Tree species delimitation in tropical forest inventories: Perspectives from a taxonomically challenging case study

Abstract

Forest conservation and management programs rely on precise knowledge of local tree diversity. Distinguishing tree species is a major concern in tropical forest inventories as morphology alone may not be sufficient for species identifications in those highly diverse environments. Therefore, a multi-evidence approach is desirable to avoid subjective interpretations. We tested the contribution of a low-cost integrative protocol for delimiting sympatric species of the taxonomically challenging genus Myrcia (Myrtaceae) in the central Amazon region. We classified specimens into local species by morphotyping them and then validated our hypotheses using two additional sources of evidence: i) near-infrared spectroscopic data, consisting of 1557 absorbance values obtained from dried leaves; and, ii) morphometric data of leaf blade contours, translated into Fourier descriptors and summarized by principal component analysis (PCA), together with four additional leaf characters. The morphotypes were validated using linear discriminant analyses (LDAs) of the raw NIR spectra, stepwise-selected spectral regions, PCA-reduced spectra, as well as the morphometric data. A taxonomic name was assigned to each validated species based on comparisons with herbarium specimens and/or the technical literature. We delimited 38 species of Myrcia, and our hypotheses were well-supported by the LDAs (81–99% accuracy), evidencing that inexpensive tools can be effectively used to discriminate species in large-scale projects, and that integrative approaches are fundamental in that regard. Although vegetative traits were sufficient for species discrimination, fertile samples were crucial for obtaining taxonomic names for them. Thirteen species delimited by us belong to four species complexes, each treated under single species names in the current systematics of the genus. We therefore argue that species delimitations prior to individual identifications are essential for robust species definition in forest plot research because local species do not necessarily match global circumscriptions. This type of approach may greatly alter how local community structures are viewed and consequently modify the results of downstream analyses of ecological studies. We also emphasize that local species delimitations may contribute to the field of taxonomy by identifying potential inconsistencies in global species definitions.