Morphological and genetic evidence for the Stigmatodon goniorachis complex (Tillandsioideae, Bromeliaceae) reveals a new species on inselbergs in the Brazilian Atlantic Forest

This study evaluated the morphological variation and genetic structure of seven natural populations of the Stigmatodon goniorachis complex (Bromeliaceae) endemic to vertical faces of inselbergs in the Brazilian Atlantic Forest. A cluster analysis and Principal Component Analysis (PCA) corroborated the existence of four clearly separated species, S. brassicoides, S. costae, S. francae, and S. goniorachis, and unveiled two distinct morphological groups in the S. goniorachis populations. The main components of variation revealed by PCA analysis were: inflorescence peduncle diameter, floral bract length, and hypanthium length. Both morphological and genetic analyses indicated two distinct groups among the four populations initially identified as S. goniorachis, which led to the recognition of a new species described here, Stigmatodon carioca.  Besides the description of the new species, we present a comparative study that allows the morphological differentiation of the five species.

Methods
Five individuals were collected at each site, according to the availability of fertile individuals, for a total of 35 individuals from seven sites. The dimensions of fully developed leaves and floral structures were measured and analyzed for presence/absence, shape and type of indumentum. The colors of inflorescences, flowers and leaves observed in the field were included in the analyses. The four species that compose the S. goniorachis complex were analyzed together. Data were tabulated in matrices of operational taxonomic units (OTUs) × quantitative (19) and qualitative (7) morphological variables in order to conduct ordination and clustering analyses. For cluster analysis, the Gower index was used as a measure of similarity (Legendre and Legendre 1998), and the criterion of average linkage between groups (Unweighted Pair Group Method with Arithmetic mean – UPGMA). For the ordination of OTUs and quantitative variables, a Detrended Correspondence Analysis (DCA) was used to determine the gradient size, and a Principal Component Analysis (PCA) to analyze which morphological characters were responsible for the formation of the possible groups (Legendre and Legendre 1998). For all analyses, Fitopac 1.6 (Shepherd 2010), PC-Ord 4.10 (McCune and Mefford 1999) and Past v3.0 (Hammer et al. 2001) were used. Clustering and ordination analyses of quantitative and qualitative variables were performed based on the character.