Fine-tuned Head Weight Estimation in Globe Artichoke (Cynara scolymus L.)

Abstract

A novel and nondestructive method for head weight estimation in globe artichoke was described. Linear measurements on head height (h) and head diameter (2r) were performed on head samples of five cultivars and one developed clone having cylindrical, conical or spherical head forms. The measurements on 2r were performed at the base in conical- and cylindrical-headed cultivars, while they were taken equatorially in spherically headed cultivars. Correlation and regression analyses were performed between single true head weights and eight different models [h, r, 2r, r3, r2h, πr2h (cylinder volume), ⅓πr2h (cone volume) and \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\frac{4}{3}{\pi}r^{3}\) \end{document} (sphere volume)]. Since πr2h and ⅓πr2h are folds of r2h, and \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\frac{4}{3}{\pi}r^{3}\) \end{document} of r3; r2h or r3 had completely the same correlation coefficients as their folds, and hence were equally effective in the statistical analyses. Head weights were more precisely estimated when the r2h model (or their folds) was used for cylindrical and conical heads and the r3 model (or their folds) for spherical heads then any other model. Coefficients of determination (R2) explained the highest variability observed for true head weights when the r2h model was used as the independent variable in the regression analysis for cylindrical and conical cultivars (96.6% to 97.5%) and the r3 model for the spherical cultivars (96.6% to 98.4%). Even though all correlation coefficients and regression F values were very highly significant (p < 0.001), 81% of estimated cases had <10% deviations when one of the appropriate models were used in comparison to 66% of the estimated cases having <10% deviations when, for example, the 2r model was used, the second most effective model for all types of heads. The agreement between true and estimated head weights was also tested, where the bias was lower for a cultivar-specific model vs. the 2r model. Discussions on applying the results to the selection procedures were made.