摘要
以黑龙江省凉水国家级自然保护区阔叶红松林为研究对象,基于14块标准地中红松(Pinus koraiensis)、红皮云杉(Picea koraiensis)、冷杉(Abies nephrolepis)、色木槭(Acer mono)等4个主要树种448棵幼树的连年树高生长量数据,从5个备选模型中分别为4个树种选择最优树高生长模型作为基础模型,并采用R语言对4个树种分别构建了样地水平和样木水平的幼树树高生长混合效应模型并进行独立检验。结果表明:4个树种幼树树高随年龄增加而增大,且总体趋势较为明显,在幼树阶段其树高的年增长量差别不大。柯列尔方程或幂函数能较好地拟合4个树种幼树的树高生长过程。将柯列尔方程作为最优基础模型并建立混合效应模型,通过比较赤池信息准则(A_(I,C))、贝叶斯信息准则(B_(I,C))以及对数似然值(L_(L))得出结果,样木水平混合效应模型优于样地水平混合效应模型。与基础模型相比,4个树种样木水平混合模型的调整后确定系数(R_(adj)^(2))提高4.16%~20.31%(平均11.13%);均方根误差(R_(M,S,E))降幅较大,为83.92%~89.66%(平均86.32%)。模型检验结果显示,相对于基础模型,混合效应模型的平均绝对误差(M_(A,E))减小6.8~13.4 cm,平均减小9.78 cm;平均预测误差百分比(M_(P,S,E))降幅较大,为14.4%~49.5%(平均30.85%)。说明样木水平混合模型既能提高模型的拟合效果,又能提高模型的预测能力,证明了树高生长的差异主要源于随机效应(即幼树个体差异)。本研究所构建的幼树树高生长模型可以为阔叶红松林的生长模拟和森林演替研究提供基础。
According to 14 sample plots measured in broad-leaved Korean pine forest of Heilongjiang Liangshui National Nature Reserve,four main tree species(Pinus koraiensis,Picea koraiensis,Abies neprolepis and Acer mono)were chosen to analyze.With the height growth data of 448 saplings of main tree species measured in these plots,the best model of tree height growth was selected as the basic model for four main regeneration tree species in five alternative models.The R Languagewas used to develop the height growth mixed-effect model of saplings for four tree species at plot level and at tree level respectively.The results showed that the height of the saplings increased with tree age and the general trend was obvious.The annual growth of the height of saplings was little difference.Height growth process of saplings of four tree species can be well fitted by Rолясрfunction or power function.The Rолясрfunction was selected as the optimum basic model for four tree species and mixed effect models were established on it.By comparing AIC,BIC and LL,the fitting results showed that tree-level mixed effect models were better than plot-level mixed effect models.Comparing with basic models,the adjusted coefficient of determination(R_(adj)^(2))of the four tree species was increased by 4.16%-20.31%(11.13%on average),and the root mean square error(R M,S,E)was decreased by 83.92%-89.66%(86.32%on average).Model test results showed that,comparing with basic models,the average absolute error(M A,E)of the mixed effect models decreased by 6.8-13.4 cm and 9.78 cm on average,and the average prediction error percentage(M P,S,E)decreased by 14.4%-49.5%(30.85%on average).Tree-level mixed effect models can not only improve the fitting effect of the model,but also improve the prediction ability of the models.It is proved that the difference of height growth is mainly due to the random effect(individual difference of saplings).The height growth models of saplings constructed in this study can provide the basis for the growth simulation and
作者
李武赫
董利虎
李凤日
Li Wuhe;Dong Lihu;Li Fengri(Northeast Forestry University,Harbin 150040,P.R.China)
出处
《东北林业大学学报》
CAS
CSCD
北大核心
2021年第3期1-8,共8页
Journal of Northeast Forestry University
基金
黑龙江省森林可持续经营试验示范区建设项目(201522)。
关键词
阔叶红松林
幼树
天然更新
树高生长
混合效应模型
Broad-leaved Korean pine forest
Saplings
Natural regeneration
Tree height growth
Mixed-effect model
