The context dependency of species interactions may, at least in part, also explain why negative or no effects of fluctuating environmental conditions on ecosystem functioning are often reported (e.g. [10, 17, 18, 35, 36]); thus, while there are environmental combinations that select one species over another, as in our study, there may also be conditions that exert neutral or negative selective pressures on other species. The fact that few congruent results have been documented to date may reflect the use of simplified model communities, where the selection of species with particularly traits may bias the effects that are observed. We therefore speculate that the performance enhancing effect may be limited to situations where environmental change and disturbances lead to the increasing dominance of generalists. Since natural ecosystems are increasingly dominated by generalists , our results might point towards an important mechanism that has hitherto not received much attention. In general, emphasis needs to be placed on investigation of the effects of the environmental context and perturbations on natural communities, since this knowledge will be essential in attempts to predict the effects of future environmental conditions on the delivery and magnitude of ecosystem services (e.g. [34, 38–40]).
It is important to consider how communities respond to changes in the frequency of environmental fluctuation as well as the magnitude of change. Disturbance frequency has been found to affect diversity alone  and in combination with intensity [42, 43], and therefore has the potential to affect functional properties of ecosystems indirectly. Our results showed that the intensity of a disturbance, i.e. the amplitude of the temperature change, was more important in determining respiratory activity than the frequency (= rate) of that change. This was probably because species richness effects were buffered by varying responses of individual species to changes in temperature frequency, i.e. strains were either slightly positively or negatively affected by higher disturbance frequency, depending on the magnitude of the temperature change. The fact that we observed interactive effects between species composition and different properties of the environmental regimes imposed here, supports the idea that insurance effects are not necessarily mediated by bulk properties of environmental change, but by multiple subtle properties of environmental forcing that are not necessarily immediately obvious . However, since the change of the temperature amplitude imposed here does not necessarily match temperature profiles typically observed in nature, future studies should investigate whether similar effects and underlying mechanisms exist and are of generic importance in naturally assembled communities.
It is important to consider the methodological limitations of our study within the context of ecological theory. Theory predicts that the performance enhancing effect of biodiversity will lead to an increase of the mean of an ecosystem function over time , yet our study considered a point measurement of ecosystem functioning at the end of the experiment rather than monitored temporal changes in ecosystem functioning throughout the duration of the experiment. Thus, whilst our design did not include multiple time points required to demonstrate performance enhancing effects as a result of negative covariances among species, our results do incorporate multiple generations and are consistent with what would be predicted when such effects are present.
Bacteria have very short generation times and can rapidly adjust their physiology in response to changes in environmental conditions. This also means, however, that it was necessary to set the initial bacterial abundance close to that of the carrying capacity of the different bacterial strains to avoid a batch culture situation; an initial period of rapid growth of the most competitive strain would have lead to strong changes in evenness. By using relatively high initial abundances we ensured that the shift in dominance that we observed resulted from more realistic changes in the densities of species (e.g. due to release of nutrients when cells were increasingly inhibited at higher temperature fluctuations) or were due to differences in the rates at which species contribute to ecosystem functioning. One limitation of our study is that we did not measure the realized diversity in the community at the end of the experiment, hence, strictly speaking our results refer to relationships between initial diversity and respiratory activities. However, it is unlikely that our findings reflect changes in diversity caused by expirations of individual strains given the short incubations time (42 hours), high initial abundances of all components strains, and that all strains showed activity across all treatments, including those with the highest level of temperature fluctuations (Figure 6). Another limitation is the low diversity compared to natural bacterial communities that greatly exceeds the richness levels that are possible to include in experiments with tractable model communities . Thus, while we have used a model system to identify the plausibility of a specific mechanism , an important next step will be to test whether performance enhancing effects of diversity also operate in natural bacterial communities exposed to environmental change and perturbations.
Increasing temperature can decrease species richness of communities , induce changes in community composition and facilitate temporal species turnover  and decrease community evenness , but it is clear that there are multiple factors that, in concert with biodiversity, will determine how climate change affects the provision and stability of ecosystem functioning [34, 38, 50]. Our study also confirms that subtleties in individual factors, in particular the amplitude and frequency of the change, may affect community structure and, subsequently, ecosystem functioning . This may result in an increasing dominance of particular key species that facilitate ecosystem functioning under certain alternative environmental conditions. Collectively, our findings emphasise the need to consider how species, alone and in combination, respond to and interact with, multiple properties (variance, extremes, cycles etc.) of a changing environment in order to reduce the uncertainty associated with predicting the functional consequences of biodiversity-environmental futures.