Study On Individual Tree Growth Model For Chinese Fir Plantation In Fujian

Individual tree growth model is an important part of stand growth and yield prediction model. This article took Cunninghamia lanceolata plantation located in national forest farm of Jiangle in FuJian province as object. With the data of sample plot and stem analysis, traditional nonlinear model, BP neural network and nonlinear mixed effects model were used to construct height-diameter model and the relation model of diameter-age, height-age and volume-age respectively. Three methods were compared for their application in construction of the individual tree growth model. The result showed that:1、In the process of constructing model of height-diameter, diameter-age, height-age and volume-age, based on related indicators of fitting and validation, three methods all performed well. And nonlinear mixed effects model was best, BP neural network model was second.2、When using height-diameter model to predict tree height, the prediction accuracy of traditional model was better than that of the two other models when diameter at breast height was less than 10 cm. Therefore, the advise that using traditional method to construct height-diameter equation for the young forest which diameter at breast height is less than 10 cm was proposed. In other conditions, the two other models should be considered.3、When simulating five theoretical growth equations, the Richards model was selected as the best growth model to model diameter-age and height-age relationship, the Gompertz model was elected as the best growth model to model volume-age relationship. And two models were took as basic model in constructing mixed models. The effect of mixed model was much better than that of others, and the effect of BP neural network model were very similar to traditional nonlinear model. Therefore, the advise that using mixed model to construct the relation model of diameter-age, height-age and volume-age was proposed.4、The traditional model was easy to established and operated, but the related accuracy was low. BP neural network could simulate the relationship that the mathematical function couldn’t express, however, its optimal model was not easy to determine and it couldn’t show its advantage when input was few. The nonlinear mixed effects model developed with consideration of plot effect or individual tree effect significantly improved the precision of four models. On the other hand, the structure of mixed model was complicated, and the size and number of subsample was hard to unify when testing the established model.