Additionally, relationships between traits and evolutionary history also result in mutual correlation among traits; thus, we also constructed a Spearman rank correlation matrix of traits . In constructing nested linear models and the correlation matrix, lecty was converted from a categorical variable into a quantitative variable corresponding to diet niche breadth to aid in model fitting.Across two years of sampling, we found that study plots in fragments harbored bee assemblages with reduced plot-level functional diversity and distinct functional composition compared to those in reserves. Changes in functional diversity and composition were closely related to declines and shifts in taxonomic diversity and composition. While we found strong evidence for non-random patterns of species loss, such patterns of loss was insufficient to cause landscape-level taxonomic or functional homogenization in the fragments. Null model analyses and correlational analyses also demonstrate that the loss of bee functional diversity can be explained by loss of bee taxonomic diversity. Taken together, these findings suggest that ecological filtering contributes to the restructuring of bee assemblages, large plastic gardening pots but is not the main driving force of bee diversity loss in habitat fragments in our system.
The strongest support for the importance of ecological filtering in our system is the detection of multiple indicator species and functional groups that appear particularly susceptible to fragmentation, typical of “winner-loser” dynamics found in modified landscapes . Also typical of “winner-loser” dynamics, we found a number of species that are present at all study plots, most of which are eusocial species in the tribe Halictini, which are known to be tolerant of habitat fragmentation . However, unlike systems where small numbers of “winner” taxa or functional groups dominate modified landscapes , indicator analyses revealed no such “winner” species or functional groups associated with fragments. Our finding only indicator taxa associated with reserves suggests that ecological filtering indeed leads to the exclusion of certain “loser” taxa and functional groups from fragments, but not to such an extent that the bee assemblages become numerically dominated by groups of disturbance-tolerant species that thrive in altered habitats. In fact, the loss of “loser” taxa seems to largely underlie the detected directional shifts in both taxonomic and functional measures of assemblage composition; simply removing the 12 indicator species from the analyses nullifies the significant differences detected between reserves and fragments with respect to both taxonomic and functional composition .
Evaluating differences between bee faunas in reserves and fragments one trait at a time revealed several differences between reserves and fragments, but only two that remained statistically significant after correction for multiple comparisons . Preferential loss of specialists in modified environments has been documented in many taxa , including bees . The increased relative abundance of late-season active bees observed in the present study has also been reported in at least one other system in which bees in modified landscapes have enhanced access to anthropogenic sources of floral resources during periods of relative resource dearth . In our system, it is likewise plausible that late-season bees in fragments are able to thrive by foraging on floral resources in the irrigated urban matrix surrounding fragments. The increase in average range size of bees inhabiting fragments also reveals the role of ecological filtering in structuring bee assemblages our system. Range size is not a functional trait per se, but it does serve as a proxy for an important ecological function that remains difficult to quantify: overall niche breadth . While many studies on bees focus on lecty as the main metric for niche breadth , selectivity of nesting substrates , phenological flexibility , and physiological tolerance to abiotic conditions may all influence how bee species respond to the addition of novel ecological filters. Our results suggest that bees in fragments tend to be those that are capable of surviving in a greater number of ecological contexts compared to bees in reserves, consistent with the view ecological filters present in habitat fragments exclude species that are more narrowly adapted to the unique local ecosystems.
Such replacement of endemics by geographically widespread species has been observed in other systems impacted by habitat alterations , and may be an important force driving reductions in ecological complexity across large spatial scales. Given that bee assemblages in fragments exhibited strong reductions in both taxonomic and functional alpha diversity as well as distinct taxonomic and functional composition compared to reserves, it is noteworthy that reserves and fragments did not differ with respect to either taxonomic or functional beta diversity. Reduced beta diversity is associated with biotic homogenization, which is a hallmark of ecological filtering resulting from anthropogenic disturbance . Biotic homogenization resulting from land use change has been found across many taxa , including pollinators . However, unlike other systems in which anthropogenic impact is dominant and pervasive, such as in cases where intensive agriculture generated highly simplified landscapes , the habitat fragments we selected in our study were comparable to our natural reserve sites with respect to both the diversity and the composition of native, insect-pollinated plant assemblages, at least at the scale of our one-hectare study plots . Thus, given that our fragment plots retained relatively intact plant assemblages, it is perhaps unsurprising that bee assemblages therein had not converged to a subset of taxa that thrive in altered habitats . As with the findings of , robust beta diversity among fragments may result from underlying heterogeneity in the habitat characteristics of our fragment plots. Taxonomic and functional diversity are often positively related to each other , but the two measures of diversity are related to each other in complex ways and may be independently impacted by habitat modifications . These complex relationships may explain our null model analysis, wherein the reduction in functional diversity in fragments did not differ from expectation under stochastic species loss despite our detecting multiple “loser” functional groups that suffer declines in fragments. While the parallel declines in taxonomic and functional diversity we detected in fragments via both null model and correlation analysis may indeed indicate stochastic loss of species , such a pattern could also arise from non-random loss of species whose functional traits have dispersions comparable to those lost due to random removal of species in the null model. Our finding that the “loser” species and functional groups associated with reserves varied with respect to every functional trait measured lends support to the latter mechanism. Alternatively, the apparent non-uniformity in the functional traits of “loser” taxa may result from our not measuring some other functional traits that may be shared among these taxa. For example, if dispersal is the main driver of bee assemblage composition in fragments, a functional trait that strongly influences the likelihood of dispersal across the urban matrix may be largely responsible for interspecific variation in likelihood of local extirpation from fragments. Irrespective of the mechanism underlying the parallel declines in taxonomic and functional diversity, our null model and correlational analysis results suggest that in our system, managing habitats in such a way as to preserve taxonomic diversity may be an effective way to preserve functional diversity . We uncovered significant phylogenetic conservatism in the functional traits we measured , large plastic growing pots which likely contributed to the numerous correlations detected among traits , a pattern also reported in other studies involving bee functional traits .
