Biodiversity change in anthropogenically transformed habitats is often nonrandom, yet the nature and importance of the different mechanisms shaping community structure are unclear. Here, we extend the classic Theory of Island Biogeography (TIB) to account for nonrandom processes by incorporating species traits and phylogenetic relationships into a study of faunal relaxation following habitat loss and fragmentation. Two possible mechanisms can create nonrandom community patterns on fragment islands. First, small and isolated islands might consist of similar or closely related species because they are environmentally homogeneous or select for certain shared traits, such as dispersal ability. Alternatively, communities on small islands might contain more dissimilar or distantly related species than on large islands because limited space and resource availability result in greater competitive exclusion among species with high niche overlap. Breeding birds were surveyed on 36 islands and two mainland sites annually from 2010 to 2014 in the Thousand Island Lake region, China. We assessed community structure of breeding birds on these subtropical land-bridge islands by integrating species' trait and evolutionary distances. We additionally analysed habitat heterogeneity and variance in size ratios to distinguish biotic and abiotic processes of community assembly. Results showed that functional-phylogenetic diversity increased with island area, and decreased with isolation. Bird communities on the mainland were more diverse and generally less clustered than island bird communities and not different than randomly assembled communities. Bird communities on islands tend to be functionally similar and phylogenetically clustered, especially on small and isolated islands. The nonrandom decline in species diversity and change in bird community structure with island area and isolation, along with the relatively homogeneous habitats on small islands, support the environmental filtering hypothesis. Our study demonstrates the importance of integrating multiple forms of diversity for understanding the effects of habitat loss and fragmentation, and further reveals that TIB could be extended to community measures by moving beyond assumptions of species equivalency in colonisation rates and extinction susceptibilities.

China is one of the countries with the richest bird biodiversity in the world. Among the 1372 Chinese birds, 146 species are considered threatened and three species are regionally extinct according to the officially released China Biodiversity Red List in 2015. Here, we conducted the first extensive analysis to systematically investigate the patterns and processes of extinction and threat in Chinese birds. We addressed the following four questions. First, is extinction risk randomly distributed among avian families in Chinese birds? Second, which families contain more threatened species than would be expected by chance? Third, which species traits are important in determining the extinction risk in Chinese birds using a multivariate phylogenetic comparative approach? Finally, is the form of the relationship between traits additive or nonadditive (synergistic)? We found that the extinction risk of Chinese birds was not randomly distributed among taxonomic families. The families that contained significantly more threatened species than expected were the hornbills, cranes, pittas, pheasants and hawks and eagles. We obtained eleven species traits that are commonly hypothesized to influence extinction risk from the literature: body size, clutch size, trophic level, mobility, habitat specificity, geographical range size, nest type, nest site, flocking tendency, migrant status and hunting vulnerability. After phylogenetic correction, model selection based on Akaike's information criterion identified the synergistic interaction between body size and hunting vulnerability as the single best correlate of extinction risk in Chinese birds. Our results suggest that, in order to be effective, priority management efforts should be given both to certain extinction‐prone families, particularly the hornbills, pelicans, cranes, pittas, pheasants and hawks and eagles, and to bird species with large body size and high hunting vulnerability.

Incorporating imperfect detection when estimating species richness has become commonplace in the past decade. However, the question of how imperfect detection of species affects estimates of functional and phylogenetic community structure remains untested. We used long-term counts of breeding bird species that were detected at least once on islands in a land-bridge island system, and employed multi-species occupancy models to assess the effects of imperfect detection of species on estimates of bird diversity and community structure by incorporating species traits and phylogenies. Our results showed that taxonomic, functional, and phylogenetic diversity were all underestimated significantly as a result of species' imperfect detection, with taxonomic diversity showing the greatest bias. The functional and phylogenetic structure calculated from observed communities were both more clustered than those from the detection-corrected communities due to missed distinct species. The discrepancy between observed and estimated diversity differed according to the measure of biodiversity employed. Our study demonstrates the importance of accounting for species' imperfect detection in biodiversity studies, especially for functional and phylogenetic community ecology, and when attempting to infer community assembly processes. With datasets that allow for detection-corrected community structure, we can better estimate diversity and infer the underlying mechanisms that structure community assembly, and thus make reliable management decisions for the conservation of biodiversity.

