Inversion of aboveground biomass of Dendrocalamus giganteus by using spaceborne LiDAR and optical remote sensing data

doi:10.13466/j.cnki.lczyyj.2026.02.010

Abstract

Abstract:

Investigating the potential of combining Global Ecosystem Dynamics Investigation(GEDI)multi-beam LiDAR data(referred to as GEDI data)with Sentinel-2 optical remote sensing data(referred to as Sentinel-2 data)for estimating aboveground biomass in bamboo forests plays a crucial role in assessing the state of forest ecosystems.This estimation is essential for evaluating the condition of forest ecosystems.With Xinping Yi and Dai Autonomous County,Yuxi City,Yunnan Province as the research area and by utilizing GEDI data alongside Sentinel-2 data as primary information sources,a total of 51 661 point parameters were extracted from the GEDI data.Utilizing Kriging interpolation,continuous and comprehensive surface information was generated.In conjunction with Sentinel-2 data and 52 sample images,two modeling techniques,involving Random Forest(RF)and Gradient Boosting Regression Trees(GBRT),were employed to estimate the biomass of Dendrocalamus giganteus in Xinping County.The results indicated that:1)The four parameters provided by GEDI L2B data,namely modis_treecover,landsat_treecover,digital_elevation_mode,and modis_nonvegetated,along with the four vegetation indices calculated from Sentinel-2 Level-2A surface reflectance imagery,including DVI,ARVI,RVI,and NDVI,exhibited a high correlation with the above-ground biomass of Dendrocalamus giganteus.They can be used to estimate the above-ground biomass of Dendrocalamus giganteus at the county scale.2)In single-source modeling scenarios,GBRT demonstrates superior performance compared to RF:With exclusive use of Sentinel-2 data vegetation indices,GBRT achieves estimation accuracy metrics of the coefficient of determination with R² of 0.84 and the root mean square error(RMSE)of 11.00 t/hm²,whereas RF yields R² is 0.74 and RMSE is 14.17 t/hm².This performance trend persists when utilizing GEDI data parameters,where GBRT attains R² of 0.92 and RMSE of 7.77 t/hm²,contrasting with RF’s results of R² of 0.87 and RMSE of 9.95 t/hm².3)When combining GEDI data parameters and vegetation indices,GBRT achieves higher predictive accuracy with R² of 0.93 and RMSE of 7.47 t/hm²,whereas RF shows slightly lower accuracy with R² of 0.88 and RMSE of 9.60 t/hm².4)Based on the evaluation results of GEDI data and Sentinel-2 data,the biomass of Dendrocalamus giganteus,Xinping County is estimated to be 1.45×10⁷ t,with an average biomass of 61.91 t/hm².This demonstrates that combining GEDI data and Sentinel-2 data optical remote sensing to estimate the biomass of Dendrocalamus giganteus,Xinping County would provide a reference for multi-source remote sensing inversion in county-level data integration.

Key words: Global Ecosystem Dynamics Investigation(GEDI), Sentinel-2, machine learning, aboveground biomass, Dendrocalamus giganteus

CLC Number:

S771.8

WANG Mingxing, SHU Qingtai, XIA Cuifen, WU Zaikun, ZHOU Sai, DUAN Dandan, LI Shengjiao. Inversion of aboveground biomass of Dendrocalamus giganteus by using spaceborne LiDAR and optical remote sensing data[J]. Forest and Grassland Resources Research, 2026, (2): 102-114.

