Individual-tree diameter growth model for fir plantations based on multi-level linear mixed effects models across southeast China

Abstract

An individual-tree diameter increment model is developed for fir plantations using a multi-level linear mixed effect model approach. The dataset came from National Forest Inventory plots. Stochastic variability is broken down among sites, blocks, plots, and within-tree components to account for repeated measurements and the hierarchical structure imposed by the sampling scheme. In addition, within-tree heteroscedasticity and correlation were taken into account. The dataset consisted of 583 plots, 62,831 trees, and 251,324 observations. The dataset was randomly split into ten parts and 80 % was used for initial model development while 20 % was used for model validation. Statistically significant predictors were total number of stems per hectare, the natural logarithm of initial diameter, basal area of trees larger than the subject tree, elevation, and thinning intensity. Both the fitting model and the validation dataset showed a substantial improvement compared with the classical approach widely used in forest management. The model was developed using autoregressive moving average [ARMA(1,1)] and constpower function covariance structures, and it performed better than the model developed using only random-intercept effects.