EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

A simple method for discriminating Schizolobium parahyba (Vell.) Blake (Fabaceae) seeds population by dimensional morphometry and machine learning algorithms

Michele A.S. Nobrega, Maria Luisa H.P. Nabuco, Débora de Araújo, Simone Y. Fernandes, Emilio C. Miguel, Gilberto J. Arruda, Etenaldo F. Santiago, Jaqueline S. Santos, Montcharles S. Pontes

Flora · 2027-01 · 346 · 153108

DOI: 10.1016/j.flora.2026.153108 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Flora, 2027-01. Verfasst von Michele A.S. Nobrega und 8 weiteren Autor:innen. Die Arbeit nennt 51 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: CNPq.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Seed morphometry provides a robust framework for assessing intraspecific variation and population differentiation in tropical tree species. This study investigated whether dimensional morphometric traits and derived physical properties of Schizolobium parahyba (Vell.) Blake (Fabaceae) seeds are sufficient to discriminate between two geographically distinct populations occurring in the Cerrado and the Atlantic Forest domains in Brazil. For each population, 100 seeds were measured along three principal axes (length, width, and thickness), and arithmetic mean diameter, geometric mean diameter, surface area, volume, aspect ratio, and sphericity index were derived. Mean trait values did not differ significantly between populations (Mann-Whitney U test, all p > 0.22), but Pearson correlation analyses revealed divergent patterns of trait association: Cerrado seeds exhibited a weak negative correlation between length and both width (r = -0.17) and thickness (r = -0.08), whereas Atlantic Forest seeds displayed strong positive correlations (r = 0.68 and r = 0.65, respectively). Principal component analysis captured 89.9% of total variance on the first two axes (PC1: 66.2%, PC2: 23.7%), with partial, non-exclusive separation between populations. Supervised machine learning classifiers, evaluated with a leakage-free, repeated 10-fold cross-validation protocol, achieved discrimination accuracies of up to 82.6% (SVM with a radial basis function kernel), with linear classifiers already performing well above chance (69.7-74.4%). Unsupervised clustering (k-means, hierarchical clustering, Gaussian mixture models) did not recover the two populations as distinct natural groups, whereas a Self-Organizing Map revealed statistically significant spatial segregation between populations (permutation test, p = 0.0005). These findings demonstrate that S. parahyba seed populations from contrasting biomes possess subtle but detectable and statistically robust morphometric signatures rooted in their variance-covariance structure rather than in mean trait values, though whether this pattern reflects genetic differentiation, maternal effects, or environmental plasticity remains to be tested.

Förderung

  • CNPq

Zitieren

APA

Nobrega, M. A., Nabuco, M. L. H., Araújo, D. D., Fernandes, S. Y., Miguel, E. C., Arruda, G. J., Santiago, E. F., Santos, J. S., & Pontes, M. S. (2027). A simple method for discriminating Schizolobium parahyba (Vell.) Blake (Fabaceae) seeds population by dimensional morphometry and machine learning algorithms. Flora, 346, 153108. https://doi.org/10.1016/j.flora.2026.153108

Vancouver

Nobrega MA, Nabuco MLH, Araújo DD, Fernandes SY, Miguel EC, Arruda GJ, et al. A simple method for discriminating Schizolobium parahyba (Vell.) Blake (Fabaceae) seeds population by dimensional morphometry and machine learning algorithms. Flora. 2027;346:153108. doi:10.1016/j.flora.2026.153108

BibTeX
@article{Nobrega2027,
  author = {Nobrega, Michele A.S. and Nabuco, Maria Luisa H.P. and Araújo, Débora de and Fernandes, Simone Y. and Miguel, Emilio C. and Arruda, Gilberto J. and Santiago, Etenaldo F. and Santos, Jaqueline S. and Pontes, Montcharles S.},
  title = {A simple method for discriminating Schizolobium parahyba (Vell.) Blake (Fabaceae) seeds population by dimensional morphometry and machine learning algorithms},
  journal = {Flora},
  year = {2027},
  volume = {346},
  pages = {153108},
  doi = {10.1016/j.flora.2026.153108}
}
Kostenlos starten →