文献类型：期刊文献

中文题名：基于无人机激光雷达的人工林碳储量线性与非线性估测模型比较

英文题名：Comparison in linear and nonlinear estimation models of carbon storage of plantations based on UAV LiDAR

作者：陈中超[1] 刘清旺[2,3] 李春干[1] 李梅[2,3] 周相贝[1] 余铸[1] 苏凯[1]

第一作者：陈中超

机构：[1]广西大学林学院,广西南宁530004;[2]中国林业科学研究院资源信息研究所,北京100091;[3]国家林业和草原局林业遥感与信息技术重点实验室,北京100091

年份：2021

卷号：43

期号：12

起止页码：9-16

中文期刊名：北京林业大学学报

外文期刊名：Journal of Beijing Forestry University

收录：CSTPCD;;北大核心:【北大核心2020】；CSCD:【CSCD2021_2022】；

基金：国家重点研发计划(2017YFD0600904)。

语种：中文

中文关键词：碳储量;激光雷达;垂直结构参数;回归模型

外文关键词：carbon storage;LiDAR;vertical structure parameter;regression model

分类号：S757.2

摘要：【目的】森林碳储量是生态系统结构与功能的重要指标,掌握森林碳储量现状有利于森林资源管理。激光雷达能够用于监测森林资源,但是存在森林参数估测的模型多、变量不确定和缺乏林分三维结构解析意义的变量等问题,因此,需要选择合适的林分解析变量和模型。【方法】借助无人机激光雷达点云数据与样地调查数据,以内蒙古自治区赤峰市喀喇沁旗旺业甸人工林为研究对象,分别使用多元线性模型与多元乘幂模型以不同变量对林分碳储量进行估测,选出最优模型并进行精度评价。【结果】研究表明:(1)模型方法而言,非线性模型的检验效果优于线性模型的检验效果:非线性模型(R2为0.66~0.86,rRMSE为23.51%~9.91%),线性模型(R2为0.52~0.85,rRMSE为27.70%~12.38%)。(2)模型使用平均高、郁闭度为基础变量,以穷举法筛选出来的变量组合,估算森林参数得出最佳模型,其中非线性模型以激光点云平均高、郁闭度、高度变动系数和叶面积变动系数的估算精度最高(R2=0.86,rRMSE=9.91%)。【结论】通过激光雷达估测人工林碳储量时,加入垂直结构变量可以提高模型拟合效果,非线性模型比线性模型更适合人工林碳储量的估测。
[Objective]Forest carbon storage is an important indicator of the composition and function of ecosystem.It will be benefit to forest resource management by investigating the state of forest carbon storage.The LiDAR can be used to monitor forest resources,however,there are many problems in forest parameter estimation,such as multi variable model,uncertainty and lack of variables with analytical significance of stand three-dimensional structure.Therefore,it is necessary to select appropriate stand analysis variables and models.[Method]This paper uses UAV LiDAR point cloud and sample plot data to analyze the plantation in Wangyedian Forest Farm of Kalaqin Banner,Chifeng City,Inner Mongolia of northern China.The multiple linear model and the multiple power models were used to estimate the forest carbon storage using different variables,and select the optimal model.[Result](1)The nonlinear models(R2 ranged in 0.66?0.86,rRMSE ranged in 23.51%?9.91%)were better than linear models(R2 ranged in 0.52?0.85,rRMSE ranged in 27.70%?12.38%).(2)The mean height of point cloud and canopy cover were used as basic variables.The combinations of different variables were emulated to select the best model of forest parameters.The nonlinear model based on average height of the laser point cloud,canopy cover,height variation coefficient and leaf area variation coefficient(R2=0.86,rRMSE=9.91%)had the highest estimating accuracy.[Conclusion]The vertical structure variables could improve the estimating accuracy of carbon storage of plantation using LiDAR.The non-linear model is more suitable for the estimation of carbon storage of plantation.