运用机载激光雷达数据和立地质量分级对亚热带森林蓄积量遥感反演  

Improving Lidar-based Stand Volume Estimation of Subtropical Forest with Site Conditions Classification

作　　者：张国飞[1] 岳彩荣[1] 王雷光[1] 李春干 Zhang Guofei;Yue Cairong;Wang Leiguang;Li Chungan(Southwest Forestry University,Kunming 650224,P.R.China;Guangxi University)

机构地区：[1]西南林业大学,昆明650224 [2]广西大学

出　　处：《东北林业大学学报》2020年第7期60-65,共6页Journal of Northeast Forestry University

基　　金：国家自然科学基金项目(31860182、31260156、41571372);云南省教育厅科学研究基金项目(2018JS330);云南省中青年学术和技术带头人后备人才项目(2018HB026)。

摘　　要：为了探讨不同的立地条件下,应用机载激光雷达(Lidar)数据建立的林分蓄积量模型的反演精度。以高峰林场2016年森林资源二类调查小班数据为研究对象,运用随机森林对Lidar分位数高度、水平结构参数等变量重要性排序,构建蓄积量估测基础模型(不考虑立地质量分级)和优化模型(考虑立地质量分级)。结果表明:105块样地中有23块低立地生产力样地(LSP),有82块高立地生产力样地(HSP);在不同立地质量分级样地中,对树高和胸径进行logistic最小二乘回归曲线,胸径-树高之间存在着显著差异;基础模型和优化-A_B模型估测的蓄积量在总体样地和HSP上没有显著差异(P>0.05),但在LSP上差异极显著(P<0.001);优化模型估测的蓄积量和实测蓄积量在两种立地质量的样地中差异不显著(P>0.05);基础模型估测的蓄积量和实测蓄积量在HSP没有显著差异(P>0.05),但在LSP差异极显著(P<0.001)。由分位高度(P H90)和实测蓄积量生成(含HSP和LSP)的两个伪模型曲线表明,在HPS中,实测蓄积量与P H90呈近似线性关系,在LSP中,蓄积量与P H90呈非线性关系;分位数高度(P H)、郁闭度与森林蓄积量表现出较强的相关性,并且提供了森林水平结构和垂直的变化。森林蓄积量估测最优模型的平均R 2值大于0.8,模型的可靠性和鲁棒性为亚热带森林蓄积量估测提供了可能。The experiment was conducted to explore the inversion accuracy of the forest stand volume estimation model based on airborne Lidar data under different site conditions.With the data of the second class survey of forest resources in the research area of 2016,the importance order of variables such as Lidar percentage height and horizontal structure parameters was conducted by using random forest model(RF),and the basic models(without site quality classification)and the optimization models(considering site quality classification)of stand volume estimation were constructed.Among the 105 plots,there were 23 plots of low site productivity(LSP)and 82 plots of high site productivity(HSP).By logistic least square regression of the relationship between height and DBH,there were significant differences between DBH and tree height in HSP and LSP.There was no significant difference between the estimated stand volumes by basic models and optimal models in the total samples(TS)and HSP(P>0.05),but there was significant difference in LSP(P<0.001).There was no significant difference between the measured stand volumes and the estimated stand volumes by optimal models in HSP and LSP(P>0.05).There was no significant difference between the measured stand volumes and the estimated stand volumes by basic models in HSP(P>0.05),but there is significant difference in LSP(P<0.001).The two pseudo-model curves generated by P H90 and the measured stand volumes(including HSP and LSP)show that,in HPS,the measured stand volumes have an approximate linear relationship with P H,while in LSP,the measured stand volumes have a nonlinear relationship with P H90.The correlation between(PH),Canopy closure(CC)and forest stand volume was strong,which provided the change of forest horizontal structure and vertical.The average R 2 value of the models was greater than 0.8,showing the reliability and robustness of the model provided a possibility for the estimation of subtropical forest stock.

关 键 词：森林蓄积量 激光雷达 立地质量 亚热带森林 

分 类 号：S757.3[农业科学—森林经理学] P237[农业科学—林学]