Based On Mixed-Effects Model And Empirical Best Linear Unbiased Predictor Predicting Growth Profile Of Chinese Fir

The stand growth model is an important part of stand growth and yield prediction model.and the predicted reliability will directly affect the forest management.This study was done in Jiangle state-owned forest in Fujian Province,using data of sample plot and stem analysis of Cunninghamia lanceolate,the least square method,Bayesian method and nonlinear mixed effects model were used to construct height growth model,stand basal area growth model and individual tree volume growth model.Studying the characteristics of Empirical Best Linear Unbiased Predictor(EBLUP).Three methods were compared for their application in construction of the model.(1)Height growth model:Based on the data of 30 sample trees from 15 permanent plots of Chinese fir,Firstly,5 growth profile equations was fitted by using the least square method,afterwards,using the best model format,along with stem analysis data,nonlinear mixed model was constructed.We use the R for model fitting.Via changing number of parameters,the model which had the minimum value of AIC,BIC and the maximum value of Loglik was selected as the best fitted one.Using the selected one to predict growth profile of height and studying the characteristics of Empirical Best Linear Unbiased Predictor(EBLUP).Fitting results showed the simulation's precision of Weibull's including three random effect parameters was maximal.In the analysis of prediction,prediction accuracy decreased as age interval of observations extended with the same number of previous observations.MSE decreased as the number of previous observations increased.EBLUP prediction could fully predict individual growth process,given that there were multiple previous observations with long-enough age intervals.(2)Stand basal area growth model:Two different forms of Schumacher equation were selected to simulate the fir basal area growth model.Studying the best function as the base model among two growth profile equations.Select the mixed model with the minimum value of AIC,BIC and the maximum value of Loglik as the best fitted one.Meantime,Bayesian method were used to estimate the stand basal area growth model.The results showed the simulation's precision of Schumacher's second equation including three random effect parameters of a0,a2 and a5 was maximal.the precision of Bayesian method was a little better than that of the classical method,Bayesian method was better than classical one about the simulation's precision.(3)Individual tree volume growth model:Based on the data of 30 sample trees.At first 5 growth profile equations was fitted by using the least square method,therefore,the best model format was selected,along with stem analysis data,nonlinear mixed model was constructed for individual tree volume growth.Using the format of selected equation and stem analysis data,nonlinear mixed model was fitted.Via changing number of parameters,the best fitted model with the minimum value of AIC,BIC and the maximum value of Loglik was selected.At the same time,Bayesian method were used to estimate the individual tree volume growth model.The results showed the simulation's precision of Korf's including two random effect parameters of ?2 and ?3 was maximal.However,the precision of Bayesian method was nearly equal to that of the classical method.(4)As for the three methods,the nonlinear mixed effects model significantly improved the precision of three models,but the mixed model was a little complicated.Because of the parameter using Bayesian method are treated as random variables,the precision of Bayesian method was a little better than the classical method,and the informative prior was slightly better than non-informative prior.Traditional model was easy to simulated and operated,but the model reliability was low.