Abstract. The relationship between omnivory and stability has been the subject of a longstanding debate in ecology. Early theory predicted that omnivory would decrease the probability of food webs being stable. While early empirical data appeared to support the prediction that omnivory should be rare, detailed study of food webs later revealed that omnivory is ubiquitous across ecosystems and taxa. Recent years have seen renewed interest in the omnivory-stability debate, and advances in mechanistic non-equilibrium models demonstrated that omnivory can both increase and decrease stability. Current efforts have therefore focused on identifying biological mechanisms that promote the persistence of food webs with omnivory. We synthesize recent evidence that omnivory often stabilizes food webs when it occurs as life-history omnivory, when prey experience reduced predation rates due to refuges or adaptive antipredator defences, and when omnivores interfere with each other or feed adaptively. Empirical research has lagged behind theory and there remains a shortage of studies directly measuring the stability of diverse natural communities that vary in the number and strength of omnivorous interactions. Early microcosm experiments indicated a narrow range of conditions for the persistence of simple omnivorous modules, while studies of omnivory embedded within larger natural networks have demonstrated its stabilizing effects. These new findings alter our view of food web dynamics and show that rather than looking for a simple and general omnivory-stability relationship, we should focus on identifying conditions under which omnivory is a stabilizing feature of more complex natural systems.

The effects of habitat connectivity on food webs have been studied both empirically and theoretically, yet the question of whether empirical results support theoretical predictions for any food web metric other than species richness has received little attention. Our synthesis brings together theory and empirical evidence for how habitat connectivity affects both food web stability and complexity. Food web stability is often predicted to be greatest at intermediate levels of connectivity, representing a compromise between the stabilizing effects of dispersal via rescue effects and prey switching, and the destabilizing effects of dispersal via regional synchronization of population dynamics. Empirical studies of food web stability generally support both this pattern and underlying mechanisms. Food chain length has been predicted to have both increasing and unimodal relationships with connectivity as a result of predators being constrained by the patch occupancy of their prey. Although both patterns have been documented empirically, the underlying mechanisms may differ from those predicted by models. In terms of other measures of food web complexity, habitat connectivity has been empirically found to generally increase link density but either reduce or have no effect on connectance, whereas a unimodal relationship is expected. In general, there is growing concordance between empirical patterns and theoretical predictions for some effects of habitat connectivity on food webs, but many predictions remain to be tested over a full connectivity gradient, and empirical metrics of complexity are rarely modeled. Closing these gaps will allow a deeper understanding of how natural and anthropogenic changes in connectivity can affect real food webs.

Habitat size is known to affect community structure and ecosystem function, but few studies have examined the underlying mechanisms over sufficient size gradients or in enough geographic contexts to determine their generality. Our goal in this study was to determine if the relationship between habitat size and leaf decomposition varied across geographic sites, and which factors may be driving the differences. We conducted replicated observations in a coastal forest in Brazil, and in rainforests in Costa Rica and Puerto Rico. We used leaf litter decomposition and macroinvertebrate composition in bromeliad phytotelmata of varying sizes to determine the relationships between habitat size, trophic structure and decomposition over a wide geographical range. We experimentally disentangled the effects of site and litter quality by quantifying invertebrate control of decomposition of a native and a transplanted litter type within one site. We found that the relationship between bromeliad size and decomposition rates differed among study sites. In rainforests in Costa Rica and Puerto Rico, decomposition was strongly linked to macroinvertebrate trophic structure, which varies with bromeliad size, driving strong bromeliad size‐decomposition relationships. However, in Brazil there was no relationship between bromeliad size and decomposition. Our manipulative experiment suggests that within coastal forest in Brazil, the poor quality of native litter resulted in little invertebrate control of decomposition. Furthermore, the key detritivore in this site builds a predator‐resistant case, which likely prevented effects of bromeliad size on trophic structure from being transmitted to decomposition even when litter quality was increased. We conclude that differences in both leaf litter quality and macroinvertebrate traits among sites determine the link between decomposition and macroinvertebrates, and consequently the decomposition‐bromeliad size relationship. These results show that the response of decomposition to habitat size is context‐dependent, and depends on which component of the food web is the main driver of the function.

