随机截距效应在模拟杉木人工林单木胸径生长量中的应用

Using random intercept effects in DBH increment model of individual trees for fir plantation

以江西省杉木人工林4次调查数据为例,建立单木胸径生长量随机截距效应混合模型,分别考虑样地层次、区域层次及多层次的随机截距效应,并在考虑多层次效应时考虑了存在的异方差和自相关问题,最后利用独立的抽样验证数据对模拟结果进行验证。结果表明:林分断面积、对象木胸径、林分内大于对象木的断面积之和与对象木胸径的比值以及海拔对单木胸径生长量有显著影响;与林业中常用的传统最小二乘法相比,采用混合效应模型方法后模型的模拟效果和验证精度均有提高;选择适合的异方差和自相关函数后,模拟结果比只考虑随机截距效应有更好的适应性;实际应用中考虑多层效应要比单层效应模拟效果好。

In the paper, we developed a mixed model of random intercept effects for individual DBH increment, taking the example of four survey data of fir plantations in Jiangxi Province, eastern China. The model was defined as a linear mixed model with intercept effects of plot, area or plot and area simultaneous. The heteroscedasticity and autocorrelation matrix were added to the model when taking into account multilevel effects. In the end, mixed model calibration of DBH increment was carried out using the independent sampling. The results showed that total stand basal area, DBH of target tree, the ratio of basal area of larger trees to target tree DBH, and altitude were found to be significant effect on DBH increment. Both the simulation results and verified accuracy of the model indicated a substantial improvement compared with the conventional approach widely used in forest management. After adding to a reasonable variance function of heteroscedasticity and autocorrelation, the model showed better goodness of fit than taking into account random intercept effects only. The goodness of multilevel effects was better than that of individual-level effect in forest actual applied process.