Taxonomic revision of the Malagasy members of the Nesomyrmex angulatus species group using the automated morphological species delineation protocol NC-PART-clustering

Background. Applying quantitative morphological approaches in systematics research is a promising way to discover cryptic biological diversity. Information obtained through twenty-first century science poses new challenges to taxonomy by offering the possibility of increased objectivity in independent and automated hypothesis formation. In recent years a number of promising new algorithmic approaches have been developed to recognize morphological diversity among insects based on multivariate morphometric analyses. These algorithms objectively delimit components in the data by automatically assigning objects into clusters. Method. In this paper, hypotheses on the diversity of the Malagasy Nesomyrmex angulatus group are formulated via a highly automated protocol involving a fusion of two algorithms, (1) Nest Centroid clustering (NC clustering) and (2) Partitioning Algorithm based on Recursive Thresholding (PART). Both algorithms assign samples into clusters, making the class assignment results of different algorithms readily inferable. The results were tested by confirmatory cross-validated Linear Discriminant Analysis (LOOCV-LDA). Results. Here we reveal the diversity of a unique and largely unexplored fragment of the Malagasy ant fauna using NC-PART-clustering on continuous morphological data, an approach that brings increased objectivity to taxonomy. We describe eight morphologically distinct species, including seven new species: Nesomyrmex angulatus (Mayr, 1862), N. bidentatus sp. n., N. clypeatus sp. n., N. devius sp. n., N. exiguus sp. n., N. fragilis sp. n., N. gracilis sp. n., and N. hirtellus sp. n.. An identification key for their worker castes using morphometric data is provided. Conclusions. Combining the dimensionality reduction feature of NC clustering with the assignment of samples into clusters by PART advances the automatization of morphometry-based alpha taxonomy