Figures/Tables 13

Fig.1

Tab.1

Tab.2

Fig.2

Tab.3

Fig.3

Tab.4

Tab.5

Fig.4

Fig.5

Tab.6

Fig.6

Fig.7

References 35

[1]	Food and Agriculture Organization of the United Nations. Global forest resources assessment 2010 main report[R]. Rome:FAO, 2010.
[2]	冯仲科, 罗旭, 石丽萍. 森林生物量研究的若干问题及完善途径[J]. 世界林业研究, 2005, 18(3):25-28.
[3]	Ebermeyer E. Die gesamte lehre der waldstreu mit rücksicht auf die chemische static des waldbaues[M]. Berlin: J springer, 1876.
[4]	郝晴, 王昌. 森林地上生物量遥感估算研究综述[J]. 植物生态学报, 2023, 47(10):1356-1374.
[5]	Fedorov N, Bikbaev I, Shirokikh P, et al. Estimation of carbon stocks of birch forests on abandoned arable lands in the Cis-Ural using unmanned aerial vehicle-mounted LiDAR camera[J]. Forests, 2023, 14(12):2392-2406. doi: 10.3390/f14122392
[6]	陈舒怀, 谭宏超. 龙竹的生物学和生态学特性[J]. 云南林业科技, 1997(4):36-39.
[7]	王玲. 竹类植物的工业利用途径综述[J]. 安徽农业科学, 2011, 39(17):10294-10296.
[8]	田芸, 安胜足, 孙照斌. 龙竹竹材的吸水膨胀特性[J]. 福建林业科技, 2007(2):148-150.
[9]	李云海, 陈丽华, 吕诗涵, 等. 云南龙竹试管无性系的建立及快繁研究[J]. 西部林业科学, 2020, 49(1):16-19.
[10]	倪惠菁, 储昊煜, 苏文会, 等. 经营强度对毛竹林土壤团聚体稳定性和碳氮磷分布的影响[J]. 应用生态学报, 2023, 34(4):928-936. doi: 10.13287/j.1001-9332.202304.002
[11]	Du Huaqiang, Cui Ruirui, Zhou Guomo, et al. The responses of Moso bamboo(Phyllostachys heterocycla var.pubescens)forest aboveground biomass to Landsat TM spectral reflectance and NDVl[J]. Aeta Eeologica Sinica, 2010, 30(5):257-263.
[12]	温雨笑, 吕杰, 马庆勋, 等. 高光谱和LiDAR联合反演森林生物量研究[J]. 测绘通报, 2022(7):38-42. doi: 10.13474/j.cnki.11-2246.2022.0200
[13]	杨可乐, 谭艳, 郭孝玉, 等. 集成多源遥感与极限梯度提升的竹林地上生物量估测[J]. 森林与环境学报, 2024, 44(4):431-440.
[14]	王柯人, 罗文秀, 舒清态, 等. 龙竹人工林的含水率分析及地上生物量回归模型构建[J]. 西南林业大学学报, 2021, 41(6):168-174.
[15]	王路君, 蔡春菊, 唐晓鹿, 等. 硬头黄竹地上生物量分配特征及模型构建[J]. 南京林业大学学报(自然科学版), 2021, 45(1):189-196. doi: 10.12302/j.issn.1000-2006.202005008
[16]	张宇, 岳祥华, 漆良华, 等. 利用异速生长关系和地统计方法估算武夷山南麓毛竹林生物量[J]. 生态学杂志, 2016, 35(7):1957-1962.
[17]	Zhang Lingjun, Zhao Yinyin, Chen Chao, et al. UAV-LiDAR integration with Sentinel-2 enhances precision in AGB estimation forbamboo forests[J]. Remote Sensing, 2024, 16(4):705-727. doi: 10.3390/rs16040705
[18]	Hemando A, Puerto L, Mola-yudego B, et al. Estimation of forest biomass components using air bome LiDAR and multispectral sensors[J]. iForest-Biogeosciences and Forestry, 2019, 12(2):207-213. doi: 10.3832/ifor2735-012
[19]	潘钰涵, 段兴武, 赵文娟. 新平县干热河谷地区农业规模化对农户生计多样化的影响研究[J]. 云南地理环境研究, 2021, 33(5):41-48.
