Individual tree mortality model for hybrid larch young plantations based on mixed effects

ObjectiveTo study the individual tree mortality of hybrid larch young plantations, using fixed intervals re-measured data and different methods to establish hybrid larch (Larix kaempferi × Larix olgensis) individual tree mortality model, this paper aims to provide the basis for the determination of sustainable management and promotion of hybrid larch.

MethodBased on the re-measured data of the 48 permanent sample plots from 2003 to 2015 in Jiangshanjiao Experimental Forest Farm in Heilongjiang Province of northeastern China, the Logistic model was used to predict the probability of individual tree mortality by the method of all sub-set and maximum likelihood estimation. Contingency table analysis, scatter plot of classification rate and threshold were used to determine the best threshold when the model estimated. Adding random parameter at plot level aims to make the mixed model with maximum likelihood estimation based on adaptive quadrature. Selected criteria of models were Akaike information criterion (AIC), Bayesian information criterion (BIC) and negative double of logarithmic likelihood ratio. The model test criteria was the absolute average deviation (Bias). The ROC curve and histogram was used for prediction of mortality rate of the models and actual mortality rate was drawn to examine the performance of model.

ResultThe results showed that the best fitting result appeared when the model contained the combination of individual level (DBH, DBH²), stand level (stand basal area, BA) and competition level (deformation of basal area of the trees greater than the subject tree, BALD). The mortality of the hybrid larch occurred when the DBH class was small, and competition was more intensive. The probability of individual mortality decreased with the increase of DBH, and increased with the increase of BALD and BA.The optimal threshold improved the prediction effect of the model. When the unstructured matrix was the variance-covariance structure, four random parameters of the mixed model had the best fitting result. The prediction of mortality rate of mixed model was closer to the actual mortality rate.

ConclusionMixed model is more effective to describe individual tree mortality for hybrid larch. Threshold analysis is an effective method to improve the prediction accuracy of the dichotomous model. Hybrid larch is fast-growing species and thinning treatment should be appropriately implemented to reduce the probability of tree mortality in young forests.