论文部分内容阅读
【目的】利用广义线性混合模型模拟人工林红松二级枝条分布数量,建立二级枝条分布数量广义线性混合模型,并选出最优模型。【方法】基于黑龙江省孟家岗林场人工林65棵红松955个一级枝上的二级枝条数量,通过传统Poisson回归方法选出模拟精度最高的基础模型,考虑树木效应与树木内枝条观测间的相关性,构建二级枝条分布数量广义线性混合模型,并利用R2、标准误差、平均绝对误差、相对平均绝对误差和Vuong检验对收敛模型进行比较。【结果】考虑树木效应的混合模型模拟精度均高于传统回归模型,最终将含有截距、lnR_(DINC)(R_(DINC)为着枝深度)、R_(DINC)~2和C_L(冠长)4个随机效应参数以及自相关矩阵AR(1)的广义线性混合模型选为二级枝条分布数量最优预测模型。在模型固定效应参数估计结果中,lnR_(DINC)、CL和DBH(胸径)前的系数为正值,R_(DINC)~2、H_(DR)(高径比)前的系数为负值,树冠内二级枝条分布数量存在最大值。最优模型的R~2为0.896 1,标准误差为5.15,平均绝对误差为3.83,相对平均绝对误差为23.25%。【结论】广义线性混合模型不仅提高了模型的拟合精度,在反映总体二级枝条分布数量变化趋势的同时,还可以反映每棵树木之间的差异。
【Objective】 The generalized linear mixed model was used to simulate the distribution of Pinus koraiensis second-order branches in the plantations, and the general linear mixed model of the second-order branches distribution was established and the optimal model was selected. 【Method】 Based on the number of secondary branches on 955 primary branches of 65 Korean pine plants in Mengjiagang Forest Farm, Heilongjiang Province, the basic model with the highest simulation accuracy was selected by the traditional Poisson regression method. Considering the relationship between the tree effect and the observation of branches in trees The generalized linear mixed model of secondary branches was constructed and the convergence models were compared by using R2, standard error, mean absolute error, relative average absolute error and Vuong test. 【Result】 The results show that the simulation accuracy of mixed model considering the tree effect is higher than that of the traditional regression model and will eventually include the intercept, lnR_ (DINC) (R_ (DINC) is the shoot depth), R_ (DINC) ~ 2 and C_L ) Four random effects parameters and the generalized linear mixed model of autocorrelation matrix AR (1) were selected as the optimal prediction model for the number of secondary branches. The coefficients before lnR_ (DINC), CL and DBH (DBH) were positive, and the coefficients before R_ (DINC) ~ 2 and H_ (DR) There is a maximum number of secondary branches in the crown. The optimal model has a R ~ 2 of 0.896 1, a standard error of 5.15, an average absolute error of 3.83 and a relative average absolute error of 23.25%. 【CONCLUSION】 The generalized linear mixed model not only improves the fitting precision of the model, but also reflects the difference of each tree while reflecting the changing trend of the total secondary branch distribution.