[20]	付小勇, 孙茂盛, 杨宇明, 等. 遥感技术在德宏州大型丛生竹资源监测中的应用研究[J]. 西部林业科学, 2012, 41(4):88-92.
[21]	谢栋平, 李国元, 赵严铭, 等. 美国GEDI天基激光测高系统及其应用[J]. 国际太空, 2018(12):6.
[22]	Liang M, Duncansun L, Silva J A, et al. Quantifying aboveground biomass dynamics from charcoal degradation in Mozambique using GEDI Lidar and Landsat[J]. Remote Sensing of Environment:An Interdisciplinary Journal, 2023, 284:113367.
[23]	Xu Li, Lai Hongyan, Yu Jinge, et al. Carbon storage estimation of quercus aquifolioides based on GEDI spaceborne LiDAR data and Landsat 9 images in Shangri-La[J]. Sustainability, 2023, 15(15):11525. doi: 10.3390/su151511525
[24]	Languille F, Gaudel A, Dechoz C, et al. Sentinel-2A image quality commissioning phase final results:geometric calibration and performances[J]. Sensors,Systems,and Next-Generation Satellites XX, 2016, 10000:64-84.
[25]	Yu Xiaohui, Ge Hongli, Lu Dengsheng, et al. Comparative study on variable selection approaches in establishment of remote sensing model for forest biomass estimation[J]. Remote Sensing, 2019, 11(12):500. doi: 10.3390/rs11050500
[26]	刘晓梅, 布仁仓, 邓华卫, 等. 基于地统计学丰林自然保护区森林生物量估测及空间格局分析[J]. 生态学报, 2011, 31(16):4783-4790.
[27]	Xu Li, Shu Qingtai, Fu Huyan, et al. Estimation of quercus biomass in Shangri-La based on GEDI spaceborne LiDAR data[J]. Forests, 2023, 14(5):876. doi: 10.3390/f14050876
[28]	刘爱利, 王培法, 丁园圆. 地统计学概论[M]. 北京: 科学出版社, 2012.
[29]	Breiman L. Random forests[J]. Machine Learning. 2001, 45:5-32. doi: 10.1023/A:1010933404324
[30]	Liaw A, Wiener M. Classifification and regression by random forest[J]. R News, 2002, 2:18-22.
[31]	Li Yingchang, Li Mingyang, Liu Zhenzhen, et al. Combining kriging interpolation to improve the accuracy of forest aboveground biomass estimation using remote sensing data[J]. IEEE Access, 2020, 8:128124-128139. doi: 10.1109/Access.6287639
[32]	Du Huaqiang, Zhou Guomo, Fan Wenyi, et al. Spatial heterogeneity and carbon contribution of aboveground biomass of moso bamboo by using geostatistical theory[J]. Plant Ecology, 2010, 207:131-139. doi: 10.1007/s11258-009-9659-3
[33]	孟露. 基于地统计学的大兴安岭塔河林业局主要森林类型生物量空间分布[D]. 哈尔滨: 东北林业大学, 2017.
[34]	Luo Hongbin, Shu Qingtai, Wang Qiang, et al. Estimation of above ground biomass of rubber forest withairborne lidar and Landsat8/OLI data[J]. Northeast Forestry University. 2019, 47:56-61.
[35]	李小文, 曹春香, 常超一. 地理学第一定律与时空邻近度的提出[J]. 自然杂志, 2007(2):69-71.