Given that phylogenetic conservatism in functional traits can shape the ecology and distribution of bee species in a landscape , our findings must be interpreted in the context of fragmentation impacting bees at the level of higher taxa. However, since analyses at the level of genera yielded qualitatively similar results , the overall patterns we report are unlikely to be driven by a few species-rich groups that respond especially strongly to fragmentation. The detection of indicator species belonging to three families and indicator functional group members belonging to five families also suggests that impacts of fragmentation are not limited to certain clades of bees. Phylogenetic relationships among bee taxa are a subject of ongoing research, even at the level of higher taxa . Once accepted phylogenies become available for bee taxa occurring in our system, it would be instructive to quantify the extent to which evolutionary relationships among taxa contribute to our findings, and the implications fragmentation may have on the evolutionary trajectory of bee faunas as time progresses.The maintenance of both taxonomic and functional beta diversity in our studied fragments argues for the preservation of each individual fragments of CSS habitat, despite the fact that fragments as a whole share the absence of sensitive “loser” bee taxa and functional groups. Our results suggest that each fragment preserves its own distinctive subset of the bee faunas formerly present in the regional species pool, and thus by extension, their ecological interactions with other taxa such as plants, parasitic or commensal invertebrates , and microbes . High levels of heterogeneity in assemblage composition among fragmented habitat remnants have also been documented in other systems ; in such systems, the cumulative species pool of compositionally divergent fragments may equal or exceed the species pools of unfragmented habitat. Beta diversity as a result of habitat heterogeneity is a strong driver of local and regional diversity of pollinators and organisms in general . In our system, bee faunas occupying habitat fragments embedded in a heterogeneous landscape do not exceed or equal those in larger natural reserves with respect to taxonomic or functional diversity, but nevertheless represent valuable units of conservation that may each exhibit unique community-level evolutionary trajectories with time if properly preserved. In fact, of the 216 species collected in the study , 40 were unique to fragments , while 74 were unique to reserves . That said, the decrease in plot-level functional diversity in habitat fragments still represents a conservation challenge with respect to both the functionality and the resilience of bee faunas , highlighting the importance of preserving large, intact areas of scrub habitat.We demonstrated that ecological filtering in fragmented scrub habitats caused shifts in the taxonomic and functional composition of bee faunas as a result of a loss of sensitive bee taxa and an increase in the relative abundance of geographically widespread bee species. However, filtering was not sufficiently strong to reduce functional diversity beyond that expected under random species loss, and bee faunas in fragments retained taxonomic and functional beta diversity among plots. Future studies that can quantitatively partition the relative contribution of deterministic and stochastic processes in driving taxonomic and functional diversity loss will shed light on the factors influencing community reassembly in structurally intact but isolated fragments of well preserved natural habitat.Animal-mediated pollination of angiosperms represents a vital ecosystem function in terrestrial ecosystems . Thus, reported declines in pollinator abundance and diversity worldwide could harm the integrity of terrestrial ecosystems. For this reason, documenting how environmental change impacts the structure and function of plant-pollinator interactions has been identified as an important research priority . In the last two decades, the bipartite network approach has become widely favored for examining interactions between communities of flower visiting animals and plants. To construct plant-pollinator interaction networks, researchers document the frequency with which each pollinator species visits each plant species within a predefined area. The resulting topology of interaction patterns provides information regarding the manner in which species are connected to one another, and the number of interactions documented between two species is often used as a surrogate for the strength of the relationship between the two putative mutualists with respect to pollination services or food provision . While studies on the structure of plant-pollinator interaction networks provide no direct information on the fitness of organisms involved, general patterns in network structure across plant and pollinator communities have shed light on the function of these networks. For example, nestedness and asymmetry , two properties common to most networks studied, result from the presence of groups of numerically abundant, ubiquitous generalist species that interact with large numbers of partner taxa. Having such generalized species at the “core” of networks may cause the ecological function of these networks to be robust to the loss of species and habitat . However, if network structure indeed predicts the resiliency of ecological relationships between plants and pollinators, then perturbations to network structure resulting from anthropogenic impacts may have strong consequences on ecological function Given the link between network structure and ecological function, a number of studies have investigated how plant-pollinator interaction networks are impacted by different sources of anthropogenic impact such as habitat fragmentation , land use intensification , grazing , and biological invasions .