Objective Tree size, competition, tree species diversity and other factors in the forest stand affect the growth of forest trees, while the individual basal area increment(BAI) is often used to describe tree growth. A random forest model is established for the individual tree BAI of the main tree species in the mixed forest of Changbai Mountain to study and quantify the environmental mechanisms that affect tree growth, to provide a theoretical basis for growth projections.

Method Data were investigated for 24 consecutive years (1987—2010) in a fixed sample plot with a total sample size of 6 903 trees. Random forest algorithm was used to build individual tree BAI model with 11 independent variables as individual tree, competition factor, diversity, and climate for 6 main tree species in mixed forest. And 10-fold cross-validation was used to optimize hyper parameter mtry and evaluate these model.

Result (1) The coefficients of determination for the model of BAI for the six major species of Abies nephrolepis, Picea koraiensis, Pinus koraiensis, Tilia amurensis, Betula costata, and Betula platyphylla were 0.663, 0.683, 0.695, 0.459, 0.384, and 0.568, respectively. (2) The basal areas(BA) of individual tree was the most important independent variable, and had a strong positive effect on the growth of trees. Competitive factors and tree size diversity were the main factors affecting tree growth, and individual BAI decreased with increasing competitive factors and tree size diversity. (3) The effects of species diversity on tree growth were relatively limited, with increases in species diversity index and mingling degrees accelerating the growth of Abies nephrolepis, Picea koraiensis and Pinus koraiensis to some extent; whereas the effects of climatic factors on tree growth were relatively small.

Conclusion Tree growth is largely depends on its own growth potential, which is mainly inhibited by competition and tree size heterogeneity in the external environment, while increased species diversity also promotes the growth of dominant species within the forest to some extent. The random forest model can well quantify and express the complex relationship between the variables and the BAI of individual tree. It can be used as a tool for forest management practice and provide a new method for forest growth and yield prediction.