Abstract:

Objective: This study aims to solve the problems of easy confusion and difficult identification of chestnut varieties in production, through using the geometric morphometry（GMMs）to digitally analyze the leaf morphology of different chestnut varieties, so as to establish a method of leaf morphology identification of chestnut varieties. Method: The leaves of 80 varieties from different chestnut producing areas in China were used as materials, a total of 6 400 leaves were collected repeatedly in two years. Images were obtained by scanning, geometric morphometric method and Image J software was used to select 24 identification points combined with chestnut leaf characteristics, and the leaf morphology coordinate data were obtained in a consistent order. Morpho J software was used to classify leaf morphology data by production regions and varieties, and a generalized Procrustes analysis was performed to separate the leaf size and the morphology factor, and further form symmetric and asymmetric components. Principal component analysis and partial least squares allometric analysis were performed on the data, and the difference in leaf morphology among different varieties was visualized with grid change diagram. The 24 identification points were classified according to the discrimination contribution rate. The varieties were identified with canonical variables (CVs), discriminant analysis and significance test. Result: 1) The difference in leaf morphology of different varieties was mainly affected by the symmetric components, and the difference was not obvious in the asymmetric components. The cumulative contribution rate of the first two principal components (PCs) in the symmetric components reached 80.6%, which could be used as a CV for the difference analysis of chestnut varieties. The grid change diagram of the symmetric components showed that there were significant differences between varieties. 2) Principal component and allometric analysis showed that the top 14 identification marks with the highest contribution rate were the same, and could be used as the first-level identification marks. 3) The cumulative contribution rate of the first two CVs in the symmetric components reached 81.4%. The scatter plots showed that except for the similarity between 'Huaijiu' and 'Yanfeng', 'Yanshanhongli' and 'Yanchang', 'Yanlong' and 'Yanming', 'Liuyuebao' and 'Yelizang' higher, the other varieties could be accurately distinguished. 4) Discriminant analysis (DA) of chestnut production regions showed that, except Hubei and Anhui (97.5% vs. 96.9%), the correct discrimination rates of other regions reached 100.0%. The DA among varieties showed that 99.3% of the varieties had a discrimination rate of 100.0%, a few of varieties were lower discrimination rate but all above 95.0%, and the discrimination results were significantly different (P<0.05). 5) Cluster analysis reflected the similarity of leaf morphology among production regions and varieties, and the classification results were mostly consistent with the geographical distribution of provenance. Conclusion: The GMMs based on 24 identification marks can accurately identify different chestnut varieties. The screened 14 primary identification marks, 3 secondary identification marks, and 7 supplementary identification marks can accurately reflect the main differences in leaf morphology of chestnut varieties, and the correct discrimination rate reaches 95.3%~100.0%. The established chestnut variety leaf morphology identification database and the chestnut variety identification method based on the digital analysis of leaf morphology will provide technical support for the accurate identification of chestnut varieties.

Key words: Castanea mollissima, leaf, geometric morphology, variety identification