A 3D framework for delimiting a polyploid complex in Rorippa (Brassicaceae): combining trait evolution, herbarium records, and machine learning

Abstract

Species delimitation in polyploid complexes remains a fundamental challenge due to pervasive morphological overlap and genomic redundancy. We examined the Rorippa dubia-indica complex (Brassicaceae), a polyploid lineage comprising tetraploid and hexaploid taxa. We developed an integrative 3D framework (Delimitation, Distribution, and Decoding) that synthesizes spatiotemporal data from field (2017-2020; n = 3,136) and herbarium (1893-2021; n = 2,015) collections to diagnose misidentification, model distributions, and reconstruct classification criteria used in polyploid complexes. Morphological traits with varying degrees of plasticity were evaluated under controlled conditions to identify stable diagnostic characters. Seed arrangement, petal number, and genome size or ploidy level exhibited clear interspecific differentiation. Phylogenomic analyses based on chloroplast genomes further defined species boundaries clarified by these taxonomic traits. We then revised herbarium specimens and applied machine learning classification models to assess the extent of specimen misidentification and to recover the trait-based rationale behind species assignments. Initial misidentification rates reached 12-50% across virtual or physical specimens, largely due to reliance on plastic traits. These errors substantially distorted spatial distribution models and future climate projections. Our findings underscore the need for secondary specimen evaluation and demonstrate the importance of integrating morphologic and phylogenetic inference with machine learning tools to resolve morphologically overlapping polyploid complexes. This approach offers direct applications for biodiversity assessment, evolutionary research, and conservation planning.