Assemblages of black flies in flowing waters of the southeastern United States are dynamic, changing with season across a spatial patchwork of environmental factors. Distributions of the component species are predictable over multiple scales from an in-stream substrate
[22, 46], across streams, to ecoregion, even in the absence of physical barriers to dispersal
. The factors responsible for distributions at smaller scales (within a stream reach) can be manipulated experimentally and, therefore, are reasonably well understood for their role in driving assemblage patterns. Our study addresses the more elusive drivers of species assemblages at larger scales—within and across ecoregions. Accordingly, we demonstrated that shifts in abiotic conditions across ecoregions were met with corresponding shifts in local species assemblages. Other studies also have found abiotic gradients in freshwater habitats across ecoregions, with accompanying differences in faunal characteristics
 used presence/absence data from the bird community of the Bismark Archipelago to infer a number of assemblage rules, starting a debate that has lasted more than 35 years. One sticking point has been whether null-model approaches can be used to infer causal mechanisms of species distributions over large scales. Nonetheless, the use of well-behaved statistical null models by a number of authors has lead to a deeper understanding of the processes structuring communities
[28, 40, 43, 50]. Inference is typically made by proposing expected outcomes, given that the mechanism under consideration has no influence on distribution, and comparing these outcomes to observational data. This approach is essential, as traditional manipulative experiments at spatial scales required to document regional patterns and processes are often not possible, or in some cases, even ethical
. Furthermore, experiments at the local level (e.g., stream reach) might not accurately reflect the nature of species patterns at the regional level (e.g., among streams)
We examined patterns of species co-occurrences at regional scales, with the idea that patterns of co-occurrence across habitats reflect both community structure and the mechanisms responsible for the structure. Null-model analyses reveal a strong pattern of negative co-occurrences for the spring and summer FULL data sets. In most communities, species co-occurrences are fewer than expected by chance, a major exception being some invertebrate assemblages
. These segregated patterns of co-occurrence usually result either from ecological checkerboards (e.g., competitively structured assemblages)
[15, 36, 43, 53] or habitat checkerboards (e.g., assemblages structured by abiotic variables)
[28, 39, 40]. For a given assemblage, the difficulty has been in determining the dominant causal mechanism
. To provide insights into potential causal mechanisms, we controlled for abiotic factors and then re-examined the patterns of co-occurrence. By dividing streams into two groups for which variation in their conditions was either low or high, we controlled, at least partially, for the influence of abiotic conditions on species distributions.
When we controlled for the influence of abiotic conditions (LOW), species during the spring remained strongly segregated, consistent with a community structured by competition. Predation also could produce negative patterns of co-occurrence
. However, given that we considered only a single guild (filter-feeding) of closely related species (79% in the genus Simulium), competition would seem a likely explanation. Considerable evidence shows competitive displacement between species of black flies and between black flies and other insects at local scales, with the limiting resource being suitable substrates for attachment and locations on those substrates for optimal food delivery
[55–57]. Although a variety of organisms feed on larval black flies
[58, 59], their influence on displacement is poorly understood. Black flies might be able to minimize the effects of predation by a preference for high stream currents
. Although predation cannot be dismissed, our study suggests that competition in local stream habitats can scale up to produce competitively driven co-occurrence patterns at regional scales. Recent evidence suggests that local competitive interactions among birds can scale up to influence broad-scale distributions
In contrast to the spring, the highly segregated community in the FULL and HIGH data sets shifted to an unsegregated community in the LOW data set during the summer, suggesting that stream conditions were responsible, consistent with a community structured largely by abotic factors
. Our data suggest that the causal mechanisms accounting, at least partly, for the patterns of species co-occurrences can shift seasonally from biotic to abiotic factors. Of particular interest with regard to biotic factors is that scaling-up from the local assemblage to the regional level is seasonally dependent. The finding that causal mechanisms of species co-occurrences can shift seasonally from biotic to abiotic drivers reinforces the view that a comprehensive understanding of community ecology requires appreciation of temporal, as well as spatial, scales
. The reasons behind seasonal shifts in causal mechanisms of the pattern of species distributions are not known. However, species composition of black flies in temperate regions changes through the year, with richness peaking in the spring and declining through the summer and fall
. As the number of species increases, competition would be expected to play a greater role in determining species distributions.
