Objective The study was aimed at the screening of superior germplasmls of Michelia chapensis with an objective to lay a foundation of the selection and breeding project of this species.

Method A total of 213 M. chapensis grafted clones derived from different regions of Guangdong, Hunan, Jiangxi were tested in National Forest Tree Germplasm Resource Bank of Magnoliaceae of Longshan in Guangdong (China) for their growth performance and form traits. The phenotypic and genetic parameters, correlation coefficients, principal components of variation and breeding values of different traits were estimated at age 3 and therefore the more superior clones were selected herein.

Result Highly significant differences (P < 0.01) were detected in tree height, ground diameter, crown width, height-diameter ratio, and crown-diameter ratio among the tested M. chapensis clones. Phenotypic coefficients of variation exceed 16% (16.59%~23.39%), the genetic coefficients of variation above 10% (10.27%~14.83%), and the repeatability is above 0.60 (0.64~0.71) among all traits. Significant (P < 0.05) to highly significant (P < 0.01) phenotypic and genetic correlations were detected among the traits (except between tree height and crown-to-diameter ratio). The breeding values exhibited normal distributions using the Best Linear Unbiased Prediction (BLUP) method of all traits, representing as -0.63 to 0.82 for tree height, -1.87 to 2.09 for ground diameter, -0.62 to 0.67 for crown width, -11.00 to 9.51 for height-diameter ratio, and -7.22 to 9.90 for crown-diameter ratio respectively. Based on breeding values for growth traits and membership function values fifteen superior clones with comprehensive growth performance and thirteen superior clones with demonstrated vigorous growth with compact crown width were identified as more super clones.

Conclusion The tested M. chapensis clones showed extensive variation and high repeatability in growth performance and form traits, and the re-elected ones could be regarded as candidate elite varieties based on the cluster analysis, breeding values, and membership function values.