Premium
Characteristics of leaf shapes among two parental Drosera species and a hybrid examined by canonical discriminant analysis and a hierarchical Bayesian model
Author(s) -
Hoyo Yuri,
Tsuyuzaki Shiro
Publication year - 2013
Publication title -
american journal of botany
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.218
H-Index - 151
eISSN - 1537-2197
pISSN - 0002-9122
DOI - 10.3732/ajb.1200510
Subject(s) - biology , identification (biology) , linear discriminant analysis , botany , species identification , statistics , zoology , mathematics
• Premise of the study: Although nondestructive, convenient species identification is desirable for follow‐up research and species conservation, species identification is often confusing, particularly when an interspecific hybrid shows intermediate morphological characteristics between the parental species. • Methods: Drosera anglica Hudson (2 n = 40) and D. rotundifolia L. (20) bear the hybrid Drosera obovata Mert. et Koch (30). The samples were identified based on seed fertility and a cytological investigation (DNA amount) before examination. Then, 13 measured morphological traits—including leaf size, leaf shape, and flowering—were used in a canonical discriminant analysis (CDA). Leaf length and width were used in a hierarchical Bayesian model (HBM). • Key results: The majority of the traits of D. obovata were intermediate between the two parental species. However, D. obovata developed larger leaves than the parental species. The identification error of the CDA based on the 13 morphological traits was 4.9%. Errors occurred more often with smaller leaves. When the CDA was used for blade length and width only, the error increased to 6.2%. The HBM, based on the relationships between blade length and width, showed the lowest identification error—4.7%—by improving the identification of small leaves. • Conclusions: The HBM enabled convenient, nondestructive measurements for species identification by considering nonlinear relationships between morphological traits and measurement error. The HBM is likely to be applicable to various follow‐up studies, as well as species conservation.