Ecosystems are often arranged in naturally patchy landscapes with habitat patches linked by dispersal of species in a metacommunity. The size of a metacommunity, or number of patches, is predicted to influence community dynamics and therefore the structure and function of local communities. However, such predictions have yet to be experimentally tested using full food webs in natural metacommunities. We used the natural mesocosm system of aquatic macroinvertebrates in bromeliad phytotelmata to test the effect of the number of patches in a metacommunity on species richness, abundance, and community composition. We created metacommunities of varying size using fine mesh cages to enclose a gradient from a single bromeliad up to the full forest. We found that species richness, abundance, and biomass increased from enclosed metacommunities to the full forest size and that diversity and evenness also increased in larger enclosures. Community composition was affected by metacommunity size across the full gradient, with a more even detritivore community in larger metacommunities, and taxonomic groups such as mosquitoes going locally extinct in smaller metacommunities. We were able to divide the effects of metacommunity size into aquatic and terrestrial habitat components and found that the importance of each varied by species; those with simple life cycles were only affected by local aquatic habitat whereas insects with complex life cycles were also affected by the amount of terrestrial matrix. This differential survival of obligate and non‐obligate dispersers allowed us to partition the beta‐diversity between metacommunities among functional groups. Our study is one of the first tests of metacommunity size in a natural metacommunity landscape and shows that both diversity and community composition are significantly affected by metacommunity size.

Summary 1. Our objective was to investigate the associations between benthic macro‐invertebrate communities and environmental factors described at three spatial scales: local, reach and catchment. Differences in these associations, because of local topography, were determined by categorising sites into those having a large or small ‘reach contributing area’ (RCA), which is the lateral area of land contributing surface and subsurface flow of water, nutrients, organic and inorganic materials to a stream reach independent of catchment size and upstream contributions, and created by local topography. 2. Twenty‐three sites were sampled in the Lake Nipigon Forest in north‐western Ontario. Local variables included hydrological metrics and stream morphology characteristics. Terrestrial reach and catchment variables included forest characteristics, topography and geomorphology. Invertebrates were identified and classified into functional feeding guilds. Canonical correspondence analysis and redundancy analysis were used to determine which variables contributed most to macro‐invertebrate community structure and how the RCA influenced those associations. 3. Overall, local‐scale variables explained most variation in taxonomic and functional community structure (51.6 and 59.1%). Catchment‐scale variables explained more variation in the community (43.9 and 43.5%) than reach‐scale variables. In sites with a large RCA, the reach‐scale variables accounted for almost 10% more variability in taxonomic structure than the catchment‐scale variables, and in sites with a small RCA, catchment‐scale variables explained almost 23% more variation in feeding guild structure than reach‐scale variables. 4. Topography was important at both the reach and the catchment scales. Variables at the reach scale most associated with invertebrate community tended to be related to forest cover, whereas important variables at the catchment scale were related to topography, wetlands and lakes. 5. The difference in associations between invertebrate communities and environmental factors based on RCA indicates the influence of local topography on the linkage between aquatic habitats and the surrounding terrestrial environment. Understanding how these linkages affect aquatic communities can help to develop more effective predictive models of invertebrate community and more prescriptive conservation and management strategies for small streams.

Ecosystem functions and the biomass of lower trophic levels are frequently controlled by predators. The strength of top‐down control in these trophic cascades can be affected by the identity and diversity of predators, prey, and resources, as well as environmental conditions such as temperature, moisture, and nutrient loading, which can all impact interaction strength between trophic levels. Few studies have been able to replicate a complete community over a large geographic area to compare the full trophic cascade in a manipulative experiment. Here, we identify geographic dependency in trophic cascade strength, and the driving factors and specific mechanisms behind it, by combining geographically replicated experiments with a novel approach of community analogues of common garden and transplant experiments. We studied a predator–detritivore–detritus food web in bromeliads in Puerto Rico, Costa Rica, and Brazil. We found that interaction strengths between resources, consumers, and predators were strongly site‐specific, but the exact mechanism differed between trophic levels. Large bodied predators created strong interaction strengths between predator and consumer trophic levels, reducing consumer abundance regardless of the geographic location, whereas small‐bodied predators created weak interactions with no impact on consumer abundances in any site. In contrast, the interaction strength between consumers and resources varied among sites, depending on the dominant species of leaf detritus. More labile leaf species in Costa Rica created a strong consumer–resource interaction and therefore strong trophic cascade, whereas tougher leaf species in Brazil created a weak consumer–resource interaction, and an overall weaker trophic cascade. Our study highlights the importance of replicating experiments over geographic scales to understand general patterns of ecological processes.