名称	胸径/cm	地上生物量/(t/hm²)
最大值	12.50	132.13
最小值	5.60	23.46
平均值	9.04	66.91
标准差	1.52	27.86

参数	描述	参数	描述
cover	总冠层覆盖率	fhd_normal	叶高多样性指数
pai	植被面积指数	solar_elevation	太阳高度角
degrade_flag	卫星地理位置信息降级状态标识	modis_treecover	MODIS数据树木覆盖率
pgap_theta_aN	垂直冠层方向植被间隙	modis_nonvegetated	MODIS数据的无植被标准差百分比
digital_elevation_mode	WGS84椭球之上的数字高程模型高度	leaf_on_doy	展叶期的儒略日
leaf_off_doy	无叶期的儒略日	rv_aN	波形中植被分量积分
rg_aN	波形中地面分量的积分	sensitivity	可穿透林冠中最大树木的波形
rh100	接收波形信号开始时的离地高度	quality_flag	简化选择最有用数据的标志
landsat_treecover	Landsat数据树木覆盖度	rx_energy_aN	去除平均噪声后接收到的波形总能量

植被指数名称	计算公式	变量释义
归一化植被指数 (Normalized Difference Vegetation Index)	I_NDV=(ρ_NIR-ρ_Red)/(ρ_NIR+ρ_Red)	I_NDV为归一化植被指数;I_RGV为红绿植被指数;I_GNDV为绿色归一化植被指数;I_DV为差值植被指数;I_EV为增强植被指数;I_ARV为大气抗阻植被指数;I_VAR为大气阻力指数;I_IPV为红外百分比植被指数;I_GC为绿色叶绿素指数;I_RV为比值植被指数;ρ为某个特定波段的反射率值;NIR为近红外波段;Red为红色波段;Green为绿色波段;Blue为蓝色波段。
红绿植被指数 (Red-Green Vegetation Index)	I_RGV=(ρ_Red-ρ_Green)/(ρ_Red+ρ_Green)
绿色归一化植被指数 (Green Normalized Difference Vegetation Index)	I_GNDV=(ρ_NIR-ρ_Green)/(ρ_NIR+ρ_Green)
差值植被指数 (Difference Vegetation Index)	I_DV=ρ_NIR-ρ_Red
增强植被指数 (Enhanced Vegetation Index)	I_EV=[2.5×(ρ_NIR-ρ_Red)]/(ρ_NIR+6×ρ_Red-7.5×ρ_Blue+1)
大气阻抗植被指数 (Atmospherically Resistant Vegetation Index)	I_ARV=[ρ_NIR-(2×ρ_Red-ρ_Blue)]/[ρ_NIR+(2×ρ_Red-ρ_Blue)]
可见光大气阻抗指数 (Visible Atmospherically Resistant Index)	I_VAR=(ρ_Green-ρ_Red)/(ρ_Green+ρ_Red-ρ_Blue)
红外百分比植被指数 (Infrared Vegetation Index)	I_IPV=ρ_NIR/ρ_NIR+ρ_Red
绿色叶绿素指数 (Green Chlorophyll Index)	I_GC=ρ_NIR/ρ_Green-1
比值植被指数 (Ratio Vegetation Index)	I_RV=ρ_NIR/ρ_Red

参数名称	模型	块金值 (C₀)	偏基台值 (C)	基台值 (C₀+C)	结构比 [C/(C₀+C)]	决定系数 (R²)	残差平方和 (RSS)	变程 (A)
modis_treecover	球状模型	237.00	454.10	691.10	65.71%	0.85	24 822.00	15 300.00
landsat_treecover	指数模型	194.00	1 161.00	1 355.00	85.68%	0.81	32 700.00	6 600.00
digital_elevation_model	球状模型	34 000.00	317.00	351 900.00	90.34%	0.81	0.90	15 000.00
modis_nonvegetated	指数模型	37.90	37.91	75.81	50.01%	0.77	214.00	22 500.00

参数名称	模型	平均误差 (ME)	均方根误差 (RMSE)	标准均方误差 (MSE)	标准均方根误差 (SRMSE)	平均标准误差 (ASE)	决定系数 (R²)
modis_treecover	球状模型	0.00	6.24	0.00	0.67	7.03	0.95
landsat_treecover	指数模型	0.07	17.65	0.01	0.77	18.10	0.73
digital_elevation_mode	球状模型	0.32	40.65	0.01	0.44	42.32	0.91
modis_nonvegetated	指数模型	0.01	2.56	0.00	0.69	3.62	0.88