| 引用本文： |
-
余铸,李春干,石程远,韦立权,韦羽茜.样地数量对机载激光雷达大区域亚热带森林地上生物量估测精度的影响[J].广西科学,2026,33(3):533-544. [点击复制]
- YU Zhu,LI Chungan,SHI Chengyuan,WEI Liquan,WEI Yuxi.Influence of Number of Sample Plots on the Accuracy of Aboveground Biomass Estimation in Large-Scale Airborne LiDAR Surveys of Subtropical Forests[J].Guangxi Sciences,2026,33(3):533-544. [点击复制]
|
|
| 摘要: |
| 利用机载激光雷达(LiDAR)准确估测森林地上生物量，对于理解全球碳循环、评估森林碳汇和碳源、制定森林管理和保护策略具有重要意义。本研究通过目的抽样(典型抽样)，在面积为23.76万km2的亚热带研究区选取1 003个样地，采用重复抽样法构建各个数量级样地的数据集，分析4个森林类型(杉木林、桉树林、松树林和阔叶林)中不同数量级样地LiDAR变量和地上生物量的差异，采用多元乘幂模型，建立各个森林类型地上生物量估测的回归模型，通过50次重复抽样分析不同样地数量对4个森林类型的地上生物量估测模型表现的影响，借助半变异函数进一步探讨机载LiDAR大区域地上生物量估测中各个森林类型需要的样地数量。研究结果表明，不同数量级样地之间，LiDAR变量和地上生物量的均值十分接近，它们的变异系数的变化幅度大于其均值的变化幅度，且随着样地数量的增加均呈逐渐减小的趋势；当样地数量由30个增加至全数量时，不同森林类型地上生物量估测模型的相对均方根误差(rRMSE)均呈逐渐减小的趋势，LiDAR变量对森林参数变化的解释率(R2)均呈逐渐增大的趋势；在样地数量较少时,地上生物量估测模型的R2和rRMSE的变化幅度很大，R2均值较小、rRMSE均值较大,随着样地数量的增加，两者的变化幅度呈逐渐减小的趋势，R2均值逐渐增大、rRMSE均值逐渐减小。随着样地数量的增加，各个森林类型地上生物量估测模型的精度逐渐提高，样地数量及构造模型LiDAR变量的变动是影响模型精度的关键因素。在机载LiDAR大区域亚热带地上生物量估测应用中，杉木林、松树林、桉树林和阔叶林的最适样地数量分别为96、122、72、140个。 |
| 关键词: 森林资源 林分参数 多元乘幂模型 半变异函数 |
| DOI：10.13656/j.cnki.gxkx.20260708.008 |
| 投稿时间：2025-10-21修订日期：2025-12-08 |
| 基金项目:广西林业科技推广示范项目(桂林科研〔2022〕第8号，GL2020KT02)资助。 |
|
| Influence of Number of Sample Plots on the Accuracy of Aboveground Biomass Estimation in Large-Scale Airborne LiDAR Surveys of Subtropical Forests |
|
YU Zhu1, LI Chungan2, SHI Chengyuan1, WEI Liquan1, WEI Yuxi1
|
| (1.Guangxi Forest Inventory & Planning Institute, Nanning, Guangxi, 530011, China;2.School of Forestry, Guangxi University, Nanning, Guangxi, 530004, China) |
| Abstract: |
| Accurate estimation of forest aboveground biomass through airborne Light Detection And Ranging (LiDAR) is of great significance for understanding the global carbon cycle,assessing forest carbon sinks and carbon sources,and formulating forest management and conservation strategies.In this study,a total of 1 003 sample plots were selected via typical sampling across the subtropical study area covering 237 600 km2,and repeated sampling was adopted to construct datasets with different numbers of sample plots.The differences in LiDAR variables and aboveground biomass under different numbers of sample plots across four distinct forest types-Chinese fir,Eucalyptus,Masson pine,and broadleaf forests-were analyzed.Regression models for estimating aboveground biomass were developed for each forest type via a multivariate power model.Fifty repeated sampling was conducted to assess the performance of the aboveground biomass estimation models for different forest types under different numbers of sample plots.Additionally,semivariable function was adopted to explore the optimal number of sample plots required for accurate aboveground biomass estimation across the large study area.The results showed that the mean values of LiDAR variables and aboveground biomass were close across different numbers of sample plots.The coefficients of change for LiDAR variables exhibited a wider range than the mean values,gradually diminishing as the number of sample plots increased.As the number of sample plots increased from 30 to the maximum,the relative root-mean-square error (rRMSE) of aboveground biomass estimation models for different forest types progressively decreased,while the explanatory power (R2) of LiDAR variables in capturing forest parameter variations steadily increased.In the case where the number of sample plots is limited,the R2 and rRMSE of aboveground biomass estimation models exhibited significant variability,with a small mean R2 value and a large mean rRMSE value.However,as the number of sample plots increased,the range of variations gradually decreased,with the mean R2 value gradually increasing and the mean rRMSE value gradually decreasing.Consequently,the accuracy of aboveground biomass estimation models for each forest type improves as the number of sample plots increases.Notably,the number of sample plots and the LiDAR variables employed in model construction are key factors influencing model accuracy.When airborne LiDAR is employed for assessing aboveground biomass across a wide subtropical area,the optimal numbers of sample plots would be 96,122,72,and 140 for Chinese fir,Masson pine,Eucalyptus,and broadleaf forests,respectively. |
| Key words: forest resources stand parameters multivariate power model semivariable function |
