Phylogenetic structure of Brazilian savannas under different fire regimes

Questions Fire is a strong filter in fire-prone communities and is expected to assemble closely related species when functional traits are conserved in plant lineages. Do frequent fires assemble savannas with closely related species (phylogenetic clustering)? If so, what are the clades pruned by fire in the phylogenetic trees? Are species of semi-deciduous seasonal forests, where fires are not frequent, less related than expected by chance (phylogenetic over-dispersion)? Are life forms conserved in the phylogeny of the species? Location Central and SE Brazilian savannas (Emas National Park, 18 degrees 18'S, 52 degrees 54'W; Brasilia, 15 degrees 56'-15 degrees 57'S, 47 degrees 53'-47 degrees 56'W and Corumbatai-Itirapina, 22 degrees 13'-22 degrees 15'S, 47 degrees 37'-47 degrees 39'W); and close semi-deciduous seasonal forests (in Pirenopolis, 15 degrees 45'S, 49 degrees 04'W; Brasilia, 15 degrees 33'S, 47 degrees 51'W; and Sao Carlos, 21 degrees 55'S, 47 degrees 48'W). Methods We recorded woody species in savannas under different fire regimes and in semi-deciduous seasonal forests. We obtained data from the literature and from field sampling. We compared mean phylogenetic distance of species of savanna and of nearby semi-deciduous seasonal forest sites. We obtained significance by randomizing the species among the tips of phylogenetic trees. We also assessed whether life forms were evolutionary conserved across phylogeny of the studied plants (phylogenetic signal) with tests based on the variance of phylogenetic independent contrasts. Results Some sites of savanna under high fire frequency were characterized by phylogenetic over-dispersion of woody species whereas, in contrast, some sites of semi-deciduous seasonal forest were characterized by phylogenetic clustering. We found phylogenetic signals in the traits across the phylogeny of the 801 species investigated. Conclusion Fire may have different roles in assembling plant species in Brazilian savannas than in other fire-prone communities. We postulate that the absence of phylogenetic clustering in the cerrado is mainly due to the persistence of long-lived resprouting species from different plant lineages.