文献类型：期刊文献

中文题名：基于两层次线性混合效应模型的杉木林单木胸径生长量模型

英文题名：Individual Tree Diameter Increment Model for Chinese Fir Plantation Based onTwo-Level Linear Mixed Effects Models

作者：李春明[1]

第一作者：李春明

机构：[1]中国林业科学研究院资源信息研究所,北京100091

年份：2012

卷号：48

期号：3

起止页码：66-73

中文期刊名：林业科学

外文期刊名：Scientia Silvae Sinicae

收录：CSTPCD;;Scopus;北大核心:【北大核心2011】；CSCD:【CSCD2011_2012】；

基金：国家自然科学基金"基于混合模型的森林生长模拟方法研究"(31170589)

语种：中文

中文关键词：单木胸径生长量;异方差;自相关结构;线性混合模型;杉木

外文关键词：individual diameter increment; heteroscedasticity; the correlation structure; linear mixed model; fir

分类号：S758.1

摘要：基于两层次线性混合效应模型方法,建立江西省杉木人工林单木胸径生长量模型。研究所用数据来自于长期观测的固定样地数据,数据库包括82个区域、365个样地、5416株树木共计16248条记录。为了解决不同区域及不同样地之间的差异,本文构建的混合模型分别考虑样地层次、区域层次及两层次的随机参数效应。针对数据存在的重复测量及嵌套结构特性,在模拟时选择合适的异方差和自相关模型矩阵来解决此类问题。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方法相比,采用混合效应模型方法后模型的模拟精度和验证精度均有提高。选择适合的异方差和自相关函数后,模型比只考虑参数的随机效应有更好的适应性,并体现出了混合效应模型的灵活性和准确性。
Based on a multilevel linear mixed model approach,an individual diameter increment model was developed for fir plantation trees growing in Jiangxi Province. The data set used in this study came from longterm permanent research plots. The database consists of total of 82 counties,365 plots,5 416 trees and 16 248 observations. The paper chose mixed effects models instead of regression analysis approach because it allows for proper treatment of error terms and correlation in a repeated measures analysis framework. The model was defined as a mixed linear model with parameter random effect of plot, area or plot and area simultaneous. In addition the heteroscedasticity and correlation was taking into account model. Mixed model calibration of diameter increment was carried out with the independent data using a different sample of complementary observations. The result showed total stand basal area,the diameter of target tree,the ratio of basal area of larger trees to target tree diameter,and altitude were found to be significant predictors. Both the fitting model and the calibrated model mean a substantial improvement compared with the classical approach widely used in forest management. After taking into account reasonable variance function of heteroscedasticity and correlation, the model shows better of goodness of fit than taking into account parameter random effects only. This type of modeling methodology showed to be flexible,precise and accurate.