ELECTRONIC SUPPLEMENTAL MATERIAL Materials and methods Statistical analysisIn order to determine the most important traits and demonstrate the contribution share of different traits for the formation and the determination of PY, the method of variable selection and multiple regressions was used. The stepwise regression analysis with forward selection method was performed. Through this method, the relationship between the PY and all the traits were quantitatively determined; from the 38 investigated traits. After that, the best models of one to ten variables were specified. Moreover, the measures of minimum, maximum, average, and the level of significance for the six traits were selected and the effect on the PY was presented. All the traits influencing the PY were considered as independent variables, and the PY as a dependent variable and, eventually, the best regression models or production models, including one-to-p variables (six variables in here), were specified by forward selection (abbreviated as maxr). The best regression model was selected based on the highest R 2 . Furthermore, for multiple alignments, the models were controlled by investigating the variance inflation factor (Soltani et al. 2000). Then, the best regression models of the six variables were selected in the step seven. The reason for selection this step was that by increasing the number of variables from one to six, the changes in R 2 remained significantly constant (R 2 = 0.64 ** ). Later, the mentioned equation was investigated and analyzed, and, by deriving the component correlation between the equation components, the positive and the negative relationship, and correlation of the components with each other were evaluated. Finally, the traits' specifications, in the form of average and best models, which can be placed in the PY regression model, entered the production model of six variables. In order to determine the PY model (production model), the relationships between all the variables were measured and the PY was evaluated using the regression method (Soltani et al. 2016). The final model was obtained through the controlled trialand-error method, which can quantify the effect of the PY limitations. The average PY was calculated by the model by putting the observed average variables (Xs) in the fields under study in the PY model. By placing the best observed value of the variables in the PY model, the maximum obtainable PY was calculated. The difference between these two variables has been considered PY changes. The