[Branch Growth of Korean Pine Plantation Based on Nonlinear Mixed Model].

Chun-Hong Wang,Feng-Ri Li,Wei-Wei Jia,Li-Hu Dong
DOI: https://doi.org/10.13287/j.1001-9332.2013.0430
2013-01-01
Abstract:Based on the branch analysis data from 36 sample trees in a Korean pine plantation in Mengjiagang Forest Farm of Heilongjiang Province, Northeast China, and by using Mitcherlich and Richards equations as the models of branch diameter and branch length growth, respectively, the effects of sampling plot and sample tree were investigated, and the nonlinear mixed models of branch diameter and branch length growth were established by the PROC NLMIXED procedure of SAS software. The evaluation statistics such as Akaike information criterion (AIC), Bayesian information criterion (BIC), -2Log likelihood, and likelihood ratio test (LRT) were used to compare the prediction precisions of the models. When considering plot effect, and taking alpha1 and alpha3 and beta1 and beta3 as the random parameters, respectively, the models of branch diameter and branch length growth had the best performance. When considering tree effect, and taking alpha2 and alpha3 and beta2 and beta3 as the random parameters, respectively, the models of branch diameter and branch length growth had the best performance. The nonlinear mixed model could not only reflect the mean variation of branch growth, but also show the differences among the individual trees. No matter considering plot effect or tree effect, the fitting precision of the nonlinear mixed model was better than that of the ordinary regression analysis model. Moreover, the fitting precision of the nonlinear mixed model was better when considering tree effect than considering plot effect.
What problem does this paper attempt to address?