兴安落叶松天然林3个主要树种单木断面积生长模型  

Individual-tree basal area growth models of three main tree species in Larix gmelinii natural forests

作　　者：王建军 陈科屹 何友均[1] 吕延杰 许单云 何亚婷[1] 谢和生[1] 卢佶 WANG Jianjun;CHEN Keyi;HE Youjun;LÜYanjie;XU Danyun;HE Yating;XIE Hesheng;LU Ji(Research Institute of Forestry Policy and Information,Chinese Academy of Forestry,Beijing 100091,China;East China Academy of Inventory and Planning of NFGA,Hangzhou 310019,Zhejiang,China)

机构地区：[1]中国林业科学研究院林业科技信息研究所,北京100091 [2]国家林业和草原局华东调查规划院,浙江杭州310019

出　　处：《中南林业科技大学学报》2024年第9期1-10,共10页Journal of Central South University of Forestry & Technology

基　　金：国家自然科学基金项目(31570633)。

摘　　要：【目的】构建兴安落叶松Larix gmelinii天然林3个主要树种组的断面积生长混合效应模型,提高断面积生长模型的预估精度,为兴安落叶松天然林的林分生长动态研究提供理论参考。【方法】以大兴安岭林区兴安落叶松天然林的主要树种组(兴安落叶松、白桦、栎类)为研究对象,采用4期森林资源连续清查的626块复测样地数据,首先使用多元逐步回归方法建立单木断面积生长模型,其次在最优基础模型的基础上,引入样地随机效应,建立单木断面积生长混合效应模型,最后对模型进行检验。【结果】期初胸径(D)的倒数(1/D)、每公顷株数(N)、大于对象木的胸高断面积和(B_(L))、对象木胸径与林分算数平均胸径之比(B_(L)/D)、海拔(E)以及坡度和坡向的组合(T_(c))对兴安落叶松天然林单木断面积的生长有显著影响(P<0.05)。考虑样地随机效应的混合效应模型的赤池信息准则(AIC)、贝叶斯信息准则(BIC)明显降低,对数似然值(LogLik)明显增大。引入异方差函数和自相关结构后,提高了混合效应模型预估的精度;混合效应模型的平均绝对误差(Bias)和均方根误差(RMSE)均比基础模型的低,兴安落叶松、白桦、栎类最优混合效应模型的R^(2)分别由0.435提高到0.658、由0.354提高到0.489、由0.307提高到0.379。【结论】与基础模型相比,考虑层次结构的混合效应模型能显著地改善模型的拟合效果,为内蒙古大兴安岭地区兴安落叶松天然林的林分生长动态预测提供了具有生物学意义和统计学可靠性的断面积生长模型。【Objective】Developing individual-tree basal area growth models for Larix gmelinii,Betula platyphylla,Quercus spp.based on mixed-effects model method,in order to improve the prediction accuracy of models and provide some reference of forest dynamics of L.gmelinii mixed-species natural forests.【Method】Based on the data from 626 sample plots of L.gmelinii mixed-species natural forests(Chinese National Forest Inventory in 2003,2008,2013 and 2018)in the Daxing’an mountains,Inner Mongolia Autonomous Region,study subjects were divided into three main species groups,i.e.L.gmelinii,B.platyphylla and Quercus spp..The individualtree basal area growth models were constructed by using the multiple stepped-regression method.Then,the optimal based model was selected,and based on that,the random effects of within-plot were introduced to mixed-effects model for basal area prediction.Finally,the fitting effect was compared with the based model.【Result】The reciprocal transformation of initial diameter at breast height(1/D),the number of trees per hectare(N),the sum of the basal area in trees with DS larger than the subject tree’s D(B_(L)),the ratio of B_(L)and D(BL/D),elevation(E)and the combination of slope and aspect(T_(c))had significant effects on the growth of individual-tree basal area in the L.gmelinii mixed-species natural forests(P<0.05).The AIC and BIC of the mixed-effects models were obviously decreased,while LogLik were increased.Mixed-effects models,which were considering heteroscedasticity function and autocorrelation structure,showed an improved prediction accuracy with the Bias and RMSE were lower than that of based model.The R^(2) of optimal mixed-effects model of L.gmelinii,B.platyphylla,Quercus spp.increased from 0.435 to 0.658,from 0.354 to 0.489,and from 0.307 to 0.379,respectively.【Conclusion】Compared with the based models,the mixed-effects models considering the hierarchical structure show an improved performance,and provide biological significance and statistical reliability,which help t

关 键 词：断面积 异方差 自相关结构 混合效应模型 兴安落叶松天然林 

分 类 号：S758.55[农业科学—森林经理学]