基于修正Kozak方程的人工樟子松树冠轮廓预估模型  

Crown Profile Prediction Model for Pinus sylvestris var. mongolica Plantation Based on Modified Kozak Model

作　　者：高慧淋 董利虎[1] 李凤日[1] Gao Huilin;Dong Lihu;Li Fengri(Key Laboratory of Sustainable Forest Ecosystem Management,Ministry of Education School of Forestry,Northeast Forestry University,Harbin 150040;College of Forestry,Shenyang Agricultural University,Shenyang 110161)

机构地区：[1]东北林业大学林学院,森林生态系统可持续经营教育部重点实验室,哈尔滨150040 [2]沈阳农业大学林学院,沈阳110161

出　　处：《林业科学》2019年第8期84-94,共11页Scientia Silvae Sinicae

基　　金：国家自然科学基金项目(31570626);沈阳农业大学博士科研启动经费(880418014)

摘　　要：【目的】对Kozak方程进行修正,采用树木易测因子为预测变量,构建人工樟子松树冠外部轮廓预估模型,为研究树木生理和树木竞争提供依据,为模拟单木树冠表面积和树冠体积奠定基础。【方法】基于黑龙江省14块固定样地70株人工樟子松解析木907个最大枝条数据,以Kozak方程基本形式为基础并对其进行修正,选出构建人工樟子松树冠外部轮廓基础模型的最优模型形式。在最优模型基础上,建立分别考虑样地效应、样木效应及同时考虑样地和样木效应两水平的非线性混合效应模型。利用R软件的nlme软件包求解非线性混合效应模型参数,采用AIC、BIC、-2LL对混合效应模型中不同随机效应参数组合形式、不同随机效应矩阵、方差-协方差矩阵和方差函数进行比较,选出最优模型形式,并对人工樟子松外部轮廓随树木因子的变化规律进行探讨。以林分密度为哑变量,构建不同密度的人工樟子松树冠外部轮廓预估模型。【结果】人工樟子松树冠外部轮廓预估模型因子包含胸径(DBH)、冠长率(CR)和高径比(HD)。与基础模型相比,分别考虑样地效应、样木效应的混合模型能够显著提高模型拟合效果,外部轮廓模型差异主要来源于样木效应。以样木为单水平的混合效应模型中,a2、a6为随机参数,对角矩阵为方差-协方差矩阵形式,ARMA(1,1)为解释组内方差的矩阵,采用幂函数消除异方差的模型形式为最优模型。同时考虑样地和样木效应两水平混合模型的拟合效果较单水平混合模型有所提高。以两水平混合模型的固定效应部分模拟外部轮廓与树木因子之间的关系,在分别固定另外2个变量的情况下,树冠半径随着DBH、CR增大均逐渐增大,树冠上半部分半径随着HD增大而增大,下半部分半径随着HD增大而减小。外部轮廓拐点的变化范围为0.6250~0.9170,拐点平均位置为0.8413,随着林木在林分中被�【Objective】Based on the modification for the Kozak equation,the easily measurable individual tree variables were used as the independent variables to develop the maximum outer crown profile model for Pinus sylvestris var.mongolica plantation.The purposes were to provide reference to study tree physiology,competition and crown surface area and volume of individual tree.【Method】The data of 907 largest branches which were selected from 70 sample trees of Pinus sylvestris var.mongolica by established 14 permanent sample plots in Heilongjiang Province were used.The Kozak model was modified to be used as the basic model to describe the crown profile of Pinus sylvestris var.mongolica,and the best model was selected to construct the nonlinear mixed effect model.The plot effect,tree effect and the effect of plot and tree were considered separately and two level nonlinear mixed effect techniques were used to develop the crown profile model.The package of nlme from R package was used to estimate the parameters of the model.AIC,BIC and-2LL were used to evaluate and compare models with different random parameters combinations,different variance-covariance structure and variance function.The best model was selected and the relationships between crown profile and tree variables were also studied.The dummy variable crown profile model for Pinus sylvestris var.mongolica was developed using forest density as the indicator.【Result】The results showed that the finial crown profile model of Pinus sylvestris var.mongolica included DBH,CR and HD.Compared with the basic model,the nonlinear mixed effect model included only plot effect or tree effect,showed significantly increased the fitting effect and the difference mainly sourced from individual tree.For the model employing only sample tree as the 1 level mixed effect,the model taking a 2,a 6 as the random parameters,diagonal matrix as the variance-covariance matrix,ARMA(1,1)to explain the variance of within-group,power function to eliminate the heteroscedasticity was select

关 键 词：人工樟子松 树冠 外部轮廓 Kozak模型 非线性混合模型 

分 类 号：S757[农业科学—森林经理学]