1. Habitat loss and fragmentation often leads to defaunation of large-bodied mammals, and their loss could trigger release from top-down control or food resource competition for small mammal seed dispersers, which in turn may affect the effectiveness of seed dispersal by altering the number of dispersed seeds or the manner in which they are dispersed. Although rodents are primary seed dispersers in habitat subjected to defaunation, changes in seed dispersal effectiveness of rodents along mammalian defaunation gradients, and empirical support for mechanisms underlying alteration of this ecological process, are unclear.2. We assessed the direct and indirect effects of forested area and isolation on seed dispersal effectiveness of rodents on 21 study islands with varying levels of defaunation in the Thousand Island Lake, China. We used camera sampling, live traps and semi-quantitative acorn counts to assess occurrence of large-bodied mammal species, relative abundance of small rodent species and seed crop size respectively. Seed dispersal, post-dispersal seed survival, seedling emergence, and seedling survival were estimated by tracking fates of tagged acorns and by planting acorns in exclosures. 3. Forested area had positive indirect effects on seed dispersal effectiveness through defaunation and rodent competition for acorns, whereas isolation had negative direct and weaker positive indirect effects on seed dispersal effectiveness mediated by loss of large-bodied mammals and rodent competition for acorns. Loss of large-bodied mammals negatively affected seed dispersal effectiveness indirectly by virtue of its impact on rodent competition for acorns. Seed dispersal effectiveness exhibited a unimodal relationship with intensity of rodent competition for acorns, peaking at intermediate levels.4. Synthesis. Indirect effects of island attributes mediated by defaunation of largebodied mammals on small or isolated islands appear to drive altered competition for food among rodents and decreased seed dispersal effectiveness. Altered interactions between acorns and their rodent consumers/dispersers can substantially affect oak population demography in the Thousand Island Lake system.More broadly, our findings highlight the importance to the seed dispersal process of multiple interwoven effects between habitat fragmentation and defaunation of large-bodied mammals.

Although arboreal camera trapping is a growing field, it has rarely been used for monitoring plant-frugivore interactions in the trees. Frugivore foraging behavior generally occurs in trees, hence arboreal camera trapping can be a potentially useful tool for frugivory research. We developed a camera trap sampling method to monitor plant-frugivore interactions during mature fruiting periods. We used this method to monitor 318 individuals (camera sites) of 18 fleshy-fruit plant species on 22 subtropical land-bridge islands in the Thousand Island Lake, China. We recorded a total of at least 52 frugivorous animals, including a ground-foraging bird species (Lophura nycthemera) and several mammals with foraging behaviors in the trees. We also recorded 4399 independent interaction events, including 275 unique plant-bird interactions. We proposed a framework to classify interaction types and performed a sampling completeness test. We found that a sampling strategy that covered approximately a third of the fruit maturation period when most fruits were ripe was sufficient to sample plant-frugivore interactions. Our results demonstrated that our sampling method with camera transects is reliable to monitor plantfrugivore interactions in a fragmented landscape. This study helps to lay the methodological foundation for building networks of plant-frugivore interactions with arboreal camera trapping on large spatial/temporal scales. As a noninvasive, labor-saving, and largely unbiased sampling method, the field application of arboreal camera trapping in different regions can advance the technology of biodiversity monitoring and lead to more accurate biodiversity inventories in arboreal environments.

Density‐dependent non‐monotonic species interactions are important in maintaining ecosystem stability and function, but empirical evidences are still rare. Rodents, as both seed dispersers and seed predators, have dual effects on plant regeneration and may result in non‐monotonic rodent‐plant interactions. According to the non‐monotonic models, the relative positive or negative effects of rodents on seedling establishment can be measured based on the positive or negative association of seedling recruitment rate and rodent abundance. In this study, we investigated the fates of acorns of Quercus serrata by tracking tagged seeds on 21 fragmented subtropical islands in the Thousand Island Lake, China. We found that the proportion of germinated seeds of all released seeds showed a dome‐shaped association with rodent abundance per seed. The proportion of removed seeds and cached seeds showed a saturated‐ and a weak dome‐shaped association with rodent abundance per seed, respectively. Our results demonstrated a clear empirical evidence that rodent abundance per seed triggered a switch between the relative mutualism and predation in a rodent–seed system. Our study implied that the observed non‐monotonic interactions between plants and animals may play a significant role in maintaining biodiversity and ecosystem function. We appeal for more investigations of the complex non‐monotonic interactions in various ecosystems.

Habitat fragmentation impacts seed dispersal processes that are important in maintaining biodiversity and ecosystem functioning. However, it is still unclear how habitat fragmentation affects frugivorous interactions due to the lack of high-quality data on plant-frugivore networks. Here we recorded 10,117 plant-frugivore interactions from 22 reservoir islands and six nearby mainland sites using the technology of arboreal camera trapping to assess the effects of island area and isolation on the diversity, structure, and stability of plant-frugivore networks. We found that network simplification under habitat fragmentation reduces the number of interactions involving specialized species and large-bodied frugivores. Small islands had more connected, less modular, and more nested networks that consisted mainly of small-bodied birds and abundant plants, as well as showed evidence of interaction release (i.e., dietary expansion of frugivores). Our results reveal the importance of preserving large forest remnants to support plant-frugivore interaction diversity and forest functionality.

Edge effects often exacerbate the negative effects of habitat loss on biodiversity. In forested ecosystems, however, many pollinators actually prefer open sunny conditions created by edge disturbances. We tested the hypothesis that forest edges have a positive buffering effect on plant-pollinator interaction networks in the face of declining forest area. In a fragmented land-bridge island system, we recorded ~20,000 plant-pollinator interactions on 41 islands over 3 yr. We show that plant richness and floral resources decline with decreasing forest area at both interior and edge sites, but edges maintain 10-fold higher pollinator abundance and richness regardless of area loss. Edge networks contain highly specialized species, with higher nestedness and lower modularity than interior networks, maintaining high robustness to extinction following area loss while forest interior networks collapse. Anthropogenic forest edges benefit community diversity and network robustness to extinction in the absence of natural gap-phase dynamics in small degraded forest remnants.