Using Gaussian Mixture Models in plant morphometrics

Morphometrics provides a rigorous quantitative-statistical framework for assessing morphological independence among taxa in plant systematics. Despite its importance, current methods for analyzing morphological data are often not appropriate. A new workflow to conduct linear morphometric analyses in plant systematics is presented here. We introduce a Bayesian framework for species circumscription using Gaussian Mixture Models (GMMs), which enables rigorous testing of alternative taxonomic hypotheses. In addition, we present a set of algorithms for morphometric analyses: a lumping-splitting algorithm, methods for computing class-wise morphometric distances, and tools for visualising admixture patterns in morphometric data. We also developed a comprehensive guide for performing linear morphometric analyses in plant systematics and exemplified the new workflow using the Juniperus oxycedrus group. This framework creates a meaningful link between morphology-based taxonomy and formal statistical methods, aligning with the probabilistic concept of evolutionary lineages (UPCEL).