Quantitative evaluation of apple (Malus × domestica Borkh.) fruit shape by principal component analysis of Fourier descriptors

Abstract

Apples from 1253 genotypes representing 82 open pollinated families planted at three sites were cut along the stem-calyx axis at the widest point to analyse fruit shape. An image analysis program was used to extract calliper measurements of the fruit outline and calculated Fourier descriptors for each fruit outline. Five independent shape traits were identified from a principal component analysis of the Fourier descriptors. The shape traits and the proportion of the total phenotypic variation they accounted for were: fruit aspect (76.8%), asymmetric-crown (7.8%), fruit conicity (6.0%), asymmetric-sides (4.3%), and fruit squareness (2.0%). Genetic and residual variance components were estimated with data from two sites using restricted maximum-likelihood techniques to select genetically-inherited apple shape traits. Combined sites heritability was estimated to be 0.79 for aspect, 0.38 for conicity and 0.35 for squareness. Multiple regression between calliper measurements and aspect, conicity, and squareness traits showed firstly that aspect was best predicted by fruit length/width ratio (R² = 0.97), secondly conicity could be described by both the distance of the maximum width from the base of the fruit/fruit length ratio and the calyx basin width/fruit width ratio (R² =0.44), and finally squareness was best described by ratio of the product of calyx basin width and distance of the maximum width from the calyx end of the fruit by the product of fruit length and fruit width(R² = 0.19). A chart based on the aspect, conicity and squareness principal component values was drawn to allow visual assessment of shape.