Seed image analysis provides evidence of taxonomic differentiation within the Medicago L. sect. Dendrotelis (Fabaceae)

Morpho-colorimetric quantitative variables describing seed size, shape, colour and texture were analysed using image analysis techniques, in order to evaluate the variability among Medicago taxa sect. Dendrotelis and verify the current taxonomic treatment which divides this section into three species: M. arborea L., M. citrina (Font Quer) Greuter and M. strasseri Greuter, Matthäs & H. Risse. Further comparisons were conducted to discriminate among populations and regions of provenance. Data obtained were statistically analysed applying stepwise Linear Discriminant Analysis method (LDA), recording an overall cross-validated classification performance of 100% at species level. With regard to inter-population comparisons, percentages of correct discrimination above 98% were achieved and high performance was recorded in the discrimination among M. arborea taxa distinguished by region of provenance. For each of these statistical comparisons, the best discriminant variables chosen by the stepwise LDA were related to colour and textural information. Finally, the obtained results confirmed the validity of the proposed method to be highly diagnostic in the statistical assessment of the morpho-colorimetric traits variability of Medicago taxa seeds, both for the taxonomic differentiation at species level and regional and population groups.