This study aimed to systematically analyze the phenotypic traits of Xanthoceras sorbifolium seeds from different provenances, to construct a scientific germplasm evaluation system, and to provide a theoretical basis for genetic improvement, germplasm innovation, and industrial development of Xanthoceras sorbifolium.

Seeds from 12 provenances (including Gansu, Shanxi, Shandong, etc.) were used as experimental materials. Ten traits, including longitudinal diameter, transverse diameter, and single seed weight, were measured. Multivariate analyses, including one-way ANOVA, Pearson correlation analysis, cluster analysis, and principal component analysis (PCA), were combined to reveal differences in phenotypic traits, internal correlations, and comprehensive performance among provenances.

Significant differences in seed phenotypic traits were observed among provenances (P<0.05). Single seed weight showed coefficient of variation, while seed shape index exhibited the lowest. 100-seed weight was significantly and positively correlated with transverse diameter, longitudinal diameter, kernel weight, and single seed weight (P<0.05), whereas seed volume was significantly negatively correlated with kernel rate (P<0.05). Cluster analysis divided the provenances into three groups: Group I (provenances 2, 3, 9, 11, 12), termed the “small-seed low-weight type”; Group II (provenances 1, 7, 8, 10), the “medium-seed medium-weight type”; and Group III (provenances 4, 5, 6), the “large-seed high-weight type”. PCA indicated that the first three principal components explained 74.05% of the cumulative variance, with provenance 4 achieving the highest comprehensive score (164.89 points), demonstrating the most superior germplasm characteristics.

The established “morphological trait analysis—statistical model construction—germplasm classification” three-level evaluation system provides a new framework for scientific evaluation and genetic improvement of Xanthoceras sorbifolium germplasm. Due to its outstanding comprehensive performance, Provenance 4 is recommended as an excellent parental material for stress-resistant breeding.