Mantel tests also show seasonal shifts between correlates of community structure. Given the results of our analyses, we argue that distances between sites (dispersal) are at least as important as local stream conditions during the spring. In contrast, during the summer, only local stream conditions are significantly associated with changes in species composition across sites. Dispersal abilities of the females, thus, might have a seasonal or climatic, as well as a taxonomic or phylogenetic, component. The females of most species of black flies typically disperse from their natal streams in search of hosts
. Species in the genus Simulium, especially those active in warmer months (e.g., summer) and hotter climates (e.g., Africa), generally disperse greater distances than spring-active species
. Thus, the importance of distance between sites could be less important in the summer than in the spring, which might account for the seasonal differences in the results of our Mantel tests.
Integrating results of null-model analyses and Mantel tests can be accomplished within a metacommunity framework. A metacommunity may be viewed as a group of communities (assemblages) of potentially interacting species interconnected by dispersal
. Given that black flies leave their natal waters in search of blood meals and can disperse many kilometers
, the idea of a metacommunity framework in which to conceptualize simuliid assemblages in the lotic environment is appealing. Although studies in metacommunity dynamics often seek to determine if particular communities are structured by factors that are either stochastic (e.g., neutral models) or deterministic (e.g., niche models), such approaches are not useful; numerous factors likely operate simultaneously in a particular community
. Given this caveat, our empirical results are consistent with two conceptual aspects of metacommunity dynamics—species sorting and mass effects.
Species sorting is a niche-based view of the community in which species occur in local habitats where abiotic and biotic conditions are favorable
. Mass effects consider not only local conditions but also how dispersal can affect local community dynamics
. A strong association of local communities with local stream conditions would indicate species sorting. A significant association between site location and community composition would suggest that dispersal among sites (mass effects) also can influence local communities.
Both species sorting and mass effects could be expected to operate partly through oviposition behavior
. Although rarely considered in detail, the idea is simple. A stream insect is often in a particular location because the adult female placed the egg in that location. Although larval black flies drift within a stream reach, no evidence has suggested that drift can influence the composition of local species assemblages
. In other words, a critical initial filter that determines the presence of species at a stream site is the oviposition decision of the female. Understanding local stream assemblages, therefore, involves a strong behavioral component. The idea that oviposition behavior is a key factor in stream assemblages dovetails with the idea of cross-ecosystem interactions
. For example, variation in stream conditions might have little influence on differences in local larval assemblages if a significant number of adult females failed to find appropriate terrestrial hosts. Thus, changes in bird or mammal abundances or distributions could influence local assemblages of aquatic insects see
We know little about the effects of oviposition preferences of black flies (or most other stream insects) on the structure of stream communities. What is apparent is that oviposition behaviors differ among species. Females of some species, for example, deposit their eggs on vegetation, whereas others drop them into the water
, and some species return to the natal waters to oviposit, whereas others do not
[66, 67]. Furthermore, the distribution of blood hosts might influence dispersal patterns and the distances traveled. We predict that generalist blood feeders have more general distributions than do host specialists. The occurrence of the immature stages of S. annulus at lake outflows in Sweden, for instance, has been related to the presence of loons (Gavia), preferred hosts of the female flies
An understanding of community assemblages of larval black flies has been building over the past 30 years. Patterns at local and regional scales, and their changes across seasons, are reasonably well known, with a good degree of predictability
. Causal factors underlying these patterns become more elusive as scale increases, although we have shown that null models and Mantel tests can provide insights.