Assessing the Discriminatory Power of Leaf Morphometric Traits in Capsicum annuum by means of a Linear Discriminant Analysis

Leaf morphometric traits are widely utilized to identify plants, yet their effectiveness when used to discriminate large Capsicum collections remains unclear.This study assessed whether eight leaf size and shape traits could reliably classify 371 and 352 C. annuum accessions collected in 2021 and 2022.Comprehensive diagnostic testing revealed major violations of the assumptions required for linear discriminant analysis (LDA), including non-normal trait distributions, unequal covariance matrices, and severe multicollinearity among predictors.Shape descriptors further exhibited nonlinear dependencies incompatible with LDA.As a result, LDA models produced unstable discriminant functions and extremely low agreement between predicted and true accession labels.Given these limitations, unsupervised methods were employed.K-means clustering identified eight phenotypic groups in 2021 and three in 2022, and a principal component analysis (PCA) supported these patterns, with the first two components capturing the primary axes of the leaf size and shape variation.However, strong overlap among accessions in the PCA space indicated weak morphological differentiation, likely reflecting strong environmental influence and limited genetic divergence with regard to the leaf form.These findings indicate that basic leaf morphometric traits alone are insufficient for accession-level classification in C. annuum.Robust discrimination will require integrative phenotyping frameworks that combine geometric, physiological, biochemical, and genomic traits, supported by measurement technologies.