摘要
建立江西杉木人工林基于非线性混合效应方法的林分优势木平均高和断面积模型以及基于对数形式线性混合效应蓄积模型的联立方程组,利用验证数据与传统模型回归方法的模拟结果进行比较分析。结果表明,优势木平均高是联立方程组最基本的组成部分。通过考虑优势木平均高和林分断面积模型中参数的随机效应以及3个因变量间的相关性,则蓄积模型中参数的随机效应可以忽略。优势木平均高决定着林分断面积预测的准确性,而优势木平均高和林分断面积又是预测蓄积的主要误差来源。基于混合效应模型方法的模拟结果明显好于传统回归估计方法。进行预测时,通过解释联立方程组中因变量相互间的相关性,利用已被观测的变量能够提高未观测变量的估计精度。
A simultaneously interdependent model system in which stand characteristics such as dominant height,basal area,and total volume are included is very common in forest growth and yield model.Permanent plots are usually the main source of data.Such data include two basic sources of errors.One is within-plot error,the other is the variation from plot to plot.Within-plot error is usually modeled by a reasonable variance function that accounts for within-plot heteroscedasticity and correlations,between-plot variation can be described by random effects of parameter to be varied from plot to plot in the models.A simultaneous system containing components of dominant height and basal area growth model based on nonlinear mixed effects,a log-linear total volume model is developed for fir plantation.In the end the result of mixed effect model is compared with that of ordinary regression analysis with validation data.The result showed that dominant height was the fundamental component in the three-component system.With random effects for dominant height and basal area and contemporaneous correlation of three components considered,random effects of total volume model were proved to be unnecessary.Dominant height itself dominated the precision of basal area prediction,dominant height and basal area were main source of error for total volume prediction.The fitted effects of simultaneous equation system based on mixed model approach were better than that of based on ordinary regression analysis.The observed components could improve the prediction of the unobserved components by accounting for the contemporaneous correlation among the components in prediction.
出处
《林业科学》
EI
CAS
CSCD
北大核心
2012年第6期80-88,共9页
Scientia Silvae Sinicae
基金
国家自然科学基金项目"基于混合模型的森林生长模拟方法研究"(31170589)
关键词
联立方程组
全林蓄积
混合效应模型
预测
simultaneous equation system
total volume
mixed effect model
prediction
