Differentiation of sweet chestnut (Castanea sativa Mill.) cultivars by leaf, nut and burr dimensions

Morphological traits including leaf, nut and burr dimensions were measured in chestnut stands in Fonsagrada (Galicia, north-west Spain). A total of 5760 data (720 measurements of eight variables) from 72 chestnuts belonging to the three main cultivars growing in the area (Lemos, Parede and Portuguesa) were available for analysis. Two discriminant linear rules/models for classifying chestnut specimens as one of the three cultivars were developed. The models differed in terms of the sampling effort involved - the 'full' model included all the analysed variables (leaf-, nut- and burr-related variables) and the 'leaf and nut' model included only variables related to leaves and nuts. The full model performed best and correctly classified 79 per cent of the Lemos, 96 per cent of the Parede and 100 per cent of the Portuguesa cultivars, according to the results of the cross-validation method used to assess the misclassified cultivars. High accuracy was also achieved with the leaf and nut model, with only a slight increase in misclassification compared with the full model. Therefore, when time consumption is an important concern for users, the leaf and nut model is recommended as a good compromise between accuracy and sampling effort.