Modelling the production of pine seeds and nuts in Pinus koraiensis plantation

We collected data on fruit and tree factors in Pinus koraiensis plantations in the Linkou Forestry Bureau of Heilongjiang Province. Pine cones were divided into three grades based on fresh weight. We analyzed the correlations between pine cones, pine seeds, and pine nuts, and constructed the foundational models for pine seed weight, pine nut quantity, and pine nut weight. Then, we introduced the effects of cone grade and random effects of sampling sites into the foundational models, and selected the optimal mixed-effects model by comparing the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC). The results showed that the weights of pine seeds and nuts were significantly negatively correlated with the number of empty seeds and the diameter at the base of pine cones, and were significantly positively correlated with other traits. The mixed-effects models that introduced cone grade and random effects of sampling sites had better fitness than the foundational models. Among them, the mixed-effects model that included the site effect in the optimal foundational model showed better fitness than the model that included the cone grade effect. Compared to the optimal foundational models, the R² values of the optimal mixed-effects models for pine seed weight, pine nut quantity, and pine nut weight with the inclusion of the site effects was improved by 20.8%, 29.5%, and 32.8%, respectively. The optimal mixed-effects models for pine seed weight, pine nut quantity, and pine nut weight had prediction accuracies (F_P) of 98.3%, 97.9%, and 97.8%, respectively. All those values surpassed the predictive accuracy of the optimal foundational model. Our results indicated that the mixed-effects models could better predict seed yield of P. koraiensis.