Morphological diversity of Chinese chinquapin (Castanea henryi (Skan) Rehder & E.H. Wilson) germplasm.

Chinese chinquapin (Castanea henryi (Skan) Rehder & E.H. Wilson) is an important woody plant producing nuts native to southern China. Phenotypic characterization of this species provides essential insights into its traits, which are fundamental for breeding programs and germplasm resource management. To provide a comprehensive phenotypic profile, 32 morphological traits were analyzed using a combination of variation and correlation analyses, principal component analysis (PCA), and multiple correspondence analysis (MCA). Furthermore, cluster analysis was performed to classify the accessions into distinct groups based on phenotypic similarity, while random forest (RF) algorithms were employed to identify key diagnostic characteristics. The coefficients of variation for most descriptive traits of leaves and fruits are greater than 20%, and the diversity indices range from 0.12 to 1.28. The pubescence on the underside of the leaves is relatively conserved. Regarding qualitative variation, MCA resolved complex interdependencies among descriptive characteristics, identifying significant character syndromes such as the tight coordination between involucre morphology and seed shape. Strong positive correlations were detected between leaf and seed characteristics. The results of PCA revealed that the first five PCs accounted for a cumulative variance of 80.298%. The results showed that while most germplasm resources clustered centrally with high phenotypic similarity, several accessions, including Xinli-1, Xinli-2, Bailuzi, Daguohuangzhen, Huali-1, Wenzhen, and Huangyou, were distributed peripherally, reflecting their distinct morphological characteristics. A hierarchical cluster analysis, based on Euclidean distance and Ward’s method, successfully resolved the 38 accessions into four distinct morphotypes. Group I is characterized by the largest leaf dimensions and superior fruit metrics, which makes it a prime candidate for large-fruit breeding programs. In contrast, Group II exhibits a unique “slender” leaf architecture coupled with high kernel moisture content. Group III, despite having the smallest leaves and fruits, achieves the highest seed rate. Finally, Group IV (36.84%) constitutes the largest cluster and represents an intermediate phenotype. Furthermore, most of the ‘Deli’ series can be used to breed late-maturing varieties; Bailuzi and Huali-1 could be used to breed for early maturing. For quantitative traits, a Random Forest Regression (RFR) model was implemented to evaluate the predictive importance of numerical variables. The model achieved high precision (test set R2 = 0.802), identifying involucre vertical diameter (35.80%) and horizontal diameter (14.60%) as the primary determinants of the target scores. These findings provide a theoretical basis for the conservation and genetic improvement of Chinese chinquapin accessions.