洪奕丰,张守攻,陈伟,陈东升,项伟波,庞勇.基于机载激光雷达的落叶松组分生物量反演[J].林业科学研究,2019,32(5):83-90
基于机载激光雷达的落叶松组分生物量反演
Inversion of Biomass Components for Larix olgensis Plantation Using Airborne LiDAR
投稿时间:2019-01-20  修订日期:2019-04-16
DOI:10.13275/j.cnki.lykxyj.2019.05.011
中文关键词:  生物量反演  机载激光雷达  组分生物量  落叶松
英文关键词:biomass inversion  airborne LiDAR  biomass component  Larix olgensis
基金项目:国家自然基金重点项目(31430017)
作者单位E-mail
洪奕丰 中国林业科学研究院林业研究所 国家林业和草原局林木培育重点实验室, 北京 100091
国家林业和草原局华东调查规划设计院, 浙江 杭州 310019 
 
张守攻 中国林业科学研究院林业研究所 国家林业和草原局林木培育重点实验室, 北京 100091  
陈伟 国家林业和草原局华东调查规划设计院, 浙江 杭州 310019  
陈东升 中国林业科学研究院林业研究所 国家林业和草原局林木培育重点实验室, 北京 100091  
项伟波 中国林业科学研究院林业研究所 国家林业和草原局林木培育重点实验室, 北京 100091  
庞勇 中国林业科学研究院资源信息研究所, 北京 100091 pangy@ifrit.ac.cn 
摘要点击次数: 102
全文下载次数: 86
中文摘要:
      [目的]构建基于机载LiDAR的落叶松组分生物量反演模型,讨论不同方法对模型构建的影响。[方法]以地面实测样地数据和同步获取的机载LiDAR点云数据为数据源,分别采用多元线性回归(MLR)和随机森林(RF)方法,估测了长白落叶松的组分生物量,利用"刀切法"评价了模型的泛化能力。[结果]表明:(1)MLR筛选得到的HintervalH80D10D20与各组分生物量普遍表现为显著(P<0.05)或极显著水平(P<0.01)。(2)MLR模型的R2高于0.82(枝、叶除外);RF模型的R2均高于0.91,且均拥有较小rRMSE、TRE值。(3)MDI和MDA方法的变量相对重要值排序均能较好地体现LiDAR变量与生物量之间的关系,MDI在趋势性判断和阈值设定方面更具优势。[结论]LiDAR变量与组分生物量具有显著的相关性。RF拥有更好的拟合效果和泛化能力,MLR则对LiDAR和组分生物量的关系有更明确的解释能力。反演模型能较好地反映林分的现势特征,生物量被低估的现象会随着林龄的增加而逐步增多。
英文摘要:
      [Objective] To establish the inversion models of biomass components for Larix olgensis plantation using Airborne LiDAR.[Method] Compatible models were established by combining dummy variable and nonlinear seemingly unrelated regression based on the biomass data of 64 trees in 40 sample plots of L. olgensis plantation. The canopy height indices and density indices were calculated from LiDAR point cloud data. Then the models based on these biomass components from field data and LiDAR indices were built using multivariate linear regression(MLR) and random forest regression (RF). The biomass model validation was accomplished by Jackknifing technique.[Result] (1) There was a significant (P<0.05) or extremely significant (P<0.01) correlation between biomass components and Hinterval, H80, D10, and D20 which were screened using MLR. (2) The models had good estimation precision with R2>0.82 and R2>0.91 for MLR and RF, respectively. (3) The relationship between LiDAR variables and biomass could be reflected well by relative importance ranking of variables by both MDI and MDA method.[Conclusion] There is a good correlations between biomass components and LiDAR indices. RF shows stronger ability about fitting goodness and generalization, while MLR can make clearer interpretation of the relationship between LiDAR indices and biomass components. The present situation of L. olgensis plantation can be accurately reflected using inversion models, and the underestimation of component biomass will increase with the increase of stand age.
查看全文  查看/发表评论  下载PDF阅读器
关闭
京ICP备09064894号-3
版权所有:《林业科学研究》编辑部
地址:北京颐和园后中国林科院林业所 电话:010-62889680 E-mail:lykxyj@vip.163.com
技术支持:北京勤云科技发展有限公司