Discussion
Whilst we already knew from studies on particular subsets of communities that non-trophic indirect effects must be important for understanding community structure and dynamics this is thought to be the first truly empirical study to demonstrate, for an entire community, how numerous HOIs are acting at the same time. Here, we not only explored community-wide effects but experimentally tested each of the interaction modifications identified in situ in the field. By performing several manipulations concomitantly under natural field conditions, we were able to study the same interactions in different contexts and detect how multiple non-trophic interactions can interfere with or modify a single pairwise interaction. Thus, HOIs are not particular to certain sets of species but rather an integral part of communities. The manipulations revealed a hidden network of HOIs modifying the direct interactions, as well as modifying each other, a level of complexity unexplored in empirical studies and seldom mentioned in theoretical studies. Whilst our study system focused on an insect community on a tropical shrub, a hidden network of HOIs will occur in all ecological communities, and therefore our results are of huge ecological relevance. The removal of different groups indirectly affected the interactions between several other pairs of groups with which the manipulated groups did not directly interact. That indicated that the species are indirectly connected and pairwise interactions are context-dependent. We acknowledge that the three density effect links in the network may represent species association and not true links (it would be difficult to differentiate between the two), which would mean that the groups do not actually affect one another directly and may be both affected by a third group. However, we include these links, first, because it would be difficult to clarify whether that is the case and, second, because the fact that a node changes in the absence of other groups shows that they are somehow connected, therefore there is potential for indirect interaction, albeit via additional intermediate groups. Also, the HOIs are represented by straight arrows in the network, but in reality, multiple steps may be involved in their effect propagation pathways. What the arrow represents is that the pairwise interaction is affected by the third group. This level of detail on indirect non-trophic interactions, and in particular for HOIs, for such a species-rich multi-trophic system, is truly unsurpassed.
It seems almost impossible to determine how a species affects another when we consider that all those non-trophic indirect effects may be taking place at the same time and interfering with each other. Besides, the links represented in the network may vary in magnitude through time, or may even be transient . In the community studied, population densities at a given time are influenced by the sum of several indirect effects taking place at that time. If this is the case for natural systems in general, no inference can be made on the magnitude or direction of a given interaction in nature by studying it in isolation such as in an experimental setting or in computational models. Thus, depicting direct links between species in a network may be a misrepresentation of species effects or roles at the community level.
The existence of this hidden network suggests that, in natural conditions, it is not very likely that one species alone can determine the persistence of any other, such as in competitive exclusion. The role of competition in species coexistence has been widely demonstrated for focal species , but rarely in species-rich communities . Here we show that HOIs can hugely increase the context-dependency of pairwise interactions, and by modifying interactions and offsetting or complementing each other, can allow a flexible modulation of species coexistence. Thus, the hidden network of HOIs very likely plays a crucial role in diversity maintenance in multispecies communities.
It is imperative to devise methods to harmonize pairwise interaction networks with the hidden network of HOIs. This would allow us to recognise fundamental mechanisms involving non-trophic interaction modulation that, for instance, allow the community as a whole to respond to a specific manipulation such as the removal of a species. Advancing our understanding of such mechanisms is likely to elucidate the underlying causes of stability and persistence of ecological communities and increase our ability to predict how they might respond to perturbations. Computational simulations that account for HOIs are helpful, but the results presented here suggest that empirical data will be key to our understanding because of the numerous and often unpredictable opportunities for indirect effects via a variety of mechanisms and pathways.
Community ecology research rarely studies entire communities, in part due to the logistical problems involved in studying multispecies systems, but mostly due to the lack of tools to approach the complex intricacies of ecological communities. The results gathered here are unique, as the experimental exclusion of more than one group was performed concomitantly in the field in a self-contained but diverse system, and the densities of almost all groups of species were monitored. The challenges of replicating this study in a larger or less self-contained community are evident, but should not be a barrier to further empirical investigation into the generality of the findings presented here. We hope that this study will instigate new methodologies for more holistic approaches to studying ecological communities in general. The changes identified here could not be predicted from analyses of pairwise networks. Unlike pairwise interactions, HOIs cannot be predicted based on species functional traits or from previously recorded interactions (e.g., plant pollination, predation, parasitism). HOIs can be transient and involve effect propagation through several intermediaries. Thus, through HOIs, species have global effects on the community context. At the same time, these effects can vary in intensity and direction depending on the community context (species composition and abundances mainly) at each moment. This circular relationship between the structure of the global system and the local interactions among the components is typical of complex adaptive systems . Thus, developing new methods of applying complexity theory to ecosystems could be one way forward.