Contrasting lineage-specific patterns conceal community phylogenetic structure in larger clades

Question Community phylogenetic approaches can provide information on the ecological forces structuring plant community composition. For example, assuming evolutionary conservatism of ecological traits, environmental filtering has been suggested to result in phylogenetic clustering, whereas competition has been suggested to lead to phylogenetic over-dispersion. However, current approaches report aggregate community-wide metrics and typically assume that all lineages respond similarly. Here, we question this assumption and evaluate evidence for lineage-specific patterns of co-occurrence. Location Eastern Canadian subarctic. Methods We use a novel individual-based, neighbourhood approach to evaluate lineage-specific co-occurrence patterns among 35 focal species of Cyperaceae (sedges). Cyperaceae is a species-rich clade, with many species sharing similar niche preferences and environmental tolerances, making this a model clade for evaluating interspecific interaction strengths. Results We reveal striking differences in co-occurrence patterns between clades. Within the Cyperaceae, the two species of the genus Trichophorum tended to cluster with each other, whereas species within Eriophorum co-occurred more often with less closely related neighbours. These contrasting patterns repeated themselves within Carex, with species in the Core Carex clade more often co-occurring with close relatives, while phylogenetic over-dispersion was more prominent in the Vignea clade. As a consequence of these opposing lineage-specific relationships, the overall phylogenetic structure of sedge communities appears random. Conclusions We show opposing lineage-specific patterns in phylogenetic structure that are obscured when species are aggregated into larger clades. Our results reveal cryptic complexities in plant community assembly across taxonomic scales, and indicate that different plant lineages may vary in their response to community-level processes. We suggest that the processes underlying community composition can be better understood by using alternative sampling and analytical approaches combined with thoughtfully-created null models.