混合模型在福建省粗叶榕生物量模型中的应用

摘要

The objective of this study was to establish the biomass mixed-effect models for Ficus simplicissima and to estimate shrub biomass accurately.Models with the highest fitting precision were selected as basic ones from six traditional models.Biomass of different parts (aboveground,underground and total biomass) were taken as dependent variables,and ground diameter,plant height,crown width were independent variables.The data were extracted from 114 groups of F.simplicissima shrubs in four types of forests (pure Chinese fir,forest of Pinus massoniana and Cunningghamia lanceolata,pure Pinus massoniana,pure Phyllostachys pubescens) occurring in Sanming,Fujian Province,China.The linear and nonlinear mixedeffect models were used to establish the F.simnplicissima biomass models with the forest types as randomeffect.Meanwhile,the power function,exponential function or the constant plus power function were introduced,respectively to remove the heteroscedasticity of the data.AIC,BIC and negative double logarithmic likelihood value were calculated to make the model accuracy comparison.The absolute average error,root mean square error and adjusted coefficient of determination were used to test the model.The results showed that the nonlinear biomass model with crown had the higher fitting precision than the linear model.Compared to the basic model,the precision of the nonlinear mixed-effect (NLME) model had improved significantly.Validation data showed the coefficients of determination of the ground biomass NLME model,underground biomass NLME mode and total biomass NLME model were improved by 12.17％,21.01％,and 20.24％,respectively compared to the basic nonlinear models,which could estimate the shrub biomass accurately.The shrub biomass mixed-effect models,considering heteroscedasticity structure could predict the F.simplicissima biomass accurately.%为建立粗叶榕灌木生物量的混合模型,精确估算灌木生物量,以福建省将乐县4种林分类型(杉木纯林、杉木马尾松混交林、马尾松纯林、毛竹林)的114组粗叶榕灌木生物量数据为基础,以粗叶榕各部分生物量(地上、地下、总生物量)为因变量,以地径、株高、冠幅等为自变量,从6种常用的灌木生物量模型中选择拟合精度最高的线性及非线性模型作为基础模型.以林分类型为随机效应,采用混合效应模型方法,建立粗叶榕的线性及非线性混合模型.利用幂函数、指数函数、常数加幂函数3种结构消除数据异方差对模型精度的影响.采用AIC、BIC和负2倍的对数似然值对模型进行精度比较,并用绝对平均误差、均方根误差和调整后的决定系数对模型进行检验.结果表明,考虑冠幅建立的非线性生物量模型拟合精度较高;以幂函数作为异方差结构建立的非线性混合模型在精度上有显著提高,检验数据显示地上、地下、总生物量模型的决定系数分别提高12.17％、21.01％、20.24％.利用混合模型,并考虑异方差结构建立的灌木生物量模型可以精确地预估粗叶榕的生物量.