基于分位数回归模型的阔叶次生林单木胸高断面积生长研究  

Study on growth of basal area increment of individual trees in broad-leaved secondary forest based on quantile regression model

作　　者：宁金魁[1,2] 王光玉 许诺[2] 黄锦程 欧阳勋志 刘洪生[3] 臧颢 NING Jinkui;WANG Guangyu;XU Nuo;HUANG Jincheng;OUYANG Xunzhi;LIU Hongsheng;ZANG Hao(College of Forestry,Jiangxi Agricultural University,Nanchang 330045,Jiangxi,China;Faculty of Forestry,University of British Columbia,Vancouver BC,V6T 1Z4,BC,Canada;Forestry Bureau of Chongyi County,Ganzhou 341300,Jiangxi,China)

机构地区：[1]江西农业大学林学院,江西南昌330045 [2]加拿大不列颠哥伦比亚大学林学院,加拿大温哥华BC V6T 1Z4 [3]江西省崇义县林业局,江西赣州341300

出　　处：《中南林业科技大学学报》2023年第12期24-34,共11页Journal of Central South University of Forestry & Technology

基　　金：国家自然科学基金项目(31960313);江西省林业局科技创新专项(201901);国家留学基金项目(202008360244)。

摘　　要：【目的】阔叶次生林是一种复杂的异龄复层林,林木的胸高断面积生长量决定了林分的林分质量和碳汇能力。本研究旨在探讨阔叶复层次生林胸高断面积年均生长量(BAI)受个体大小、试验林、林层等因素的影响。【方法】以赣南地区3个林分结构各异的阔叶次生林试验林为研究对象,根据树高及树种组成的差异,分别将3个试验林划分3个林层和6种经营类型,采用分位数回归模型,设19个分位数(τ∈{0.05,0.10,0.15,…,0.90,0.95}),建立了胸高断面积年均生长量与胸径(DBH_2021)、高径比(H_D_ratio)、林层、试验林类型、经营类型等之间的非线性回归关系,并采用了AIC、Rτ2、MAD、MD等指标评价各分位数回归模型。【结果】1)3个试验林之间的BAI差异显著,但同一试验林内作业样地和对照样地的BAI差异在统计上表现不显著;2)AIC值最低时的分位数回归模型为log(BAI)τ~log(DBH_2021)+DBH_2021+H_D_ratio,分位数τ=0.45时,BAI的分位数模型最优,此时BAI的最佳分位数回归模型的拟合系数Rτ2为0.5353,考虑林层、试验林类型、经营类型时,拟合系数又分别提高了10%,分别为0.6383(考虑林层、试验林类型)、0.6389(考虑林层、经营类型);3)log(DBH_2021)、H_D_ratio、林层、发育阶段、经营类型对BAI的影响都是正效应。【结论】基于胸径、高径比、林层、试验林类型、经营类型构建的单木胸高断面积生长分位数模型可为阔叶次生林的提质增效技术提供量化依据。【Objective】The broad-leaved secondary forest is a complex multi-layered mixed-aged forests,and the growth of the basal area determines the stand quality and carbon sink capacity of the forest.The purpose of this study was to investigate the influence of the annual basal area increment(BAI)of individuals in the broad-leaved stratified forests on individual sizes,experimental forests,and forest vertical levels(forest stories).【Method】Three different experimental forests of the broad-leaved secondary forest in southern Jiangxi in China were taken as the research objects.According to the differences in individual-tree height and stand species composition,three experimental forests were divided into three forest stories and six different silvicultural types.The article was adopted the method of quantile regression model that was set 19 quantiles(τ∈{0.05,0.10,0.15,...,0.90,0.95}),and established the nonlinear regression relationship between the annual basal area increment(BAI)and diameter at breast height(DBH_2021),height-to-diameter ratio(H_D_ratio),forest stories,experimental forests,silvicultural types,etc.,and used some indicator,such as AIC,Rτ2,MAD and MD,to evaluate each quantile time regression model.【Result】1)There were significant differences in BAI between three experimental forests,but the difference in BAI between the silvicultural plots and the unsilvicultural plots in the same experimental forest was not statistically significant;2)The quantile regression model when the AIC value was the lowest was log(BAI)τ-log(DBH_2021)+DBH_2021+H_D_ratio;when the quantileτ=0.45,the quantile regression model of BAI was most optimal,the fitting coefficient Rτ2 of the best quantile regression model of BAI was 0.5353.When forest stories,experimental forest types,and silvicultural types were considered,the fitting coefficient increased respectively 10%,respectively 0.6383(considering forest stories,experimental forest types),0.6389(considering forest stories,silvicultural types);3)log(DBH_2021),H_D_ratio,

关 键 词：分位数回归模型 胸高断面积生长量 阔叶次生林 林层 个体大小 经营类型 

分 类 号：S791.222[农业科学—林木遗传育种]